application got interrupted, thread_at_kernel_exit() was called - but
that expected interrupts to be enabled for signal handling.
However, only exceptions 3 (breakpoint) and 99 (syscall) are trap
gates, and thus, only those actually had interrupt enabled at that
point.
If a KILL signal was pending when a hardware interrupt interrupted a
user space thread you were entering KDL before ("acquire_sem() called
with interrupts turned off").
Thanks to mouse interrupts I finally got this often enough to find
this...
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15143 a95241bf-73f2-0310-859d-f6bbb57e9c96
categories:
* Missing includes (like <stdlib.h> and <string.h>).
* Linking against $(TARGET_LIBSTDC++) instead of libstdc++.r4.so.
* Local variables shadowing parameters.
* Default parameters in function definitions (as opposed to function
declarations).
* All C++ stuff (nothrow, map, set, vector, min, max,...) must be imported
explicitly from the std:: namespace now.
* "new (sometype)[...]" must read "new sometype[...]", even if sometype is
something like "const char *".
* __FUNCTION__ is no longer a string literal (but a string expression), i.e.
'printf(__FUNCTION__ ": ...\n")' is invalid code.
* A type cast results in a non-lvalue. E.g. "(char *)buffer += bytes"
is an invalid expression.
* "friend class SomeClass" only works when SomeClass is known before.
Otherwise the an inner class with that name is considered as friend.
gcc 4 is much pickier about scopes.
* gcc 4 is generally stricter with respect to type conversions in C.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14878 a95241bf-73f2-0310-859d-f6bbb57e9c96
The stack frame that got interrupted were missing from the stack output
(and thus the call that got us into the kernel).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14706 a95241bf-73f2-0310-859d-f6bbb57e9c96
to be able to follow the stack trace into userland.
No symbols there, yet, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14697 a95241bf-73f2-0310-859d-f6bbb57e9c96
to the parameters nor the number of parameters; the kernel is able to
get both without problems.
The syscall functions in libroot (_kern_*()) are now self-contained
(they don't "jmp" to separate code anymore), which at least
theoretically allows gdb to print a stack trace with the syscall
function on the top when a thread is currently performing a syscall.
Practically it doesn't work yet, though, since those functions are
frameless (i.e. create no stack frame) which needs special support
I haven't implemented yet.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14661 a95241bf-73f2-0310-859d-f6bbb57e9c96
* there is now a "Disable Hyper-Threading" safemode in the boot loader
* the SMP & HT menu items are now added in smp.cpp - and are only added
if the system supports one of them.
* more cleanup to smp_apic.h
* removed cpuid() from the boot loader's support.S - instead, it will now
use the one from the kernel.
* added a very weak HT detection: if the MP config only listed one CPU,
and this CPU supports HT, we enable the other logic processor manually -
as this currently doesn't work, it's disabled, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14536 a95241bf-73f2-0310-859d-f6bbb57e9c96
The "data_ptr" parameter is now only freed if you specify SMP_MSG_FLAG_FREE_ARG
when sending the ICI message.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14519 a95241bf-73f2-0310-859d-f6bbb57e9c96
Extracted scheduler_init() from start_scheduler() (which is now called scheduler_start()).
Moved scheduler related function prototypes from thread.h to the new scheduler.h.
Cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14518 a95241bf-73f2-0310-859d-f6bbb57e9c96
having the thread lock held and interrupts disabled.
Cleaned up the signal handling code, and fixed some minor bugs with blockable
vs. non-blockable signals.
thread_debug_info was using uint64 for signals sets instead of sigset_t.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14457 a95241bf-73f2-0310-859d-f6bbb57e9c96
boot message - unfortunately, it crashed when used this way until now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14399 a95241bf-73f2-0310-859d-f6bbb57e9c96
longer called - instead, it get its own mini page fault handler: if the thread
has a fault handler, that one will be executed.
Exiting the debugger (with the "continue" command) no longer causes the
vm_page_fault() routine to be called if interrupts were disabled.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14070 a95241bf-73f2-0310-859d-f6bbb57e9c96
removed the locking it did, as that doesn't work anymore in the kernel debugger.
The function was not thought to be used outside the debugger, anyway.
Improved usage message from debugger command "ls".
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14050 a95241bf-73f2-0310-859d-f6bbb57e9c96
even if "kernel_startup" was "true". Page faults are now no longer allowed
during kernel startup (and could be easily avoided). The only situation where
we accept page faults with interrupts turned off now is during a kernel debugger
session.
