From 88430ebcbc0ec637b710b947738839848c20feff Mon Sep 17 00:00:00 2001 From: Luo Meng Date: Tue, 29 Nov 2022 10:48:47 +0800 Subject: dm thin: Fix UAF in run_timer_softirq() When dm_resume() and dm_destroy() are concurrent, it will lead to UAF, as follows: BUG: KASAN: use-after-free in __run_timers+0x173/0x710 Write of size 8 at addr ffff88816d9490f0 by task swapper/0/0 Call Trace: dump_stack_lvl+0x73/0x9f print_report.cold+0x132/0xaa2 _raw_spin_lock_irqsave+0xcd/0x160 __run_timers+0x173/0x710 kasan_report+0xad/0x110 __run_timers+0x173/0x710 __asan_store8+0x9c/0x140 __run_timers+0x173/0x710 call_timer_fn+0x310/0x310 pvclock_clocksource_read+0xfa/0x250 kvm_clock_read+0x2c/0x70 kvm_clock_get_cycles+0xd/0x20 ktime_get+0x5c/0x110 lapic_next_event+0x38/0x50 clockevents_program_event+0xf1/0x1e0 run_timer_softirq+0x49/0x90 __do_softirq+0x16e/0x62c __irq_exit_rcu+0x1fa/0x270 irq_exit_rcu+0x12/0x20 sysvec_apic_timer_interrupt+0x8e/0xc0 One of the concurrency UAF can be shown as below: use free do_resume | __find_device_hash_cell | dm_get | atomic_inc(&md->holders) | | dm_destroy | __dm_destroy | if (!dm_suspended_md(md)) | atomic_read(&md->holders) | msleep(1) dm_resume | __dm_resume | dm_table_resume_targets | pool_resume | do_waker #add delay work | dm_put | atomic_dec(&md->holders) | | dm_table_destroy | pool_dtr | __pool_dec | __pool_destroy | destroy_workqueue | kfree(pool) # free pool time out __do_softirq run_timer_softirq # pool has already been freed This can be easily reproduced using: 1. create thin-pool 2. dmsetup suspend pool 3. dmsetup resume pool 4. dmsetup remove_all # Concurrent with 3 The root cause of this UAF bug is that dm_resume() adds timer after dm_destroy() skips cancelling the timer because of suspend status. After timeout, it will call run_timer_softirq(), however pool has already been freed. The concurrency UAF bug will happen. Therefore, cancelling timer again in __pool_destroy(). Cc: stable@vger.kernel.org Fixes: 991d9fa02da0d ("dm: add thin provisioning target") Signed-off-by: Luo Meng Signed-off-by: Mike Snitzer --- drivers/md/dm-thin.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/md') diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index e76c96c760a9..dc271c107fb5 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -2889,6 +2889,8 @@ static void __pool_destroy(struct pool *pool) dm_bio_prison_destroy(pool->prison); dm_kcopyd_client_destroy(pool->copier); + cancel_delayed_work_sync(&pool->waker); + cancel_delayed_work_sync(&pool->no_space_timeout); if (pool->wq) destroy_workqueue(pool->wq); -- cgit v1.2.3 From e4b5957c6f749a501c464f92792f1c8e26b61a94 Mon Sep 17 00:00:00 2001 From: Luo Meng Date: Tue, 29 Nov 2022 10:48:48 +0800 Subject: dm clone: Fix UAF in clone_dtr() Dm_clone also has the same UAF problem when dm_resume() and dm_destroy() are concurrent. Therefore, cancelling timer again in clone_dtr(). Cc: stable@vger.kernel.org Fixes: 7431b7835f554 ("dm: add clone target") Signed-off-by: Luo Meng Signed-off-by: Mike Snitzer --- drivers/md/dm-clone-target.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/md') diff --git a/drivers/md/dm-clone-target.c b/drivers/md/dm-clone-target.c index 2f1cc66d2641..29e0b85eeaf0 100644 --- a/drivers/md/dm-clone-target.c +++ b/drivers/md/dm-clone-target.c @@ -1958,6 +1958,7 @@ static void clone_dtr(struct dm_target *ti) mempool_exit(&clone->hydration_pool); dm_kcopyd_client_destroy(clone->kcopyd_client); + cancel_delayed_work_sync(&clone->waker); destroy_workqueue(clone->wq); hash_table_exit(clone); dm_clone_metadata_close(clone->cmd); -- cgit v1.2.3 From 6a459d8edbdbe7b24db42a5a9f21e6aa9e00c2aa Mon Sep 17 00:00:00 2001 From: Luo Meng Date: Tue, 29 Nov 2022 10:48:49 +0800 Subject: dm cache: Fix UAF in destroy() Dm_cache also has the same UAF problem when dm_resume() and dm_destroy() are concurrent. Therefore, cancelling timer again in destroy(). Cc: stable@vger.kernel.org Fixes: c6b4fcbad044e ("dm: add cache target") Signed-off-by: Luo Meng Signed-off-by: Mike Snitzer --- drivers/md/dm-cache-target.