* fix block device size update serialization
@ 2020-08-21 8:55 ` Christoph Hellwig
0 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-21 8:55 UTC (permalink / raw)
To: Jens Axboe
Cc: linux-s390, Jan Hoeppner, Justin Sanders, linux-nvme,
Josef Bacik, Xianting Tian, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd
Hi Jens,
this series fixes how we update i_size for the block device inodes (and
thus the block device). Different helpers use two different locks
(bd_mutex and i_rwsem) to protect the update, and it appears device
mapper uses yet another internal lock. A lot of the drivers do the
update handcrafted in often crufty ways. And in addition to that mess
it turns out that the "main" lock, bd_mutex is pretty dead lock prone
vs other spots in the block layer that acquire it during revalidation
operations, as reported by Xianting.
Fix all that by adding a dedicated spinlock just for the size updates.
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH 1/2] block: replace bd_set_size with bd_set_nr_sectors
2020-08-21 8:55 ` Christoph Hellwig
@ 2020-08-21 8:55 ` Christoph Hellwig
-1 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-21 8:55 UTC (permalink / raw)
To: Jens Axboe
Cc: Justin Sanders, Josef Bacik, Xianting Tian, linux-block,
dm-devel, Stefan Haberland, Jan Hoeppner, linux-kernel, nbd,
linux-nvme, linux-s390
Replace bd_set_size with a version that takes the number of sectors
instead, as that fits most of the current and future callers much better.
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
drivers/block/loop.c | 4 ++--
drivers/block/nbd.c | 7 ++++---
drivers/block/pktcdvd.c | 2 +-
drivers/nvme/host/nvme.h | 2 +-
fs/block_dev.c | 10 +++++-----
include/linux/genhd.h | 2 +-
6 files changed, 14 insertions(+), 13 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 2f137d6ce169d5..7069899a94903e 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -253,7 +253,7 @@ static void loop_set_size(struct loop_device *lo, loff_t size)
{
struct block_device *bdev = lo->lo_device;
- bd_set_size(bdev, size << SECTOR_SHIFT);
+ bd_set_nr_sectors(bdev, size);
set_capacity_revalidate_and_notify(lo->lo_disk, size, false);
}
@@ -1248,7 +1248,7 @@ static int __loop_clr_fd(struct loop_device *lo, bool release)
set_capacity(lo->lo_disk, 0);
loop_sysfs_exit(lo);
if (bdev) {
- bd_set_size(bdev, 0);
+ bd_set_nr_sectors(bdev, 0);
/* let user-space know about this change */
kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE);
}
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 3ff4054d6834d2..f07243335472a4 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -300,6 +300,7 @@ static void nbd_size_update(struct nbd_device *nbd)
{
struct nbd_config *config = nbd->config;
struct block_device *bdev = bdget_disk(nbd->disk, 0);
+ sector_t nr_sectors = config->bytesize >> 9;
if (config->flags & NBD_FLAG_SEND_TRIM) {
nbd->disk->queue->limits.discard_granularity = config->blksize;
@@ -308,10 +309,10 @@ static void nbd_size_update(struct nbd_device *nbd)
}
blk_queue_logical_block_size(nbd->disk->queue, config->blksize);
blk_queue_physical_block_size(nbd->disk->queue, config->blksize);
- set_capacity(nbd->disk, config->bytesize >> 9);
+ set_capacity(nbd->disk, nr_sectors);
if (bdev) {
if (bdev->bd_disk) {
- bd_set_size(bdev, config->bytesize);
+ bd_set_nr_sectors(bdev, nr_sectors);
set_blocksize(bdev, config->blksize);
} else
bdev->bd_invalidated = 1;
@@ -1138,7 +1139,7 @@ static void nbd_bdev_reset(struct block_device *bdev)
{
if (bdev->bd_openers > 1)
return;
- bd_set_size(bdev, 0);
+ bd_set_nr_sectors(bdev, 0);
}
static void nbd_parse_flags(struct nbd_device *nbd)
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index 4becc1efe775fc..015fe128fa8a35 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -2192,7 +2192,7 @@ static int pkt_open_dev(struct pktcdvd_device *pd, fmode_t write)
set_capacity(pd->disk, lba << 2);
set_capacity(pd->bdev->bd_disk, lba << 2);
- bd_set_size(pd->bdev, (loff_t)lba << 11);
+ bd_set_nr_sectors(pd->bdev, lba << 2);
q = bdev_get_queue(pd->bdev);
