1
0
Fork 0
mirror of https://gitlab.com/sortix/sortix.git synced 2023-02-13 20:55:38 -05:00
sortix--sortix/kernel/process.cpp

1234 lines
32 KiB
C++
Raw Normal View History

Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
/*******************************************************************************
Copyright(C) Jonas 'Sortie' Termansen 2011, 2012, 2013.
This file is part of Sortix.
Sortix is free software: you can redistribute it and/or modify it under the
terms of the GNU General Public License as published by the Free Software
Foundation, either version 3 of the License, or (at your option) any later
version.
Sortix is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
details.
You should have received a copy of the GNU General Public License along with
Sortix. If not, see <http://www.gnu.org/licenses/>.
process.cpp
A named collection of threads.
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
*******************************************************************************/
#include <assert.h>
#include <errno.h>
#include <stdint.h>
2013-08-19 20:23:53 -04:00
#include <stdlib.h>
#include <string.h>
#include <sortix/clock.h>
#include <sortix/fcntl.h>
#include <sortix/fork.h>
#include <sortix/mman.h>
2013-08-30 17:55:09 -04:00
#include <sortix/resource.h>
#include <sortix/signal.h>
#include <sortix/stat.h>
#include <sortix/unistd.h>
#include <sortix/wait.h>
#include <sortix/kernel/copy.h>
#include <sortix/kernel/descriptor.h>
#include <sortix/kernel/dtable.h>
#include <sortix/kernel/ioctx.h>
2013-10-26 20:42:10 -04:00
#include <sortix/kernel/kernel.h>
#include <sortix/kernel/kthread.h>
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
#include <sortix/kernel/memorymanagement.h>
#include <sortix/kernel/mtable.h>
#include <sortix/kernel/process.h>
#include <sortix/kernel/refcount.h>
2013-01-09 04:47:22 -05:00
#include <sortix/kernel/scheduler.h>
#include <sortix/kernel/sortedlist.h>
#include <sortix/kernel/string.h>
2013-05-22 16:11:29 -04:00
#include <sortix/kernel/symbol.h>
#include <sortix/kernel/syscall.h>
#include <sortix/kernel/thread.h>
#include <sortix/kernel/time.h>
#include <sortix/kernel/worker.h>
2013-01-08 18:41:35 -05:00
2011-08-27 14:57:39 -04:00
#include "elf.h"
#include "initrd.h"
namespace Sortix {
Process::Process()
{
string_table = NULL;
string_table_length = 0;
symbol_table = NULL;
symbol_table_length = 0;
addrspace = 0;
segments = NULL;
parent = NULL;
prevsibling = NULL;
nextsibling = NULL;
firstchild = NULL;
zombiechild = NULL;
2013-06-11 19:02:01 -04:00
group = NULL;
groupprev = NULL;
groupnext = NULL;
groupfirst = NULL;
program_image_path = NULL;
parentlock = KTHREAD_MUTEX_INITIALIZER;
childlock = KTHREAD_MUTEX_INITIALIZER;
2013-06-11 19:02:01 -04:00
groupparentlock = KTHREAD_MUTEX_INITIALIZER;
groupchildlock = KTHREAD_MUTEX_INITIALIZER;
groupchildleft = KTHREAD_COND_INITIALIZER;
zombiecond = KTHREAD_COND_INITIALIZER;
zombiewaiting = 0;
iszombie = false;
nozombify = false;
2013-06-11 19:02:01 -04:00
grouplimbo = false;
firstthread = NULL;
threadlock = KTHREAD_MUTEX_INITIALIZER;
ptrlock = KTHREAD_MUTEX_INITIALIZER;
idlock = KTHREAD_MUTEX_INITIALIZER;
user_timers_lock = KTHREAD_MUTEX_INITIALIZER;
2013-08-19 20:23:53 -04:00
segments = NULL;
segments_used = 0;
segments_length = 0;
segment_lock = KTHREAD_MUTEX_INITIALIZER;
exitstatus = -1;
pid = AllocatePID();
uid = euid = 0;
gid = egid = 0;
umask = 0022;
2013-08-30 11:35:30 -04:00
nicelock = KTHREAD_MUTEX_INITIALIZER;
nice = 0;
2013-08-30 17:55:09 -04:00
resource_limits_lock = KTHREAD_MUTEX_INITIALIZER;
for ( size_t i = 0; i < RLIMIT_NUM_DECLARED; i++ )
{
resource_limits[i].rlim_cur = RLIM_INFINITY;
resource_limits[i].rlim_max = RLIM_INFINITY;
}
Time::InitializeProcessClocks(this);
alarm_timer.Attach(Time::GetClock(CLOCK_MONOTONIC));
Put(this);
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
Process::~Process()
{
delete[] string_table;
delete[] symbol_table;
if ( alarm_timer.IsAttached() )
alarm_timer.Detach();
if ( program_image_path )
delete[] program_image_path;
assert(!zombiechild);
assert(!firstchild);
assert(!addrspace);
assert(!segments);
assert(!dtable);
assert(!mtable);
assert(!cwd);
assert(!root);
Remove(this);
}
void Process::BootstrapTables(Ref<DescriptorTable> dtable, Ref<MountTable> mtable)
{
ScopedLock lock(&ptrlock);
assert(!this->dtable);
assert(!this->mtable);
this->dtable = dtable;
this->mtable = mtable;
}
void Process::BootstrapDirectories(Ref<Descriptor> root)
{
ScopedLock lock(&ptrlock);
assert(!this->root);
assert(!this->cwd);
this->root = root;
this->cwd = root;
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
void Process__OnLastThreadExit(void* user);
void Process::OnThreadDestruction(Thread* thread)
{
assert(thread->process == this);
kthread_mutex_lock(&threadlock);
if ( thread->prevsibling )
thread->prevsibling->nextsibling = thread->nextsibling;
if ( thread->nextsibling )
thread->nextsibling->prevsibling = thread->prevsibling;
if ( thread == firstthread )
firstthread = thread->nextsibling;
if ( firstthread )
firstthread->prevsibling = NULL;
thread->prevsibling = thread->nextsibling = NULL;
bool threadsleft = firstthread;
kthread_mutex_unlock(&threadlock);
// We are called from the threads destructor, let it finish before we
// we handle the situation by killing ourselves.