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/md') diff --git a/drivers/md/dm-cache-target.c b/drivers/md/dm-cache-target.c index 54a8d5c9a44e..624a6335c832 100644 --- a/drivers/md/dm-cache-target.c +++ b/drivers/md/dm-cache-target.c @@ -1887,6 +1887,7 @@ static void destroy(struct cache *cache) if (cache->prison) dm_bio_prison_destroy_v2(cache->prison); + cancel_delayed_work_sync(&cache->waker); if (cache->wq) destroy_workqueue(cache->wq); -- cgit v1.2.3 From f50cb2cbabd6c4a60add93d72451728f86e4791c Mon Sep 17 00:00:00 2001 From: Luo Meng Date: Tue, 29 Nov 2022 10:48:50 +0800 Subject: dm integrity: Fix UAF in dm_integrity_dtr() Dm_integrity also has the same UAF problem when dm_resume() and dm_destroy() are concurrent. Therefore, cancelling timer again in dm_integrity_dtr(). Cc: stable@vger.kernel.org Fixes: 7eada909bfd7a ("dm: add integrity target") Signed-off-by: Luo Meng Signed-off-by: Mike Snitzer --- drivers/md/dm-integrity.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/md') diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c index e97e9f97456d..1388ee35571e 100644 --- a/drivers/md/dm-integrity.c +++ b/drivers/md/dm-integrity.c @@ -4558,6 +4558,8 @@ static void dm_integrity_dtr(struct dm_target *ti) BUG_ON(!RB_EMPTY_ROOT(&ic->in_progress)); BUG_ON(!list_empty(&ic->wait_list)); + if (ic->mode == 'B') + cancel_delayed_work_sync(&ic->bitmap_flush_work); if (ic->metadata_wq) destroy_workqueue(ic->metadata_wq); if (ic->wait_wq) -- cgit v1.2.3 From 8111964f1b8524c4bb56b02cd9c7a37725ea21fd Mon Sep 17 00:00:00 2001 From: Zhihao Cheng Date: Wed, 30 Nov 2022 21:31:34 +0800 Subject: dm thin: Fix ABBA deadlock between shrink_slab and dm_pool_abort_metadata MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Following concurrent processes: P1(drop cache) P2(kworker) drop_caches_sysctl_handler drop_slab shrink_slab down_read(&shrinker_rwsem) - LOCK A do_shrink_slab super_cache_scan prune_icache_sb dispose_list evict ext4_evict_inode ext4_clear_inode ext4_discard_preallocations ext4_mb_load_buddy_gfp ext4_mb_init_cache ext4_read_block_bitmap_nowait ext4_read_bh_nowait submit_bh dm_submit_bio do_worker process_deferred_bios commit metadata_operation_failed dm_pool_abort_metadata down_write(&pmd->root_lock) - LOCK B __destroy_persistent_data_objects dm_block_manager_destroy dm_bufio_client_destroy unregister_shrinker down_write(&shrinker_rwsem) thin_map | dm_thin_find_block ↓ down_read(&pmd->root_lock) --> ABBA deadlock , which triggers hung task: [ 76.974820] INFO: task kworker/u4:3:63 blocked for more than 15 seconds. [ 76.976019] Not tainted 6.1.0-rc4-00011-g8f17dd350364-dirty #910 [ 76.978521] task:kworker/u4:3 state:D stack:0 pid:63 ppid:2 [ 76.978534] Workqueue: dm-thin do_worker [ 76.978552] Call Trace: [ 76.978564] __schedule+0x6ba/0x10f0 [ 76.978582] schedule+0x9d/0x1e0 [ 76.978588] rwsem_down_write_slowpath+0x587/0xdf0 [ 76.978600] down_write+0xec/0x110 [ 76.978607] unregister_shrinker+0x2c/0xf0 [ 76.978616] dm_bufio_client_destroy+0x116/0x3d0 [ 76.978625] dm_block_manager_destroy+0x19/0x40 [ 