linux-stable.git/drivers/md/bcache, branch linux-5.8.y

bcache: use disk_{start,end}_io_acct() to count I/O for bcache device

2020-08-21T11:14:56+00:00

commit c5be1f2c5bab1538aa29cd42e226d6b80391e3ff upstream.

This patch is a fix to patch "bcache: fix bio_{start,end}_io_acct with
proper device". The previous patch uses a hack to temporarily set
bi_disk to bcache device, which is mistaken too.

As Christoph suggests, this patch uses disk_{start,end}_io_acct() to
count I/O for bcache device in the correct way.

Fixes: 85750aeb748f ("bcache: use bio_{start,end}_io_acct")
Signed-off-by: Coly Li 
Cc: Christoph Hellwig 
Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

bcache: fix bio_{start,end}_io_acct with proper device

2020-08-21T11:14:56+00:00

commit a2f32ee8fd853cec8860f883d98afc3a339546de upstream.

Commit 85750aeb748f ("bcache: use bio_{start,end}_io_acct") moves the
io account code to the location after bio_set_dev(bio, dc->bdev) in
cached_dev_make_request(). Then the account is performed incorrectly on
backing device, indeed the I/O should be counted to bcache device like
/dev/bcache0.

With the mistaken I/O account, iostat does not display I/O counts for
bcache device and all the numbers go to backing device. In writeback
mode, the hard drive may have 340K+ IOPS which is impossible and wrong
for spinning disk.

This patch introduces bch_bio_start_io_acct() and bch_bio_end_io_acct(),
which switches bio->bi_disk to bcache device before calling
bio_start_io_acct() or bio_end_io_acct(). Now the I/Os are counted to
bcache device, and bcache device, cache device and backing device have
their correct I/O count information back.

Fixes: 85750aeb748f ("bcache: use bio_{start,end}_io_acct")
Signed-off-by: Coly Li 
Cc: Christoph Hellwig 
Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

bcache: avoid nr_stripes overflow in bcache_device_init()

2020-08-21T11:14:56+00:00

commit 65f0f017e7be8c70330372df23bcb2a407ecf02d upstream.

For some block devices which large capacity (e.g. 8TB) but small io_opt
size (e.g. 8 sectors), in bcache_device_init() the stripes number calcu-
lated by,
	DIV_ROUND_UP_ULL(sectors, d->stripe_size);
might be overflow to the unsigned int bcache_device->nr_stripes.

This patch uses the uint64_t variable to store DIV_ROUND_UP_ULL()
and after the value is checked to be available in unsigned int range,
sets it to bache_device->nr_stripes. Then the overflow is avoided.

Reported-and-tested-by: Ken Raeburn 
Signed-off-by: Coly Li 
Cc: stable@vger.kernel.org
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1783075
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

bcache: fix overflow in offset_to_stripe()

2020-08-21T11:14:56+00:00

commit 7a1481267999c02abf4a624515c1b5c7c1fccbd6 upstream.

offset_to_stripe() returns the stripe number (in type unsigned int) from
an offset (in type uint64_t) by the following calculation,
	do_div(offset, d->stripe_size);
For large capacity backing device (e.g. 18TB) with small stripe size
(e.g. 4KB), the result is 4831838208 and exceeds UINT_MAX. The actual
returned value which caller receives is 536870912, due to the overflow.

Indeed in bcache_device_init(), bcache_device->nr_stripes is limited in
range [1, INT_MAX]. Therefore all valid stripe numbers in bcache are
in range [0, bcache_dev->nr_stripes - 1].

This patch adds a upper limition check in offset_to_stripe(): the max
valid stripe number should be less than bcache_device->nr_stripes. If
the calculated stripe number from do_div() is equal to or larger than
bcache_device->nr_stripe, -EINVAL will be returned. (Normally nr_stripes
is less than INT_MAX, exceeding upper limitation doesn't mean overflow,
therefore -EOVERFLOW is not used as error code.)