if ( !threadsleft )
ScheduleDeath();
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
void Process::ScheduleDeath()
{
// All our threads must have exited at this point.
assert(!firstthread);
Worker::Schedule(Process__OnLastThreadExit, this);
}
2013-05-11 19:24:42 -04:00
// Useful for killing a partially constructed process without waiting for
// it to die and garbage collect its zombie. It is not safe to access this
// process after this call as another thread may garbage collect it.
void Process::AbortConstruction()
{
nozombify = true;
ScheduleDeath();
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
void Process__OnLastThreadExit(void* user)
{
return ((Process*) user)->OnLastThreadExit();
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
void Process::OnLastThreadExit()
{
LastPrayer();
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
static void SwitchCurrentAddrspace(addr_t addrspace, void* user)
{
((Thread*) user)->SwitchAddressSpace(addrspace);
}
2013-10-13 17:56:58 -04:00
void Process::DeleteTimers()
{
for ( timer_t i = 0; i < PROCESS_TIMER_NUM_MAX; i++ )
{
if ( user_timers[i].timer.IsAttached() )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
user_timers[i].timer.Cancel();
user_timers[i].timer.Detach();
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
}
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
void Process::LastPrayer()
{
assert(this);
// This must never be called twice.
assert(!iszombie);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// This must be called from a thread using another address space as the
// address space of this process is about to be destroyed.
Thread* curthread = CurrentThread();
assert(curthread->process != this);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// This can't be called if the process is still alive.
assert(!firstthread);
// Disarm and detach all the timers in the process.
DeleteTimers();
if ( alarm_timer.IsAttached() )
{
alarm_timer.Cancel();
alarm_timer.Detach();
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
// We need to temporarily reload the correct addrese space of the dying
// process such that we can unmap and free its memory.
addr_t prevaddrspace = curthread->SwitchAddressSpace(addrspace);
ResetAddressSpace();
if ( dtable ) dtable.Reset();
if ( cwd ) cwd.Reset();
if ( root ) root.Reset();
if ( mtable ) mtable.Reset();
// Destroy the address space and safely switch to the replacement
// address space before things get dangerous.
Memory::DestroyAddressSpace(prevaddrspace,
SwitchCurrentAddrspace,
curthread);
addrspace = 0;
// Unload the process symbol and string tables.
delete[] symbol_table; symbol_table = NULL;
delete[] string_table; string_table = NULL;
// Init is nice and will gladly raise our orphaned children and zombies.
Process* init = Scheduler::GetInitProcess();
assert(init);
kthread_mutex_lock(&childlock);
while ( firstchild )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
ScopedLock firstchildlock(&firstchild->parentlock);
ScopedLock initlock(&init->childlock);
Process* process = firstchild;
firstchild = process->nextsibling;
process->parent = init;
process->prevsibling = NULL;
process->nextsibling = init->firstchild;
if ( init->firstchild )
init->firstchild->prevsibling = process;
init->firstchild = process;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
// Since we have no more children (they are with init now), we don't
// have to worry about new zombie processes showing up, so just collect
// those that are left. Then we satisfiy the invariant !zombiechild that
// applies on process termination.
bool hadzombies = zombiechild;
while ( zombiechild )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
ScopedLock zombiechildlock(&zombiechild->parentlock);
ScopedLock initlock(&init->childlock);
Process* zombie = zombiechild;
zombiechild = zombie->nextsibling;
zombie->prevsibling = NULL;
zombie->nextsibling = init->zombiechild;
if ( init->zombiechild )
init->zombiechild->prevsibling = zombie;
init->zombiechild = zombie;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
kthread_mutex_unlock(&childlock);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
if ( hadzombies )
init->NotifyNewZombies();
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
iszombie = true;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
bool zombify = !nozombify;
// Remove ourself from our process group.
2013-06-11 19:02:01 -04:00
kthread_mutex_lock(&groupchildlock);
if ( group )
group->NotifyMemberExit(this);
2013-06-11 19:02:01 -04:00
kthread_mutex_unlock(&groupchildlock);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// This class instance will be destroyed by our parent process when it
// has received and acknowledged our death.
kthread_mutex_lock(&parentlock);
if ( parent )
parent->NotifyChildExit(this, zombify);
kthread_mutex_unlock(&parentlock);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// If nobody is waiting for us, then simply commit suicide.
