On 11/24/21 5:52 AM, Pankaj Raghav wrote:
> Hi Jens,
>
> On Tue, Nov 23, 2021 at 12:15:18PM -0700, Jens Axboe wrote:
>> +bool blk_stats_alloc_enable(struct request_queue *q)
>> +{
>> + struct blk_rq_stat *poll_stat;
>> +
>> + poll_stat = kcalloc(BLK_MQ_POLL_STATS_BKTS, sizeof(*poll_stat),
>> + GFP_ATOMIC);
>> + if (!poll_stat)
>> + return false;
>> +
>> + if (cmpxchg(&q->poll_stat, poll_stat, NULL) != poll_stat) {
> Isn't the logic inverted here? As we already check for non-NULL
> q->poll_stat at the caller side, shouldn't it be:
>
> if (cmpxchg(&q->poll_stat, NULL, poll_stat) != NULL) {
Yes, I did fix that one up right after sending this out, here's the
patch:
https://git.kernel.dk/cgit/linux-block/commit/?h=for-5.17/block&id=987567fece249eabb14406e4e52f427e679d8461
--
Jens Axboe