76.978629] __destroy_persistent_data_objects+0x5e/0x70 [ 76.978636] dm_pool_abort_metadata+0x8e/0x100 [ 76.978643] metadata_operation_failed+0x86/0x110 [ 76.978649] commit+0x6a/0x230 [ 76.978655] do_worker+0xc6e/0xd90 [ 76.978702] process_one_work+0x269/0x630 [ 76.978714] worker_thread+0x266/0x630 [ 76.978730] kthread+0x151/0x1b0 [ 76.978772] INFO: task test.sh:2646 blocked for more than 15 seconds. [ 76.979756] Not tainted 6.1.0-rc4-00011-g8f17dd350364-dirty #910 [ 76.982111] task:test.sh state:D stack:0 pid:2646 ppid:2459 [ 76.982128] Call Trace: [ 76.982139] __schedule+0x6ba/0x10f0 [ 76.982155] schedule+0x9d/0x1e0 [ 76.982159] rwsem_down_read_slowpath+0x4f4/0x910 [ 76.982173] down_read+0x84/0x170 [ 76.982177] dm_thin_find_block+0x4c/0xd0 [ 76.982183] thin_map+0x201/0x3d0 [ 76.982188] __map_bio+0x5b/0x350 [ 76.982195] dm_submit_bio+0x2b6/0x930 [ 76.982202] __submit_bio+0x123/0x2d0 [ 76.982209] submit_bio_noacct_nocheck+0x101/0x3e0 [ 76.982222] submit_bio_noacct+0x389/0x770 [ 76.982227] submit_bio+0x50/0xc0 [ 76.982232] submit_bh_wbc+0x15e/0x230 [ 76.982238] submit_bh+0x14/0x20 [ 76.982241] ext4_read_bh_nowait+0xc5/0x130 [ 76.982247] ext4_read_block_bitmap_nowait+0x340/0xc60 [ 76.982254] ext4_mb_init_cache+0x1ce/0xdc0 [ 76.982259] ext4_mb_load_buddy_gfp+0x987/0xfa0 [ 76.982263] ext4_discard_preallocations+0x45d/0x830 [ 76.982274] ext4_clear_inode+0x48/0xf0 [ 76.982280] ext4_evict_inode+0xcf/0xc70 [ 76.982285] evict+0x119/0x2b0 [ 76.982290] dispose_list+0x43/0xa0 [ 76.982294] prune_icache_sb+0x64/0x90 [ 76.982298] super_cache_scan+0x155/0x210 [ 76.982303] do_shrink_slab+0x19e/0x4e0 [ 76.982310] shrink_slab+0x2bd/0x450 [ 76.982317] drop_slab+0xcc/0x1a0 [ 76.982323] drop_caches_sysctl_handler+0xb7/0xe0 [ 76.982327] proc_sys_call_handler+0x1bc/0x300 [ 76.982331] proc_sys_write+0x17/0x20 [ 76.982334] vfs_write+0x3d3/0x570 [ 76.982342] ksys_write+0x73/0x160 [ 76.982347] __x64_sys_write+0x1e/0x30 [ 76.982352] do_syscall_64+0x35/0x80 [ 76.982357] entry_SYSCALL_64_after_hwframe+0x63/0xcd Function metadata_operation_failed() is called when operations failed on dm pool metadata, dm pool will destroy and recreate metadata. So, shrinker will be unregistered and registered, which could down write shrinker_rwsem under pmd_write_lock. Fix it by allocating dm_block_manager before locking pmd->root_lock and destroying old dm_block_manager after unlocking pmd->root_lock, then old dm_block_manager is replaced with new dm_block_manager under pmd->root_lock. So, shrinker register/unregister could be done without holding pmd->root_lock. Fetch a reproducer in [Link]. Link: https://bugzilla.kernel.org/show_bug.cgi?id=216676 Cc: stable@vger.kernel.org #v5.2+ Fixes: e49e582965b3 ("dm thin: add read only and fail io modes") Signed-off-by: Zhihao Cheng Signed-off-by: Mike Snitzer --- drivers/md/dm-thin-metadata.c | 51 ++++++++++++++++++++++++++++++++++++------- 1 file changed, 43 insertions(+), 8 deletions(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c index a27395c8621f..1a62226ac34e 100644 --- a/drivers/md/dm-thin-metadata.c +++ b/drivers/md/dm-thin-metadata.c @@ -776,13 +776,15 @@ static int __create_persistent_data_objects(struct dm_pool_metadata *pmd, bool f return r; } -static void __destroy_persistent_data_objects(struct dm_pool_metadata *pmd) +static void __destroy_persistent_data_objects(struct dm_pool_metadata *pmd, + bool destroy_bm) { dm_sm_destroy(pmd->data_sm); dm_sm_destroy(pmd->metadata_sm); dm_tm_destroy(pmd->nb_tm); dm_tm_destroy(pmd->tm); - dm_block_manager_destroy(pmd->bm); + if (destroy_bm) + dm_block_manager_destroy(pmd->bm); } static