| Message ID | 20230918044739.29782-1-mirsad.todorovac@alu.unizg.hr (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v1,1/1] xarray: fix the data-race in xas_find_chunk() by using READ_ONCE() | expand |
On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: > KCSAN has discovered the following data-race: > > [ 206.510010] ================================================================== > [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked > > [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: > [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) > [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) > [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) > [ 206.510128] folio_end_writeback (mm/filemap.c:1616) > [ 206.510143] end_page_writeback (mm/folio-compat.c:28) > [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs > [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs > [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs > [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs > [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs > [ 206.514350] bio_endio (block/bio.c:1617) > [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) > [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme > [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme > [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) > [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) > [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) > [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) > [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) > [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) > [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) > [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) > [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs > [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs > [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs > [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs > [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs > [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs > [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs > [ 206.520643] process_one_work (kernel/workqueue.c:2600) > [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) > [ 206.520672] kthread (kernel/kthread.c:389) > [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) > [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) > > [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: > [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) > [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) > [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) > [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) > [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs > [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs > [ 206.522495] vfs_fsync_range (fs/sync.c:188) > [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) > [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) > [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) > > [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 > > [ 206.522574] Reported by Kernel Concurrency Sanitizer on: > [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 > [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 > [ 206.522608] ================================================================== Thanks for working on this. I guess the full KCSAN warning isn't that useful in the changelog. Rather I'd spend more time explaining the real problem here ... > As Jan Kara explained, the problem is in the function xas_find_chuck(): > > /* Private */ > static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > xa_mark_t mark) > { > unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; > unsigned int offset = xas->xa_offset; > > if (advance) > offset++; > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > if (offset < XA_CHUNK_SIZE) { > → unsigned long data = *addr & (~0UL << offset); > if (data) > return __ffs(data); ... which is that xas_find_chunk() is called only under RCU protection and thus the two uses of 'data' in the above code can yield different results. > } > return XA_CHUNK_SIZE; > } > > return find_next_bit(addr, XA_CHUNK_SIZE, offset); > } > > In particular, the line > > unsigned long data = *addr & (~0UL << offset); > > contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN > data-race warning completely. Yes, this improves the situation for xarray use on 64-bit architectures but doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I mentioned in my previous reply, I'd rather: 1) Fix find_next_bit(), find_first_bit() and related functions in lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is quite some churn but I don't see how else to make these functions safe when the underlying contents can change. 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days because find_next_bit() is inline and does small_const_nbits(size) check. Honza > Reported-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> > Suggested-by: Jan Kara <jack@suse.cz> > Fixes: b803b42823d0d ("xarray: Add XArray iterators") > Matthew Wilcox <willy@infradead.org> > Cc: Chris Mason <clm@fb.com> > Cc: Andrew Morton <akpm@linux-foundation.org> > Cc: Josef Bacik <josef@toxicpanda.com> > Cc: David Sterba <dsterba@suse.com> > Cc: linux-btrfs@vger.kernel.org > Cc: linux-fsdevel@vger.kernel.org > Cc: linux-mm@kvack.org > Signed-off-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> > --- > v1: the proposed fix (RFC) > > include/linux/xarray.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/xarray.h b/include/linux/xarray.h > index cb571dfcf4b1..1715fd322d62 100644 > --- a/include/linux/xarray.h > +++ b/include/linux/xarray.h > @@ -1720,7 +1720,7 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > offset++; > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > if (offset < XA_CHUNK_SIZE) { > - unsigned long data = *addr & (~0UL << offset); > + unsigned long data = READ_ONCE(*addr) & (~0UL << offset); > if (data) > return __ffs(data); > } > -- > 2.34.1 >
