Re: [CHECKER] repetitive/contradictory comparison bugs for 2.4.7

[Alex Bligh - linux-kernel]

<[e-mail]> wrote:
>> How will this be guaranteed to help handle a race, when gcc is
>> likely either to have tmp_buf in a register (not declared
>> volatile), or perhaps even optimize out the second reference.
>
> The function call is a synchronization pointAlan,

Ooops - indeed, yes. Though the cli()/restore_flags() doesn't seem
like the right way to perform a lock. My naive interpretation is
that either it needs a (faster) real lock that doesn't disable IRQs on
multiple CPUs etc., or lock_kernel does the job in which
case cli()/restore_flags() is redundant, and, indeed, so is the
check itself. May be I have missed something (again).

I am presuming even an inline cli() is a synchronization point too,
else there would still be a race if the read of tmp_buf in
if(tmp_buf) and the cli() were reordered.

--
Alex Bligh