if (write) {
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index ebb8c3ed388554..ae5cad5a08f411 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -664,7 +664,7 @@ static inline void nvme_mpath_update_disk_size(struct gendisk *disk)
struct block_device *bdev = bdget_disk(disk, 0);
if (bdev) {
- bd_set_size(bdev, get_capacity(disk) << SECTOR_SHIFT);
+ bd_set_nr_sectors(bdev, get_capacity(disk));
bdput(bdev);
}
}
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 8ae833e004439b..f52597172c8b79 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1371,13 +1371,13 @@ int check_disk_change(struct block_device *bdev)
EXPORT_SYMBOL(check_disk_change);
-void bd_set_size(struct block_device *bdev, loff_t size)
+void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors)
{
inode_lock(bdev->bd_inode);
- i_size_write(bdev->bd_inode, size);
+ i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
inode_unlock(bdev->bd_inode);
}
-EXPORT_SYMBOL(bd_set_size);
+EXPORT_SYMBOL(bd_set_nr_sectors);
static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
@@ -1514,7 +1514,7 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
}
if (!ret) {
- bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
+ bd_set_nr_sectors(bdev, get_capacity(disk));
set_init_blocksize(bdev);
}
@@ -1542,7 +1542,7 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
ret = -ENXIO;
goto out_clear;
}
- bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
+ bd_set_nr_sectors(bdev, bdev->bd_part->nr_sects);
set_init_blocksize(bdev);
}
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index 4ab853461dff25..39025dc0397c04 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -375,7 +375,7 @@ void unregister_blkdev(unsigned int major, const char *name);
int revalidate_disk(struct gendisk *disk);
int check_disk_change(struct block_device *bdev);
int __invalidate_device(struct block_device *bdev, bool kill_dirty);
-void bd_set_size(struct block_device *bdev, loff_t size);
+void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors);
/* for drivers/char/raw.c: */
int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long);
--
2.28.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH 1/2] block: replace bd_set_size with bd_set_nr_sectors
@ 2020-08-21 8:55 ` Christoph Hellwig
0 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-21 8:55 UTC (permalink / raw)
To: Jens Axboe
Cc: linux-s390, Jan Hoeppner, Justin Sanders, linux-nvme,
Josef Bacik, Xianting Tian, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd
Replace bd_set_size with a version that takes the number of sectors
instead, as that fits most of the current and future callers much better.
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
drivers/block/loop.c | 4 ++--
drivers/block/nbd.c | 7 ++++---
drivers/block/pktcdvd.c | 2 +-
drivers/nvme/host/nvme.h | 2 +-
fs/block_dev.c | 10 +++++-----
include/linux/genhd.h | 2 +-
6 files changed, 14 insertions(+), 13 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index 2f137d6ce169d5..7069899a94903e 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -253,7 +253,7 @@ static void loop_set_size(struct loop_device *lo, loff_t size)
{
struct block_device *bdev = lo->lo_device;
- bd_set_size(bdev, size << SECTOR_SHIFT);
+ bd_set_nr_sectors(bdev, size);
set_capacity_revalidate_and_notify(lo->lo_disk, size, false);
}
@@ -1248,7 +1248,7 @@ static int __loop_clr_fd(struct loop_device *lo, bool release)
set_capacity(lo->lo_disk, 0);
loop_sysfs_exit(lo);
if (bdev) {
- bd_set_size(bdev, 0);
+ bd_set_nr_sectors(bdev, 0);
/* let user-space know about this change */
kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, KOBJ_CHANGE);
}
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 3ff4054d6834d2..f07243335472a4 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -300,6 +300,7 @@ static void nbd_size_update(struct nbd_device *nbd)
{
struct nbd_config *config = nbd->config;
struct block_device *bdev = bdget_disk(nbd->disk, 0);
+ sector_t nr_sectors = config->bytesize >> 9;