if ( !zombify )
{
kthread_mutex_lock(&groupparentlock);
bool in_limbo = groupfirst || (grouplimbo = true);
kthread_mutex_unlock(&groupparentlock);
if ( !in_limbo )
delete this;
}
}
void Process::ResetAddressSpace()
{
2013-08-19 20:23:53 -04:00
ScopedLock lock(&segment_lock);
assert(Memory::GetAddressSpace() == addrspace);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
2013-08-19 20:23:53 -04:00
for ( size_t i = 0; i < segments_used; i++ )
Memory::UnmapRange(segments[i].addr, segments[i].size);
Memory::Flush();
segments_used = segments_length = 0;
free(segments);
segments = NULL;
}
void Process::NotifyMemberExit(Process* child)
{
assert(child->group == this);
kthread_mutex_lock(&groupparentlock);
if ( child->groupprev )
child->groupprev->groupnext = child->groupnext;
else
groupfirst = child->groupnext;
if ( child->groupnext )
child->groupnext->groupprev = child->groupprev;
kthread_cond_signal(&groupchildleft);
kthread_mutex_unlock(&groupparentlock);
child->group = NULL;
NotifyLeftProcessGroup();
}
void Process::NotifyLeftProcessGroup()
{
ScopedLock parentlock(&groupparentlock);
if ( !grouplimbo || groupfirst )
return;
grouplimbo = false;
delete this;
}
void Process::NotifyChildExit(Process* child, bool zombify)
{
kthread_mutex_lock(&childlock);
if ( child->prevsibling )
child->prevsibling->nextsibling = child->nextsibling;
if ( child->nextsibling )
child->nextsibling->prevsibling = child->prevsibling;
if ( firstchild == child )
firstchild = child->nextsibling;
if ( firstchild )
firstchild->prevsibling = NULL;
if ( zombify )
{
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
if ( zombiechild )
zombiechild->prevsibling = child;
child->prevsibling = NULL;
child->nextsibling = zombiechild;
zombiechild = child;
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
kthread_mutex_unlock(&childlock);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
if ( zombify )
NotifyNewZombies();
}
void Process::NotifyNewZombies()
{
ScopedLock lock(&childlock);
// TODO: Send SIGCHLD here?
if ( zombiewaiting )
kthread_cond_broadcast(&zombiecond);
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
pid_t Process::Wait(pid_t thepid, int* user_status, int options)
{
// TODO: Process groups are not supported yet.
if ( thepid < -1 || thepid == 0 ) { errno = ENOSYS; return -1; }
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
ScopedLock lock(&childlock);
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// A process can only wait if it has children.
if ( !firstchild && !zombiechild )
return errno = ECHILD, -1;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// Processes can only wait for their own children to exit.
if ( 0 < thepid )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
// TODO: This is a slow but multithread safe way to verify that the
// target process has the correct parent.
bool found = false;
for ( Process* p = firstchild; !found && p; p = p->nextsibling )
if ( p->pid == thepid )
found = true;
for ( Process* p = zombiechild; !found && p; p = p->nextsibling )
if ( p->pid == thepid )
found = true;
if ( !found )
return errno = ECHILD, -1;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
Process* zombie = NULL;
while ( !zombie )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
for ( zombie = zombiechild; zombie; zombie = zombie->nextsibling )
if ( thepid == -1 || thepid == zombie->pid )
break;
if ( zombie )
break;
if ( options & WNOHANG )
return 0;
zombiewaiting++;
unsigned long r = kthread_cond_wait_signal(&zombiecond, &childlock);
zombiewaiting--;
if ( !r )
return errno = EINTR, -1;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
2013-06-11 19:02:01 -04:00
// Remove from the list of zombies.
if ( zombie->prevsibling )
zombie->prevsibling->nextsibling = zombie->nextsibling;
if ( zombie->nextsibling )
zombie->nextsibling->prevsibling = zombie->prevsibling;
if ( zombiechild == zombie )
zombiechild = zombie->nextsibling;
if ( zombiechild )
zombiechild->prevsibling = NULL;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
thepid = zombie->pid;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// It is safe to access these clocks directly as the child process is no
// longer running at this point and the values are nicely frozen.
child_execute_clock.Advance(zombie->child_execute_clock.current_time);
child_system_clock.Advance(zombie->child_system_clock.current_time);
2011-08-27 14:57:39 -04:00
int status = zombie->exitstatus;
if ( status < 0 )
status = W_EXITCODE(128 + SIGKILL, SIGKILL);
kthread_mutex_lock(&zombie->groupparentlock);
bool in_limbo = zombie->groupfirst || (zombie->grouplimbo = true);
kthread_mutex_unlock(&zombie->groupparentlock);
// And so, the process was fully deleted.
if ( !in_limbo )
delete zombie;
if ( user_status && !CopyToUser(user_status, &status, sizeof(status)) )
return -1;
return thepid;
}
static pid_t sys_waitpid(pid_t pid, int* user_status, int options)
{
return CurrentProcess()->Wait(pid, user_status, options);
}
void Process::Exit(int status)
{
ScopedLock lock(&threadlock);
// Status codes can only contain 8 bits according to ISO C and POSIX.
if ( exitstatus == -1 )
exitstatus = W_EXITCODE(status & 0xFF, 0);
// Broadcast SIGKILL to all our threads which will begin our long path
// of process termination. We simply can't stop the threads as they may
// be running in kernel mode doing dangerous stuff. This thread will be
// destroyed by SIGKILL once the system call returns.
for ( Thread* t = firstthread; t; t = t->nextsibling )
t->DeliverSignal(SIGKILL);
}
2013-09-23 15:58:46 -04:00
static int sys_exit(int status)
{
CurrentProcess()->Exit(status);
return 0;
}
bool Process::DeliverSignal(int signum)
{
// TODO: How to handle signals that kill the process?
