2013-07-10 09:26:01 -04:00
|
|
|
/*******************************************************************************
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2014-01-31 07:19:27 -05:00
|
|
|
Copyright(C) Jonas 'Sortie' Termansen 2011, 2012, 2013, 2014.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
This file is part of Sortix.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
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.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
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.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
You should have received a copy of the GNU General Public License along with
|
|
|
|
|
Sortix. If not, see <http://www.gnu.org/licenses/>.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
signal.cpp
|
2013-08-04 14:24:59 -04:00
|
|
|
Asynchronous user-space thread interruption.
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
*******************************************************************************/
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sys/types.h>
|
|
|
|
|
|
2013-10-26 20:42:10 -04:00
|
|
|
#include <assert.h>
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <errno.h>
|
2013-10-26 20:42:10 -04:00
|
|
|
#include <string.h>
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <stdint.h>
|
|
|
|
|
#include <signal.h>
|
2013-10-26 20:42:10 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sortix/sigaction.h>
|
2013-05-12 18:52:58 -04:00
|
|
|
#include <sortix/signal.h>
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sortix/sigset.h>
|
|
|
|
|
#include <sortix/stack.h>
|
|
|
|
|
#include <sortix/ucontext.h>
|
2013-05-12 18:52:58 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sortix/kernel/copy.h>
|
2013-01-09 17:30:36 -05:00
|
|
|
#include <sortix/kernel/interrupt.h>
|
2013-10-26 20:42:10 -04:00
|
|
|
#include <sortix/kernel/kernel.h>
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sortix/kernel/process.h>
|
2014-03-19 18:34:09 -04:00
|
|
|
#include <sortix/kernel/ptable.h>
|
2013-05-12 18:52:58 -04:00
|
|
|
#include <sortix/kernel/signal.h>
|
2013-08-04 14:24:59 -04:00
|
|
|
#include <sortix/kernel/syscall.h>
|
2013-05-12 18:41:30 -04:00
|
|
|
#include <sortix/kernel/thread.h>
|
2013-01-09 17:30:36 -05:00
|
|
|
|
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
|
|
|
namespace Sortix {
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
sigset_t default_ignored_signals;
|
|
|
|
|
sigset_t default_stop_signals;
|
|
|
|
|
sigset_t unblockable_signals;
|
|
|
|
|
|
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 per-cpu value whether a signal is pending in the running task.
|
|
|
|
|
extern "C" { volatile unsigned long asm_signal_is_pending = 0; }
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
static
|
2013-08-04 14:24:59 -04:00
|
|
|
void UpdatePendingSignals(Thread* thread) // thread->process->signal_lock held
|
|
|
|
|
{
|
|
|
|
|
struct sigaction* signal_actions = thread->process->signal_actions;
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
// Determine which signals wouldn't be ignored if received.
|
|
|
|
|
sigset_t handled_signals;
|
|
|
|
|
sigemptyset(&handled_signals);
|
|
|
|
|
for ( int i = 1; i < SIG_MAX_NUM; i++ )
|
|
|
|
|
{
|
|
|
|
|
if ( signal_actions[i].sa_handler == SIG_IGN )
|
|
|
|
|
continue;
|
|
|
|
|
if ( signal_actions[i].sa_handler == SIG_DFL &&
|
|
|
|
|
sigismember(&default_ignored_signals, i) )
|
|
|
|
|
continue;
|
|
|
|
|
// TODO: A process that is a member of an orphaned process group shall
|
|
|
|
|
// not be allowed to stop in response to the SIGTSTP, SIGTTIN, or
|
|
|
|
|
// SIGTTOU signals. In cases where delivery of one of these
|
|
|
|
|
// signals would stop such a process, the signal shall be
|
|
|
|
|
// discarded.
|
|
|
|
|
if ( /* is member of an orphaned process group */ false &&
|
|
|
|
|
signal_actions[i].sa_handler == SIG_DFL &&
|
|
|
|
|
sigismember(&default_stop_signals, i) )
|
|
|
|
|
continue;
|
|
|
|
|
sigaddset(&handled_signals, i);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// TODO: Handle that signals can be pending process-wide!
|
|
|
|
|
|
|
|
|
|
// Discard all requested signals that would be ignored if delivered.
|
|
|
|
|
sigandset(&thread->signal_pending, &thread->signal_pending, &handled_signals);
|
|
|
|
|
|
|
|
|
|
// Determine which signals are not blocked.
|
|
|
|
|
sigset_t permitted_signals;
|
|
|
|
|
signotset(&permitted_signals, &thread->signal_mask);
|
|
|
|
|
sigorset(&permitted_signals, &permitted_signals, &unblockable_signals);
|
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
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
// Determine which signals can currently be delivered to this thread.
|
|
|
|
|
sigset_t deliverable_signals;
|
|
|
|
|
sigandset(&deliverable_signals, &permitted_signals, &thread->signal_pending);
|
|
|
|
|
|
|
|
|
|
// Determine whether any signals can be delivered.
|
|
|
|
|
unsigned long is_pending = !sigisemptyset(&deliverable_signals) ? 1 : 0;
|
|
|
|
|
|
|
|
|
|
// Store whether a signal is pending in the virtual register.
|
|
|
|
|
if ( thread == CurrentThread() )
|
|
|
|
|
asm_signal_is_pending = is_pending;
|
|
|
|
|
else
|
|
|
|
|
thread->registers.signal_pending = is_pending;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int sys_sigaction(int signum,
|
|
|
|
|
const struct sigaction* user_newact,
|
|
|
|
|
struct sigaction* user_oldact)
|
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
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
if ( signum < 0 || signum == 0 /* null signal */ || SIG_MAX_NUM <= signum )
|
|
|
|
|
return errno = EINVAL;
|
|
|
|
|
|
|
|
|
|
Process* process = CurrentProcess();
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
struct sigaction* kact = &process->signal_actions[signum];
|
|
|
|
|
|
|
|
|
|
// Let the caller know the previous action.
