BMC/Intel-BMC/linux.git/drivers/md, branch v5.4.20

bcache: avoid unnecessary btree nodes flushing in btree_flush_write()

2020-02-14T21:34:19+00:00

commit 2aa8c529387c25606fdc1484154b92f8bfbc5746 upstream. the commit 91be66e1318f ("bcache: performance improvement for btree_flush_write()") was an effort to flushing btree node with oldest btree node faster in following methods, - Only iterate dirty btree nodes in c->btree_cache, avoid scanning a lot of clean btree nodes. - Take c->btree_cache as a LRU-like list, aggressively flushing all dirty nodes from tail of c->btree_cache util the btree node with oldest journal entry is flushed. This is to reduce the time of holding c->bucket_lock. Guoju Fang and Shuang Li reported that they observe unexptected extra write I/Os on cache device after applying the above patch. Guoju Fang provideed more detailed diagnose information that the aggressive btree nodes flushing may cause 10x more btree nodes to flush in his workload. He points out when system memory is large enough to hold all btree nodes in memory, c->btree_cache is not a LRU-like list any more. Then the btree node with oldest journal entry is very probably not- close to the tail of c->btree_cache list. In such situation much more dirty btree nodes will be aggressively flushed before the target node is flushed. When slow SATA SSD is used as cache device, such over- aggressive flushing behavior will cause performance regression. After spending a lot of time on debug and diagnose, I find the real condition is more complicated, aggressive flushing dirty btree nodes from tail of c->btree_cache list is not a good solution. - When all btree nodes are cached in memory, c->btree_cache is not a LRU-like list, the btree nodes with oldest journal entry won't be close to the tail of the list. - There can be hundreds dirty btree nodes reference the oldest journal entry, before flushing all the nodes the oldest journal entry cannot be reclaimed. When the above two conditions mixed together, a simply flushing from tail of c->btree_cache list is really NOT a good idea. Fortunately there is still chance to make btree_flush_write() work better. Here is how this patch avoids unnecessary btree nodes flushing, - Only acquire c->journal.lock when getting oldest journal entry of fifo c->journal.pin. In rested locations check the journal entries locklessly, so their values can be changed on other cores in parallel. - In loop list_for_each_entry_safe_reverse(), checking latest front point of fifo c->journal.pin. If it is different from the original point which we get with locking c->journal.lock, it means the oldest journal entry is reclaim on other cores. At this moment, all selected dirty nodes recorded in array btree_nodes[] are all flushed and clean on other CPU cores, it is unncessary to iterate c->btree_cache any longer. Just quit the list_for_each_entry_safe_reverse() loop and the following for-loop will skip all the selected clean nodes. - Find a proper time to quit the list_for_each_entry_safe_reverse() loop. Check the refcount value of orignial fifo front point, if the value is larger than selected node number of btree_nodes[], it means more matching btree nodes should be scanned. Otherwise it means no more matching btee nodes in rest of c->btree_cache list, the loop can be quit. If the original oldest journal entry is reclaimed and fifo front point is updated, the refcount of original fifo front point will be 0, then the loop will be quit too. - Not hold c->bucket_lock too long time. c->bucket_lock is also required for space allocation for cached data, hold it for too long time will block regular I/O requests. When iterating list c->btree_cache, even there are a lot of maching btree nodes, in order to not holding c->bucket_lock for too long time, only BTREE_FLUSH_NR nodes are selected and to flush in following for-loop. With this patch, only btree nodes referencing oldest journal entry are flushed to cache device, no aggressive flushing for unnecessary btree node any more. And in order to avoid blocking regluar I/O requests, each time when btree_flush_write() called, at most only BTREE_FLUSH_NR btree nodes are selected to flush, even there are more maching btree nodes in list c->btree_cache. At last, one more thing to explain: Why it is safe to read front point of c->journal.pin without holding c->journal.lock inside the list_for_each_entry_safe_reverse() loop ? Here is my answer: When reading the front point of fifo c->journal.pin, we don't need to know the exact value of front point, we just want to check whether the value is different from the original front point (which is accurate value because we get it while c->jouranl.lock is held). For such purpose, it works as expected without holding c->journal.lock. Even the front point is changed on other CPU core and not updated to local core, and current iterating btree node has identical journal entry local as original fetched fifo front point, it is still safe. Because after holding mutex b->write_lock (with memory barrier) this btree node can be found as clean and skipped, the loop will quite latter when iterate on next node of list c->btree_cache. Fixes: 91be66e1318f ("bcache: performance improvement for btree_flush_write()") Reported-by: Guoju Fang Reported-by: Shuang Li Signed-off-by: Coly Li Signed-off-by: Jens Axboe Signed-off-by: Greg Kroah-Hartman