int __begin_transaction(struct dm_pool_metadata *pmd) @@ -989,7 +991,7 @@ int dm_pool_metadata_close(struct dm_pool_metadata *pmd) } pmd_write_unlock(pmd); if (!pmd->fail_io) - __destroy_persistent_data_objects(pmd); + __destroy_persistent_data_objects(pmd, true); kfree(pmd); return 0; @@ -1860,19 +1862,52 @@ static void __set_abort_with_changes_flags(struct dm_pool_metadata *pmd) int dm_pool_abort_metadata(struct dm_pool_metadata *pmd) { int r = -EINVAL; + struct dm_block_manager *old_bm = NULL, *new_bm = NULL; + + /* fail_io is double-checked with pmd->root_lock held below */ + if (unlikely(pmd->fail_io)) + return r; + + /* + * Replacement block manager (new_bm) is created and old_bm destroyed outside of + * pmd root_lock to avoid ABBA deadlock that would result (due to life-cycle of + * shrinker associated with the block manager's bufio client vs pmd root_lock). + * - must take shrinker_rwsem without holding pmd->root_lock + */ + new_bm = dm_block_manager_create(pmd->bdev, THIN_METADATA_BLOCK_SIZE << SECTOR_SHIFT, + THIN_MAX_CONCURRENT_LOCKS); pmd_write_lock(pmd); - if (pmd->fail_io) + if (pmd->fail_io) { + pmd_write_unlock(pmd); goto out; + } __set_abort_with_changes_flags(pmd); - __destroy_persistent_data_objects(pmd); - r = __create_persistent_data_objects(pmd, false); + __destroy_persistent_data_objects(pmd, false); + old_bm = pmd->bm; + if (IS_ERR(new_bm)) { + DMERR("could not create block manager during abort"); + pmd->bm = NULL; + r = PTR_ERR(new_bm); + goto out_unlock; + } + + pmd->bm = new_bm; + r = __open_or_format_metadata(pmd, false); + if (r) { + pmd->bm = NULL; + goto out_unlock; + } + new_bm = NULL; +out_unlock: if (r) pmd->fail_io = true; - -out: pmd_write_unlock(pmd); + dm_block_manager_destroy(old_bm); +out: + if (new_bm && !IS_ERR(new_bm)) + dm_block_manager_destroy(new_bm); return r; } -- cgit v1.2.3 From 352b837a5541690d4f843819028cf2b8be83d424 Mon Sep 17 00:00:00 2001 From: Mike Snitzer Date: Wed, 30 Nov 2022 13:26:32 -0500 Subject: dm cache: Fix ABBA deadlock between shrink_slab and dm_cache_metadata_abort Same ABBA deadlock pattern fixed in commit 4b60f452ec51 ("dm thin: Fix ABBA deadlock between shrink_slab and dm_pool_abort_metadata") to DM-cache's metadata. Reported-by: Zhihao Cheng Cc: stable@vger.kernel.org Fixes: 028ae9f76f29 ("dm cache: add fail io mode and needs_check flag") Signed-off-by: Mike Snitzer --- drivers/md/dm-cache-metadata.c | 54 ++++++++++++++++++++++++++++++++++++------ 1 file changed, 47 insertions(+), 7 deletions(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-cache-metadata.c b/drivers/md/dm-cache-metadata.c index ab13b7380265..83a5975bcc72 100644 --- a/drivers/md/dm-cache-metadata.c +++ b/drivers/md/dm-cache-metadata.c @@ -551,11 +551,13 @@ static int __create_persistent_data_objects(struct dm_cache_metadata *cmd, return r; } -static void __destroy_persistent_data_objects(struct dm_cache_metadata *cmd) +static void __destroy_persistent_data_objects(struct dm_cache_metadata *cmd, + bool destroy_bm) { dm_sm_destroy(cmd->metadata_sm); dm_tm_destroy(cmd->tm); - dm_block_manager_destroy(cmd->bm); + if (destroy_bm) + dm_block_manager_destroy(cmd->bm); } typedef unsigned long (*flags_mutator)(unsigned long); @@ -826,7 +828,7 @@ static struct dm_cache_metadata *lookup_or_open(struct block_device *bdev, cmd2 = lookup(bdev); if (cmd2) { mutex_unlock(&table_lock); - __destroy_persistent_data_objects(cmd); + __destroy_persistent_data_objects(cmd, true); kfree(cmd); return cmd2; } @@ -874,7 +876,7 @@ void dm_cache_metadata_close(struct dm_cache_metadata *cmd) mutex_unlock(&table_lock); if (!cmd->fail_io) - __destroy_persistent_data_objects(cmd); + __destroy_persistent_data_objects(cmd, true); kfree(cmd); } } @@ -1807,14 +1809,52 @@ int dm_cache_metadata_needs_check(struct dm_cache_metadata *cmd, bool *result) int dm_cache_metadata_abort(struct dm_cache_metadata *cmd) { - int r; + int r = -EINVAL; + struct dm_block_manager *old_bm = NULL, *new_bm = NULL; + + /* fail_io is double-checked with cmd->root_lock held below */ + if (unlikely(cmd->fail_io)) + return r; + + /* + * Replacement block manager (new_bm) is created and old_bm destroyed outside of + * cmd root_lock to avoid ABBA deadlock that would result (due to life-cycle of + * shrinker associated with the block manager's bufio client vs cmd root_lock). + * - must take shrinker_rwsem without holding cmd->root_lock + */ + new_bm = dm_block_manager_create(cmd->bdev, DM_CACHE_METADATA_BLOCK_SIZE << SECTOR_SHIFT, + CACHE_MAX_CONCURRENT_LOCKS); WRITE_LOCK(cmd); - __destroy_persistent_data_objects(cmd); - r = __create_persistent_data_objects(cmd, false); + if (cmd->fail_io) { + WRITE_UNLOCK(cmd); + goto out; + } + + __destroy_persistent_data_objects(cmd, false); + old_bm = cmd->bm; + if (IS_ERR(new_bm)) { + DMERR("could not create block manager during abort"); + cmd->bm = NULL; + r = PTR_ERR(new_bm); + goto out_unlock; + } + + cmd->bm = new_bm; + r = __open_or_format_metadata(cmd, false); + if (r) { + cmd->bm = NULL; + goto out_unlock; + } + new_bm = NULL; +out_unlock: if (r) cmd->fail_io = true; WRITE_UNLOCK(cmd); + dm_block_manager_destroy(old_bm); +out: + if (new_bm && !IS_ERR(new_bm)) + dm_block_manager_destroy(new_bm); return r; } -- cgit v1.2.3 From 6b9973861cb2e96dcd0bb0f1baddc5c034207c5c Mon Sep 17 00:00:00 2001 From: Mike Snitzer Date: Wed, 30 Nov 2022 14:02:47 -0500 Subject: dm cache: set needs_check flag after aborting metadata Otherwise the commit that will be aborted will be associated with the metadata objects that will be torn down. Must write needs_check flag to metadata with a reset block manager. Found through code-inspection (and compared against dm-thin.c). Cc: stable@vger.kernel.org Fixes: 028ae9f76f29 ("dm cache: add fail io mode and needs_check flag") Signed-off-by: Mike Snitzer --- drivers/md/dm-cache-target.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-cache-target.c b/drivers/md/dm-cache-target.c index 624a6335c832..5e92fac90b67 100644 --- a/drivers/md/dm-cache-target.c +++ b/drivers/md/dm-cache-target.c @@ -907,16 +907,16 @@ static void abort_transaction(struct cache *cache) if (get_cache_mode(cache) >= CM_READ_ONLY) return; - if (dm_cache_metadata_set_needs_check(cache->cmd)) { - DMERR("%s: failed to set 'needs_check' flag in metadata", dev_name); - set_cache_mode(cache, CM_FAIL); - } - DMERR_LIMIT("%s: aborting current metadata transaction", dev_name); if (dm_cache_metadata_abort(cache->cmd)) { DMERR("%s: failed to abort metadata transaction", dev_name); set_cache_mode(cache, CM_FAIL); } + + if (dm_cache_metadata_set_needs_check(cache->cmd)) { + DMERR("%s: failed to set 'needs_check' flag in metadata", dev_name); + set_cache_mode(cache, CM_FAIL); + } } static void metadata_operation_failed(struct cache *cache, const char *op, int r) -- cgit v1.2.3 From 19eb1650afeb1aa86151f61900e9e5f1de5d8d02 Mon Sep 17 00:00:00 2001 From: Luo Meng Date: Wed, 30 Nov 2022 10:09:45 +0800 Subject: dm thin: resume even if in FAIL mode If a thinpool set fail_io while suspending, resume will fail with: device-mapper: resume ioctl on vg-thinpool failed: Invalid argument The thin-pool also can't be