|
|
|
|
|
if ( user_oldact )
|
|
|
|
|
{
|
|
|
|
|
if ( !CopyToUser(user_oldact, kact, sizeof(struct sigaction)) )
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Retrieve and validate the new signal action.
|
|
|
|
|
if ( user_newact )
|
|
|
|
|
{
|
|
|
|
|
struct sigaction newact;
|
|
|
|
|
if ( !CopyFromUser(&newact, user_newact, sizeof(struct sigaction)) )
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
if ( newact.sa_flags & ~__SA_SUPPORTED_FLAGS )
|
|
|
|
|
return errno = EINVAL, -1;
|
|
|
|
|
|
|
|
|
|
if ( newact.sa_handler == SIG_ERR )
|
|
|
|
|
return errno = EINVAL, -1;
|
|
|
|
|
|
|
|
|
|
memcpy(kact, &newact, sizeof(struct sigaction));
|
|
|
|
|
|
|
|
|
|
// Signals may become discarded because of the new handler.
|
|
|
|
|
ScopedLock threads_lock(&process->threadlock);
|
|
|
|
|
for ( Thread* t = process->firstthread; t; t = t->nextsibling )
|
|
|
|
|
UpdatePendingSignals(t);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-16 18:04:47 -04:00
|
|
|
int sys_sigaltstack(const stack_t* user_newstack, stack_t* user_oldstack)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
Thread* thread = CurrentThread();
|
|
|
|
|
|
|
|
|
|
if ( user_oldstack )
|
|
|
|
|
{
|
|
|
|
|
if ( !CopyToUser(user_oldstack, &thread->signal_stack, sizeof(stack_t)) )
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if ( user_newstack )
|
|
|
|
|
{
|
|
|
|
|
stack_t newstack;
|
|
|
|
|
if ( !CopyFromUser(&newstack, user_newstack, sizeof(stack_t)) )
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
if ( newstack.ss_flags & ~__SS_SUPPORTED_FLAGS )
|
|
|
|
|
return errno = EINVAL, -1;
|
|
|
|
|
|
|
|
|
|
memcpy(&thread->signal_stack, &newstack, sizeof(stack_t));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-16 18:04:47 -04:00
|
|
|
int sys_sigpending(sigset_t* set)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
Process* process = CurrentProcess();
|
|
|
|
|
Thread* thread = CurrentThread();
|
|
|
|
|
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
// TODO: What about process-wide signals?
|
|
|
|
|
|
|
|
|
|
return CopyToUser(set, &thread->signal_pending, sizeof(sigset_t)) ? 0 : -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
int sys_sigprocmask(int how, const sigset_t* user_set, sigset_t* user_oldset)
|
|
|
|
|
{
|
|
|
|
|
Process* process = CurrentProcess();
|
|
|
|
|
Thread* thread = CurrentThread();
|
|
|
|
|
|
|
|
|
|
// TODO: Signal masks are a per-thread property, perhaps this should be
|
|
|
|
|
// locked in another manner?
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
// Let the caller know the previous signal mask.
|
|
|
|
|
if ( user_oldset )
|
|
|
|
|
{
|
|
|
|
|
if ( !CopyToUser(user_oldset, &thread->signal_mask, sizeof(sigset_t)) )
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Update the current signal mask according to how.
|
|
|
|
|
if ( user_set )
|
|
|
|
|
{
|
|
|
|
|
sigset_t set;
|
|
|
|
|
if ( !CopyFromUser(&set, user_set, sizeof(sigset_t)) )
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
switch ( how )
|
|
|
|
|
{
|
|
|
|
|
case SIG_BLOCK:
|
|
|
|
|
sigorset(&thread->signal_mask, &thread->signal_mask, &set);
|
|
|
|
|
break;
|
|
|
|
|
case SIG_UNBLOCK:
|
|
|
|
|
signotset(&set, &set);
|
|
|
|
|
sigandset(&thread->signal_mask, &thread->signal_mask, &set);
|
|
|
|
|
break;
|
|
|
|
|
case SIG_SETMASK:
|
|
|
|
|
memcpy(&thread->signal_mask, &set, sizeof(sigset_t));
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
return errno = EINVAL, -1;
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
UpdatePendingSignals(thread);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-16 18:04:47 -04:00
|
|
|
int sys_sigsuspend(const sigset_t* set)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
Process* process = CurrentProcess();
|
|
|
|
|
Thread* thread = CurrentThread();
|
|
|
|
|
|
|
|
|
|
sigset_t old_signal_mask;
|
|
|
|
|
sigset_t new_signal_mask;
|
|
|
|
|
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
// Only accept signals from the user-provided set if given.
|
|
|
|
|
if ( set )
|
|
|
|
|
{
|
|
|
|
|
if ( !CopyFromUser(&new_signal_mask, set, sizeof(sigset_t)) )
|
|
|
|
|
return -1;
|
|
|
|
|
memcpy(&old_signal_mask, &thread->signal_mask, sizeof(sigset_t));
|
|
|
|
|
memcpy(&thread->signal_mask, &new_signal_mask, sizeof(sigset_t));
|
|
|
|
|
UpdatePendingSignals(thread);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Wait for a signal to happen or otherwise never halt.
|
|
|
|
|
kthread_cond_t never_triggered = KTHREAD_COND_INITIALIZER;
|
|
|
|
|
while ( !Signal::IsPending() )
|
|
|
|
|
kthread_cond_wait_signal(&never_triggered, &process->signal_lock);
|
|
|
|
|
|
|
|
|
|
// Restore the previous signal mask if the user gave its own set to wait on.