if (config->flags & NBD_FLAG_SEND_TRIM) {
nbd->disk->queue->limits.discard_granularity = config->blksize;
@@ -308,10 +309,10 @@ static void nbd_size_update(struct nbd_device *nbd)
}
blk_queue_logical_block_size(nbd->disk->queue, config->blksize);
blk_queue_physical_block_size(nbd->disk->queue, config->blksize);
- set_capacity(nbd->disk, config->bytesize >> 9);
+ set_capacity(nbd->disk, nr_sectors);
if (bdev) {
if (bdev->bd_disk) {
- bd_set_size(bdev, config->bytesize);
+ bd_set_nr_sectors(bdev, nr_sectors);
set_blocksize(bdev, config->blksize);
} else
bdev->bd_invalidated = 1;
@@ -1138,7 +1139,7 @@ static void nbd_bdev_reset(struct block_device *bdev)
{
if (bdev->bd_openers > 1)
return;
- bd_set_size(bdev, 0);
+ bd_set_nr_sectors(bdev, 0);
}
static void nbd_parse_flags(struct nbd_device *nbd)
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index 4becc1efe775fc..015fe128fa8a35 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -2192,7 +2192,7 @@ static int pkt_open_dev(struct pktcdvd_device *pd, fmode_t write)
set_capacity(pd->disk, lba << 2);
set_capacity(pd->bdev->bd_disk, lba << 2);
- bd_set_size(pd->bdev, (loff_t)lba << 11);
+ bd_set_nr_sectors(pd->bdev, lba << 2);
q = bdev_get_queue(pd->bdev);
if (write) {
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index ebb8c3ed388554..ae5cad5a08f411 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -664,7 +664,7 @@ static inline void nvme_mpath_update_disk_size(struct gendisk *disk)
struct block_device *bdev = bdget_disk(disk, 0);
if (bdev) {
- bd_set_size(bdev, get_capacity(disk) << SECTOR_SHIFT);
+ bd_set_nr_sectors(bdev, get_capacity(disk));
bdput(bdev);
}
}
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 8ae833e004439b..f52597172c8b79 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -1371,13 +1371,13 @@ int check_disk_change(struct block_device *bdev)
EXPORT_SYMBOL(check_disk_change);
-void bd_set_size(struct block_device *bdev, loff_t size)
+void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors)
{
inode_lock(bdev->bd_inode);
- i_size_write(bdev->bd_inode, size);
+ i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
inode_unlock(bdev->bd_inode);
}
-EXPORT_SYMBOL(bd_set_size);
+EXPORT_SYMBOL(bd_set_nr_sectors);
static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part);
@@ -1514,7 +1514,7 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
}
if (!ret) {
- bd_set_size(bdev,(loff_t)get_capacity(disk)<<9);
+ bd_set_nr_sectors(bdev, get_capacity(disk));
set_init_blocksize(bdev);
}
@@ -1542,7 +1542,7 @@ static int __blkdev_get(struct block_device *bdev, fmode_t mode, void *holder,
ret = -ENXIO;
goto out_clear;
}
- bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9);
+ bd_set_nr_sectors(bdev, bdev->bd_part->nr_sects);
set_init_blocksize(bdev);
}
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index 4ab853461dff25..39025dc0397c04 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -375,7 +375,7 @@ void unregister_blkdev(unsigned int major, const char *name);
int revalidate_disk(struct gendisk *disk);
int check_disk_change(struct block_device *bdev);
int __invalidate_device(struct block_device *bdev, bool kill_dirty);
-void bd_set_size(struct block_device *bdev, loff_t size);
+void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors);
/* for drivers/char/raw.c: */
int blkdev_ioctl(struct block_device *, fmode_t, unsigned, unsigned long);
--
2.28.0
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH 1/2] block: replace bd_set_size with bd_set_nr_sectors
2020-08-21 8:55 ` Christoph Hellwig
@ 2020-08-21 9:24 ` Hannes Reinecke
-1 siblings, 0 replies; 13+ messages in thread
From: Hannes Reinecke @ 2020-08-21 9:24 UTC (permalink / raw)
To: Christoph Hellwig, Jens Axboe
Cc: Justin Sanders, Josef Bacik, Xianting Tian, linux-block,
dm-devel, Stefan Haberland, Jan Hoeppner, linux-kernel, nbd,
linux-nvme, linux-s390
On 8/21/20 10:55 AM, Christoph Hellwig wrote:
> Replace bd_set_size with a version that takes the number of sectors
> instead, as that fits most of the current and future callers much better.