removed if an in-flight bio is in the deferred list. This can be easily reproduced using: echo "offline" > /sys/block/sda/device/state dd if=/dev/zero of=/dev/mapper/thin bs=4K count=1 dmsetup suspend /dev/mapper/pool mkfs.ext4 /dev/mapper/thin dmsetup resume /dev/mapper/pool The root cause is maybe_resize_data_dev() will check fail_io and return error before called dm_resume. Fix this by adding FAIL mode check at the end of pool_preresume(). Cc: stable@vger.kernel.org Fixes: da105ed5fd7e ("dm thin metadata: introduce dm_pool_abort_metadata") Signed-off-by: Luo Meng Signed-off-by: Mike Snitzer --- drivers/md/dm-thin.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index dc271c107fb5..196f82559ad6 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -3542,20 +3542,28 @@ static int pool_preresume(struct dm_target *ti) */ r = bind_control_target(pool, ti); if (r) - return r; + goto out; r = maybe_resize_data_dev(ti, &need_commit1); if (r) - return r; + goto out; r = maybe_resize_metadata_dev(ti, &need_commit2); if (r) - return r; + goto out; if (need_commit1 || need_commit2) (void) commit(pool); +out: + /* + * When a thin-pool is PM_FAIL, it cannot be rebuilt if + * bio is in deferred list. Therefore need to return 0 + * to allow pool_resume() to flush IO. + */ + if (r && get_pool_mode(pool) == PM_FAIL) + r = 0; - return 0; + return r; } static void pool_suspend_active_thins(struct pool *pool) -- cgit v1.2.3 From d043f9a1ca929eee76dac940b4ccfb5429e6098f Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Tue, 1 Nov 2022 16:54:01 -0400 Subject: dm ioctl: a small code cleanup in list_version_get_info No need to modify info->vers if we overwrite it immediately. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer --- drivers/md/dm-ioctl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index 3bfc1583c20a..cdadc77fa4fc 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -681,7 +681,7 @@ static void list_version_get_info(struct target_type *tt, void *param) strcpy(info->vers->name, tt->name); info->old_vers = info->vers; - info->vers = align_ptr(((void *) ++info->vers) + strlen(tt->name) + 1); + info->vers = align_ptr((void *)(info->vers + 1) + strlen(tt->name) + 1); } static int __list_versions(struct dm_ioctl *param, size_t param_size, const char *name) -- cgit v1.2.3 From b52c3de84bd02083f2bbaae130adb243a689dd63 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Tue, 1 Nov 2022 16:54:27 -0400 Subject: dm ioctl: fix a couple ioctl codes Change the ioctl codes from DM_DEV_ARM_POLL to DM_DEV_ARM_POLL_CMD and from DM_GET_TARGET_VERSION to DM_GET_TARGET_VERSION_CMD. Note that the "cmd" field of "struct _ioctls" is never used, thus this commit doesn't fix any bug, it just makes the code consistent. Signed-off-by: Mikulas Patocka Signed-off-by: Mike Snitzer --- drivers/md/dm-ioctl.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/md') diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index cdadc77fa4fc..36fc6ae4737a 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -1788,8 +1788,8 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) {DM_TARGET_MSG_CMD, 0, target_message}, {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, - {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, - {DM_GET_TARGET_VERSION, 0, get_target_version}, + {DM_DEV_ARM_POLL_CMD, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, + {DM_GET_TARGET_VERSION_CMD, 0, get_target_version}, }; if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) -- cgit v1.2.3 From 035641b01e72af4f6c6cf22a4bdb5d7dfc4e8e8e Mon Sep 17 00:00:00 