bcache: add readahead cache policy options via sysfs interface

2020-02-11T12:35:37+00:00

commit 038ba8cc1bffc51250add4a9b9249d4331576d8f upstream. In year 2007 high performance SSD was still expensive, in order to save more space for real workload or meta data, the readahead I/Os for non-meta data was bypassed and not cached on SSD. In now days, SSD price drops a lot and people can find larger size SSD with more comfortable price. It is unncessary to alway bypass normal readahead I/Os to save SSD space for now. This patch adds options for readahead data cache policies via sysfs file /sys/block/bcache/readahead_cache_policy, the options are, - "all": cache all readahead data I/Os. - "meta-only": only cache meta data, and bypass other regular I/Os. If users want to make bcache continue to only cache readahead request for metadata and bypass regular data readahead, please set "meta-only" to this sysfs file. By default, bcache will back to cache all read- ahead requests now. Cc: stable@vger.kernel.org Signed-off-by: Coly Li Acked-by: Eric Wheeler Cc: Michael Lyle Signed-off-by: Jens Axboe Signed-off-by: Greg Kroah-Hartman

dm: fix potential for q->make_request_fn NULL pointer

2020-02-11T12:35:27+00:00

commit 47ace7e012b9f7ad71d43ac9063d335ea3d6820b upstream. Move blk_queue_make_request() to dm.c:alloc_dev() so that q->make_request_fn is never NULL during the lifetime of a DM device (even one that is created without a DM table). Otherwise generic_make_request() will crash simply by doing: dmsetup create -n test mount /dev/dm-N /mnt While at it, move ->congested_data initialization out of dm.c:alloc_dev() and into the bio-based specific init method. Reported-by: Stefan Bader BugLink: https://bugs.launchpad.net/bugs/1860231 Fixes: ff36ab34583a ("dm: remove request-based logic from make_request_fn wrapper") Depends-on: c12c9a3c3860c ("dm: various cleanups to md->queue initialization code") Cc: stable@vger.kernel.org Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm thin metadata: use pool locking at end of dm_pool_metadata_close

2020-02-11T12:35:26+00:00

commit 44d8ebf436399a40fcd10dd31b29d37823d62fcc upstream. Ensure that the pool is locked during calls to __commit_transaction and __destroy_persistent_data_objects. Just being consistent with locking, but reality is dm_pool_metadata_close is called once pool is being destroyed so access to pool shouldn't be contended. Also, use pmd_write_lock_in_core rather than __pmd_write_lock in dm_pool_commit_metadata and rename __pmd_write_lock to pmd_write_lock_in_core -- there was no need for the alias. In addition, verify that the pool is locked in __commit_transaction(). Fixes: 873f258becca ("dm thin metadata: do not write metadata if no changes occurred") Cc: stable@vger.kernel.org Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm crypt: fix benbi IV constructor crash if used in authenticated mode

2020-02-11T12:35:26+00:00

commit 4ea9471fbd1addb25a4d269991dc724e200ca5b5 upstream. If benbi IV is used in AEAD construction, for example: cryptsetup luksFormat --cipher twofish-xts-benbi --key-size 512 --integrity=hmac-sha256 the constructor uses wrong skcipher function and crashes: BUG: kernel NULL pointer dereference, address: 00000014 ... EIP: crypt_iv_benbi_ctr+0x15/0x70 [dm_crypt] Call Trace: ? crypt_subkey_size+0x20/0x20 [dm_crypt] crypt_ctr+0x567/0xfc0 [dm_crypt] dm_table_add_target+0x15f/0x340 [dm_mod] Fix this by properly using crypt_aead_blocksize() in this case. Fixes: ef43aa38063a6 ("dm crypt: add cryptographic data integrity protection (authenticated encryption)") Cc: stable@vger.kernel.org # v4.12+ Link: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=941051 Reported-by: Jerad Simpson Signed-off-by: Milan Broz Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm crypt: fix GFP flags passed to skcipher_request_alloc()

2020-02-11T12:35:26+00:00

commit 9402e959014a18b4ebf7558733076875808dd66c upstream. GFP_KERNEL is not supposed to be or'd with GFP_NOFS (the result is equivalent to GFP_KERNEL). Also, we use GFP_NOIO instead of GFP_NOFS because we don't want any I/O being submitted in the direct reclaim path. Fixes: 39d13a1ac41d ("dm crypt: reuse eboiv skcipher for IV generation") Cc: stable@vger.kernel.org # v5.4+ Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm writecache: fix incorrect flush sequence when doing SSD mode commit

