2012-04-13 11:34:17 -04:00
|
|
|
/*******************************************************************************
|
|
|
|
|
2014-03-03 18:11:13 -05:00
|
|
|
Copyright(C) Jonas 'Sortie' Termansen 2011, 2012, 2014.
|
2012-04-13 11:34:17 -04:00
|
|
|
|
2013-07-10 09:26:01 -04:00
|
|
|
This file is part of Sortix.
|
2012-04-13 11:34:17 -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.
|
2012-04-13 11:34:17 -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.
|
2012-04-13 11:34:17 -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/>.
|
2012-04-13 11:34:17 -04:00
|
|
|
|
2013-12-28 15:00:15 -05:00
|
|
|
x86/interrupt.S
|
2013-07-10 09:26:01 -04:00
|
|
|
Transfers control to interrupt handlers when interrupts happen.
|
2012-04-13 11:34:17 -04:00
|
|
|
|
|
|
|
*******************************************************************************/
|
|
|
|
|
|
|
|
.section .text
|
|
|
|
|
|
|
|
.global isr0
|
|
|
|
.type isr0, @function
|
|
|
|
isr0:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $0 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr1
|
|
|
|
.type isr1, @function
|
|
|
|
isr1:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $1 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr2
|
|
|
|
.type isr2, @function
|
|
|
|
isr2:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $2 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr3
|
|
|
|
.type isr3, @function
|
|
|
|
isr3:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $3 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr4
|
|
|
|
.type isr4, @function
|
|
|
|
isr4:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $4 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr5
|
|
|
|
.type isr5, @function
|
|
|
|
isr5:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $5 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr6
|
|
|
|
.type isr6, @function
|
|
|
|
isr6:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $6 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr7
|
|
|
|
.type isr7, @function
|
|
|
|
isr7:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $7 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr8
|
|
|
|
.type isr8, @function
|
|
|
|
isr8:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $8 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr9
|
|
|
|
.type isr9, @function
|
|
|
|
isr9:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $9 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr10
|
|
|
|
.type isr10, @function
|
|
|
|
isr10:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $10 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr11
|
|
|
|
.type isr11, @function
|
|
|
|
isr11:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $11 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr12
|
|
|
|
.type isr12, @function
|
|
|
|
isr12:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $12 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr13
|
|
|
|
.type isr13, @function
|
|
|
|
isr13:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $13 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr14
|
|
|
|
.type isr14, @function
|
|
|
|
isr14:
|
|
|
|
# pushl $0 # err_code pushed by CPU
|
|
|
|
pushl $14 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr15
|
|
|
|
.type isr15, @function
|
|
|
|
isr15:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $15 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr16
|
|
|
|
.type isr16, @function
|
|
|
|
isr16:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $16 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr17
|
|
|
|
.type isr17, @function
|
|
|
|
isr17:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $17 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr18
|
|
|
|
.type isr18, @function
|
|
|
|
isr18:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $18 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr19
|
|
|
|
.type isr19, @function
|
|
|
|
isr19:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $19 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr20
|
|
|
|
.type isr20, @function
|
|
|
|
isr20:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $20 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr21
|
|
|
|
.type isr21, @function
|
|
|
|
isr21:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $21 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr22
|
|
|
|
.type isr22, @function
|
|
|
|
isr22:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $22 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr23
|
|
|
|
.type isr23, @function
|
|
|
|
isr23:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $23 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr24
|
|
|
|
.type isr24, @function
|
|
|
|
isr24:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $24 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr25
|
|
|
|
.type isr25, @function
|
|
|
|
isr25:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $25 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr26
|
|
|
|
.type isr26, @function
|
|
|
|
isr26:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $26 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr27
|
|
|
|
.type isr27, @function
|
|
|
|
isr27:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $27 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr28
|
|
|
|
.type isr28, @function
|
|
|
|
isr28:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $28 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr29
|
|
|
|
.type isr29, @function
|
|
|
|