2001 From: Peter Korsgaard Date: Wed, 16 Nov 2022 07:16:56 +0100 Subject: dm init: add dm-mod.waitfor to wait for asynchronously probed block devices Just calling wait_for_device_probe() is not enough to ensure that asynchronously probed block devices are available (E.G. mmc, usb), so add a "dm-mod.waitfor=[,..,]" parameter to get dm-init to explicitly wait for specific block devices before initializing the tables with logic similar to the rootwait logic that was introduced with commit cc1ed7542c8c ("init: wait for asynchronously scanned block devices"). E.G. with dm-verity on mmc using: dm-mod.waitfor="PARTLABEL=hash-a,PARTLABEL=root-a" [ 0.671671] device-mapper: init: waiting for all devices to be available before creating mapped devices [ 0.671679] device-mapper: init: waiting for device PARTLABEL=hash-a ... [ 0.710695] mmc0: new HS200 MMC card at address 0001 [ 0.711158] mmcblk0: mmc0:0001 004GA0 3.69 GiB [ 0.715954] mmcblk0boot0: mmc0:0001 004GA0 partition 1 2.00 MiB [ 0.722085] mmcblk0boot1: mmc0:0001 004GA0 partition 2 2.00 MiB [ 0.728093] mmcblk0rpmb: mmc0:0001 004GA0 partition 3 512 KiB, chardev (249:0) [ 0.738274] mmcblk0: p1 p2 p3 p4 p5 p6 p7 [ 0.751282] device-mapper: init: waiting for device PARTLABEL=root-a ... [ 0.751306] device-mapper: init: all devices available [ 0.751683] device-mapper: verity: sha256 using implementation "sha256-generic" [ 0.759344] device-mapper: ioctl: dm-0 (vroot) is ready [ 0.766540] VFS: Mounted root (squashfs filesystem) readonly on device 254:0. Signed-off-by: Peter Korsgaard Signed-off-by: Mike Snitzer --- .../admin-guide/device-mapper/dm-init.rst | 8 ++++++++ drivers/md/dm-init.c | 22 +++++++++++++++++++++- 2 files changed, 29 insertions(+), 1 deletion(-) (limited to 'drivers/md') diff --git a/Documentation/admin-guide/device-mapper/dm-init.rst b/Documentation/admin-guide/device-mapper/dm-init.rst index e5242ff17e9b..981d6a907699 100644 --- a/Documentation/admin-guide/device-mapper/dm-init.rst +++ b/Documentation/admin-guide/device-mapper/dm-init.rst @@ -123,3 +123,11 @@ Other examples (per target): 0 1638400 verity 1 8:1 8:2 4096 4096 204800 1 sha256 fb1a5a0f00deb908d8b53cb270858975e76cf64105d412ce764225d53b8f3cfd 51934789604d1b92399c52e7cb149d1b3a1b74bbbcb103b2a0aaacbed5c08584 + +For setups using device-mapper on top of asynchronously probed block +devices (MMC, USB, ..), it may be necessary to tell dm-init to +explicitly wait for them to become available before setting up the +device-mapper tables. This can be done with the "dm-mod.waitfor=" +module parameter, which takes a list of devices to wait for:: + + dm-mod.waitfor=[,..,] diff --git a/drivers/md/dm-init.c b/drivers/md/dm-init.c index b0c45c6ebe0b..dc4381d68313 100644 --- a/drivers/md/dm-init.c +++ b/drivers/md/dm-init.c @@ -8,6 +8,7 @@ */ #include +#include #include #include #include @@ -18,12 +19,17 @@ #define DM_MAX_DEVICES 256 #define DM_MAX_TARGETS 256 #define DM_MAX_STR_SIZE 4096 +#define DM_MAX_WAITFOR 256 static char *create; +static char *waitfor[DM_MAX_WAITFOR]; + /* * Format: dm-mod.create=,,,,[,
+][;,,,,
[,