if ( firstthread )
return firstthread->DeliverSignal(signum);
return errno = EINIT, false;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
bool Process::DeliverGroupSignal(int signum)
{
ScopedLock lock(&groupparentlock);
if ( !groupfirst )
return errno = ESRCH, false;
for ( Process* iter = groupfirst; iter; iter = iter->groupnext )
iter->DeliverSignal(signum);
return true;
}
void Process::AddChildProcess(Process* child)
{
ScopedLock mylock(&childlock);
ScopedLock itslock(&child->parentlock);
assert(!child->parent);
assert(!child->nextsibling);
assert(!child->prevsibling);
child->parent = this;
child->nextsibling = firstchild;
child->prevsibling = NULL;
if ( firstchild )
firstchild->prevsibling = child;
firstchild = child;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
Ref<MountTable> Process::GetMTable()
{
ScopedLock lock(&ptrlock);
assert(mtable);
return mtable;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
Ref<DescriptorTable> Process::GetDTable()
{
ScopedLock lock(&ptrlock);
assert(dtable);
return dtable;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
Ref<Descriptor> Process::GetRoot()
{
ScopedLock lock(&ptrlock);
assert(root);
return root;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
Ref<Descriptor> Process::GetCWD()
{
ScopedLock lock(&ptrlock);
assert(cwd);
return cwd;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
void Process::SetRoot(Ref<Descriptor> newroot)
{
ScopedLock lock(&ptrlock);
assert(newroot);
root = newroot;
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
void Process::SetCWD(Ref<Descriptor> newcwd)
{
ScopedLock lock(&ptrlock);
assert(newcwd);
cwd = newcwd;
}
Ref<Descriptor> Process::GetDescriptor(int fd)
{
ScopedLock lock(&ptrlock);
assert(dtable);
return dtable->Get(fd);
}
Process* Process::Fork()
{
assert(CurrentProcess() == this);
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
2013-06-11 19:02:01 -04:00
// TODO: This adds the new process to the process table, but it's not ready
// and functions that access this new process will be surprised that
// it's not fully constructed and really bad things will happen.
Process* clone = new Process;
if ( !clone )
return NULL;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
2013-08-19 20:23:53 -04:00
struct segment* clone_segments = NULL;
2013-01-12 20:37:14 -05:00
// Fork the segment list.
if ( segments )
{
2013-08-19 20:23:53 -04:00
size_t segments_size = sizeof(struct segment) * segments_used;
if ( !(clone_segments = (struct segment*) malloc(segments_size)) )
{
delete clone;
return NULL;
}
memcpy(clone_segments, segments, segments_size);
}
2013-03-19 18:17:36 -04:00
// Fork address-space here and copy memory.
clone->addrspace = Memory::Fork();
if ( !clone->addrspace )
{
2013-08-19 20:23:53 -04:00
free(clone_segments);
delete clone;
return NULL;
}
2013-05-22 16:11:29 -04:00
// Now it's too late to clean up here, if anything goes wrong, we simply
// ask the process to commit suicide before it goes live.
2013-08-19 20:23:53 -04:00
clone->segments = clone_segments;
clone->segments_used = segments_used;
clone->segments_length = segments_used;
2013-05-22 16:11:29 -04:00
// Remember the relation to the child process.
AddChildProcess(clone);
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
2013-06-11 19:02:01 -04:00
// Add the new process to the current process group.
kthread_mutex_lock(&groupchildlock);
kthread_mutex_lock(&group->groupparentlock);
clone->group = group;
clone->groupprev = NULL;
if ( (clone->groupnext = group->groupfirst) )
group->groupfirst->groupprev = clone;
group->groupfirst = clone;
kthread_mutex_unlock(&group->groupparentlock);
kthread_mutex_unlock(&groupchildlock);
// Initialize everything that is safe and can't fail.
2013-08-30 17:55:09 -04:00
kthread_mutex_lock(&resource_limits_lock);
for ( size_t i = 0; i < RLIMIT_NUM_DECLARED; i++ )
clone->resource_limits[i] = resource_limits[i];
kthread_mutex_unlock(&resource_limits_lock);
2013-08-30 11:35:30 -04:00
kthread_mutex_lock(&nicelock);
clone->nice = nice;
kthread_mutex_unlock(&nicelock);
kthread_mutex_lock(&ptrlock);
clone->root = root;
clone->cwd = cwd;
kthread_mutex_unlock(&ptrlock);
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
// Initialize things that can fail and abort if needed.
bool failure = false;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
kthread_mutex_lock(&ptrlock);
if ( !(clone->dtable = dtable->Fork()) )
failure = true;
//if ( !(clone->mtable = mtable->Fork()) )
// failure = true;
clone->mtable = mtable;
kthread_mutex_unlock(&ptrlock);
2011-11-20 09:58:42 -05:00
kthread_mutex_lock(&idlock);
clone->uid = uid;
clone->gid = gid;
clone->euid = euid;
clone->egid = egid;
clone->umask = umask;
kthread_mutex_unlock(&idlock);
2013-05-22 16:11:29 -04:00
if ( !(clone->program_image_path = String::Clone(program_image_path)) )
failure = false;
2013-05-11 19:24:42 -04:00
if ( string_table && (clone->string_table = new char[string_table_length]) )
{
memcpy(clone->string_table, string_table, string_table_length);
clone->string_table_length = string_table_length;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
}
if ( clone->string_table && symbol_table &&
(clone->symbol_table = new Symbol[symbol_table_length]) )
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
{
for ( size_t i = 0; i < symbol_table_length; i++ )
{
clone->symbol_table[i].address = symbol_table[i].address;
clone->symbol_table[i].size = symbol_table[i].size;
clone->symbol_table[i].name =
(const char*)((uintptr_t) symbol_table[i].name -
(uintptr_t) string_table +
(uintptr_t) clone->string_table);
}
clone->symbol_table_length = symbol_table_length;
}
2013-03-19 18:17:36 -04:00
if ( pid == 1)
assert(dtable->Get(1));
2013-03-19 18:17:36 -04:00
if ( pid == 1)
assert(clone->dtable->Get(1));
// If the proces creation failed, ask the process to commit suicide and
// not become a zombie, as we don't wait for it to exit. It will clean
// up all the above resources and delete itself.