|
|
|
|
|
if ( set )
|
|
|
|
|
{
|
|
|
|
|
memcpy(&thread->signal_mask, &old_signal_mask, sizeof(sigset_t));
|
|
|
|
|
UpdatePendingSignals(thread);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// The system call never halts or it halts because a signal interrupted it.
|
|
|
|
|
return errno = EINTR, -1;
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-16 18:04:47 -04:00
|
|
|
int sys_kill(pid_t pid, int signum)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
// Protect the kernel process.
|
|
|
|
|
if ( !pid )
|
|
|
|
|
return errno = EPERM, -1;
|
|
|
|
|
|
|
|
|
|
// TODO: Implement that pid == -1 means all processes!
|
|
|
|
|
bool process_group = pid < 0 ? (pid = -pid, true) : false;
|
|
|
|
|
|
|
|
|
|
// TODO: Race condition: The process could be deleted while we use it.
|
2014-03-19 18:34:09 -04:00
|
|
|
Process* process = CurrentProcess()->GetPTable()->Get(pid);
|
2013-08-04 14:24:59 -04:00
|
|
|
if ( !process )
|
|
|
|
|
return errno = ESRCH, -1;
|
|
|
|
|
|
|
|
|
|
// TODO: Protect init?
|
|
|
|
|
// TODO: Check for permission.
|
|
|
|
|
// TODO: Check for zombies.
|
|
|
|
|
|
|
|
|
|
if ( process_group )
|
|
|
|
|
{
|
|
|
|
|
if ( !process->DeliverGroupSignal(signum) && errno != ESIGPENDING )
|
|
|
|
|
return -1;
|
|
|
|
|
return errno = 0, 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if ( !process->DeliverSignal(signum) && errno != ESIGPENDING )
|
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
|
|
|
return -1;
|
2013-08-04 14:24:59 -04:00
|
|
|
return errno = 0, 0;
|
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
|
|
|
}
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
bool Process::DeliverGroupSignal(int signum)
|
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
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
ScopedLock lock(&groupparentlock);
|
|
|
|
|
if ( !groupfirst )
|
|
|
|
|
return errno = ESRCH, false;
|
|
|
|
|
for ( Process* iter = groupfirst; iter; iter = iter->groupnext )
|
|
|
|
|
{
|
|
|
|
|
int saved_errno = errno;
|
|
|
|
|
if ( !iter->DeliverSignal(signum) && errno != ESIGPENDING )
|
|
|
|
|
{
|
|
|
|
|
// This is not currently an error condition.
|
|
|
|
|
}
|
|
|
|
|
errno = saved_errno;
|
|
|
|
|
}
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
bool Process::DeliverSignal(int signum)
|
|
|
|
|
{
|
|
|
|
|
ScopedLock lock(&threadlock);
|
|
|
|
|
|
|
|
|
|
if ( !firstthread )
|
|
|
|
|
return errno = EINIT, false;
|
|
|
|
|
|
|
|
|
|
// Broadcast particular signals to all the threads in the process.
|
|
|
|
|
if ( signum == SIGCONT || signum == SIGSTOP || signum == SIGKILL )
|
|
|
|
|
{
|
|
|
|
|
int saved_errno = errno;
|
|
|
|
|
for ( Thread* t = firstthread; t; t = t->nextsibling )
|
|
|
|
|
{
|
|
|
|
|
if ( !t->DeliverSignal(signum) && errno != ESIGPENDING )
|
|
|
|
|
{
|
|
|
|
|
// This is not currently an error condition.
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
errno = saved_errno;
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Route the signal to a suitable thread that accepts it.
|
|
|
|
|
// TODO: This isn't how signals should be routed to a particular thread.
|
|
|
|
|
if ( CurrentThread()->process == this )
|
|
|
|
|
return CurrentThread()->DeliverSignal(signum);
|
|
|
|
|
return firstthread->DeliverSignal(signum);
|
|
|
|
|
}
|
|
|
|
|
|
2014-10-16 18:04:47 -04:00
|
|
|
int sys_raise(int signum)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
if ( !CurrentThread()->DeliverSignal(signum) && errno != ESIGPENDING )
|
|
|
|
|
return -1;
|
|
|
|
|
return errno = 0, 0;
|
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
|
|
|
}
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
bool Thread::DeliverSignal(int signum)
|
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
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
return DeliverSignalUnlocked(signum);
|
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
|
|
|
}
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
bool Thread::DeliverSignalUnlocked(int signum) // thread->process->signal_lock held
|
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
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
if ( signum <= 0 || SIG_MAX_NUM <= signum )
|
|
|
|
|
return errno = EINVAL, false;
|
|
|
|
|
|
|
|
|
|
// Discard the null signal, which does error checking, but doesn't actually
|
|
|
|
|
// deliver a signal to the process or thread.