isr29:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $29 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr30
|
|
|
|
.type isr30, @function
|
|
|
|
isr30:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $30 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr31
|
|
|
|
.type isr31, @function
|
|
|
|
isr31:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $31 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr128
|
|
|
|
.type isr128, @function
|
|
|
|
isr128:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $128 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
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
|
|
|
.global isr130
|
|
|
|
.type isr130, @function
|
|
|
|
isr130:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $130 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global isr131
|
|
|
|
.type isr131, @function
|
|
|
|
isr131:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $131 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
2012-04-13 11:34:17 -04:00
|
|
|
.global irq0
|
|
|
|
.type irq0, @function
|
|
|
|
irq0:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $32 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq1
|
|
|
|
.type irq1, @function
|
|
|
|
irq1:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $33 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq2
|
|
|
|
.type irq2, @function
|
|
|
|
irq2:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $34 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq3
|
|
|
|
.type irq3, @function
|
|
|
|
irq3:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $35 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq4
|
|
|
|
.type irq4, @function
|
|
|
|
irq4:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $36 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq5
|
|
|
|
.type irq5, @function
|
|
|
|
irq5:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $37 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq6
|
|
|
|
.type irq6, @function
|
|
|
|
irq6:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $38 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq7
|
|
|
|
.type irq7, @function
|
|
|
|
irq7:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $39 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq8
|
|
|
|
.type irq8, @function
|
|
|
|
irq8:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $40 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq9
|
|
|
|
.type irq9, @function
|
|
|
|
irq9:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $41 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq10
|
|
|
|
.type irq10, @function
|
|
|
|
irq10:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $42 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq11
|
|
|
|
.type irq11, @function
|
|
|
|
irq11:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $43 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq12
|
|
|
|
.type irq12, @function
|
|
|
|
irq12:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $44 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq13
|
|
|
|
.type irq13, @function
|
|
|
|
irq13:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $45 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq14
|
|
|
|
.type irq14, @function
|
|
|
|
irq14:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $46 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global irq15
|
|
|
|
.type irq15, @function
|
|
|
|
irq15:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $47 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
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
|
|
|
.global yield_cpu_handler
|
|
|
|
.type yield_cpu_handler, @function
|
|
|
|
yield_cpu_handler:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $129 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
|
|
|
.global thread_exit_handler
|
|
|
|
.type thread_exit_handler, @function
|
|
|
|
thread_exit_handler:
|
|
|
|
pushl $0 # err_code
|
|
|
|
pushl $132 # int_no
|
|
|
|
jmp interrupt_handler_prepare
|
2012-04-13 11:34:17 -04:00
|
|
|
|
|
|
|
interrupt_handler_prepare:
|
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
|
|
|
movl $1, asm_is_cpu_interrupted
|
|
|
|
|
|
|
|
# Check if an interrupt happened while having kernel permissions.
|
|
|
|
testw $0x3, 12(%esp) # cs
|
|
|
|
jz fixup_relocate_stack
|
|
|
|
fixup_relocate_stack_complete:
|
|
|
|
|
2012-04-13 11:34:17 -04:00
|
|
|
pushl %eax
|
|
|
|
pushl %ecx
|
|
|
|
pushl %edx
|
|
|
|
pushl %ebx
|
|
|
|
pushl %esp
|
|
|
|
pushl %ebp
|
|
|
|
pushl %esi
|
|
|
|
pushl %edi
|
|
|
|
|
|
|
|
# Push the user-space data segment.
|
|
|
|
movl %ds, %ebp
|
|
|
|
pushl %ebp
|
|
|
|
|
|
|
|
# Load the kernel data segment.
|
|
|
|
movw $0x10, %bp
|
|
|
|
movl %ebp, %ds
|
|
|
|
movl %ebp, %es
|
|
|
|
|
|
|
|
# Push CR2 in case of page faults
|
|
|
|
movl %cr2, %ebp
|
|
|
|
pushl %ebp
|
|
|
|
|
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
|
|
|
# Push the current kernel errno value.
|
|
|
|
movl global_errno, %ebp
|
|
|
|
pushl %ebp
|
|
|
|
|
|
|
|
# Push whether a signal is pending.
|
|
|
|
movl asm_signal_is_pending, %ebp
|
|
|
|
pushl %ebp
|
|
|
|
|
2012-04-13 11:34:17 -04:00
|
|
|
# Now call the interrupt handler.
|
2014-03-03 18:11:13 -05:00
|
|
|
movl %esp, %ebx
|
|
|
|
subl $4, %esp
|
|
|
|
andl $0xFFFFFFF0, %esp
|
|
|
|
movl %ebx, (%esp)
|
2012-04-13 11:34:17 -04:00
|
|
|
call interrupt_handler
|
2014-03-03 18:11:13 -05:00
|
|
|
movl %ebx, %esp
|
2012-04-13 11:34:17 -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
|
|
|
load_interrupted_registers:
|
|
|
|
# Restore whether signals are pending.
|
|
|
|
popl %ebp
|
|
|
|
movl %ebp, asm_signal_is_pending
|
|
|
|
|
|
|
|
# Restore the previous kernel errno.
|
|
|
|
popl %ebp
|
|
|
|
movl %ebp, global_errno
|
|
|
|
|
2012-04-13 11:34:17 -04:00
|
|
|
# Remove CR2 from the stack.
|
|
|
|
addl $4, %esp
|
|
|
|
|
|
|
|
# Restore the user-space data segment.
|
|
|
|
popl %ebp
|
|
|
|
movl %ebp, %ds
|
|
|
|
movl %ebp, %es
|
|
|
|
|
|
|
|
popl %edi
|
|
|
|
popl %esi
|
|
|
|
popl %ebp
|
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
|
|
|
addl $4, %esp # Don't pop %esp, may not be defined.
|
2012-04-13 11:34:17 -04:00
|
|
|
popl %ebx
|
|
|
|
popl %edx
|
|
|
|
popl %ecx
|
|
|
|
popl %eax
|
|
|
|
|
|
|
|
# Remove int_no and err_code
|
|
|
|
addl $8, %esp
|
|
|
|
|
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
|
|
|
movl $0, asm_is_cpu_interrupted
|
|
|
|
|
|
|
|
# If interrupted with kernel permissions we may need to switch stack.
|
|
|
|
testw $0x3, 4(%esp) # int_no and err_code now gone, so cs is at 4(%esp).
|
|
|
|
jz fixup_switch_stack
|
|
|
|
fixup_switch_stack_complete:
|
|
|
|
|
2012-04-13 11:34:17 -04:00
|
|
|
# Return to where we came from.
|
|
|
|
iret
|
|
|
|
|
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
|
|
|
fixup_relocate_stack:
|
|
|
|
# Ok, so some genius at Intel decided that if the permission level does not
|
|
|
|
# change during an interrupt then the CPU won't push the stack pointer and
|
|
|
|
# it won't reload it during iret. This seriously messes up the scheduler
|
|
|
|
# that wants to preempt kernel threads each with their own stack. The
|
|
|
|
# scheduler will attempt to read (and modify) the stack value which doesn't
|
|
|
|
# exist and worse: the value at that location is likely used by the
|
|
|
|
# interrupted kernel thread. A quick and dirty solution is to simply move
|
|
|
|
# the stack 8 bytes down the stack. Right now there are the 5 elements on
|
|
|
|
# the stack (eflags, cs, eip, err_code, int_no) of 5 bytes each.
|
|
|
|
mov %eax, -4-8(%esp) # Save eax
|
|
|
|
mov 0(%esp), %eax # int_no
|
|
|
|
mov %eax, 0-8(%esp)
|
|
|
|
mov 4(%esp), %eax # err_code
|
|
|
|
mov %eax, 4-8(%esp)
|
|
|
|
mov 8(%esp), %eax # eip
|
|
|
|
mov %eax, 8-8(%esp)
|
|
|
|
mov 12(%esp), %eax # cs
|
|
|
|
mov %eax, 12-8(%esp)
|
|
|
|
mov 16(%esp), %eax # eflags
|
|
|
|
mov %eax, 16-8(%esp)
|
|
|
|
# Next up we have to fake what the CPU should have done: pushed ss and esp.
|
|
|
|
mov %esp, %eax
|
|
|
|
addl $5*4, %eax # Calculate original esp
|
|
|
|
mov %eax, 20-8(%esp)
|
|
|
|
mov %ss, %eax
|
|
|
|
mov %eax, 24-8(%esp)
|
|
|
|
# Now that we moved the stack, it's time to really handle the interrupt.
|
|
|
|
mov -4-8(%esp), %eax
|
|
|
|
subl $8, %esp
|
|
|
|
jmp fixup_relocate_stack_complete
|
|
|
|
|
|
|
|
fixup_switch_stack:
|
|
|
|
# Yup, we also have to do special processing when we return from the
|
|
|
|
# interrupt. The problem is that if the iret instruction won't load a new
|
|
|
|
# stack if interrupted with kernel permissions and that the scheduler may
|
|
|
|
# wish to change the current stack during a context switch. We will then
|
|
|
|
# switch the stack before calling iret; but iret needs the return
|
|
|
|
# information on the stack (and now it isn't), so we'll copy our stack onto
|
|
|
|
# our new stack and then fire the interrupt and everyone is happy.
|
|
|
|
|
|
|
|
# In the following code, %esp will point our fixed iret return parameters
|
|
|
|
# that has stack data. However, the processor does not expect this
|
|
|
|
# information as cs hasn't changed. %ebx will point to the new stack plus
|
|
|
|
# room for three 32-bit values (eip, cs, eflags) that will be given to the
|
|
|
|
# actual iret. We will then load the new stack and copy the eip, cs and
|
|
|
|
# eflags to the new stack. However, we have to be careful in the case that
|
|
|
|
# we are switching to the same stack (in which case stuff on the same
|
|
|
|
# horizontal location in the diagram is actually on the same memory
|
|
|
|
# location). We therefore copy to the new stack and carefully avoid
|
|
|
|
# corrupting the destination if %esp + 8 = %ebx, This diagram show the
|
|
|
|
# structure of the stacks and where temporaries will be stored:
|
|
|
|
# -12 -8 -4 %esp 4 8 12 16 20
|
|
|
|
# old: IECX IEBX IEAX EIP CS EFLAGS ESP SS ...
|
|
|
|
# new: IECX IEBX IEAX - - EIP CS EFLAGS ...
|
|
|
|
# -20 -16 -12 -8 -4 %ebx 4 8 12
|
|
|
|
|
|
|
|
mov %eax, -4(%esp) # IEAX, Clobbered as copying temporary
|
|
|
|
mov %ebx, -8(%esp) # IEBX, Clobbered as pointer to new stack
|
|
|
|
mov %ecx, -12(%esp) # IECX, Clobbered as new stack selector
|
|
|
|
|
|
|
|
mov 12(%esp), %ebx # Pointer to new stack
|
|
|
|
sub $3*4, %ebx # Point to eip on the new stack (see diagram)
|
|
|
|
movw 16(%esp), %cx # New ss
|
|
|
|
|
|
|
|
# The order of these does not matter if we are switching to the same stack,
|
|
|
|
# as the memory would be copied to the same location (see diagram).
|
|
|
|
mov -4(%esp), %eax # interrupted eax value
|
|
|
|
mov %eax, -12(%ebx)
|
|
|
|
mov -8(%esp), %eax # interrupted ebx value
|
|
|
|
mov %eax, -16(%ebx)
|
|
|
|
mov -12(%esp), %eax # interrupted ecx value
|
|
|
|
mov %eax, -20(%ebx)
|
|
|
|
|
|
|
|
# The order of these three copies matter if switching to the same stack.
|
|
|
|
mov 8(%esp), %eax # eflags
|
|
|
|
mov %eax, 8(%ebx)
|
|
|
|
mov 4(%esp), %eax # cs
|
|
|
|
mov %eax, 4(%ebx)
|
|
|
|
mov 0(%esp), %eax # eip
|
|
|
|
mov %eax, 0(%ebx)
|
|
|
|
|
|
|
|
mov %cx, %ss # Load new stack selector
|
|
|
|
mov %ebx, %esp # Load new stack pointer
|
|
|
|
|
|
|
|
mov -12(%esp), %eax # restore interrupted eax value
|
|
|
|
mov -16(%esp), %ebx # restore interrupted ebx value
|
|
|
|
mov -20(%esp), %ecx # restore interrupted ecx value
|
|
|
|
|
|
|
|
jmp fixup_switch_stack_complete
|
2013-05-22 16:06:18 -04:00
|
|
|
.size interrupt_handler_prepare, . - interrupt_handler_prepare
|
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
|
|
|
|
2012-04-13 11:34:17 -04:00
|
|
|
.global interrupt_handler_null
|
|
|
|
.type interrupt_handler_null, @function
|
|
|
|
interrupt_handler_null:
|
|
|
|
iret
|
2013-05-22 16:06:18 -04:00
|
|
|
.size interrupt_handler_null, . - interrupt_handler_null
|
2012-04-13 11:34:17 -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
|
|
|
.global load_registers
|
|
|
|
.type load_registers, @function
|
|
|
|
load_registers:
|
|
|
|
# Let the register struct become our temporary stack
|
|
|
|
movl 4(%esp), %esp
|
|
|
|
jmp load_interrupted_registers
|
2013-05-22 16:06:18 -04:00
|
|
|
.size load_registers, . - load_registers
|