Message ID | 20200712174736.9840-1-colyli@suse.de (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [1/2] bcache: avoid nr_stripes overflow in bcache_device_init() | expand |
On 2020/7/13 01:47, Coly Li wrote: > 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 an unsigned long 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. Hi Ken, Could you please to try whether these two patches may avoid the kernel panic ? I will post the overwhelm stripe_size patch later. Thanks. Coly Li
On 7/12/20 11:47 AM, Coly Li wrote: > 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 an unsigned long 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. Does that work on 32-bit, where sizeof(unsigned long) == 4?
On 7/12/20 1:49 PM, Coly Li wrote: > On 2020/7/13 01:47, Coly Li wrote: >> 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 an unsigned long 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. > Hi Ken, > > Could you please to try whether these two patches may avoid the kernel > panic ? I will post the overwhelm stripe_size patch later. > > Thanks. > > Coly Li > I will. But, from inspection: On a 32-bit system, "unsigned long" will still be 32 bits, but sector_t (u64) will still be 64 bits, so that assignment will still discard high bits before validation in that environment. I suggest "unsigned long long" or another specifically 64-bit type. Also, the VDO driver I work on doesn't support 32-bit platforms currently, so my own testing will be limited to 64-bit platforms. Ken
On 2020/7/13 04:37, Jens Axboe wrote: > On 7/12/20 11:47 AM, Coly Li wrote: >> 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 an unsigned long 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. > > Does that work on 32-bit, where sizeof(unsigned long) == 4? > I will post v2 to explicit use uint64_t. Thanks for the remind. Coly Li
On 2020/7/13 04:38, Ken Raeburn wrote: > > On 7/12/20 1:49 PM, Coly Li wrote: >> On 2020/7/13 01:47, Coly Li wrote: >>> 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 an unsigned long 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. >> Hi Ken, >> >> Could you please to try whether these two patches may avoid the kernel >> panic ? I will post the overwhelm stripe_size patch later. >> >> Thanks. >> >> Coly Li >> > I will. But, from inspection: On a 32-bit system, "unsigned long" will > still be 32 bits, but sector_t (u64) will still be 64 bits, so that > assignment will still discard high bits before validation in that > environment. I suggest "unsigned long long" or another specifically > 64-bit type. > > Also, the VDO driver I work on doesn't support 32-bit platforms > currently, so my own testing will be limited to 64-bit platforms. I will post a v2 for your test. Thanks :-) Coly Li
Hi [This is an automated email] This commit has been processed because it contains a -stable tag. The stable tag indicates that it's relevant for the following trees: all The bot has tested the following trees: v5.7.8, v5.4.51, v4.19.132, v4.14.188, v4.9.230, v4.4.230. v5.7.8: Failed to apply! Possible dependencies: 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") v5.4.51: Failed to apply! Possible dependencies: 253a99d95d5b3 ("bcache: move macro btree() and btree_root() into btree.h") 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") 49d08d596e85f ("bcache: check return value of prio_read()") 8e7102273f597 ("bcache: make bch_btree_check() to be multithreaded") b144e45fc5764 ("bcache: make bch_sectors_dirty_init() to be multithreaded") feac1a70b8063 ("bcache: add bcache_ prefix to btree_root() and btree() macros") v4.19.132: Failed to apply! Possible dependencies: 0b13efecf5f25 ("bcache: add return value check to bch_cached_dev_run()") 253a99d95d5b3 ("bcache: move macro btree() and btree_root() into btree.h") 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") 49d08d596e85f ("bcache: check return value of prio_read()") 4b6efb4bdbce2 ("bcache: more detailed error message to bcache_device_link()") 5c2a634cbfaf1 ("bcache: stop writeback kthread and kworker when bch_cached_dev_run() failed") 633bb2ce60b94 ("bcache: add more error message in bch_cached_dev_attach()") 792732d9852c0 ("bcache: use kmemdup_nul for CACHED_LABEL buffer") 88c12d42d2bb6 ("bcache: add error check for calling register_bdev()") 8e7102273f597 ("bcache: make bch_btree_check() to be multithreaded") 91be66e1318f6 ("bcache: performance improvement for btree_flush_write()") cb07ad63682ff ("bcache: introduce force_wake_up_gc()") e0faa3d7f79f7 ("bcache: improve error message in bch_cached_dev_run()") feac1a70b8063 ("bcache: add bcache_ prefix to btree_root() and btree() macros") v4.14.188: Failed to apply! Possible dependencies: 1d316e658374f ("bcache: implement PI controller for writeback rate") 1dbe32ad0a82f ("bcache: rewrite multiple partitions support") 25d8be77e1922 ("block: move bio_alloc_pages() to bcache") 27a40ab9269e7 ("bcache: add backing_request_endio() for bi_end_io") 2831231d4c3f9 ("bcache: reduce cache_set devices iteration by devices_max_used") 3b304d24a718a ("bcache: convert cached_dev.count from atomic_t to refcount_t") 3fd47bfe55b00 ("bcache: stop dc->writeback_rate_update properly") 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") 5138ac6748e38 ("bcache: fix misleading error message in bch_count_io_errors()") 539d39eb27083 ("bcache: fix wrong return