On Sat, Aug 16, 2025 at 04:05:30PM +0800, 余快 wrote:
> Hi,
>
> Ming Lei <ming.lei@xxxxxxxxxx> 于2025年8月16日周六 12:05写道：
> >
> > On Sat, Aug 16, 2025 at 10:57:23AM +0800, Yu Kuai wrote:
> > > Hi,
> > >
> > > 在 2025/8/16 3:30, Nilay Shroff 写道:
> > > >
> > > > On 8/15/25 1:32 PM, Yu Kuai wrote:
> > > > > From: Yu Kuai <yukuai3@xxxxxxxxxx>
> > > > >
> > > > > In the case user trigger tags grow by queue sysfs attribute nr_requests,
> > > > > hctx->sched_tags will be freed directly and replaced with a new
> > > > > allocated tags, see blk_mq_tag_update_depth().
> > > > >
> > > > > The problem is that hctx->sched_tags is from elevator->et->tags, while
> > > > > et->tags is still the freed tags, hence later elevator exist will try to
> > > > > free the tags again, causing kernel panic.
> > > > >
> > > > > Fix this problem by using new allocated elevator_tags, also convert
> > > > > blk_mq_update_nr_requests to void since this helper will never fail now.
> > > > >
> > > > > Meanwhile, there is a longterm problem can be fixed as well:
> > > > >
> > > > > If blk_mq_tag_update_depth() succeed for previous hctx, then bitmap depth
> > > > > is updated, however, if following hctx failed, q->nr_requests is not
> > > > > updated and the previous hctx->sched_tags endup bigger than q->nr_requests.
> > > > >
> > > > > Fixes: f5a6604f7a44 ("block: fix lockdep warning caused by lock dependency in elv_iosched_store")
> > > > > Fixes: e3a2b3f931f5 ("blk-mq: allow changing of queue depth through sysfs")
> > > > > Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
> > > > > ---
> > > > > block/blk-mq.c | 19 ++++++-------------
> > > > > block/blk-mq.h | 4 +++-
> > > > > block/blk-sysfs.c | 21 ++++++++++++++-------
> > > > > 3 files changed, 23 insertions(+), 21 deletions(-)
> > > > >
> > > > > diff --git a/block/blk-mq.c b/block/blk-mq.c
> > > > > index 11c8baebb9a0..e9f037a25fe3 100644
> > > > > --- a/block/blk-mq.c
> > > > > +++ b/block/blk-mq.c
> > > > > @@ -4917,12 +4917,12 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
> > > > > }
> > > > > EXPORT_SYMBOL(blk_mq_free_tag_set);
> > > > > -int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
> > > > > +void blk_mq_update_nr_requests(struct request_queue *q,
> > > > > + struct elevator_tags *et, unsigned int nr)
> > > > > {
> > > > > struct blk_mq_tag_set *set = q->tag_set;
> > > > > struct blk_mq_hw_ctx *hctx;
> > > > > unsigned long i;
> > > > > - int ret = 0;
> > > > > blk_mq_quiesce_queue(q);
> > > > > @@ -4946,24 +4946,17 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
> > > > > nr - hctx->sched_tags->nr_reserved_tags);
> > > > > }
> > > > > } else {
> > > > > - queue_for_each_hw_ctx(q, hctx, i) {
> > > > > - if (!hctx->tags)
> > > > > - continue;
> > > > > - ret = blk_mq_tag_update_depth(hctx, &hctx->sched_tags,
> > > > > - nr);
> > > > > - if (ret)
> > > > > - goto out;
> > > > > - }
> > > > > + blk_mq_free_sched_tags(q->elevator->et, set);
> > > > I think you also need to ensure that elevator tags are freed after we unfreeze
> > > > queue and release ->elevator_lock otherwise we may get into the lockdep splat
> > > > for pcpu_lock dependency on ->freeze_lock and/or ->elevator_lock. Please note
> > > > that blk_mq_free_sched_tags internally invokes sbitmap_free which invokes
> > > > free_percpu which acquires pcpu_lock.
> > >
> > > Ok, thanks for the notice. However, as Ming suggested, we might fix this
> > > problem
> > >
> > > in the next merge window.
> >
> > There are two issues involved:
> >
> > - blk_mq_tags double free, introduced recently
> >
> > - long-term lock issue in queue_requests_store()
> >
> > IMO, the former is a bit serious, because kernel panic can be triggered,
> > so suggest to make it to v6.17. The latter looks less serious and has
> > existed for long time, but may need code refactor to get clean fix.
> >
> > > I'll send one patch to fix this regression by
> > > replace
> > >
> > > st->tags with reallocated new sched_tags as well.
> >
> > Patch 7 in this patchset and patch 8 in your 1st post looks enough to
> > fix this double free issue.
> >
> But without previous refactor, this looks hard. Can we consider the following
I feel the following one should be enough:
diff --git a/block/blk-mq.c b/block/blk-mq.c
index b67d6c02eceb..021155bb64fc 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -4917,6 +4917,29 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
}
EXPORT_SYMBOL(blk_mq_free_tag_set);
+static int blk_mq_sched_grow_tags(struct request_queue *q, unsigned int nr)
+{
+ struct blk_mq_hw_ctx *hctx;
+ struct elevator_tags *et;
+ unsigned long i;
+
+ if (nr > MAX_SCHED_RQ)
+ return -EINVAL;
+
+ et = blk_mq_alloc_sched_tags(q->tag_set, q->nr_hw_queues, nr);
+ if (!et)
+ return -ENOMEM;
+
+ blk_mq_free_sched_tags(q->elevator->et, q->tag_set);
+ queue_for_each_hw_ctx(q, hctx, i) {
+ hctx->sched_tags = et->tags[i];
+ if (q->elevator->type->ops.depth_updated)
+ q->elevator->type->ops.depth_updated(hctx);
+ }
+ q->elevator->et = et;
+ return 0;
+}
+
int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
{
struct blk_mq_tag_set *set = q->tag_set;
@@ -4936,6 +4959,12 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
blk_mq_quiesce_queue(q);
ret = 0;
+
+ if (q->elevator && nr > q->elevator->et->nr_requests) {
+ ret = blk_mq_sched_grow_tags(q, nr);
+ goto update_nr_requests;
+ }
+
queue_for_each_hw_ctx(q, hctx, i) {
if (!hctx->tags)
continue;
@@ -4955,6 +4984,7 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
if (q->elevator && q->elevator->type->ops.depth_updated)
q->elevator->type->ops.depth_updated(hctx);
}
+update_nr_requests:
if (!ret) {
q->nr_requests = nr;
if (blk_mq_is_shared_tags(set->flags)) {
> one line patch for this merge window? just fix the first double free
> issue for now.
>
> diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
> index d880c50629d6..1e0ccf19295a 100644
> --- a/block/blk-mq-tag.c
> +++ b/block/blk-mq-tag.c
> @@ -622,6 +622,7 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
> return -ENOMEM;
>
> blk_mq_free_map_and_rqs(set, *tagsptr, hctx->queue_num);
> + hctx->queue->elevator->et->tags[hctx->queue_num]= new;
> *tagsptr = new;
It is fine if this way can work, however old elevator->et may has lower
depth, then:
- the above change cause et->tags overflow
- meantime memory leak is caused in blk_mq_free_sched_tags()
Thanks,
Ming