if ( failure )
{
clone->AbortConstruction();
return NULL;
}
2011-11-09 17:18:26 -05:00
return clone;
}
void Process::ResetForExecute()
{
string_table_length = 0;
symbol_table_length = 0;
delete[] string_table; string_table = NULL;
delete[] symbol_table; symbol_table = NULL;
2011-11-09 17:18:26 -05:00
DeleteTimers();
ResetAddressSpace();
}
int Process::Execute(const char* programname, const uint8_t* program,
size_t programsize, int argc, const char* const* argv,
int envc, const char* const* envp,
CPU::InterruptRegisters* regs)
{
assert(CurrentProcess() == this);
char* programname_clone = String::Clone(programname);
if ( !programname_clone )
return -1;
addr_t entry = ELF::Construct(CurrentProcess(), program, programsize);
if ( !entry ) { delete[] programname_clone; return -1; }
2011-11-09 17:18:26 -05:00
delete[] program_image_path;
program_image_path = programname_clone; programname_clone = NULL;
2013-08-22 10:08:08 -04:00
uintptr_t userspace_addr;
size_t userspace_size;
Memory::GetUserVirtualArea(&userspace_addr, &userspace_size);
const size_t DEFAULT_STACK_SIZE = 512UL * 1024UL;
void* const PREFERRED_STACK_LOCATION =
(void*) (userspace_addr + userspace_size - DEFAULT_STACK_SIZE);
const int STACK_PROTECTION = PROT_READ | PROT_WRITE | PROT_KREAD |
PROT_KWRITE | PROT_FORK;
// Attempt to allocate a stack for the new process.
kthread_mutex_lock(&segment_lock);
struct segment stack_segment;
if ( !PlaceSegment(&stack_segment, this, PREFERRED_STACK_LOCATION, DEFAULT_STACK_SIZE, 0) ||
!Memory::MapMemory(this, stack_segment.addr, stack_segment.size, stack_segment.prot = STACK_PROTECTION) )
{
kthread_mutex_unlock(&segment_lock);
ResetForExecute();
return errno = ENOMEM, -1;
}
kthread_mutex_unlock(&segment_lock);
2013-08-22 10:08:08 -04:00
addr_t stackpos = stack_segment.addr + stack_segment.size;
// Alright, move argv onto the new stack! First figure out exactly how
// big argv actually is.
addr_t argvpos = stackpos - sizeof(char*) * (argc+1);
char** stackargv = (char**) argvpos;
size_t argvsize = 0;
for ( int i = 0; i < argc; i++ )
{
size_t len = strlen(argv[i]) + 1;
argvsize += len;
char* dest = ((char*) argvpos) - argvsize;
stackargv[i] = dest;
memcpy(dest, argv[i], len);
}
stackargv[argc] = NULL;
if ( argvsize % 16UL )
argvsize += 16 - argvsize % 16UL;
// And then move envp onto the stack.
addr_t envppos = argvpos - argvsize - sizeof(char*) * (envc+1);
char** stackenvp = (char**) envppos;
size_t envpsize = 0;
for ( int i = 0; i < envc; i++ )
{
size_t len = strlen(envp[i]) + 1;
envpsize += len;
char* dest = ((char*) envppos) - envpsize;
stackenvp[i] = dest;
memcpy(dest, envp[i], len);
}
stackenvp[envc] = NULL;
2011-11-09 17:18:26 -05:00
if ( envpsize % 16UL )
envpsize += 16 - envpsize % 16UL;
stackpos = envppos - envpsize;
dtable->OnExecute();
ExecuteCPU(argc, stackargv, envc, stackenvp, stackpos, entry, regs);
2011-11-09 17:18:26 -05:00
return 0;
}
2011-11-09 17:18:26 -05:00
// TODO. This is a hack. Please remove this when execve is moved to another
// file/class, it doesn't belong here, it's a program loader ffs!
Ref<Descriptor> Process::Open(ioctx_t* ctx, const char* path, int flags, mode_t mode)
{
// TODO: Locking the root/cwd pointers. How should that be arranged?
Ref<Descriptor> dir = path[0] == '/' ? root : cwd;
return dir->open(ctx, path, flags, mode);
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
2013-09-23 15:58:46 -04:00
static
int sys_execve(const char* _filename, char* const _argv[], char* const _envp[])
{
char* filename;
int argc;
int envc;
char** argv;
char** envp;
ioctx_t ctx;
Ref<Descriptor> desc;
struct stat st;
size_t sofar;
size_t count;
uint8_t* buffer;
int result = -1;
Process* process = CurrentProcess();
CPU::InterruptRegisters regs;
memset(&regs, 0, sizeof(regs));
filename = String::Clone(_filename);
if ( !filename ) { goto cleanup_done; }
for ( argc = 0; _argv && _argv[argc]; argc++ );
for ( envc = 0; _envp && _envp[envc]; envc++ );
argv = new char*[argc+1];
if ( !argv ) { goto cleanup_filename; }
memset(argv, 0, sizeof(char*) * (argc+1));
for ( int i = 0; i < argc; i++ )
{
argv[i] = String::Clone(_argv[i]);
if ( !argv[i] ) { goto cleanup_argv; }
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
}
envp = new char*[envc+1];
if ( !envp ) { goto cleanup_argv; }
envc = envc;
memset(envp, 0, sizeof(char*) * (envc+1));
for ( int i = 0; i < envc; i++ )
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
{
envp[i] = String::Clone(_envp[i]);
if ( !envp[i] ) { goto cleanup_envp; }
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
SetupKernelIOCtx(&ctx);
// TODO: Somehow mark the executable as busy and don't permit writes?
