powerpc: Skip emulating & leave interrupts off for kernel program checks
authorMichael Ellerman <michael@ellerman.id.au>
Thu, 15 Aug 2013 05:22:19 +0000 (15:22 +1000)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Tue, 27 Aug 2013 04:45:09 +0000 (14:45 +1000)
commitb3f6a45925232d29519007f1ba6687e1e790fb5d
treeb76bc13ff985e56795c2f5a16de8897e7a0f8a0d
parentd671ddd6654daf93c56f6ddff827bed448e1a312
powerpc: Skip emulating & leave interrupts off for kernel program checks

In the program check handler we handle some causes with interrupts off
and others with interrupts on.

We need to enable interrupts to handle the emulation cases, because they
access userspace memory and might sleep.

For faults in the kernel we don't want to do any emulation, and
emulate_instruction() enforces that. do_mathemu() doesn't but probably
should.

The other disadvantage of enabling interrupts for kernel faults is that
we may take another interrupt, and recurse. As seen below:

  --- Exception: e40 at c000000000004ee0 performance_monitor_relon_pSeries_1
  [link register   ] c00000000000f858 .arch_local_irq_restore+0x38/0x90
  [c000000fb185dc100000000000000000 (unreliable)
  [c000000fb185dc80c0000000007d8558 .program_check_exception+0x298/0x2d0
  [c000000fb185dd00c000000000002f40 emulation_assist_common+0x140/0x180
  --- Exception: e40 at c000000000004ee0 performance_monitor_relon_pSeries_1
  [link register   ] c00000000000f858 .arch_local_irq_restore+0x38/0x90
  [c000000fb185dff000000000008b9190 (unreliable)
  [c000000fb185e060c0000000007d8558 .program_check_exception+0x298/0x2d0

So avoid both problems by checking if the fault was in the kernel and
skipping the enable of interrupts and the emulation. Go straight to
delivering the SIGILL, which for kernel faults calls die() and so on,
dropping us in the debugger etc.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/kernel/traps.c