On 9/18/23 11:41, Jan Kara wrote: > On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: >> KCSAN has discovered the following data-race: >> >> [ 206.510010] ================================================================== >> [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked >> >> [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: >> [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) >> [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) >> [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) >> [ 206.510128] folio_end_writeback (mm/filemap.c:1616) >> [ 206.510143] end_page_writeback (mm/folio-compat.c:28) >> [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs >> [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs >> [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs >> [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs >> [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs >> [ 206.514350] bio_endio (block/bio.c:1617) >> [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) >> [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme >> [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme >> [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) >> [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) >> [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) >> [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) >> [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) >> [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) >> [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) >> [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) >> [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs >> [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs >> [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs >> [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs >> [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs >> [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs >> [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs >> [ 206.520643] process_one_work (kernel/workqueue.c:2600) >> [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) >> [ 206.520672] kthread (kernel/kthread.c:389) >> [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) >> [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) >> >> [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: >> [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) >> [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) >> [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) >> [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) >> [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs >> [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs >> [ 206.522495] vfs_fsync_range (fs/sync.c:188) >> [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) >> [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) >> [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) >> >> [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 >> >> [ 206.522574] Reported by Kernel Concurrency Sanitizer on: >> [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 >> [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 >> [ 206.522608] ================================================================== > > Thanks for working on this. I guess the full KCSAN warning isn't that > useful in the changelog. Rather I'd spend more time explaining the real > problem here ... > >> As Jan Kara explained, the problem is in the function xas_find_chuck(): >> >> /* Private */ >> static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >> xa_mark_t mark) >> { >> unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; >> unsigned int offset = xas->xa_offset; >> >> if (advance) >> offset++; >> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >> if (offset < XA_CHUNK_SIZE) { >> → unsigned long data = *addr & (~0UL << offset); >> if (data) >> return __ffs(data); > > ... which is that xas_find_chunk() is called only under RCU protection and > thus the two uses of 'data' in the above code can yield different results. > >> } >> return XA_CHUNK_SIZE; >> } >> >> return find_next_bit(addr, XA_CHUNK_SIZE, offset); >> } >> >> In particular, the line >> >> unsigned long data = *addr & (~0UL << offset); >> >> contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN >> data-race warning completely. > > Yes, this improves the situation for xarray use on 64-bit architectures but > doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I > mentioned in my previous reply, I'd rather: > > 1) Fix find_next_bit(), find_first_bit() and related functions in > lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is > quite some churn but I don't see how else to make these functions safe when > the underlying contents can change. Thank you for your review. I assume you have the big picture, but just a stupid question: if (XA_CHUNK_SIZE == BITS_PER_LONG) { if (offset < XA_CHUNK_SIZE) { unsigned long data = READ_ONCE(*addr) & (~0UL << offset); if (data) return __ffs(data); } return XA_CHUNK_SIZE; } I would hate to argue, but ... Wouldn't BITS_PER_LONG simply change to 32 on 32-bit architectures? Is there something I am missing? From include/asm-generic/bitsperlong.h: ---------------------------------------- #ifdef CONFIG_64BIT #define BITS_PER_LONG 64 #else #define BITS_PER_LONG 32 #endif /* CONFIG_64BIT */ About the CONFIG_BASE_SMALL I cannot tell: ---------------------------------------- #ifndef XA_CHUNK_SHIFT #define XA_CHUNK_SHIFT (CONFIG_BASE_SMALL ? 