desc = process->Open(&ctx, filename, O_READ | O_WRITE, 0);
if ( !desc ) { goto cleanup_envp; }
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
if ( desc->stat(&ctx, &st) ) { goto cleanup_desc; }
if ( st.st_size < 0 ) { errno = EINVAL; goto cleanup_desc; }
if ( SIZE_MAX < (uintmax_t) st.st_size ) { errno = ERANGE; goto cleanup_desc; }
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
count = (size_t) st.st_size;
buffer = new uint8_t[count];
if ( !buffer ) { goto cleanup_desc; }
sofar = 0;
while ( sofar < count )
{
ssize_t bytesread = desc->read(&ctx, buffer + sofar, count - sofar);
if ( bytesread < 0 ) { goto cleanup_buffer; }
if ( bytesread == 0 ) { errno = EEOF; goto cleanup_buffer; }
sofar += bytesread;
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
result = process->Execute(filename, buffer, count, argc, argv, envc,
envp, &regs);
cleanup_buffer:
delete[] buffer;
cleanup_desc:
desc.Reset();
cleanup_envp:
for ( int i = 0; i < envc; i++)
delete[] envp[i];
delete[] envp;
cleanup_argv:
for ( int i = 0; i < argc; i++)
delete[] argv[i];
delete[] argv;
cleanup_filename:
delete[] filename;
cleanup_done:
if ( !result ) { CPU::LoadRegisters(&regs); }
return result;
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
static pid_t sys_tfork(int flags, tforkregs_t* user_regs)
{
tforkregs_t regs;
if ( !CopyFromUser(&regs, user_regs, sizeof(regs)) )
return -1;
if ( Signal::IsPending() )
return errno = EINTR, -1;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
// TODO: Properly support tfork(2).
if ( flags != SFFORK )
return errno = ENOSYS, -1;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
CPU::InterruptRegisters cpuregs;
InitializeThreadRegisters(&cpuregs, &regs);
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
// TODO: Is it a hack to create a new kernel stack here?
Thread* curthread = CurrentThread();
uint8_t* newkernelstack = new uint8_t[curthread->kernelstacksize];
if ( !newkernelstack )
return -1;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
Process* clone = CurrentProcess()->Fork();
if ( !clone ) { delete[] newkernelstack; return -1; }
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
// If the thread could not be created, make the process commit suicide
// in a manner such that we don't wait for its zombie.
Thread* thread = CreateKernelThread(clone, &cpuregs);
if ( !thread )
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
{
clone->AbortConstruction();
return -1;
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
}
thread->kernelstackpos = (addr_t) newkernelstack;
thread->kernelstacksize = curthread->kernelstacksize;
thread->kernelstackmalloced = true;
thread->sighandler = curthread->sighandler;
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
StartKernelThread(thread);
return clone->pid;
}
Multithreaded kernel and improvement of signal handling. Pardon the big ass-commit, this took months to develop and debug and the refactoring got so far that a clean merge became impossible. The good news is that this commit does quite a bit of cleaning up and generally improves the kernel quality. This makes the kernel fully pre-emptive and multithreaded. This was done by rewriting the interrupt code, the scheduler, introducing new threading primitives, and rewriting large parts of the kernel. During the past few commits the kernel has had its device drivers thread secured; this commit thread secures large parts of the core kernel. There still remains some parts of the kernel that is _not_ thread secured, but this is not a problem at this point. Each user-space thread has an associated kernel stack that it uses when it goes into kernel mode. This stack is by default 8 KiB since that value works for me and is also used by Linux. Strange things tends to happen on x86 in case of a stack overflow - there is no ideal way to catch such a situation right now. The system call conventions were changed, too. The %edx register is now used to provide the errno value of the call, instead of the kernel writing it into a registered global variable. The system call code has also been updated to better reflect the native calling conventions: not all registers have to be preserved. This makes system calls faster and simplifies the assembly. In the kernel, there is no longer the event.h header or the hacky method of 'resuming system calls' that closely resembles cooperative multitasking. If a system call wants to block, it should just block. The signal handling was also improved significantly. At this point, signals cannot interrupt kernel threads (but can always interrupt user-space threads if enabled), which introduces some problems with how a SIGINT could interrupt a blocking read, for instance. This commit introduces and uses a number of new primitives such as kthread_lock_mutex_signal() that attempts to get the lock but fails if a signal is pending. In this manner, the kernel is safer as kernel threads cannot be shut down inconveniently, but in return for complexity as blocking operations must check they if they should fail. Process exiting has also been refactored significantly. The _exit(2) system call sets the exit code and sends SIGKILL to all the threads in the process. Once all the threads have cleaned themselves up and exited, a worker thread calls the process's LastPrayer() method that unmaps memory, deletes the address space, notifies the parent, etc. This provides a very robust way to terminate processes as even half-constructed processes (during a failing fork for instance) can be gracefully terminated. I have introduced a number of kernel threads to help avoid threading problems and simplify kernel design. For instance, there is now a functional generic kernel worker thread that any kernel thread can schedule jobs for. Interrupt handlers run with interrupts off (hence they cannot call kthread_ functions as it may deadlock the system if another thread holds the lock) therefore they cannot use the standard kernel worker threads. Instead, they use a special purpose interrupt worker thread that works much like the generic one expect that interrupt handlers can safely queue work with interrupts off. Note that this also means that interrupt handlers cannot allocate memory or print to the kernel log/screen as such mechanisms uses locks. I'll introduce a lock free algorithm for such cases later on. The boot process has also changed. The original kernel init thread in kernel.cpp creates a new bootstrap thread and becomes the system idle thread. Note that pid=0 now means the kernel, as there is no longer a system idle process. The bootstrap thread launches all the kernel worker threads and then creates a new process and loads /bin/init into it and then creates a thread in pid=1, which starts the system. The bootstrap thread then quietly waits for pid=1 to exit after which it shuts down/reboots/panics the system. In general, the introduction of race conditions and dead locks have forced me to revise a lot of the design and make sure it was thread secure. Since early parts of the kernel was quite hacky, I had to refactor such code. So it seems that the risk of dead locks forces me to write better code. Note that a real preemptive multithreaded kernel simplifies the construction of blocking system calls. My hope is that this will trigger a clean up of the filesystem code that current is almost beyond repair. Almost all of the kernel was modified during this refactoring. To the extent possible, these changes have been backported to older non-multithreaded kernel, but many changes were tightly coupled and went into this commit. Of interest is the implementation of the kthread_ api based on the design of pthreads; this library allows easy synchronization mechanisms and includes C++-style scoped locks. This commit also introduces new worker threads and tested mechanisms for interrupt handlers to schedule work in a kernel worker thread. A lot of code have been rewritten from scratch and has become a lot more stable and correct. Share and enjoy!