|
|
|
|
|
if ( signum == 0 )
|
|
|
|
|
return true;
|
|
|
|
|
|
|
|
|
|
if ( sigismember(&signal_pending, signum) )
|
|
|
|
|
return errno = ESIGPENDING, false;
|
|
|
|
|
|
|
|
|
|
sigaddset(&signal_pending, signum);
|
|
|
|
|
if ( signum == SIGSTOP || signum == SIGTSTP ||
|
|
|
|
|
signum == SIGTTIN || signum == SIGTTOU )
|
|
|
|
|
sigdelset(&signal_pending, SIGCONT);
|
|
|
|
|
if ( signum == SIGCONT )
|
|
|
|
|
{
|
|
|
|
|
sigdelset(&signal_pending, SIGSTOP);
|
|
|
|
|
sigdelset(&signal_pending, SIGTSTP);
|
|
|
|
|
sigdelset(&signal_pending, SIGTTIN);
|
|
|
|
|
sigdelset(&signal_pending, SIGTTOU);
|
|
|
|
|
}
|
|
|
|
|
UpdatePendingSignals(this);
|
|
|
|
|
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int PickImportantSignal(const sigset_t* set)
|
|
|
|
|
{
|
|
|
|
|
if ( sigismember(set, SIGKILL) )
|
|
|
|
|
return SIGKILL;
|
|
|
|
|
if ( sigismember(set, SIGSTOP) )
|
|
|
|
|
return SIGSTOP;
|
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 ( int i = 1; i < SIG_MAX_NUM; i++ )
|
2013-08-04 14:24:59 -04:00
|
|
|
if ( sigismember(set, i) )
|
|
|
|
|
return i;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
static void EncodeMachineContext(mcontext_t* mctx,
|
|
|
|
|
const struct thread_registers* regs,
|
|
|
|
|
const struct interrupt_context* intctx)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
memset(mctx, 0, sizeof(*mctx));
|
|
|
|
|
#if defined(__i386__)
|
|
|
|
|
// TODO: REG_GS
|
|
|
|
|
// TODO: REG_FS
|
|
|
|
|
// TODO: REG_ES
|
|
|
|
|
// TODO: REG_DS
|
|
|
|
|
mctx->gregs[REG_EDI] = regs->edi;
|
|
|
|
|
mctx->gregs[REG_ESI] = regs->esi;
|
|
|
|
|
mctx->gregs[REG_EBP] = regs->ebp;
|
|
|
|
|
mctx->gregs[REG_ESP] = regs->esp;
|
|
|
|
|
mctx->gregs[REG_EBX] = regs->ebx;
|
|
|
|
|
mctx->gregs[REG_EDX] = regs->edx;
|
|
|
|
|
mctx->gregs[REG_ECX] = regs->ecx;
|
|
|
|
|
mctx->gregs[REG_EAX] = regs->eax;
|
|
|
|
|
mctx->gregs[REG_EIP] = regs->eip;
|
|
|
|
|
// TODO: REG_CS
|
|
|
|
|
mctx->gregs[REG_EFL] = regs->eflags & 0x0000FFFF;
|
2014-03-03 18:11:13 -05:00
|
|
|
mctx->gregs[REG_CR2] = intctx->cr2;
|
2013-08-04 14:24:59 -04:00
|
|
|
// TODO: REG_SS
|
2014-03-03 18:11:13 -05:00
|
|
|
memcpy(mctx->fpuenv, regs->fpuenv, 512);
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
|
|
|
|
mctx->gregs[REG_R8] = regs->r8;
|
|
|
|
|
mctx->gregs[REG_R9] = regs->r9;
|
|
|
|
|
mctx->gregs[REG_R10] = regs->r10;
|
|
|
|
|
mctx->gregs[REG_R11] = regs->r11;
|
|
|
|
|
mctx->gregs[REG_R12] = regs->r12;
|
|
|
|
|
mctx->gregs[REG_R13] = regs->r13;
|
|
|
|
|
mctx->gregs[REG_R14] = regs->r14;
|
|
|
|
|
mctx->gregs[REG_R15] = regs->r15;
|
|
|
|
|
mctx->gregs[REG_RDI] = regs->rdi;
|
|
|
|
|
mctx->gregs[REG_RSI] = regs->rsi;
|
|
|
|
|
mctx->gregs[REG_RBP] = regs->rbp;
|
|
|
|
|
mctx->gregs[REG_RBX] = regs->rbx;
|
|
|
|
|
mctx->gregs[REG_RDX] = regs->rdx;
|
|
|
|
|
mctx->gregs[REG_RAX] = regs->rax;
|
|
|
|
|
mctx->gregs[REG_RCX] = regs->rcx;
|
|
|
|
|
mctx->gregs[REG_RSP] = regs->rsp;
|
|
|
|
|
mctx->gregs[REG_RIP] = regs->rip;
|
|
|
|
|
mctx->gregs[REG_EFL] = regs->rflags & 0x000000000000FFFF;
|
|
|
|
|
// TODO: REG_CSGSFS.