value in bch_debug_init()") 5f2b18ec8e164 ("bcache: Fix a compiler warning in bcache_device_init()") 5fa89fb9a86bc ("bcache: don't write back data if reading it failed") 6ae63e3501c49 ("bcache: replace printk() by pr_*() routines") 6f10f7d1b02b1 ("bcache: style fix to replace 'unsigned' by 'unsigned int'") 771f393e8ffc9 ("bcache: add CACHE_SET_IO_DISABLE to struct cache_set flags") 7ba0d830dc0e4 ("bcache: set error_limit correctly") 7e027ca4b534b ("bcache: add stop_when_cache_set_failed option to backing device") 804f3c6981f5e ("bcache: fix cached_dev->count usage for bch_cache_set_error()") a8500fc816b19 ("bcache: rearrange writeback main thread ratelimit") b1092c9af9ed8 ("bcache: allow quick writeback when backing idle") bc082a55d25c8 ("bcache: fix inaccurate io state for detached bcache devices") c7b7bd07404c5 ("bcache: add io_disable to struct cached_dev") v4.9.230: Failed to apply! Possible dependencies: 1d316e658374f ("bcache: implement PI controller for writeback rate") 1dbe32ad0a82f ("bcache: rewrite multiple partitions support") 2831231d4c3f9 ("bcache: reduce cache_set devices iteration by devices_max_used") 297e3d8547848 ("blk-throttle: make throtl_slice tunable") 3fd47bfe55b00 ("bcache: stop dc->writeback_rate_update properly") 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") 4e4cbee93d561 ("block: switch bios to blk_status_t") 5138ac6748e38 ("bcache: fix misleading error message in bch_count_io_errors()") 5f2b18ec8e164 ("bcache: Fix a compiler warning in bcache_device_init()") 6ae63e3501c49 ("bcache: replace printk() by pr_*() routines") 6f10f7d1b02b1 ("bcache: style fix to replace 'unsigned' by 'unsigned int'") 7e027ca4b534b ("bcache: add stop_when_cache_set_failed option to backing device") 87760e5eef359 ("block: hook up writeback throttling") 9e234eeafbe17 ("blk-throttle: add a simple idle detection") b8c0d911ac528 ("bcache: partition support: add 16 minors per bcacheN device") c7b7bd07404c5 ("bcache: add io_disable to struct cached_dev") cf43e6be865a5 ("block: add scalable completion tracking of requests") e806402130c9c ("block: split out request-only flags into a new namespace") fbbaf700e7b16 ("block: trace completion of all bios.") v4.4.230: Failed to apply! Possible dependencies: 005411ea7ee77 ("doc: update block/queue-sysfs.txt entries") 1d316e658374f ("bcache: implement PI controller for writeback rate") 1dbe32ad0a82f ("bcache: rewrite multiple partitions support") 2831231d4c3f9 ("bcache: reduce cache_set devices iteration by devices_max_used") 297e3d8547848 ("blk-throttle: make throtl_slice tunable") 38f8baae89056 ("block: factor out chained bio completion") 3fd47bfe55b00 ("bcache: stop dc->writeback_rate_update properly") 46f5aa8806e34 ("bcache: Convert pr_<level> uses to a more typical style") 4e4cbee93d561 ("block: switch bios to blk_status_t") 511cbce2ff8b9 ("irq_poll: make blk-iopoll available outside the block layer") 5138ac6748e38 ("bcache: fix misleading error message in bch_count_io_errors()") 5f2b18ec8e164 ("bcache: Fix a compiler warning in bcache_device_init()") 6ae63e3501c49 ("bcache: replace printk() by pr_*() routines") 6f10f7d1b02b1 ("bcache: style fix to replace 'unsigned' by 'unsigned int'") 7e027ca4b534b ("bcache: add stop_when_cache_set_failed option to backing device") 87760e5eef359 ("block: hook up writeback throttling") 90706094d5be6 ("bcache: pr_err: more meaningful error message when nr_stripes is invalid") 9467f85960a31 ("blk-mq/cpu-notif: Convert to new hotplug state machine") 9e234eeafbe17 ("blk-throttle: add a simple idle detection") af3e3a5259e35 ("block: don't unecessarily clobber bi_error for chained bios") b8c0d911ac528 ("bcache: partition support: add 16 minors per bcacheN device") ba8c6967b7391 ("block: cleanup bio_endio") c7b7bd07404c5 ("bcache: add io_disable to struct cached_dev") cf43e6be865a5 ("block: add scalable completion tracking of requests") fbbaf700e7b16 ("block: trace completion of all bios.") NOTE: The patch will not be queued to stable trees until it is upstream. How should we proceed with this patch?
diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index a239fcaec70b..0c25ebc035b1 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -886,19 +886,19 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size, struct request_queue *q; const size_t max_stripes = min_t(size_t, INT_MAX, SIZE_MAX / sizeof(atomic_t)); - size_t n; + unsigned long n; int idx; if (!d->stripe_size) d->stripe_size = 1 << 31; - d->nr_stripes = DIV_ROUND_UP_ULL(sectors, d->stripe_size); - - if (!d->nr_stripes || d->nr_stripes > max_stripes) { - pr_err("nr_stripes too large or invalid: %u (start sector beyond end of disk?)\n", - (unsigned int)d->nr_stripes); + n = DIV_ROUND_UP_ULL(sectors, d->stripe_size); + if (!n || n > max_stripes) { + pr_err("nr_stripes too large or invalid: %lu (start sector beyond end of disk?)\n", + n); return -ENOMEM; } + d->nr_stripes = n; n = d->nr_stripes * sizeof(atomic_t); d->stripe_sectors_dirty = kvzalloc(n, GFP_KERNEL);
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 an unsigned long 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-by: Ken Raeburn <raeburn@redhat.com> Signed-off-by: Coly Li <colyli@suse.de> Link: https://bugzilla.redhat.com/show_bug.cgi?id=1783075 Cc: stable@vger.kernel.org --- drivers/md/bcache/super.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)