Re: [PATCH 1/3] x86_64,entry: Fix RCX for traced syscalls

From: Andy Lutomirski
Date: Tue Jan 06 2015 - 13:44:23 EST


On Jan 6, 2015 7:34 AM, "Borislav Petkov" <bp@xxxxxxxxx> wrote:
>
> On Mon, Jan 05, 2015 at 12:31:15PM -0800, Andy Lutomirski wrote:
> > Do you have context tracking on?
>
> Yap, it is enabled for whatever reason:
> CONFIG_CONTEXT_TRACKING=y
> CONFIG_CONTEXT_TRACKING_FORCE=y
> CONFIG_HAVE_CONTEXT_TRACKING=y

I'll boot a kernel like this on bare metal and see what shakes loose.

>
> > I assume that's in the historical tree?
>
> Yeah.
>
> > > [ 180.059170] ata1.00: exception Emask 0x0 SAct 0x7fffffff SErr 0x0 action 0x6 frozen
> > > [ 180.066873] ata1.00: failed command: WRITE FPDMA QUEUED
> > > [ 180.072158] ata1.00: cmd 61/08:00:a8:ac:d9/00:00:23:00:00/40 tag 0 ncq 4096 out
> > > [ 180.072158] res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout)
> >
> > That's really weird. The only thing I can think of is that somehow we
> > returned to user mode without enabling interrupts.
>
> Right, considering FIXUP_TOP_OF_STACK is used in a bunch of cases in
> entry_64.S, no wonder it corrupts something.
>
> > This leads me to wonder: why do we save eflags in the R11 pt_regs
> > slot?
>
> That: "If executed in 64-bit mode, SYSRET loads the lower-32 RFLAGS bits
> from R11[31:0] and clears the upper 32 RFLAGS bits."

Sure, but the code would be simpler if we shoved that value in the EFLAGS slot.

>
> > This seems entirely backwards, not to mention that it accounts for two
> > instructions in each of FIXUP_TOP_OF_STACK and RESTORE_TOP_OF_STACK
> > for no apparently reason whatsoever.
>
> > Can you send the full output from syscall_exit_regs_64 from here:
> >
> > https://gitorious.org/linux-test-utils/linux-clock-tests/source/34884122b6ebe81d9b96e3e5128b6d6d95082c6e:
> >
> > with the patch applied (assuming it even gets that far for you)? I
> > see results like:
> >
> > [NOTE] syscall ffff: orig RCX = 1 ss = 2b orig_ss = 6b flags =
> > 217 orig_flags = 217
> >
> > which seems fine.
>
> ./syscall_exit_regs_64
> [OK] int80 ffff: AX = ffffffffffffffda
> [OK] int80 40000000: AX = ffffffffffffffda
> [OK] syscall ffff: RCX = 400962 RIP = 400962
> [OK] syscall ffff: AX = ffffffffffffffda
> [NOTE] syscall ffff: orig RCX = 1 ss = 2b orig_ss = 6b flags = 217 orig_flags = 217
> [OK] syscall 40000000: RCX = 400962 RIP = 400962
> [FAIL] syscall 40000000: AX = fffffffffffffff7
> [NOTE] syscall 40000000: orig RCX = 1 ss = 2b orig_ss = 6b flags = 217 orig_flags = 217
> [OK] syscall(ffff): ret = -1, errno = 38
>
> > Are you seeing this with the whole series applied or with only this patch?
>
> I applied this patch only and started seeing those. Then I booted in the
> previous kernel and tried to repro but it didn't trigger.
>
> I'll try hammering on the kernel *without* your patch to see whether I
> can trigger it somehow...

Hmm. I added and pushed a test for fork, but that didn't turn
anything up. And I don't see any bugs in the code.

I booted 3.18 plus this patch with context tracking forced on on my
laptop, and something seems to have gone wrong.

--Andy

>
> --
> Regards/Gruss,
> Boris.
>
> Sent from a fat crate under my desk. Formatting is fine.
> --
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/