Added a command debug_debugger_running() to test for that situation.
kernel_debugger() no longer sets kernel_startup while it's running; there should
be no situation when this could be helpful.
Interrupts are no longer enabled when a page fault happens in the kernel
debugger.
This potentially fixes all sorts of problems, and not only in the kernel debugger,
it could also have affected SMP (will test later).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14045 a95241bf-73f2-0310-859d-f6bbb57e9c96
movement - IOW, the history of kernel debugger calls is now working also from
the on-screen KDL.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@14000 a95241bf-73f2-0310-859d-f6bbb57e9c96
then shows the stack crawl of that stack instead of the current one).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13984 a95241bf-73f2-0310-859d-f6bbb57e9c96
user memory (it got its own fault handler). IOW the stack frame is not changed
in that case anymore.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13921 a95241bf-73f2-0310-859d-f6bbb57e9c96
(there might be some more left, but it's not urgent or fatal)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13884 a95241bf-73f2-0310-859d-f6bbb57e9c96
and ignores keyboard input. Also, it now uses the definitions from the PS/2
HID driver (from ps2.h).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13881 a95241bf-73f2-0310-859d-f6bbb57e9c96
user application performs a division by zero or causes a general
protection fault. For some exceptions (e.g. machine check) I wasn't
quite sure whether they can be caused by user apps at all, so we panic()
in those cases. Wouldn't harm, if someone more knowledgable would check
this, though.
* Removed the unused fault handling stuff, respectively moved the little
that was used into x86/arch_int.c.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13795 a95241bf-73f2-0310-859d-f6bbb57e9c96
the opportunity to show the "System is now ready to reboot" alert.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@13714 a95241bf-73f2-0310-859d-f6bbb57e9c96
called by debug_init_post_vm().
Since the availability of a blue screen specific getchar() is static anyway, there
is no need for the sBlueScreenGetChar variable (only the message "only serial input
available" gets lost, but since that is platform specific anyway...).
Hello blue screen! We now have an on-screen KDL, to be enabled by the kernel
setting "bluescreen", just like on BeOS.
The blue screen does not yet support any cursor actions or backspace, though (need
to grab some stuff from our console driver).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12896 a95241bf-73f2-0310-859d-f6bbb57e9c96
kernel_debugger() didn't do enough before; panic() did all the work - but
since the former is a public function as well, I moved all the functionality
to it. Also fixed a possible buffer overrun in panic().
Renamed dbg_* to debug_*.
"serial_debug_port" setting did not ignore negative values.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12889 a95241bf-73f2-0310-859d-f6bbb57e9c96
interrupt service register bit is never set, anyway...
Thanks to Korli for reporting.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12801 a95241bf-73f2-0310-859d-f6bbb57e9c96
forwarded to the standard interrupt handler. Untested, though.
For more information on the subject: http://www.ggd.nsu.ru/~serg/bsdfaq/part3.html#316
Some cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12800 a95241bf-73f2-0310-859d-f6bbb57e9c96
Lots of cleanup:
- moved B_NO_ENABLE_COUNTER flag definition out of int.h to KernelExport.h, as it's
described in the BeBook (although it's probably not really used that often :))
- int.c no longer has any platform dependent code (+ 0x20 on interrupt numbers is gone);
it's now entirely handled in the arch/x86/ section.
- the io_vectors[] is now statically initialized, instead of allocated from the heap
- removed {install|remove}_interrupt_handler(); they weren't that useful, arch_smp_init()
is now calling install_io_interrupt_handler() correctly instead
- introduced a new arch_int.h header file that currently contains NUM_IO_VECTORS only
(though on x86, it also has ARCH_INTERRUPT_BASE == 0x20).
- changed the return type from {install|remove}_io_interrupt_handler() from "long" to
"status_t"
- rearranged and cleaned the PIC initialization code, made the PIC code more prominent
- changed comments that talk about a non existing 8239 (the PIC chip is actually 8259)
- moved arch/x86/interrupts.h to the source directory, as it's not used outside
- added BeOS compatible interrupts_enabled() function, that should replace our
equivalent (and private) are_interrupts_enabled()
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12477 a95241bf-73f2-0310-859d-f6bbb57e9c96
If anything is still broken (and was not before :)), please shout.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@12361 a95241bf-73f2-0310-859d-f6bbb57e9c96
Axel Dörfler