| Message ID | 20220122111054.1126146-10-ming.lei@redhat.com |
|---|---|
| State | New |
| Headers | show |
| Series | block: don't drain file system I/O on del_gendisk | expand |
On Sat, Jan 22, 2022 at 07:10:50PM +0800, Ming Lei wrote: > In scsi_disk_release() request queue is frozen for clearing > disk->private_data, and there can't be any FS IO issued to > this queue, and only private passthrough request will be handled, so > force unfreezing queue into atomic mode. > > Signed-off-by: Ming Lei <ming.lei@redhat.com> > --- > drivers/scsi/sd.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c > index 0e73c3f2f381..27f04c860f00 100644 > --- a/drivers/scsi/sd.c > +++ b/drivers/scsi/sd.c > @@ -3670,7 +3670,7 @@ static void scsi_disk_release(struct device *dev) > * in case multiple processes open a /dev/sd... node concurrently. > */ > blk_mq_freeze_queue(q); > - blk_mq_unfreeze_queue(q); > + __blk_mq_unfreeze_queue(q, true); I think the right thing here is to drop the freeze/unfreeze pair. Now that del_gendisk properly freezes the queue, we don't need this protection as the issue that Bart fixed with it can't happen any more.
On Tue, Jan 25, 2022 at 07:21:55AM +0800, Ming Lei wrote: > > > @@ -3670,7 +3670,7 @@ static void scsi_disk_release(struct device *dev) > > > * in case multiple processes open a /dev/sd... node concurrently. > > > */ > > > blk_mq_freeze_queue(q); > > > - blk_mq_unfreeze_queue(q); > > > + __blk_mq_unfreeze_queue(q, true); > > > > I think the right thing here is to drop the freeze/unfreeze pair. > > Now that del_gendisk properly freezes the queue, we don't need this > > protection as the issue that Bart fixed with it can't happen any more. > > As you see, the last patch removes freeze/unfreeze pair in del_gendisk(), > which looks not very useful: it can't drain IO on bio based driver, and > del_gendisk() is supposed to provide consistent behavior for both request > and bio based driver. So what is the advantage of trying to remove the freeze from where it belongs (common unregister code) while keeping it where it is a bandaid (driver specific unregister code)?
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 0e73c3f2f381..27f04c860f00 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3670,7 +3670,7 @@ static void scsi_disk_release(struct device *dev) * in case multiple processes open a /dev/sd... node concurrently. */ blk_mq_freeze_queue(q); - blk_mq_unfreeze_queue(q); + __blk_mq_unfreeze_queue(q, true); disk->private_data = NULL; put_disk(disk);
In scsi_disk_release() request queue is frozen for clearing disk->private_data, and there can't be any FS IO issued to this queue, and only private passthrough request will be handled, so force unfreezing queue into atomic mode. Signed-off-by: Ming Lei <ming.lei@redhat.com> --- drivers/scsi/sd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)