|
2014-03-03 18:11:13 -05:00
|
|
|
mctx->gregs[REG_CR2] = intctx->cr2;
|
2013-08-04 14:24:59 -04:00
|
|
|
mctx->gregs[REG_FSBASE] = 0x0;
|
|
|
|
|
mctx->gregs[REG_GSBASE] = 0x0;
|
2014-03-03 18:11:13 -05:00
|
|
|
memcpy(mctx->fpuenv, regs->fpuenv, 512);
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You need to implement conversion to mcontext"
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
static void DecodeMachineContext(const mcontext_t* mctx,
|
|
|
|
|
struct thread_registers* regs)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
#if defined(__i386__) || defined(__x86_64__)
|
|
|
|
|
unsigned long user_flags = FLAGS_CARRY | FLAGS_PARITY | FLAGS_AUX
|
|
|
|
|
| FLAGS_ZERO | FLAGS_SIGN | FLAGS_DIRECTION
|
|
|
|
|
| FLAGS_OVERFLOW;
|
|
|
|
|
#endif
|
|
|
|
|
#if defined(__i386__)
|
|
|
|
|
regs->edi = mctx->gregs[REG_EDI];
|
|
|
|
|
regs->esi = mctx->gregs[REG_ESI];
|
|
|
|
|
regs->ebp = mctx->gregs[REG_EBP];
|
|
|
|
|
regs->esp = mctx->gregs[REG_ESP];
|
|
|
|
|
regs->ebx = mctx->gregs[REG_EBX];
|
|
|
|
|
regs->edx = mctx->gregs[REG_EDX];
|
|
|
|
|
regs->ecx = mctx->gregs[REG_ECX];
|
|
|
|
|
regs->eax = mctx->gregs[REG_EAX];
|
|
|
|
|
regs->eip = mctx->gregs[REG_EIP];
|
|
|
|
|
regs->eflags &= ~user_flags;
|
|
|
|
|
regs->eflags |= mctx->gregs[REG_EFL] & user_flags;
|
2014-03-03 18:11:13 -05:00
|
|
|
memcpy(regs->fpuenv, mctx->fpuenv, 512);
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
|
|
|
|
regs->r8 = mctx->gregs[REG_R8];
|
|
|
|
|
regs->r9 = mctx->gregs[REG_R9];
|
|
|
|
|
regs->r10 = mctx->gregs[REG_R10];
|
|
|
|
|
regs->r11 = mctx->gregs[REG_R11];
|
|
|
|
|
regs->r12 = mctx->gregs[REG_R12];
|
|
|
|
|
regs->r13 = mctx->gregs[REG_R13];
|
|
|
|
|
regs->r14 = mctx->gregs[REG_R14];
|
|
|
|
|
regs->r15 = mctx->gregs[REG_R15];
|
|
|
|
|
regs->rdi = mctx->gregs[REG_RDI];
|
|
|
|
|
regs->rsi = mctx->gregs[REG_RSI];
|
|
|
|
|
regs->rbp = mctx->gregs[REG_RBP];
|
|
|
|
|
regs->rbx = mctx->gregs[REG_RBX];
|
|
|
|
|
regs->rdx = mctx->gregs[REG_RDX];
|
|
|
|
|
regs->rax = mctx->gregs[REG_RAX];
|
|
|
|
|
regs->rcx = mctx->gregs[REG_RCX];
|
|
|
|
|
regs->rsp = mctx->gregs[REG_RSP];
|
|
|
|
|
regs->rip = mctx->gregs[REG_RIP];
|
|
|
|
|
regs->rflags &= ~user_flags;
|
|
|
|
|
regs->rflags |= mctx->gregs[REG_EFL] & user_flags;
|
2014-03-03 18:11:13 -05:00
|
|
|
memcpy(regs->fpuenv, mctx->fpuenv, 512);
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You need to implement conversion to mcontext"
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#if defined(__i386__)
|
|
|
|
|
struct stack_frame
|
|
|
|
|
{
|
|
|
|
|
unsigned long sigreturn;
|
|
|
|
|
int signum_param;
|
|
|
|
|
siginfo_t* siginfo_param;
|
|
|
|
|
ucontext_t* ucontext_param;
|
|
|
|
|
void* cookie_param;
|
|
|
|
|
siginfo_t siginfo;
|
|
|
|
|
ucontext_t ucontext;
|
|
|
|
|
};
|
|
|
|
|
#elif defined(__x86_64__)
|
|
|
|
|
struct stack_frame
|
|
|
|
|
{
|
|
|
|
|
unsigned long sigreturn;
|
|
|
|
|
siginfo_t siginfo;
|
|
|
|
|
ucontext_t ucontext;
|
|
|
|
|
};
|
|
|
|
|
#else
|
|
|
|
|
#error "You need to implement struct stack_frame"
|
|
|
|
|
#endif
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
void Thread::HandleSignal(struct interrupt_context* intctx)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
assert(Interrupt::IsEnabled());
|
|
|
|
|
assert(this == CurrentThread());
|
|
|
|
|
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
assert(process->sigreturn);
|
|
|
|
|
|
|
|
|
|
retry_another_signal:
|
|
|
|
|
|
|
|
|
|
// Determine which signals are not blocked.
|
|
|
|
|
sigset_t permitted_signals;
|
|
|
|
|
signotset(&permitted_signals, &signal_mask);
|
|
|
|
|
sigorset(&permitted_signals, &permitted_signals, &unblockable_signals);
|
|
|
|
|
|
|
|
|
|
// Determine which signals can currently be delivered to this thread.
|
|
|
|
|
sigset_t deliverable_signals;
|
|
|
|
|
sigandset(&deliverable_signals, &permitted_signals, &signal_pending);
|
|
|
|
|
|
|
|
|
|
// Decide which signal to deliver to the thread.
|
|
|
|
|
int signum = PickImportantSignal(&deliverable_signals);
|
|
|
|
|
if ( !signum )
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
// Unmark the selected signal as pending.
|
|
|
|
|
sigdelset(&signal_pending, signum);
|
|
|
|
|
UpdatePendingSignals(this);
|
2014-03-03 18:11:13 -05:00
|
|
|
intctx->signal_pending = asm_signal_is_pending;
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
// Destroy the current thread if the signal is critical.
|
|
|
|
|
if ( signum == SIGKILL )
|
|
|
|
|
{
|
|
|
|
|
lock.Reset();
|
|
|
|
|
kthread_exit();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
struct sigaction* action = &process->signal_actions[signum];
|
|
|
|
|
|
|
|
|
|
// Stop the current thread upon receipt of a stop signal that isn't handled
|
|
|
|
|
// or cannot be handled (SIGSTOP).
|
|
|
|
|
if ( (action->sa_handler == SIG_DFL &&
|
|
|
|
|
sigismember(&default_stop_signals, signum) ) ||
|
|
|
|
|
signum == SIGSTOP )
|
|
|
|
|
{
|
|
|
|
|
Log::PrintF("%s:%u: `%s' FIXME SIGSTOP\n", __FILE__, __LINE__, __PRETTY_FUNCTION__);
|
|
|
|
|
// TODO: Stop the current process.
|
|
|
|
|
// TODO: Deliver SIGCHLD to the parent except if SA_NOCLDSTOP is set in
|
|
|
|
|
// the parent's SIGCHLD sigaction.
|
|
|
|
|
// TODO: SIGCHLD should not be delivered until all the threads in the
|
|
|
|
|
// process has received SIGSTOP and stopped?
|
|
|
|
|
// TODO: SIGKILL must still be deliverable to a stopped process.