2012-08-01 11:30:34 -04:00
2013-09-23 15:58:46 -04:00
static pid_t sys_getpid()
{
return CurrentProcess()->pid;
}
pid_t Process::GetParentProcessId()
{
ScopedLock lock(&parentlock);
if( !parent )
return 0;
return parent->pid;
}
2013-09-23 15:58:46 -04:00
static pid_t sys_getppid()
{
return CurrentProcess()->GetParentProcessId();
}
2013-06-11 19:02:01 -04:00
static pid_t sys_getpgid(pid_t pid)
{
Process* process = !pid ? CurrentProcess() : Process::Get(pid);
if ( !process )
return errno = ESRCH, -1;
// Prevent the process group from changing while we read it.
ScopedLock childlock(&process->groupchildlock);
assert(process->group);
return process->group->pid;
}
static int sys_setpgid(pid_t pid, pid_t pgid)
{
// TODO: Prevent changing the process group of zombies and other volatile
// things that are about to implode.
// TODO: Either prevent changing the process group after an exec or provide
// a version of this system call with a flags parameter that lets you
// decide if you want this behavior. This will fix a race condition
// where the shell spawns a child and both parent and child sets the
// process group, but the child sets the process group and execve's
// and the new program image exploits this 'bug' and also changes the
// process group, and then the shell gets around to change the process
// group. This probably unlikely, but correctness over all!
// Find the processes in question.
Process* process = !pid ? CurrentProcess() : Process::Get(pid);
if ( !process )
return errno = ESRCH, -1;
Process* group = !pgid ? process : Process::Get(pgid);
if ( !group )
return errno = ESRCH, -1;
// Prevent the current group from being changed while we also change it
ScopedLock childlock(&process->groupchildlock);
assert(process->group);
// Exit early if this is a noop.
if ( process->group == group )
return 0;
// Prevent changing the process group of a process group leader.
if ( process->group == process )
return errno = EPERM, -1;
// Remove the process from its current process group.
kthread_mutex_lock(&process->group->groupparentlock);
if ( process->groupprev )
process->groupprev->groupnext = process->groupnext;
else
process->group->groupfirst = process->groupnext;
if ( process->groupnext )
process->groupnext->groupprev = process->groupprev;
kthread_cond_signal(&process->group->groupchildleft);
kthread_mutex_unlock(&process->group->groupparentlock);
process->group->NotifyLeftProcessGroup();
process->group = NULL;
// TODO: Somehow prevent joining a zombie group, or worse yet, one that is
// currently being deleted by its parent!
// Insert the process into its new process group.
kthread_mutex_lock(&group->groupparentlock);
process->groupprev = NULL;
process->groupnext = group->groupfirst;
if ( group->groupfirst )
group->groupfirst->groupprev = process;
group->groupfirst = process;
process->group = group;
kthread_mutex_unlock(&group->groupparentlock);
return 0;
}
pid_t nextpidtoallocate;
kthread_mutex_t pidalloclock;
pid_t Process::AllocatePID()
{
ScopedLock lock(&pidalloclock);
return nextpidtoallocate++;
}
// TODO: This is not thread safe.
pid_t Process::HackGetForegroundProcess()
{
for ( pid_t i = nextpidtoallocate; 1 <= i; i-- )
{
Process* process = Get(i);
if ( !process )
continue;
if ( process->pid <= 1 )
continue;
if ( process->iszombie )
continue;
return i;
}
return 0;
}
int ProcessCompare(Process* a, Process* b)
{
if ( a->pid < b->pid )
return -1;
if ( a->pid > b->pid )
return 1;
return 0;
}
int ProcessPIDCompare(Process* a, pid_t pid)
{
if ( a->pid < pid )
return -1;
if ( a->pid > pid )
return 1;
return 0;
}
SortedList<Process*>* pidlist;
Process* Process::Get(pid_t pid)
{
ScopedLock lock(&pidalloclock);
size_t index = pidlist->Search(ProcessPIDCompare, pid);
if ( index == SIZE_MAX )
return NULL;
return pidlist->Get(index);
}
bool Process::Put(Process* process)
{
ScopedLock lock(&pidalloclock);
return pidlist->Add(process);
}
void Process::Remove(Process* process)
{
ScopedLock lock(&pidalloclock);
size_t index = pidlist->Search(process);
assert(index != SIZE_MAX);
pidlist->Remove(index);
}
2013-08-19 20:23:53 -04:00
static void* sys_sbrk(intptr_t increment)
{
Process* process = CurrentProcess();
2013-08-19 20:23:53 -04:00
ScopedLock lock(&process->segment_lock);
// Locate the heap segment.
