kernel.git/arch/x86/kernel/traps.c, branch linux-2.6.28.y

x86: fix math_emu register frame access

2009-03-17T00:32:44Z

commit d315760ffa261c15ff92699ac6f514112543d7ca upstream. do_device_not_available() is the handler for #NM and it declares that it takes a unsigned long and calls math_emu(), which takes a long argument and surprisingly expects the stack frame starting at the zero argument would match struct math_emu_info, which isn't true regardless of configuration in the current code. This patch makes do_device_not_available() take struct pt_regs like other exception handlers and initialize struct math_emu_info with pointer to it and pass pointer to the math_emu_info to math_emulate() like normal C functions do. This way, unless gcc makes a copy of struct pt_regs in do_device_not_available(), the register frame is correctly accessed regardless of kernel configuration or compiler used. This doesn't fix all math_emu problems but it at least gets it somewhat working. Signed-off-by: Tejun Heo Signed-off-by: Ingo Molnar Signed-off-by: Greg Kroah-Hartman

x86, vm86: fix preemption bug

2009-02-20T22:40:21Z

commit be716615fe596ee117292dc615e95f707fb67fd1 upstream. Commit 3d2a71a596bd9c761c8487a2178e95f8a61da083 ("x86, traps: converge do_debug handlers") changed the preemption disable logic of do_debug() so vm86_handle_trap() is called with preemption disabled resulting in: BUG: sleeping function called from invalid context at include/linux/kernel.h:155 in_atomic(): 1, irqs_disabled(): 0, pid: 3005, name: dosemu.bin Pid: 3005, comm: dosemu.bin Tainted: G W 2.6.29-rc1 #51 Call Trace: [] copy_to_user+0x33/0x108 [] save_v86_state+0x65/0x149 [] handle_vm86_trap+0x20/0x8f [] do_debug+0x15b/0x1a4 [] debug_stack_correct+0x27/0x2c [] sysenter_do_call+0x12/0x2f BUG: scheduling while atomic: dosemu.bin/3005/0x10000001 Restore the original calling convention and reenable preemption before calling handle_vm86_trap(). Reported-by: Michal Suchanek Signed-off-by: Thomas Gleixner Signed-off-by: Ingo Molnar Signed-off-by: Greg Kroah-Hartman

MCE: Don't run 32bit machine checks with interrupts on

2008-10-22T11:19:01Z

Running machine checks with interrupt on is a extremly bad idea. The machine check handler only runs when the system is broken and needs to finish as quickly as possible. Remove the respective bogus post 2.6.27 regression and call the machine check vector directly again. This removes only code. Signed-off-by: Andi Kleen [Cherry-picked from x86/mce] Signed-off-by: Thomas Gleixner

dumpstack: x86: move die_nmi to dumpstack_32.c

2008-10-13T08:33:39Z

For some reason die_nmi is still defined in traps.c for i386, but is found in dumpstack_64.c for x86_64. Move it to dumpstack_32.c Signed-off-by: Alexander van Heukelum Signed-off-by: Ingo Molnar

traps: x86: finalize unification of traps.c

2008-10-13T08:33:29Z

traps_32.c and traps_64.c are now equal. Move one to traps.c, delete the other one and change the Makefile Signed-off-by: Alexander van Heukelum Signed-off-by: Ingo Molnar