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Resume the current thread upon receipt of SIGCONT.
|
|
|
|
|
if ( signum == SIGCONT )
|
|
|
|
|
{
|
|
|
|
|
Log::PrintF("%s:%u: `%s' FIXME SIGCONT\n", __FILE__, __LINE__, __PRETTY_FUNCTION__);
|
|
|
|
|
// TODO: Resume the current process.
|
|
|
|
|
// TODO: Can SIGCONT be masked?
|
|
|
|
|
// TODO: Can SIGCONT be handled?
|
|
|
|
|
// TODO: Can SIGCONT be ignored?
|
|
|
|
|
// TODO: Deliver SIGCHLD to the parent except if SA_NOCLDSTOP is set in
|
|
|
|
|
// the parent's SIGCHLD sigaction.
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Signals that would be ignored are already filtered away at this point.
|
|
|
|
|
assert(action->sa_handler != SIG_IGN);
|
|
|
|
|
assert(action->sa_handler != SIG_DFL || !sigismember(&default_ignored_signals, signum));
|
|
|
|
|
|
|
|
|
|
// The default action must be to terminate the process. Signals that are
|
|
|
|
|
// ignored by default got discarded earlier.
|
|
|
|
|
if ( action->sa_handler == SIG_DFL )
|
|
|
|
|
{
|
|
|
|
|
kthread_mutex_unlock(&process->signal_lock);
|
|
|
|
|
process->ExitThroughSignal(signum);
|
|
|
|
|
kthread_mutex_lock(&process->signal_lock);
|
|
|
|
|
goto retry_another_signal;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// At this point we have to attempt to invoke the user-space signal handler,
|
|
|
|
|
// which will then return control to us through sigreturn. However, we can't
|
|
|
|
|
// save the kernel state because 1) we can't trust the user-space stack 2)
|
|
|
|
|
// we can't rely on the kernel stack being intact as the signal handler may
|
|
|
|
|
// invoke system calls. For those reasons, we'll have to modify the saved
|
|
|
|
|
// registers so they restore a user-space state. We can do this because
|
|
|
|
|
// threads in the kernel cannot be delivered signals except when returning
|
|
|
|
|
// from a system call, so we'll simply save the state that would have been
|
|
|
|
|
// returned to user-space had no signal occured.
|
2014-03-03 18:11:13 -05:00
|
|
|
if ( !InUserspace(intctx) )
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
#if defined(__i386__)
|
2014-03-03 18:11:13 -05:00
|
|
|
uint32_t* params = (uint32_t*) intctx->ebx;
|
|
|
|
|
intctx->eip = params[0];
|
|
|
|
|
intctx->eflags = params[2];
|
|
|
|
|
intctx->esp = params[3];
|
|
|
|
|
intctx->cs = UCS | URPL;
|
|
|
|
|
intctx->ds = UDS | URPL;
|
|
|
|
|
intctx->ss = UDS | URPL;
|
2015-05-16 13:22:46 -04:00
|
|
|
intctx->ebx = 0;
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
2014-03-03 18:11:13 -05:00
|
|
|
intctx->rip = intctx->rdi;
|
|
|
|
|
intctx->rflags = intctx->rsi;
|
|
|
|
|
intctx->rsp = intctx->r8;
|
|
|
|
|
intctx->cs = UCS | URPL;
|
|
|
|
|
intctx->ds = UDS | URPL;
|
|
|
|
|
intctx->ss = UDS | URPL;
|
2015-05-16 13:22:46 -04:00
|
|
|
intctx->rdi = 0;
|
|
|
|
|
intctx->rsi = 0;
|
|
|
|
|
intctx->r8 = 0;
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You may need to fix the registers"
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
struct thread_registers stopped_regs;
|
|
|
|
|
Scheduler::SaveInterruptedContext(intctx, &stopped_regs);
|
|
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
sigset_t new_signal_mask;
|
|
|
|
|
memcpy(&new_signal_mask, &action->sa_mask, sizeof(sigset_t));
|
|
|
|
|
sigorset(&new_signal_mask, &new_signal_mask, &signal_mask);
|
|
|
|
|
|
|
|
|
|
// Prevent signals from interrupting themselves by default.
|
|
|
|
|
if ( !(action->sa_flags & SA_NODEFER) )
|
|
|
|
|
sigaddset(&new_signal_mask, signum);
|
|
|
|
|
|
|
|
|
|
// Determine whether we use an alternate signal stack.
|
|
|
|
|
bool signal_uses_altstack = action->sa_flags & SA_ONSTACK;
|
|
|
|
|
bool usable_altstack = !(signal_stack.ss_flags & (SS_DISABLE | SS_ONSTACK));
|
|
|
|
|
bool use_altstack = signal_uses_altstack && usable_altstack;
|
|
|
|
|
|
|
|
|
|
// Determine which signal stack to use and what to save.