4 : 6) #endif #define XA_CHUNK_SIZE (1UL << XA_CHUNK_SHIFT) #define XA_CHUNK_MASK (XA_CHUNK_SIZE - 1) #define XA_MAX_MARKS 3 #define XA_MARK_LONGS DIV_ROUND_UP(XA_CHUNK_SIZE, BITS_PER_LONG) ---------------------------------------- I see why you would want find_next_bit() and find_first_bit() fixed, but I am not that deep into those bitops, so I guess I cannot make this in one step ... Probably it would require a lot of homework. _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) 5 _find_next_zero_bit ... but I am simply not certain what is the right thing to do ATM about those and whether they are false positives. AFAICS, READ_ONCE() here solves the case of 64 and 32 architectures which is an incremental step, and it works ... I am just not ready for an universal solution ATM. > 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the > special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days > because find_next_bit() is inline and does small_const_nbits(size) check. I see your point. A generalised solution would of course be better. But from the report about data-races in those functions it seems that they need a major rethink. It isn't that obvious to me what should be READ_ONCE()-ed in a bit field ... Those functions are extensively used throughout the kernel and I get the notion it is a job for someone with more experience ... Best regards, Mirsad > Honza > > >> Reported-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> >> Suggested-by: Jan Kara <jack@suse.cz> >> Fixes: b803b42823d0d ("xarray: Add XArray iterators") >> Matthew Wilcox <willy@infradead.org> >> Cc: Chris Mason <clm@fb.com> >> Cc: Andrew Morton <akpm@linux-foundation.org> >> Cc: Josef Bacik <josef@toxicpanda.com> >> Cc: David Sterba <dsterba@suse.com> >> Cc: linux-btrfs@vger.kernel.org >> Cc: linux-fsdevel@vger.kernel.org >> Cc: linux-mm@kvack.org >> Signed-off-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> >> --- >> v1: the proposed fix (RFC) >> >> include/linux/xarray.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/include/linux/xarray.h b/include/linux/xarray.h >> index cb571dfcf4b1..1715fd322d62 100644 >> --- a/include/linux/xarray.h >> +++ b/include/linux/xarray.h >> @@ -1720,7 +1720,7 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >> offset++; >> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >> if (offset < XA_CHUNK_SIZE) { >> - unsigned long data = *addr & (~0UL << offset); >> + unsigned long data = READ_ONCE(*addr) & (~0UL << offset); >> if (data) >> return __ffs(data); >> } >> -- >> 2.34.1 >>
On Mon 18-09-23 12:20:09, Mirsad Todorovac wrote: > On 9/18/23 11:41, Jan Kara wrote: > > On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: > > > KCSAN has discovered the following data-race: > > > > > > [ 206.510010] ================================================================== > > > [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked > > > > > > [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: > > > [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) > > > [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) > > > [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) > > > [ 206.510128] folio_end_writeback (mm/filemap.c:1616) > > > [ 206.510143] end_page_writeback (mm/folio-compat.c:28) > > > [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs > > > [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs > > > [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs > > > [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs > > > [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs > > > [ 206.514350] bio_endio (block/bio.c:1617) > > > [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) > > > [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme > > > [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme > > > [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) > > > [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) > > > [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) > > > [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) > > > [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) > > > [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) > > > [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) > > > [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) > > > [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs > > > [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs > > > [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs > > > [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs > > > [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs > > > [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs > > > [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs > > > [ 206.520643] process_one_work (kernel/workqueue.c:2600) > > > [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) > > > [ 206.520672] kthread (kernel/kthread.c:389) > > > [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) > > > [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) > > > > > > [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: > > > [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) > > > [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) > > > [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) > > > [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) > > > [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs > > > [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs > > > [ 206.522495] vfs_fsync_range (fs/sync.c:188) > > > [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) > > > [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) > > > [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) > > > > > > [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 > > > > > > [ 206.522574] Reported by Kernel Concurrency Sanitizer on: > > > [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 > > > [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 > > > [ 206.522608] ================================================================== > > > > Thanks for working on this. I guess the full KCSAN warning isn't that > > useful in the changelog. Rather I'd spend more time explaining the real > > problem here ... > > > > > As Jan Kara explained, the problem is in the function xas_find_chuck(): > > > > > > /* Private */ > > > static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > > > xa_mark_t mark) > > > { > > > unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; > > > unsigned int offset = xas->xa_offset; > > > > > > if (advance) > > > offset++; > > > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > > > if (offset < XA_CHUNK_SIZE) { > > > → unsigned long data = *addr & (~0UL << offset); > > > if (data) > > > return __ffs(data); > > > > ... which is that xas_find_chunk() is called only under RCU protection and > > thus the two uses of 'data' in the above code can yield different results. > > > > > } > > > return XA_CHUNK_SIZE; > > > } > > > > > > return find_next_bit(addr, XA_CHUNK_SIZE, offset); > > > } > > > > > > In particular, the line > > > > > > unsigned long data = *addr & (~0UL << offset); > > > > > > contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN > > > data-race warning completely. > > > > Yes, this improves the situation for xarray use on 64-bit architectures but > > doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I > > mentioned in my previous reply, I'd rather: > > > > 1) Fix find_next_bit(), find_first_bit() and related functions in > > lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is > > quite some churn but I don't see how else to make these functions safe when > > the underlying contents can change. > > Thank you for your review. > > I assume you have the big picture, but just a stupid question: > > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > if (offset < XA_CHUNK_SIZE) { > unsigned long data = READ_ONCE(*addr) & (~0UL << offset); > if (data) > return __ffs(data); > } > return XA_CHUNK_SIZE; > } > > I would hate to argue, but ... No problem, asking questions isn't argueing ;). > Wouldn't BITS_PER_LONG simply change to 32 on 32-bit architectures? Yes, they will. But XA_CHUNK_SIZE will still be 64 on 32-bit AFAICT so XA_CHUNK_SIZE != BITS_PER_LONG there. > Is there something I am missing? > > From include/asm-generic/bitsperlong.h: > ---------------------------------------- > #ifdef CONFIG_64BIT > #define BITS_PER_LONG 64 > #else > #define BITS_PER_LONG 32 > #endif /* CONFIG_64BIT */ > > About the CONFIG_BASE_SMALL I cannot tell: > ---------------------------------------- > #ifndef XA_CHUNK_SHIFT > #define XA_CHUNK_SHIFT (CONFIG_BASE_SMALL ? 4 : 6) > #endif > #define XA_CHUNK_SIZE (1UL << XA_CHUNK_SHIFT) > #define XA_CHUNK_MASK (XA_CHUNK_SIZE - 1) > #define XA_MAX_MARKS 3 > #define XA_MARK_LONGS DIV_ROUND_UP(XA_CHUNK_SIZE, BITS_PER_LONG) > ---------------------------------------- Again with CONFIG_BASE_SMALL we have XA_CHUNK_SIZE == 16 so it will not be equal to BITS_PER_LONG. > I see why you would want find_next_bit() and find_first_bit() fixed, but > I am not that deep into those bitops, so I guess I cannot make this in > one step ... Probably it would require a lot of homework. > > _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: > > 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) > 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) > 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) > 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) > 5 _find_next_zero_bit > > ... but I am simply not certain what is the right thing to do ATM about > those and whether they are false positives. Well, it would require some auditing to be sure but there is at least one user of these functions (xarray) where the problem is real so given the fix has no real runtime cost the fix looks justified. > AFAICS, READ_ONCE() here solves the case of 64 and 32 architectures which is > an incremental step, and it works ... I am just not ready for an > universal solution ATM. > > > 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the > > special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days > > because find_next_bit() is inline and does small_const_nbits(size) check. > > I see your point. A generalised solution would of course be better. But > from the report about data-races in those functions it seems that they > need a major rethink. It isn't that obvious to me what should be > READ_ONCE()-ed in a bit field ... Well, it's actually not that difficult. They all need a treatment like: unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, uns { - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); } > Those functions are extensively used throughout the kernel and I get the > notion it is a job for someone with more experience ... Sure, if you don't feel like doing the general change, I can look into it myself. Honza