This patch also changes nr_stripes' type of struct bcache_device from
'unsigned int' to 'int', and return value type of offset_to_stripe()
from 'unsigned int' to 'int', to match their exact data ranges.

All locations where bcache_device->nr_stripes and offset_to_stripe() are
referenced also get updated for the above type change.

Reported-and-tested-by: Ken Raeburn 
Signed-off-by: Coly Li 
Cc: stable@vger.kernel.org
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1783075
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

bcache: allocate meta data pages as compound pages

2020-08-21T11:14:56+00:00

commit 5fe48867856367142d91a82f2cbf7a57a24cbb70 upstream.

There are some meta data of bcache are allocated by multiple pages,
and they are used as bio bv_page for I/Os to the cache device. for
example cache_set->uuids, cache->disk_buckets, journal_write->data,
bset_tree->data.

For such meta data memory, all the allocated pages should be treated
as a single memory block. Then the memory management and underlying I/O
code can treat them more clearly.

This patch adds __GFP_COMP flag to all the location allocating >0 order
pages for the above mentioned meta data. Then their pages are treated
as compound pages now.

Signed-off-by: Coly Li 
Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe 
Signed-off-by: Greg Kroah-Hartman

bcache: fix super block seq numbers comparision in register_cache_set()

2020-08-19T06:26:22+00:00

[ Upstream commit 117f636ea695270fe492d0c0c9dfadc7a662af47 ]

In register_cache_set(), c is pointer to struct cache_set, and ca is
pointer to struct cache, if ca->sb.seq > c->sb.seq, it means this
registering cache has up to date version and other members, the in-
memory version and other members should be updated to the newer value.

But current implementation makes a cache set only has a single cache
device, so the above assumption works well except for a special case.
The execption is when a cache device new created and both ca->sb.seq and
c->sb.seq are 0, because the super block is never flushed out yet. In
the location for the following if() check,
2156         if (ca->sb.seq > c->sb.seq) {
2157                 c->sb.version           = ca->sb.version;
2158                 memcpy(c->sb.set_uuid, ca->sb.set_uuid, 16);
2159                 c->sb.flags             = ca->sb.flags;
2160                 c->sb.seq               = ca->sb.seq;
2161                 pr_debug("set version = %llu\n", c->sb.version);
2162         }
c->sb.version is not initialized yet and valued 0. When ca->sb.seq is 0,
the if() check will fail (because both values are 0), and the cache set
version, set_uuid, flags and seq won't be updated.

The above problem is hiden for current code, because the bucket size is
compatible among different super block version. And the next time when
running cache set again, ca->sb.seq will be larger than 0 and cache set
super block version will be updated properly.

But if the large bucket feature is enabled,  sb->bucket_size is the low
16bits of the bucket size. For a power of 2 value, when the actual
bucket size exceeds 16bit width, sb->bucket_size will always be 0. Then
read_super_common() will fail because the if() check to
is_power_of_2(sb->bucket_size) is false. This is how the long time
hidden bug is triggered.

This patch modifies the if() check to the following way,
2156         if (ca->sb.seq > c->sb.seq || c->sb.seq == 0) {
Then cache set's version, set_uuid, flags and seq will always be updated
corectly including for a new created cache device.

Signed-off-by: Coly Li 
Reviewed-by: Hannes Reinecke 
Signed-off-by: Jens Axboe 
Signed-off-by: Sasha Levin

bcache: pr_info() format clean up in bcache_device_init()

2020-06-14T22:47:56+00:00

scripts/checkpatch.pl reports following warning for patch
("bcache: check and adjust logical block size for backing devices"),
    WARNING: quoted string split across lines
    #146: FILE: drivers/md/bcache/super.c:896:
    +  pr_info("%s: sb/logical block size (%u) greater than page size "
    +	       "(%lu) falling back to device logical block size (%u)",

There are two things to fix up,
- The kernel message print should be in a single line.
- pr_info() won't automatically add new line since v5.8, a '\n' should
  be added.