>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> drivers/block/loop.c | 4 ++--
> drivers/block/nbd.c | 7 ++++---
> drivers/block/pktcdvd.c | 2 +-
> drivers/nvme/host/nvme.h | 2 +-
> fs/block_dev.c | 10 +++++-----
> include/linux/genhd.h | 2 +-
> 6 files changed, 14 insertions(+), 13 deletions(-)
>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 1/2] block: replace bd_set_size with bd_set_nr_sectors
@ 2020-08-21 9:24 ` Hannes Reinecke
0 siblings, 0 replies; 13+ messages in thread
From: Hannes Reinecke @ 2020-08-21 9:24 UTC (permalink / raw)
To: Christoph Hellwig, Jens Axboe
Cc: linux-s390, Jan Hoeppner, Justin Sanders, linux-nvme,
Josef Bacik, Xianting Tian, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd
On 8/21/20 10:55 AM, Christoph Hellwig wrote:
> Replace bd_set_size with a version that takes the number of sectors
> instead, as that fits most of the current and future callers much better.
>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> drivers/block/loop.c | 4 ++--
> drivers/block/nbd.c | 7 ++++---
> drivers/block/pktcdvd.c | 2 +-
> drivers/nvme/host/nvme.h | 2 +-
> fs/block_dev.c | 10 +++++-----
> include/linux/genhd.h | 2 +-
> 6 files changed, 14 insertions(+), 13 deletions(-)
>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH 2/2] block: fix locking for struct block_device size updates
2020-08-21 8:55 ` Christoph Hellwig
@ 2020-08-21 8:56 ` Christoph Hellwig
-1 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-21 8:56 UTC (permalink / raw)
To: Jens Axboe
Cc: Justin Sanders, Josef Bacik, Xianting Tian, linux-block,
dm-devel, Stefan Haberland, Jan Hoeppner, linux-kernel, nbd,
linux-nvme, linux-s390
Two different callers use two different mutexes for updating the
block device size, which obviously doesn't help to actually protect
against concurrent updates from the different callers. In addition
one of the locks, bd_mutex is rather prone to deadlocks with other
parts of the block stack that use it for high level synchronization.
Switch to using a new spinlock protecting just the size updates, as
that is all we need, and make sure everyone does the update through
the proper helper.
This fixeѕ a bug reported with the nvme revalidating disks during a
hot removal operation.
Reported-by: Xianting Tian <xianting_tian@126.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
block/partitions/core.c | 4 ++--
drivers/block/aoe/aoecmd.c | 4 +---
drivers/md/dm.c | 15 ++-------------
drivers/s390/block/dasd_ioctl.c | 9 ++-------
fs/block_dev.c | 18 +++++++++---------
include/linux/blk_types.h | 1 +
6 files changed, 17 insertions(+), 34 deletions(-)
diff --git a/block/partitions/core.c b/block/partitions/core.c
index e62a98a8eeb750..328a2cb7875ba1 100644
--- a/block/partitions/core.c
+++ b/block/partitions/core.c
@@ -585,8 +585,8 @@ int bdev_resize_partition(struct block_device *bdev, int partno,
if (partition_overlaps(bdev->bd_disk, start, length, partno))
goto out_unlock;
- part_nr_sects_write(part, (sector_t)length);
- i_size_write(bdevp->bd_inode, length << SECTOR_SHIFT);
+ part_nr_sects_write(part, length);
+ bd_set_nr_sectors(bdevp, length);
ret = 0;
out_unlock:
diff --git a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
index 3cf9bc5d8d9599..6ad73fe730bede 100644
--- a/drivers/block/aoe/aoecmd.c
+++ b/drivers/block/aoe/aoecmd.c
@@ -900,9 +900,7 @@ aoecmd_sleepwork(struct work_struct *work)
ssize = get_capacity(d->gd);
bd = bdget_disk(d->gd, 0);
if (bd) {
- inode_lock(bd->bd_inode);
- i_size_write(bd->bd_inode, (loff_t)ssize<<9);
- inode_unlock(bd->bd_inode);
+ bd_set_nr_sectors(bd, ssize);
bdput(bd);
}
spin_lock_irq(&d->lock);
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 32fa6499739fb9..6b21e5104e3e08 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -2097,18 +2097,6 @@ static void event_callback(void *context)
dm_issue_global_event();
}
-/*
- * Protected by md->suspend_lock obtained by dm_swap_table().
- */
-static void __set_size(struct mapped_device *md, sector_t size)
-{
- lockdep_assert_held(&md->suspend_lock);
-
- set_capacity(md->disk, size);
-
- i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
-}
-
/*
* Returns old map, which caller must destroy.