+]+] * Table format: + * Block devices to wait for to become available before setting up tables: + * dm-mod.waitfor=[,..,] * * See Documentation/admin-guide/device-mapper/dm-init.rst for dm-mod.create="..." format * details. @@ -266,7 +272,7 @@ static int __init dm_init_init(void) struct dm_device *dev; LIST_HEAD(devices); char *str; - int r; + int i, r; if (!create) return 0; @@ -286,6 +292,17 @@ static int __init dm_init_init(void) DMINFO("waiting for all devices to be available before creating mapped devices"); wait_for_device_probe(); + for (i = 0; i < ARRAY_SIZE(waitfor); i++) { + if (waitfor[i]) { + DMINFO("waiting for device %s ...", waitfor[i]); + while (!dm_get_dev_t(waitfor[i])) + msleep(5); + } + } + + if (waitfor[0]) + DMINFO("all devices available"); + list_for_each_entry(dev, &devices, list) { if (dm_early_create(&dev->dmi, dev->table, dev->target_args_array)) @@ -301,3 +318,6 @@ late_initcall(dm_init_init); module_param(create, charp, 0); MODULE_PARM_DESC(create, "Create a mapped device in early boot"); + +module_param_array(waitfor, charp, NULL, 0); +MODULE_PARM_DESC(waitfor, "Devices to wait for before setting up tables"); -- cgit v1.2.3 From 7991dbff6849f67e823b7cc0c15e5a90b0549b9f Mon Sep 17 00:00:00 2001 From: Zhihao Cheng Date: Thu, 8 Dec 2022 22:28:02 +0800 Subject: dm thin: Use last transaction's pmd->root when commit failed Recently we found a softlock up problem in dm thin pool btree lookup code due to corrupted metadata: Kernel panic - not syncing: softlockup: hung tasks CPU: 7 PID: 2669225 Comm: kworker/u16:3 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996) Workqueue: dm-thin do_worker [dm_thin_pool] Call Trace: dump_stack+0x9c/0xd3 panic+0x35d/0x6b9 watchdog_timer_fn.cold+0x16/0x25 __run_hrtimer+0xa2/0x2d0 RIP: 0010:__relink_lru+0x102/0x220 [dm_bufio] __bufio_new+0x11f/0x4f0 [dm_bufio] new_read+0xa3/0x1e0 [dm_bufio] dm_bm_read_lock+0x33/0xd0 [dm_persistent_data] ro_step+0x63/0x100 [dm_persistent_data] btree_lookup_raw.constprop.0+0x44/0x220 [dm_persistent_data] dm_btree_lookup+0x16f/0x210 [dm_persistent_data] dm_thin_find_block+0x12c/0x210 [dm_thin_pool] __process_bio_read_only+0xc5/0x400 [dm_thin_pool] process_thin_deferred_bios+0x1a4/0x4a0 [dm_thin_pool] process_one_work+0x3c5/0x730 Following process may generate a broken btree mixed with fresh and stale btree nodes, which could get dm thin trapped in an infinite loop while looking up data block: Transaction 1: pmd->root = A, A->B->C // One path in btree pmd->root = X, X->Y->Z // Copy-up Transaction 2: X,Z is updated on disk, Y write failed. // Commit failed, dm thin becomes read-only. process_bio_read_only dm_thin_find_block __find_block dm_btree_lookup(pmd->root) The pmd->root points to a broken btree, Y may contain stale node pointing to any block, for example X, which gets dm thin trapped into a dead loop while looking up Z. Fix this by setting pmd->root in __open_metadata(), so that dm thin will use the last transaction's pmd->root if commit failed. Fetch a reproducer in [Link]. Linke: https://bugzilla.kernel.org/show_bug.cgi?id=216790 Cc: stable@vger.kernel.org Fixes: 991d9fa02da0 ("dm: add thin provisioning target") Signed-off-by: Zhihao Cheng Acked-by: Joe Thornber Signed-off-by: Mike Snitzer --- drivers/md/dm-thin-metadata.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'drivers/md') diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c index 1a62226ac34e..6bcc4c4786d8 100644 --- a/drivers/md/dm-thin-metadata.c +++ b/drivers/md/dm-thin-metadata.c @@ -724,6 +724,15 @@ static int __open_metadata(struct dm_pool_metadata *pmd) goto bad_cleanup_data_sm; } + /* + * For pool metadata opening process, root setting is redundant + * because it will be set again in __begin_transaction(). But dm + * pool aborting process really needs to get last transaction's + * root to avoid accessing broken btree. + */ + pmd->root = le64_to_cpu(disk_super->data_mapping_root); + pmd->details_root = le64_to_cpu(disk_super->device_details_root); + __setup_btree_details(pmd); dm_bm_unlock(sblock); -- cgit v1.2.3