This patch just does the above cleanup in bcache_device_init().

Signed-off-by: Coly Li 
Signed-off-by: Jens Axboe

bcache: use delayed kworker fo asynchronous devices registration

2020-06-14T22:47:56+00:00

This patch changes the asynchronous registration kworker to a delayed
kworker. There is probability queue_work() queues the async registration
kworker to the same CPU (even though very little), then the process
which writing sysfs interface to reigster bcache device may won't return
immeidately. queue_delayed_work() in this patch will delay 10 jiffies
before insert the kworker to run queue, which makes sure the registering
process may always returns to user space in time.

Fixes: 9e23ccf8f0a22 ("bcache: asynchronous devices registration")
Signed-off-by: Coly Li 
Cc: Hannes Reinecke 
Signed-off-by: Jens Axboe

bcache: check and adjust logical block size for backing devices

2020-06-14T22:47:56+00:00

It's possible for a block driver to set logical block size to
a value greater than page size incorrectly; e.g. bcache takes
the value from the superblock, set by the user w/ make-bcache.

This causes a BUG/NULL pointer dereference in the path:

  __blkdev_get()
  -> set_init_blocksize() // set i_blkbits based on ...
     -> bdev_logical_block_size()
        -> queue_logical_block_size() // ... this value
  -> bdev_disk_changed()
     ...
     -> blkdev_readpage()
        -> block_read_full_page()
           -> create_page_buffers() // size = 1 << i_blkbits
              -> create_empty_buffers() // give size/take pointer
                 -> alloc_page_buffers() // return NULL
                 .. BUG!

Because alloc_page_buffers() is called with size > PAGE_SIZE,
thus it initializes head = NULL, skips the loop, return head;
then create_empty_buffers() gets (and uses) the NULL pointer.

This has been around longer than commit ad6bf88a6c19 ("block:
fix an integer overflow in logical block size"); however, it
increased the range of values that can trigger the issue.

Previously only 8k/16k/32k (on x86/4k page size) would do it,
as greater values overflow unsigned short to zero, and queue_
logical_block_size() would then use the default of 512.

Now the range with unsigned int is much larger, and users w/
the 512k value, which happened to be zero'ed previously and
work fine, started to hit this issue -- as the zero is gone,
and queue_logical_block_size() does return 512k (>PAGE_SIZE.)

Fix this by checking the bcache device's logical block size,
and if it's greater than page size, fallback to the backing/
cached device's logical page size.

This doesn't affect cache devices as those are still checked
for block/page size in read_super(); only the backing/cached
devices are not.

Apparently it's a regression from commit 2903381fce71 ("bcache:
Take data offset from the bdev superblock."), moving the check
into BCACHE_SB_VERSION_CDEV only. Now that we have superblocks
of backing devices out there with this larger value, we cannot
refuse to load them (i.e., have a similar check in _BDEV.)

Ideally perhaps bcache should use all values from the backing
device (physical/logical/io_min block size)? But for now just
fix the problematic case.

Test-case:

    # IMG=/root/disk.img
    # dd if=/dev/zero of=$IMG bs=1 count=0 seek=1G
    # DEV=$(losetup --find --show $IMG)
    # make-bcache --bdev $DEV --block 8k
      < see dmesg >

Before:

    # uname -r
    5.7.0-rc7

    [   55.944046] BUG: kernel NULL pointer dereference, address: 0000000000000000
    ...
    [   55.949742] CPU: 3 PID: 610 Comm: bcache-register Not tainted 5.7.0-rc7 #4
    ...
    [   55.952281] RIP: 0010:create_empty_buffers+0x1a/0x100
    ...
    [   55.966434] Call Trace:
    [   55.967021]  create_page_buffers+0x48/0x50
    [   55.967834]  block_read_full_page+0x49/0x380
    [   55.972181]  do_read_cache_page+0x494/0x610
    [   55.974780]  read_part_sector+0x2d/0xaa
    [   55.975558]  read_lba+0x10e/0x1e0
    [   55.977904]  efi_partition+0x120/0x5a6
    [   55.980227]  blk_add_partitions+0x161/0x390
    [   55.982177]  bdev_disk_changed+0x61/0xd0
    [   55.982961]  __blkdev_get+0x350/0x490
    [   55.983715]  __device_add_disk+0x318/0x480
    [   55.984539]  bch_cached_dev_run+0xc5/0x270
    [   55.986010]  register_bcache.cold+0x122/0x179
    [   55.987628]  kernfs_fop_write+0xbc/0x1a0
    [   55.988416]  vfs_write+0xb1/0x1a0
    [   55.989134]  ksys_write+0x5a/0xd0
    [   55.989825]  do_syscall_64+0x43/0x140
    [   55.990563]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
    [   55.991519] RIP: 0033:0x7f7d60ba3154
    ...

After:

    # uname -r
    5.7.0.bcachelbspgsz

    [   31.672460] bcache: bcache_device_init() bcache0: sb/logical block size (8192) greater than page size (4096) falling back to device logical block size (512)
    [   31.675133] bcache: register_bdev() registered backing device loop0

    # grep ^ /sys/block/bcache0/queue/*_block_size
    /sys/block/bcache0/queue/logical_block_size:512
    /sys/block/bcache0/queue/physical_block_size:8192

Reported-by: Ryan Finnie 
Reported-by: Sebastian Marsching 
Signed-off-by: Mauricio Faria de Oliveira 
Signed-off-by: Coly Li 
Signed-off-by: Jens Axboe

bcache: fix potential deadlock problem in btree_gc_coalesce

2020-06-14T22:47:56+00:00

coccicheck reports:
  drivers/md//bcache/btree.c:1538:1-7: preceding lock on line 1417

In btree_gc_coalesce func, if the coalescing process fails, we will goto
to out_nocoalesce tag directly without releasing new_nodes[i]->write_lock.
Then, it will cause a deadlock when trying to acquire new_nodes[i]->
write_lock for freeing new_nodes[i] before return.

btree_gc_coalesce func details as follows:
	if alloc new_nodes[i] fails:
		goto out_nocoalesce;
	// obtain new_nodes[i]->write_lock
	mutex_lock(&new_nodes[i]->write_lock)
	// main coalescing process
	for (i = nodes - 1; i > 0; --i)
		[snipped]
		if coalescing process fails:
			// Here, directly goto out_nocoalesce
			 // tag will cause a deadlock
			goto out_nocoalesce;
		[snipped]
	// release new_nodes[i]->write_lock
	mutex_unlock(&new_nodes[i]->write_lock)
	// coalesing succ, return
	return;
out_nocoalesce:
	btree_node_free(new_nodes[i])	// free new_nodes[i]
	// obtain new_nodes[i]->write_lock
	mutex_lock(&new_nodes[i]->write_lock);
	// set flag for reuse
	clear_bit(BTREE_NODE_dirty, &ew_nodes[i]->flags);
	// release new_nodes[i]->write_lock
	mutex_unlock(&new_nodes[i]->write_lock);

To fix the problem, we add a new tag 'out_unlock_nocoalesce' for
releasing new_nodes[i]->write_lock before out_nocoalesce tag. If
coalescing process fails, we will go to out_unlock_nocoalesce tag
for releasing new_nodes[i]->write_lock before free new_nodes[i] in
out_nocoalesce tag.

(Coly Li helps to clean up commit log format.)

Fixes: 2a285686c109816 ("bcache: btree locking rework")
Signed-off-by: Zhiqiang Liu 
Signed-off-by: Coly Li 
Signed-off-by: Jens Axboe