*/
@@ -2131,7 +2119,8 @@ static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
if (size != dm_get_size(md))
memset(&md->geometry, 0, sizeof(md->geometry));
- __set_size(md, size);
+ set_capacity(md->disk, size);
+ bd_set_nr_sectors(md->bdev, size);
dm_table_event_callback(t, event_callback, md);
diff --git a/drivers/s390/block/dasd_ioctl.c b/drivers/s390/block/dasd_ioctl.c
index 777734d1b4e58c..faaf5596e31c12 100644
--- a/drivers/s390/block/dasd_ioctl.c
+++ b/drivers/s390/block/dasd_ioctl.c
@@ -55,10 +55,7 @@ dasd_ioctl_enable(struct block_device *bdev)
dasd_enable_device(base);
/* Formatting the dasd device can change the capacity. */
- mutex_lock(&bdev->bd_mutex);
- i_size_write(bdev->bd_inode,
- (loff_t)get_capacity(base->block->gdp) << 9);
- mutex_unlock(&bdev->bd_mutex);
+ bd_set_nr_sectors(bdev, get_capacity(base->block->gdp));
dasd_put_device(base);
return 0;
}
@@ -91,9 +88,7 @@ dasd_ioctl_disable(struct block_device *bdev)
* Set i_size to zero, since read, write, etc. check against this
* value.
*/
- mutex_lock(&bdev->bd_mutex);
- i_size_write(bdev->bd_inode, 0);
- mutex_unlock(&bdev->bd_mutex);
+ bd_set_nr_sectors(bdev, 0);
dasd_put_device(base);
return 0;
}
diff --git a/fs/block_dev.c b/fs/block_dev.c
index f52597172c8b79..aa559f1f08800e 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -876,6 +876,7 @@ struct block_device *bdget(dev_t dev)
bdev = &BDEV_I(inode)->bdev;
if (inode->i_state & I_NEW) {
+ spin_lock_init(&bdev->bd_size_lock);
bdev->bd_contains = NULL;
bdev->bd_super = NULL;
bdev->bd_inode = inode;
@@ -1290,6 +1291,7 @@ static void check_disk_size_change(struct gendisk *disk,
{
loff_t disk_size, bdev_size;
+ spin_lock(&bdev->bd_size_lock);
disk_size = (loff_t)get_capacity(disk) << 9;
bdev_size = i_size_read(bdev->bd_inode);
if (disk_size != bdev_size) {
@@ -1304,6 +1306,7 @@ static void check_disk_size_change(struct gendisk *disk,
disk->disk_name);
}
bdev->bd_invalidated = 0;
+ spin_unlock(&bdev->bd_size_lock);
}
/**
@@ -1328,13 +1331,10 @@ int revalidate_disk(struct gendisk *disk)
if (!(disk->flags & GENHD_FL_HIDDEN)) {
struct block_device *bdev = bdget_disk(disk, 0);
- if (!bdev)
- return ret;
-
- mutex_lock(&bdev->bd_mutex);
- check_disk_size_change(disk, bdev, ret == 0);
- mutex_unlock(&bdev->bd_mutex);
- bdput(bdev);
+ if (bdev) {
+ check_disk_size_change(disk, bdev, ret == 0);
+ bdput(bdev);
+ }
}
return ret;
}
@@ -1373,9 +1373,9 @@ EXPORT_SYMBOL(check_disk_change);
void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors)
{
- inode_lock(bdev->bd_inode);
+ spin_lock(&bdev->bd_size_lock);
i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
- inode_unlock(bdev->bd_inode);
+ spin_unlock(&bdev->bd_size_lock);
}
EXPORT_SYMBOL(bd_set_nr_sectors);
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index 4ecf4fed171f0d..5accc2549d2259 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -38,6 +38,7 @@ struct block_device {
/* number of times partitions within this device have been opened. */
unsigned bd_part_count;
int bd_invalidated;
+ spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
struct gendisk * bd_disk;
struct backing_dev_info *bd_bdi;
--
2.28.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH 2/2] block: fix locking for struct block_device size updates
@ 2020-08-21 8:56 ` Christoph Hellwig
0 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-21 8:56 UTC (permalink / raw)
To: Jens Axboe
Cc: linux-s390, Jan Hoeppner, Justin Sanders, linux-nvme,
Josef Bacik, Xianting Tian, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd
Two different callers use two different mutexes for updating the
block device size, which obviously doesn't help to actually protect
against concurrent updates from the different callers. In addition
one of the locks, bd_mutex is rather prone to deadlocks with other
parts of the block stack that use it for high level synchronization.
Switch to using a new spinlock protecting just the size updates, as
that is all we need, and make sure everyone does the update through
the proper helper.
