[PATCH v7 1/4] syscalls: Restore address limit after a syscall

Thomas Garnier thgarnie at google.com
Wed Apr 26 10:09:14 EDT 2017


On Wed, Apr 26, 2017 at 1:12 AM, Ingo Molnar <mingo at kernel.org> wrote:
>
> * Thomas Garnier <thgarnie at google.com> wrote:
>
>> >> +#ifdef CONFIG_ARCH_NO_SYSCALL_VERIFY_PRE_USERMODE_STATE
>> >> +/*
>> >> + * This function is called when an architecture specific implementation detected
>> >> + * an invalid address limit. The generic user-mode state checker will finish on
>> >> + * the appropriate BUG_ON.
>> >> + */
>> >> +asmlinkage void address_limit_check_failed(void)
>> >> +{
>> >> +     verify_pre_usermode_state();
>> >> +     panic("address_limit_check_failed called with a valid user-mode state");
>> >
>> > It's very unconstructive to unconditionally panic the system, just because some
>> > kernel code leaked the address limit! Do a warn-once printout and kill the current
>> > task (i.e. don't continue execution), but don't crash everything else!
>>
>> The original change did not crash the kernel for this exact reason.
>> Through reviews, there was an overall agreement that the kernel should
>> not continue in this state.
>
> Ok, I guess we can try that - but the panic message is still pretty misleading:
>
>         panic("address_limit_check_failed called with a valid user-mode state");
>
> ... so it was called with a _valid_ user-mode state, and we crash due to something
> valid? Huh?

Yes the message is accurate but I agree that it is misleading and I
will improve it. The address_limit_check_failed function is called by
assembly code on different architectures once the state was detected
as invalid. Instead of crashing at different places, we redirect to
the generic handler (verify_pre_usermode_state) that will crash on the
appropriate BUG_ON line. The address_limit_check_failed function is
not supposed to comeback, the panic call is just a safe guard.

>
> ( Also, the style rule applies to kernel messages as well: function names should
>   be referred to as "function_name()". )

Will change.

>
> Thanks,
>
>         Ingo



-- 
Thomas



More information about the linux-arm-kernel mailing list