struct segment* heap_segment = NULL;
for ( size_t i = process->segments_used; !heap_segment && i != 0; i-- )
2011-12-16 07:24:49 -05:00
{
2013-08-19 20:23:53 -04:00
struct segment* candidate = &process->segments[i-1];
if ( !(candidate->prot & PROT_HEAP) )
continue;
2013-08-19 20:23:53 -04:00
heap_segment = candidate;
}
2013-08-19 20:23:53 -04:00
if ( !heap_segment )
return errno = ENOMEM, (void*) -1UL;
2013-08-19 20:23:53 -04:00
assert(IsUserspaceSegment(heap_segment));
// Decrease the size of the heap segment if requested.
if ( increment < 0 )
{
2013-08-19 20:23:53 -04:00
uintptr_t abs_amount = Page::AlignDown(- (uintptr_t) increment);
if ( heap_segment->size < abs_amount )
abs_amount = heap_segment->size;
uintptr_t new_end = heap_segment->addr + heap_segment->size - abs_amount;
Memory::UnmapRange(new_end, abs_amount);
heap_segment->size -= abs_amount;
// TODO: How do we handle that the heap shrinks to 0 bytes?
}
2013-08-19 20:23:53 -04:00
// Increase the size of the heap if requested.
if ( 0 < increment )
{
2013-08-19 20:23:53 -04:00
uintptr_t abs_amount = Page::AlignUp(increment);
uintptr_t max_growth = 0 - (heap_segment->addr + heap_segment->size);
if ( max_growth < abs_amount )
return errno = ENOMEM, (void*) -1UL;
struct segment growth;
growth.addr = heap_segment->addr + heap_segment->size;
growth.size = abs_amount;
growth.prot = heap_segment->prot;
if ( !IsUserspaceSegment(&growth) )
return errno = ENOMEM, (void*) -1UL;
if ( FindOverlappingSegment(process, &growth) )
return errno = ENOMEM, (void*) -1UL;
if ( !Memory::MapRange(growth.addr, growth.size, growth.prot) )
return errno = ENOMEM, (void*) -1UL;
heap_segment->size += growth.size;
2011-12-16 07:24:49 -05:00
}
2013-08-19 20:23:53 -04:00
assert(IsUserspaceSegment(heap_segment));
return (void*) (heap_segment->addr + heap_segment->size);
}
2011-12-16 07:24:49 -05:00
2013-09-23 15:58:46 -04:00
static size_t sys_getpagesize()
{
return Page::Size();
}
2013-09-23 15:58:46 -04:00
static mode_t sys_umask(mode_t newmask)
{
Process* process = CurrentProcess();
ScopedLock lock(&process->idlock);
mode_t oldmask = process->umask;
process->umask = newmask & 0666;
return oldmask;
}
2013-05-16 16:03:15 -04:00
void Process::Init()
{
Syscall::Register(SYSCALL_EXEC, (void*) sys_execve);
Syscall::Register(SYSCALL_EXIT, (void*) sys_exit);
Syscall::Register(SYSCALL_GET_PAGE_SIZE, (void*) sys_getpagesize);
2013-06-11 19:02:01 -04:00
Syscall::Register(SYSCALL_GETPGID, (void*) sys_getpgid);
Syscall::Register(SYSCALL_GETPID, (void*) sys_getpid);
Syscall::Register(SYSCALL_GETPPID, (void*) sys_getppid);
Syscall::Register(SYSCALL_SBRK, (void*) sys_sbrk);
2013-06-11 19:02:01 -04:00
Syscall::Register(SYSCALL_SETPGID, (void*) sys_setpgid);
Syscall::Register(SYSCALL_TFORK, (void*) sys_tfork);
Syscall::Register(SYSCALL_UMASK, (void*) sys_umask);
Syscall::Register(SYSCALL_WAITPID, (void*) sys_waitpid);
pidalloclock = KTHREAD_MUTEX_INITIALIZER;
nextpidtoallocate = 0;
pidlist = new SortedList<Process*>(ProcessCompare);
if ( !pidlist )
Panic("could not allocate pidlist\n");
}
Implemented the fork() system call and what it needed to work properly. This commit got completely out of control. Added the fork(), getpid(), getppid(), sleep(), usleep() system calls, and aliases in the Maxsi:: namespace. Fixed a bug where zero-byte allocation would fail. Worked on the DescriptorTable class which now works and can fork. Got rid of some massive print-registers statements and replaced them with the portable InterruptRegisters::LogRegisters() function. Removed the SysExecuteOld function and replaced it with Process::Execute(). Rewrote the boot sequence in kernel.cpp such that it now loads the system idle process 'idle' as PID 0, and the initization process 'init' as PID 1. Rewrote the SIGINT hack. Processes now maintain a family-tree structure and keep track of their threads. PIDs are now allocated using a simple hack. Virtual memory per-process can now be allocated using a simple hack. Processes can now be forked. Fixed the Process::Execute function such that it now resets the stack pointer to where the stack actually is - not just a magic value. Removed the old and ugly Process::_endcodesection hack. Rewrote the scheduler into a much cleaner and faster version. Debug code is now moved to designated functions. The noop kernel-thread has been replaced by a simple user-space infinite-loop program 'idle'. The Thread class has been seperated from the Scheduler except in Scheduler- related code. Thread::{Save,Load}Registers has been improved and has been moved to $(CPU)/thread.cpp. Threads can now be forked. A new CreateThread function creates threads properly and portably. Added a MicrosecondsSinceBoot() function. Fixed a crucial bug in MemoryManagement::Fork(). Added an 'idle' user-space program that is a noop infinite loop, which is used by the scheduler when there is nothing to do. Rewrote the 'init' program such that it now forks off a shell, instead of becoming the shell. Added the $$ (current PID) and $PPID (parent PPID) variables to the shell.
2011-09-21 14:52:29 -04:00
} // namespace Sortix