|
|
|
|
|
stack_t old_signal_stack, new_signal_stack;
|
|
|
|
|
uintptr_t stack_location;
|
|
|
|
|
if ( use_altstack )
|
|
|
|
|
{
|
|
|
|
|
old_signal_stack = signal_stack;
|
|
|
|
|
new_signal_stack = signal_stack;
|
|
|
|
|
new_signal_stack.ss_flags |= SS_ONSTACK;
|
|
|
|
|
#if defined(__i386__) || defined(__x86_64__)
|
|
|
|
|
stack_location = (uintptr_t) signal_stack.ss_sp + signal_stack.ss_size;
|
|
|
|
|
#else
|
|
|
|
|
#error "You need to implement getting the alternate stack pointer"
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
old_signal_stack.ss_sp = NULL;
|
|
|
|
|
old_signal_stack.ss_flags = SS_DISABLE;
|
|
|
|
|
old_signal_stack.ss_size = 0;
|
|
|
|
|
new_signal_stack = signal_stack;
|
|
|
|
|
#if defined(__i386__)
|
2014-03-03 18:11:13 -05:00
|
|
|
stack_location = (uintptr_t) stopped_regs.esp;
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
2014-03-03 18:11:13 -05:00
|
|
|
stack_location = (uintptr_t) stopped_regs.rsp;
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You need to implement getting the user-space stack pointer"
|
|
|
|
|
#endif
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
struct thread_registers handler_regs;
|
|
|
|
|
memcpy(&handler_regs, &stopped_regs, sizeof(handler_regs));
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
struct stack_frame stack_frame;
|
|
|
|
|
memset(&stack_frame, 0, sizeof(stack_frame));
|
|
|
|
|
|
|
|
|
|
void* handler_ptr = action->sa_flags & SA_COOKIE ?
|
|
|
|
|
(void*) action->sa_sigaction_cookie :
|
|
|
|
|
action->sa_flags & SA_SIGINFO ?
|
|
|
|
|
(void*) action->sa_sigaction :
|
|
|
|
|
(void*) action->sa_handler;
|
|
|
|
|
|
|
|
|
|
#if defined(__i386__)
|
|
|
|
|
stack_location -= sizeof(stack_frame);
|
|
|
|
|
stack_location &= ~(4UL-1UL);
|
|
|
|
|
struct stack_frame* stack = (struct stack_frame*) stack_location;
|
|
|
|
|
|
|
|
|
|
stack_frame.sigreturn = (unsigned long) process->sigreturn;
|
|
|
|
|
stack_frame.signum_param = signum;
|
|
|
|
|
stack_frame.siginfo_param = &stack->siginfo;
|
|
|
|
|
stack_frame.ucontext_param = &stack->ucontext;
|
|
|
|
|
stack_frame.cookie_param = action->sa_cookie;
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
handler_regs.esp = (unsigned long) stack;
|
|
|
|
|
handler_regs.eip = (unsigned long) handler_ptr;
|
|
|
|
|
handler_regs.eflags &= ~FLAGS_DIRECTION;
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
|
|
|
|
stack_location -= 128; /* Red zone. */
|
|
|
|
|
stack_location -= sizeof(stack_frame);
|
|
|
|
|
stack_location = ((stack_location - 8) & ~(16UL-1UL)) + 8;
|
|
|
|
|
struct stack_frame* stack = (struct stack_frame*) stack_location;
|
|
|
|
|
|
|
|
|
|
stack_frame.sigreturn = (unsigned long) process->sigreturn;
|
2014-03-03 18:11:13 -05:00
|
|
|
handler_regs.rdi = (unsigned long) signum;
|
|
|
|
|
handler_regs.rsi = (unsigned long) &stack->siginfo;
|
|
|
|
|
handler_regs.rdx = (unsigned long) &stack->ucontext;
|
|
|
|
|
handler_regs.rcx = (unsigned long) action->sa_cookie;
|
|
|
|
|
|
|
|
|
|
handler_regs.rsp = (unsigned long) stack;
|
|
|
|
|
handler_regs.rip = (unsigned long) handler_ptr;
|
|
|
|
|
handler_regs.rflags &= ~FLAGS_DIRECTION;
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You need to format the stack frame"
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
// Format the siginfo into the stack frame.
|
|
|
|
|
stack_frame.siginfo.si_signo = signum;
|
|
|
|
|
#if defined(__i386__) || defined(__x86_64__)
|
|
|
|
|
// TODO: Is this cr2 value trustworthy? I don't think it is.
|
|
|
|
|
if ( signum == SIGSEGV )
|
2014-03-03 18:11:13 -05:00
|
|
|
stack_frame.siginfo.si_addr = (void*) intctx->cr2;
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#warning "You need to tell user-space where it crashed"
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
// Format the ucontext into the stack frame.
|
|
|
|
|
stack_frame.ucontext.uc_link = NULL;
|
|
|
|
|
memcpy(&stack_frame.ucontext.uc_sigmask, &signal_mask, sizeof(signal_mask));
|
|
|
|
|
memcpy(&stack_frame.ucontext.uc_stack, &signal_stack, sizeof(signal_stack));
|
2014-03-03 18:11:13 -05:00
|
|
|
EncodeMachineContext(&stack_frame.ucontext.uc_mcontext, &stopped_regs, intctx);
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
if ( !CopyToUser(stack, &stack_frame, sizeof(stack_frame)) )
|
|
|
|
|
{
|
|
|
|
|
// Self-destruct if we crashed during delivering the crash signal.
|
|
|
|
|
if ( signum == SIGSEGV )
|
2011-09-15 16:38:40 -04:00
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
kthread_mutex_unlock(&process->signal_lock);
|
|
|
|
|
process->ExitThroughSignal(signum);
|
|
|
|
|
kthread_mutex_lock(&process->signal_lock);
|
|
|
|
|
goto retry_another_signal;
|
2011-09-15 16:38:40 -04:00
|
|
|
}
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
// Deliver SIGSEGV if we could not deliver the signal on the stack.