On 9/18/23 13:38, Jan Kara wrote: > On Mon 18-09-23 12:20:09, Mirsad Todorovac wrote: >> On 9/18/23 11:41, Jan Kara wrote: >>> On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: >>>> KCSAN has discovered the following data-race: >>>> >>>> [ 206.510010] ================================================================== >>>> [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked >>>> >>>> [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: >>>> [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) >>>> [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) >>>> [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) >>>> [ 206.510128] folio_end_writeback (mm/filemap.c:1616) >>>> [ 206.510143] end_page_writeback (mm/folio-compat.c:28) >>>> [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs >>>> [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs >>>> [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs >>>> [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs >>>> [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs >>>> [ 206.514350] bio_endio (block/bio.c:1617) >>>> [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) >>>> [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme >>>> [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme >>>> [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) >>>> [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) >>>> [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) >>>> [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) >>>> [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) >>>> [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) >>>> [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) >>>> [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) >>>> [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs >>>> [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs >>>> [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs >>>> [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs >>>> [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs >>>> [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs >>>> [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs >>>> [ 206.520643] process_one_work (kernel/workqueue.c:2600) >>>> [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) >>>> [ 206.520672] kthread (kernel/kthread.c:389) >>>> [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) >>>> [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) >>>> >>>> [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: >>>> [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) >>>> [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) >>>> [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) >>>> [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) >>>> [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs >>>> [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs >>>> [ 206.522495] vfs_fsync_range (fs/sync.c:188) >>>> [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) >>>> [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) >>>> [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) >>>> >>>> [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 >>>> >>>> [ 206.522574] Reported by Kernel Concurrency Sanitizer on: >>>> [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 >>>> [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 >>>> [ 206.522608] ================================================================== >>> >>> Thanks for working on this. I guess the full KCSAN warning isn't that >>> useful in the changelog. Rather I'd spend more time explaining the real >>> problem here ... >>> >>>> As Jan Kara explained, the problem is in the function xas_find_chuck(): >>>> >>>> /* Private */ >>>> static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >>>> xa_mark_t mark) >>>> { >>>> unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; >>>> unsigned int offset = xas->xa_offset; >>>> >>>> if (advance) >>>> offset++; >>>> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >>>> if (offset < XA_CHUNK_SIZE) { >>>> → unsigned long data = *addr & (~0UL << offset); >>>> if (data) >>>> return __ffs(data); >>> >>> ... which is that xas_find_chunk() is called only under RCU protection and >>> thus the two uses of 'data' in the above code can yield different results. >>> >>>> } >>>> return XA_CHUNK_SIZE; >>>> } >>>> >>>> return find_next_bit(addr, XA_CHUNK_SIZE, offset); >>>> } >>>> >>>> In particular, the line >>>> >>>> unsigned long data = *addr & (~0UL << offset); >>>> >>>> contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN >>>> data-race warning completely. >>> >>> Yes, this improves the situation for xarray use on 64-bit architectures but >>> doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I >>> mentioned in my previous reply, I'd rather: >>> >>> 1) Fix find_next_bit(), find_first_bit() and related functions in >>> lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is >>> quite some churn but I don't see how else to make these functions safe when >>> the underlying contents can change. >> >> Thank you for your review. >> >> I assume you have the big picture, but just a stupid question: >> >> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >> if (offset < XA_CHUNK_SIZE) { >> unsigned long data = READ_ONCE(*addr) & (~0UL << offset); >> if (data) >> return __ffs(data); >> } >> return XA_CHUNK_SIZE; >> } >> >> I would hate to argue, but ... > > No problem, asking questions isn't argueing ;). > >> Wouldn't BITS_PER_LONG simply change to 32 on 32-bit architectures? > > Yes, they will. But XA_CHUNK_SIZE will still be 64 on 32-bit AFAICT