This fixeѕ a bug reported with the nvme revalidating disks during a
hot removal operation.
Reported-by: Xianting Tian <xianting_tian@126.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
block/partitions/core.c | 4 ++--
drivers/block/aoe/aoecmd.c | 4 +---
drivers/md/dm.c | 15 ++-------------
drivers/s390/block/dasd_ioctl.c | 9 ++-------
fs/block_dev.c | 18 +++++++++---------
include/linux/blk_types.h | 1 +
6 files changed, 17 insertions(+), 34 deletions(-)
diff --git a/block/partitions/core.c b/block/partitions/core.c
index e62a98a8eeb750..328a2cb7875ba1 100644
--- a/block/partitions/core.c
+++ b/block/partitions/core.c
@@ -585,8 +585,8 @@ int bdev_resize_partition(struct block_device *bdev, int partno,
if (partition_overlaps(bdev->bd_disk, start, length, partno))
goto out_unlock;
- part_nr_sects_write(part, (sector_t)length);
- i_size_write(bdevp->bd_inode, length << SECTOR_SHIFT);
+ part_nr_sects_write(part, length);
+ bd_set_nr_sectors(bdevp, length);
ret = 0;
out_unlock:
diff --git a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
index 3cf9bc5d8d9599..6ad73fe730bede 100644
--- a/drivers/block/aoe/aoecmd.c
+++ b/drivers/block/aoe/aoecmd.c
@@ -900,9 +900,7 @@ aoecmd_sleepwork(struct work_struct *work)
ssize = get_capacity(d->gd);
bd = bdget_disk(d->gd, 0);
if (bd) {
- inode_lock(bd->bd_inode);
- i_size_write(bd->bd_inode, (loff_t)ssize<<9);
- inode_unlock(bd->bd_inode);
+ bd_set_nr_sectors(bd, ssize);
bdput(bd);
}
spin_lock_irq(&d->lock);
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 32fa6499739fb9..6b21e5104e3e08 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -2097,18 +2097,6 @@ static void event_callback(void *context)
dm_issue_global_event();
}
-/*
- * Protected by md->suspend_lock obtained by dm_swap_table().
- */
-static void __set_size(struct mapped_device *md, sector_t size)
-{
- lockdep_assert_held(&md->suspend_lock);
-
- set_capacity(md->disk, size);
-
- i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
-}
-
/*
* Returns old map, which caller must destroy.
*/
@@ -2131,7 +2119,8 @@ static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
if (size != dm_get_size(md))
memset(&md->geometry, 0, sizeof(md->geometry));
- __set_size(md, size);
+ set_capacity(md->disk, size);
+ bd_set_nr_sectors(md->bdev, size);
dm_table_event_callback(t, event_callback, md);
diff --git a/drivers/s390/block/dasd_ioctl.c b/drivers/s390/block/dasd_ioctl.c
index 777734d1b4e58c..faaf5596e31c12 100644
--- a/drivers/s390/block/dasd_ioctl.c
+++ b/drivers/s390/block/dasd_ioctl.c
@@ -55,10 +55,7 @@ dasd_ioctl_enable(struct block_device *bdev)
dasd_enable_device(base);
/* Formatting the dasd device can change the capacity. */
- mutex_lock(&bdev->bd_mutex);
- i_size_write(bdev->bd_inode,
- (loff_t)get_capacity(base->block->gdp) << 9);
- mutex_unlock(&bdev->bd_mutex);
+ bd_set_nr_sectors(bdev, get_capacity(base->block->gdp));
dasd_put_device(base);
return 0;
}
@@ -91,9 +88,7 @@ dasd_ioctl_disable(struct block_device *bdev)
* Set i_size to zero, since read, write, etc. check against this
* value.