|
|
|
|
|
// TODO: Is it possible to block SIGSEGV here?
|
|
|
|
|
kthread_mutex_unlock(&process->signal_lock);
|
|
|
|
|
DeliverSignal(SIGSEGV);
|
|
|
|
|
kthread_mutex_lock(&process->signal_lock);
|
|
|
|
|
goto retry_another_signal;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Update the current signal mask.
|
|
|
|
|
memcpy(&signal_mask, &new_signal_mask, sizeof(sigset_t));
|
|
|
|
|
|
|
|
|
|
// Update the current alternate signal stack.
|
|
|
|
|
signal_stack = new_signal_stack;
|
|
|
|
|
|
|
|
|
|
// Update the current registers.
|
2014-03-03 18:11:13 -05:00
|
|
|
Scheduler::LoadInterruptedContext(intctx, &handler_regs);
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
// TODO: SA_RESETHAND:
|
|
|
|
|
// "If set, the disposition of the signal shall be reset to SIG_DFL
|
|
|
|
|
// and the SA_SIGINFO flag shall be cleared on entry to the signal
|
|
|
|
|
// handler. Note: SIGILL and SIGTRAP cannot be automatically reset
|
|
|
|
|
// when delivered; the system silently enforces this restriction."
|
|
|
|
|
|
|
|
|
|
// Run the signal handler by returning to user-space.
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
void Thread::HandleSigreturn(struct interrupt_context* intctx)
|
2013-08-04 14:24:59 -04:00
|
|
|
{
|
|
|
|
|
assert(Interrupt::IsEnabled());
|
|
|
|
|
assert(this == CurrentThread());
|
|
|
|
|
|
|
|
|
|
ScopedLock lock(&process->signal_lock);
|
|
|
|
|
|
|
|
|
|
struct stack_frame stack_frame;
|
|
|
|
|
const struct stack_frame* user_stack_frame;
|
|
|
|
|
#if defined(__i386__)
|
2014-03-03 18:11:13 -05:00
|
|
|
user_stack_frame = (const struct stack_frame*) (intctx->esp - 4);
|
2013-08-04 14:24:59 -04:00
|
|
|
#elif defined(__x86_64__)
|
2014-03-03 18:11:13 -05:00
|
|
|
user_stack_frame = (const struct stack_frame*) (intctx->rsp - 8);
|
2013-08-04 14:24:59 -04:00
|
|
|
#else
|
|
|
|
|
#error "You need to locate the stack we passed the signal handler"
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
if ( CopyFromUser(&stack_frame, user_stack_frame, sizeof(stack_frame)) )
|
|
|
|
|
{
|
|
|
|
|
memcpy(&signal_mask, &stack_frame.ucontext.uc_sigmask, sizeof(signal_mask));
|
|
|
|
|
memcpy(&signal_stack, &stack_frame.ucontext.uc_stack, sizeof(signal_stack));
|
|
|
|
|
signal_stack.ss_flags &= __SS_SUPPORTED_FLAGS;
|
2014-03-03 18:11:13 -05:00
|
|
|
struct thread_registers resume_regs;
|
|
|
|
|
Scheduler::SaveInterruptedContext(intctx, &resume_regs);
|
|
|
|
|
DecodeMachineContext(&stack_frame.ucontext.uc_mcontext, &resume_regs);
|
|
|
|
|
Scheduler::LoadInterruptedContext(intctx, &resume_regs);
|
2013-08-04 14:24:59 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
UpdatePendingSignals(this);
|
2014-03-03 18:11:13 -05:00
|
|
|
intctx->signal_pending = asm_signal_is_pending;
|
2013-08-04 14:24:59 -04:00
|
|
|
|
|
|
|
|
lock.Reset();
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
HandleSignal(intctx);
|
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
|
|
|
}
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2013-08-04 14:24:59 -04:00
|
|
|
namespace Signal {
|
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
void DispatchHandler(struct interrupt_context* intctx, void* /*user*/)
|
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
|
|
|
{
|
2014-03-03 18:11:13 -05:00
|
|
|
return CurrentThread()->HandleSignal(intctx);
|
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
|
|
|
}
|
2011-09-15 16:38:40 -04:00
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
void ReturnHandler(struct interrupt_context* intctx, void* /*user*/)
|
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
|
|
|
{
|
2014-03-03 18:11:13 -05:00
|
|
|
return CurrentThread()->HandleSigreturn(intctx);
|
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
|
|
|
}
|
2011-09-15 16:38:40 -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 Init()
|
|
|
|
|
{
|
2013-08-04 14:24:59 -04:00
|
|
|
sigemptyset(&default_ignored_signals);
|
|
|
|
|
sigaddset(&default_ignored_signals, SIGCHLD);
|
|
|
|
|
sigaddset(&default_ignored_signals, SIGURG);
|
|
|
|
|
sigaddset(&default_ignored_signals, SIGPWR);
|
|
|
|
|
sigaddset(&default_ignored_signals, SIGWINCH);
|
|
|
|
|
sigemptyset(&default_stop_signals);
|
|
|
|
|
sigaddset(&default_stop_signals, SIGTSTP);
|
|
|
|
|
sigaddset(&default_stop_signals, SIGTTIN);
|
|
|
|
|
sigaddset(&default_stop_signals, SIGTTOU);
|
|
|
|
|
sigemptyset(&unblockable_signals);
|
|
|
|
|
sigaddset(&unblockable_signals, SIGKILL);
|
|
|
|
|
sigaddset(&unblockable_signals, SIGSTOP);
|
2011-09-15 16:38:40 -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
|
|
|
} // namespace Signal
|
2013-08-04 14:24:59 -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
|
|
|
} // namespace Sortix
|