so > XA_CHUNK_SIZE != BITS_PER_LONG there. Ah, I see. This is definitely not good. But I managed to fix and test the find_next_bit() family, but this seems that simply ------------------------------------------- include/linux/xarray.h | 8 -------- 1 file changed, 8 deletions(-) diff --git a/include/linux/xarray.h b/include/linux/xarray.h index 1715fd322d62..89918b65b00d 100644 --- a/include/linux/xarray.h +++ b/include/linux/xarray.h @@ -1718,14 +1718,6 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, if (advance) offset++; - if (XA_CHUNK_SIZE == BITS_PER_LONG) { - if (offset < XA_CHUNK_SIZE) { - unsigned long data = READ_ONCE(*addr) & (~0UL << offset); - if (data) - return __ffs(data); - } - return XA_CHUNK_SIZE; - } return find_next_bit(addr, XA_CHUNK_SIZE, offset); } seems too good to be true. According to what you explained, the performance impact would be negligent or non-existing, and the CONFIG_BASE_SMALL problem would disappear? I did not even try to run that, as I am not 100% confident in the logic. Am I doing something very wrong? >> Is there something I am missing? >> >> From include/asm-generic/bitsperlong.h: >> ---------------------------------------- >> #ifdef CONFIG_64BIT >> #define BITS_PER_LONG 64 >> #else >> #define BITS_PER_LONG 32 >> #endif /* CONFIG_64BIT */ >> >> About the CONFIG_BASE_SMALL I cannot tell: >> ---------------------------------------- >> #ifndef XA_CHUNK_SHIFT >> #define XA_CHUNK_SHIFT (CONFIG_BASE_SMALL ? 4 : 6) >> #endif >> #define XA_CHUNK_SIZE (1UL << XA_CHUNK_SHIFT) >> #define XA_CHUNK_MASK (XA_CHUNK_SIZE - 1) >> #define XA_MAX_MARKS 3 >> #define XA_MARK_LONGS DIV_ROUND_UP(XA_CHUNK_SIZE, BITS_PER_LONG) >> ---------------------------------------- > > Again with CONFIG_BASE_SMALL we have XA_CHUNK_SIZE == 16 so it will not be > equal to BITS_PER_LONG. > >> I see why you would want find_next_bit() and find_first_bit() fixed, but >> I am not that deep into those bitops, so I guess I cannot make this in >> one step ... Probably it would require a lot of homework. >> >> _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: >> >> 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) >> 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) >> 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) >> 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) >> 5 _find_next_zero_bit >> >> ... but I am simply not certain what is the right thing to do ATM about >> those and whether they are false positives. > > Well, it would require some auditing to be sure but there is at least one > user of these functions (xarray) where the problem is real so given the fix > has no real runtime cost the fix looks justified. > >> AFAICS, READ_ONCE() here solves the case of 64 and 32 architectures which is >> an incremental step, and it works ... I am just not ready for an >> universal solution ATM. >> >>> 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the >>> special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days >>> because find_next_bit() is inline and does small_const_nbits(size) check. >> >> I see your point. A generalised solution would of course be better. But >> from the report about data-races in those functions it seems that they >> need a major rethink. It isn't that obvious to me what should be >> READ_ONCE()-ed in a bit field ... > > Well, it's actually not that difficult. They all need a treatment like: > > unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, uns > { > - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); > } > > >> Those functions are extensively used throughout the kernel and I get the >> notion it is a job for someone with more experience ... > > Sure, if you don't feel like doing the general change, I can look into it > myself. > > Honza Hi, I tried this patch and the >> 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) >> 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) >> 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) >> 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) >> 5 _find_next_zero_bit data-races do not seem to appear any longer. -------------------------------------------------------- lib/find_bit.c | 33 +++++++++++++++++---------------- 1 file changed, 17 insertions(+), 16 deletions(-) diff --git a/lib/find_bit.c b/lib/find_bit.c index 32f99e9a670e..56244e4f744e 100644 --- a/lib/find_bit.c +++ b/lib/find_bit.c @@ -18,6 +18,7 @@ #include <linux/math.h> #include <linux/minmax.h> #include <linux/swab.h> +#include <asm/rwonce.h> /* * Common helper for find_bit() function family @@ -98,7 +99,7 @@ out: \ */ unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) { - return FIND_FIRST_BIT(addr[idx], /* nop */, size); + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); } EXPORT_SYMBOL(_find_first_bit); #endif @@ -111,7 +112,7 @@ unsigned long _find_first_and_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long size) { - return FIND_FIRST_BIT(addr1[idx] & addr2[idx], /* nop */, size); + return FIND_FIRST_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, size); } EXPORT_SYMBOL(_find_first_and_bit); #endif @@ -122,7 +123,7 @@ EXPORT_SYMBOL(_find_first_and_bit); */ unsigned long _find_first_zero_bit(const unsigned long *addr, unsigned long size) { - return FIND_FIRST_BIT(~addr[idx], /* nop */, size); + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), /* nop */, size); } EXPORT_SYMBOL(_find_first_zero_bit); #endif @@ -130,28 +131,28 @@ EXPORT_SYMBOL(_find_first_zero_bit); #ifndef find_next_bit unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, unsigned long start) { - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); } EXPORT_SYMBOL(_find_next_bit); #endif