2020-02-11T12:35:26+00:00

commit aa9509209c5ac2f0b35d01a922bf9ae072d0c2fc upstream. When committing state, the function writecache_flush does the following: 1. write metadata (writecache_commit_flushed) 2. flush disk cache (writecache_commit_flushed) 3. wait for data writes to complete (writecache_wait_for_ios) 4. increase superblock seq_count 5. write the superblock 6. flush disk cache It may happen that at step 3, when we wait for some write to finish, the disk may report the write as finished, but the write only hit the disk cache and it is not yet stored in persistent storage. At step 5 we write the superblock - it may happen that the superblock is written before the write that we waited for in step 3. If the machine crashes, it may result in incorrect data being returned after reboot. In order to fix the bug, we must swap steps 2 and 3 in the above sequence, so that we first wait for writes to complete and then flush the disk cache. Fixes: 48debafe4f2f ("dm: add writecache target") Cc: stable@vger.kernel.org # 4.18+ Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm space map common: fix to ensure new block isn't already in use

2020-02-11T12:35:26+00:00

commit 4feaef830de7ffdd8352e1fe14ad3bf13c9688f8 upstream. The space-maps track the reference counts for disk blocks allocated by both the thin-provisioning and cache targets. There are variants for tracking metadata blocks and data blocks. Transactionality is implemented by never touching blocks from the previous transaction, so we can rollback in the event of a crash. When allocating a new block we need to ensure the block is free (has reference count of 0) in both the current and previous transaction. Prior to this fix we were doing this by searching for a free block in the previous transaction, and relying on a 'begin' counter to track where the last allocation in the current transaction was. This 'begin' field was not being updated in all code paths (eg, increment of a data block reference count due to breaking sharing of a neighbour block in the same btree leaf). This fix keeps the 'begin' field, but now it's just a hint to speed up the search. Instead the current transaction is searched for a free block, and then the old transaction is double checked to ensure it's free. Much simpler. This fixes reports of sm_disk_new_block()'s BUG_ON() triggering when DM thin-provisioning's snapshots are heavily used. Reported-by: Eric Wheeler Cc: stable@vger.kernel.org Signed-off-by: Joe Thornber Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm zoned: support zone sizes smaller than 128MiB

2020-02-11T12:35:26+00:00

commit b39962950339912978484cdac50069258545d753 upstream. dm-zoned is observed to log failed kernel assertions and not work correctly when operating against a device with a zone size smaller than 128MiB (e.g. 32768 bits per 4K block). The reason is that the bitmap size per zone is calculated as zero with such a small zone size. Fix this problem and also make the code related to zone bitmap management be able to handle per zone bitmaps smaller than a single block. A dm-zoned-tools patch is required to properly format dm-zoned devices with zone sizes smaller than 128MiB. Fixes: 3b1a94c88b79 ("dm zoned: drive-managed zoned block device target") Cc: stable@vger.kernel.org Signed-off-by: Dmitry Fomichev Reviewed-by: Damien Le Moal Signed-off-by: Mike Snitzer Signed-off-by: Greg Kroah-Hartman

dm thin: fix use-after-free in metadata_pre_commit_callback

2020-02-05T21:22:53+00:00

[ Upstream commit a4a8d286586d4b28c8517a51db8d86954aadc74b ] dm-thin uses struct pool to hold the state of the pool. There may be multiple pool_c's pointing to a given pool, each pool_c represents a loaded target. pool_c's may be created and destroyed arbitrarily and the pool contains a reference count of pool_c's pointing to it. Since commit 694cfe7f31db3 ("dm thin: Flush data device before committing metadata") a pointer to pool_c is passed to dm_pool_register_pre_commit_callback and this function stores it in pmd->pre_commit_context. If this pool_c is freed, but pool is not (because there is another pool_c referencing it), we end up in a situation where pmd->pre_commit_context structure points to freed pool_c. It causes a crash in metadata_pre_commit_callback. Fix this by moving the dm_pool_register_pre_commit_callback() from pool_ctr() to pool_preresume(). This way the in-core thin-pool metadata is only ever armed with callback data whose lifetime matches the active thin-pool target. In should be noted that this fix preserves the ability to load a thin-pool table that uses a different data block device (that contains the same data) -- though it is unclear if that capability is still useful and/or needed. Fixes: 694cfe7f31db3 ("dm thin: Flush data device before committing metadata") Cc: stable@vger.kernel.org Reported-by: Zdenek Kabelac Reported-by: Mikulas Patocka Signed-off-by: Mike Snitzer Signed-off-by: Sasha Levin