lkml.org 
[lkml]   [2016]   [Jun]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/2] ipc/sem.c: Fix complex_count vs. simple op race
On Thu, 23 Jun 2016, Manfred Spraul wrote:

>What I'm not sure yet is if smp_load_acquire() is sufficient:
>
>Thread A:
>> if (!READ_ONCE(sma->complex_mode)) {
>The code is test_and_test, no barrier requirements for first test

Yeah, it would just make us take the big lock unnecessarily, nothing fatal
and I agree its probably worth the optimization. It still might be worth
commenting.

>> /*
>> * It appears that no complex operation is around.
>> * Acquire the per-semaphore lock.
>> */
>> spin_lock(&sem->lock);
>>
>> if (!smp_load_acquire(&sma->complex_mode)) {
>> /* fast path successful! */
>> return sops->sem_num;
>> }
>> spin_unlock(&sem->lock);
>> }
>
>Thread B:
>> WRITE_ONCE(sma->complex_mode, true);
>>
>> /* We need a full barrier:
>> * The write to complex_mode must be visible
>> * before we read the first sem->lock spinlock state.
>> */
>> smp_mb();
>>
>> for (i = 0; i < sma->sem_nsems; i++) {
>> sem = sma->sem_base + i;
>> spin_unlock_wait(&sem->lock);
>> }
>
>If thread A is allowed to issue read_spinlock;read complex_mode;write
>spinlock, then thread B would not notice that thread A is in the
>critical section

Are you referring to the sem->lock word not being visibly locked before we
read complex_mode (for the second time)? This issue was fixed in 2c610022711
(locking/qspinlock: Fix spin_unlock_wait() some more). So smp_load_acquire
should be enough afaict, or are you referring to something else?

Thanks,
Davidlohr

\
 
 \ /
  Last update: 2016-06-28 08:01    [W:1.092 / U:1.532 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site