*/
- mutex_lock(&bdev->bd_mutex);
- i_size_write(bdev->bd_inode, 0);
- mutex_unlock(&bdev->bd_mutex);
+ bd_set_nr_sectors(bdev, 0);
dasd_put_device(base);
return 0;
}
diff --git a/fs/block_dev.c b/fs/block_dev.c
index f52597172c8b79..aa559f1f08800e 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -876,6 +876,7 @@ struct block_device *bdget(dev_t dev)
bdev = &BDEV_I(inode)->bdev;
if (inode->i_state & I_NEW) {
+ spin_lock_init(&bdev->bd_size_lock);
bdev->bd_contains = NULL;
bdev->bd_super = NULL;
bdev->bd_inode = inode;
@@ -1290,6 +1291,7 @@ static void check_disk_size_change(struct gendisk *disk,
{
loff_t disk_size, bdev_size;
+ spin_lock(&bdev->bd_size_lock);
disk_size = (loff_t)get_capacity(disk) << 9;
bdev_size = i_size_read(bdev->bd_inode);
if (disk_size != bdev_size) {
@@ -1304,6 +1306,7 @@ static void check_disk_size_change(struct gendisk *disk,
disk->disk_name);
}
bdev->bd_invalidated = 0;
+ spin_unlock(&bdev->bd_size_lock);
}
/**
@@ -1328,13 +1331,10 @@ int revalidate_disk(struct gendisk *disk)
if (!(disk->flags & GENHD_FL_HIDDEN)) {
struct block_device *bdev = bdget_disk(disk, 0);
- if (!bdev)
- return ret;
-
- mutex_lock(&bdev->bd_mutex);
- check_disk_size_change(disk, bdev, ret == 0);
- mutex_unlock(&bdev->bd_mutex);
- bdput(bdev);
+ if (bdev) {
+ check_disk_size_change(disk, bdev, ret == 0);
+ bdput(bdev);
+ }
}
return ret;
}
@@ -1373,9 +1373,9 @@ EXPORT_SYMBOL(check_disk_change);
void bd_set_nr_sectors(struct block_device *bdev, sector_t sectors)
{
- inode_lock(bdev->bd_inode);
+ spin_lock(&bdev->bd_size_lock);
i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
- inode_unlock(bdev->bd_inode);
+ spin_unlock(&bdev->bd_size_lock);
}
EXPORT_SYMBOL(bd_set_nr_sectors);
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index 4ecf4fed171f0d..5accc2549d2259 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -38,6 +38,7 @@ struct block_device {
/* number of times partitions within this device have been opened. */
unsigned bd_part_count;
int bd_invalidated;
+ spinlock_t bd_size_lock; /* for bd_inode->i_size updates */
struct gendisk * bd_disk;
struct backing_dev_info *bd_bdi;
--
2.28.0
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH 2/2] block: fix locking for struct block_device size updates
2020-08-21 8:56 ` Christoph Hellwig
@ 2020-08-21 9:26 ` Hannes Reinecke
-1 siblings, 0 replies; 13+ messages in thread
From: Hannes Reinecke @ 2020-08-21 9:26 UTC (permalink / raw)
To: Christoph Hellwig, Jens Axboe
Cc: Justin Sanders, Josef Bacik, Xianting Tian, linux-block,
dm-devel, Stefan Haberland, Jan Hoeppner, linux-kernel, nbd,
linux-nvme, linux-s390
On 8/21/20 10:56 AM, Christoph Hellwig wrote:
> Two different callers use two different mutexes for updating the
> block device size, which obviously doesn't help to actually protect
> against concurrent updates from the different callers. In addition
> one of the locks, bd_mutex is rather prone to deadlocks with other
> parts of the block stack that use it for high level synchronization.
>
> Switch to using a new spinlock protecting just the size updates, as
> that is all we need, and make sure everyone does the update through
> the proper helper.
>
> This fixeѕ a bug reported with the nvme revalidating disks during a
> hot removal operation.
>
> Reported-by: Xianting Tian <xianting_tian@126.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> block/partitions/core.c | 4 ++--
> drivers/block/aoe/aoecmd.c | 4 +---
> drivers/md/dm.c | 15 ++-------------
> drivers/s390/block/dasd_ioctl.c | 9 ++-------
> fs/block_dev.c | 18 +++++++++---------
> include/linux/blk_types.h | 1 +
> 6 files changed, 17 insertions(+), 34 deletions(-)
>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] block: fix locking for struct block_device size updates
@ 2020-08-21 9:26 ` Hannes Reinecke
0 siblings, 0 replies; 13+ messages in thread
From: Hannes Reinecke @ 2020-08-21 9:26 UTC (permalink / raw)
To: Christoph Hellwig, Jens Axboe
Cc: linux-s390, Jan Hoeppner, Justin Sanders, linux-nvme,
Josef Bacik, Xianting Tian, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd
On 8/21/20 10:56 AM, Christoph Hellwig wrote:
> Two different callers use two different mutexes for updating the
> block device size, which obviously doesn't help to actually protect
> against concurrent updates from the different callers. In addition
> one of the locks, bd_mutex is rather prone to deadlocks with other
> parts of the block stack that use it for high level synchronization.
