Re: [PATCH 1/3] arm/syscalls: Move address limit check in loop

[Thomas Garnier]

On Wed, Jul 26, 2017 at 11:25 AM, Russell King - ARM Linux
<linux@xxxxxxxxxxxxxxx> wrote:
> On Wed, Jul 26, 2017 at 07:20:22AM -0700, Thomas Garnier wrote:
>> On Wed, Jul 26, 2017 at 5:02 AM, Will Deacon <will.deacon@xxxxxxx> wrote:
>> > I looked to see what you've done for x86, but it looks like you check/clear
>> > the flag before the work pending loop (exit_to_usermode_loop), which
>> > subsequently re-enables interrupts and exits when
>> > EXIT_TO_USERMODE_LOOP_FLAGS are all clear. Since TIF_FSCHECK isn't included
>> > in those flags, what stops it being set again by an irq and remaining set
>> > for the return to userspace?
>>
>> Nothing, I plan to improve the x86 logic later. I focused on ARM/ARM64
>> right now based on Leonard report.
>
> Hmm.  In this case, I'd suggest concentrating on x86 and getting the
> implementation correct there before porting it to other architectures.

I think the ARM architecture is different.

>
> If x86 were to check TIF_FSCHECK in the loop, and repeat until clear,
> would x86 also end up in these infinite loops that have been reported
> on ARM as well?

If for every loop set_fs was called. I think the idea is that at some
point only the TIF_FSCHECK remain. I don't think x86 suffers from the
same issue than ARM.

>
> I strongly suggest testing the behaviour with kprobes/tracing enabled
> for a function called from the work pending loop, and checking how
> that behaves.
>
> --
> RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
> FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
> according to speedtest.net.



-- 
Thomas