On Nov 21, 2022 / 08:58, Christoph Hellwig wrote: > The fix looks good, but could be simplified a bit more: > > diff --git a/block/blk-mq.c b/block/blk-mq.c > index ee16b4c34c6a5..cdd8efcec1916 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -4381,7 +4381,7 @@ static int blk_mq_realloc_tag_set_tags(struct blk_mq_tag_set *set, > struct blk_mq_tags **new_tags; > > if (set->nr_hw_queues >= new_nr_hw_queues) > - return 0; > + goto done; > > new_tags = kcalloc_node(new_nr_hw_queues, sizeof(struct blk_mq_tags *), > GFP_KERNEL, set->numa_node); > @@ -4393,8 +4393,8 @@ static int blk_mq_realloc_tag_set_tags(struct blk_mq_tag_set *set, > sizeof(*set->tags)); > kfree(set->tags); > set->tags = new_tags; > +done: > set->nr_hw_queues = new_nr_hw_queues; > - > return 0; > } Thanks, this fix is the better. And I reconfirmed it avoids the block/029 and block/030 failures. I guess it is too late for Jens to fold-in this fix in the for-next branch. Christoph, would you prepare a formal fix patch? Or if it helps, I can send out the patch with your authorship and SoB tag (with Co-developed-by tag of mine). -- Shin'ichiro Kawasaki