>
> Switch to using a new spinlock protecting just the size updates, as
> that is all we need, and make sure everyone does the update through
> the proper helper.
>
> This fixeѕ a bug reported with the nvme revalidating disks during a
> hot removal operation.
>
> Reported-by: Xianting Tian <xianting_tian@126.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> block/partitions/core.c | 4 ++--
> drivers/block/aoe/aoecmd.c | 4 +---
> drivers/md/dm.c | 15 ++-------------
> drivers/s390/block/dasd_ioctl.c | 9 ++-------
> fs/block_dev.c | 18 +++++++++---------
> include/linux/blk_types.h | 1 +
> 6 files changed, 17 insertions(+), 34 deletions(-)
>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] block: fix locking for struct block_device size updates
2020-08-21 9:26 ` Hannes Reinecke
(?)
@ 2020-08-21 10:02 ` 田
2020-08-23 9:09 ` Christoph Hellwig
-1 siblings, 1 reply; 13+ messages in thread
From: 田 @ 2020-08-21 10:02 UTC (permalink / raw)
To: Hannes Reinecke
Cc: Jens Axboe, linux-s390, Jan Hoeppner, Justin Sanders,
Josef Bacik, linux-nvme, linux-kernel, linux-block, dm-devel,
Stefan Haberland, nbd, Christoph Hellwig
[-- Attachment #1.1: Type: text/plain, Size: 1546 bytes --]
thanks Hellwig for your kindly reply and your fix and add report by me :)
On 08/21/2020 17:26, Hannes Reinecke wrote:
On 8/21/20 10:56 AM, Christoph Hellwig wrote:
> Two different callers use two different mutexes for updating the
> block device size, which obviously doesn't help to actually protect
> against concurrent updates from the different callers. In addition
> one of the locks, bd_mutex is rather prone to deadlocks with other
> parts of the block stack that use it for high level synchronization.
>
> Switch to using a new spinlock protecting just the size updates, as
> that is all we need, and make sure everyone does the update through
> the proper helper.
>
> This fixeѕ a bug reported with the nvme revalidating disks during a
> hot removal operation.
>
> Reported-by: Xianting Tian <xianting_tian@126.com>
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
> block/partitions/core.c | 4 ++--
> drivers/block/aoe/aoecmd.c | 4 +---
> drivers/md/dm.c | 15 ++-------------
> drivers/s390/block/dasd_ioctl.c | 9 ++-------
> fs/block_dev.c | 18 +++++++++---------
> include/linux/blk_types.h | 1 +
> 6 files changed, 17 insertions(+), 34 deletions(-)
>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@suse.de +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
[-- Attachment #1.2: Type: text/html, Size: 2934 bytes --]
[-- Attachment #2: Type: text/plain, Size: 0 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] block: fix locking for struct block_device size updates
2020-08-21 10:02 ` 田
@ 2020-08-23 9:09 ` Christoph Hellwig
0 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-23 9:09 UTC (permalink / raw)
To: 田
Cc: Hannes Reinecke, Christoph Hellwig, Jens Axboe, Justin Sanders,
Josef Bacik, linux-block, dm-devel, Stefan Haberland,
Jan Hoeppner, linux-kernel, nbd, linux-nvme, linux-s390
On Fri, Aug 21, 2020 at 06:02:32PM +0800, 田 wrote:
> thanks Hellwig for your kindly reply and your fix and add report by me :)
I found an issue with the patch, and will send a new version in a bit.
If this works for your reproducer, can you add a Tested-by: tag?
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 2/2] block: fix locking for struct block_device size updates
@ 2020-08-23 9:09 ` Christoph Hellwig
0 siblings, 0 replies; 13+ messages in thread
From: Christoph Hellwig @ 2020-08-23 9:09 UTC (permalink / raw)
To: 田
Cc: Jens Axboe, linux-s390, Jan Hoeppner, Justin Sanders,
Josef Bacik, linux-nvme, linux-kernel, linux-block, dm-devel,
Hannes Reinecke, nbd, Christoph Hellwig, Stefan Haberland
On Fri, Aug 21, 2020 at 06:02:32PM +0800, 田 wrote:
> thanks Hellwig for your kindly reply and your fix and add report by me :)
I found an issue with the patch, and will send a new version in a bit.
If this works for your reproducer, can you add a Tested-by: tag?
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 13+ messages in thread