unsigned long __find_nth_bit(const unsigned long *addr, unsigned long size, unsigned long n) { - return FIND_NTH_BIT(addr[idx], size, n); + return FIND_NTH_BIT(READ_ONCE(addr[idx]), size, n); } EXPORT_SYMBOL(__find_nth_bit); unsigned long __find_nth_and_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long size, unsigned long n) { - return FIND_NTH_BIT(addr1[idx] & addr2[idx], size, n); + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), size, n); } EXPORT_SYMBOL(__find_nth_and_bit); unsigned long __find_nth_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long size, unsigned long n) { - return FIND_NTH_BIT(addr1[idx] & ~addr2[idx], size, n); + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), size, n); } EXPORT_SYMBOL(__find_nth_andnot_bit); @@ -160,7 +161,7 @@ unsigned long __find_nth_and_andnot_bit(const unsigned long *addr1, const unsigned long *addr3, unsigned long size, unsigned long n) { - return FIND_NTH_BIT(addr1[idx] & addr2[idx] & ~addr3[idx], size, n); + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]) & ~READ_ONCE(addr3[idx]), size, n); } EXPORT_SYMBOL(__find_nth_and_andnot_bit); @@ -168,7 +169,7 @@ EXPORT_SYMBOL(__find_nth_and_andnot_bit); unsigned long _find_next_and_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long nbits, unsigned long start) { - return FIND_NEXT_BIT(addr1[idx] & addr2[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, nbits, start); } EXPORT_SYMBOL(_find_next_and_bit); #endif @@ -177,7 +178,7 @@ EXPORT_SYMBOL(_find_next_and_bit); unsigned long _find_next_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long nbits, unsigned long start) { - return FIND_NEXT_BIT(addr1[idx] & ~addr2[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), /* nop */, nbits, start); } EXPORT_SYMBOL(_find_next_andnot_bit); #endif @@ -186,7 +187,7 @@ EXPORT_SYMBOL(_find_next_andnot_bit); unsigned long _find_next_or_bit(const unsigned long *addr1, const unsigned long *addr2, unsigned long nbits, unsigned long start) { - return FIND_NEXT_BIT(addr1[idx] | addr2[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) | READ_ONCE(addr2[idx]), /* nop */, nbits, start); } EXPORT_SYMBOL(_find_next_or_bit); #endif @@ -195,7 +196,7 @@ EXPORT_SYMBOL(_find_next_or_bit); unsigned long _find_next_zero_bit(const unsigned long *addr, unsigned long nbits, unsigned long start) { - return FIND_NEXT_BIT(~addr[idx], /* nop */, nbits, start); + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), /* nop */, nbits, start); } EXPORT_SYMBOL(_find_next_zero_bit); #endif @@ -208,7 +209,7 @@ unsigned long _find_last_bit(const unsigned long *addr, unsigned long size) unsigned long idx = (size-1) / BITS_PER_LONG; do { - val &= addr[idx]; + val &= READ_ONCE(addr[idx]); if (val) return idx * BITS_PER_LONG + __fls(val); @@ -242,7 +243,7 @@ EXPORT_SYMBOL(find_next_clump8); */ unsigned long _find_first_zero_bit_le(const unsigned long *addr, unsigned long size) { - return FIND_FIRST_BIT(~addr[idx], swab, size); + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), swab, size); } EXPORT_SYMBOL(_find_first_zero_bit_le); @@ -252,7 +253,7 @@ EXPORT_SYMBOL(_find_first_zero_bit_le); unsigned long _find_next_zero_bit_le(const unsigned long *addr, unsigned long size, unsigned long offset) { - return FIND_NEXT_BIT(~addr[idx], swab, size, offset); + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), swab, size, offset); } EXPORT_SYMBOL(_find_next_zero_bit_le); #endif @@ -261,7 +262,7 @@ EXPORT_SYMBOL(_find_next_zero_bit_le); unsigned long _find_next_bit_le(const unsigned long *addr, unsigned long size, unsigned long offset) { - return FIND_NEXT_BIT(addr[idx], swab, size, offset); + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), swab, size, offset); } EXPORT_SYMBOL(_find_next_bit_le); --
On Mon 18-09-23 14:46:02, Mirsad Todorovac wrote: > On 9/18/23 13:38, Jan Kara wrote: > > On Mon 18-09-23 12:20:09, Mirsad Todorovac wrote: > > > On 9/18/23 11:41, Jan Kara wrote: > > > > On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: > > > > > KCSAN has discovered the following data-race: > > > > > > > > > > [ 206.510010] ================================================================== > > > > > [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked > > > > > > > > > > [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: > > > > > [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) > > > > > [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) > > > > > [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) > > > > > [ 206.510128] folio_end_writeback (mm/filemap.c:1616) > > > > > [ 206.510143] end_page_writeback (mm/folio-compat.c:28) > > > > > [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs > > > > > [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs > > > > > [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs > > > > > [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs > > > > > [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs > > > > > [ 206.514350] bio_endio (block/bio.c:1617) > > > > > [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) > > > > > [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme > > > > > [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme > > > > > [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) > > > > > [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) > > > > > [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) > > > > > [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) > > > > > [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) > > > > > [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) > > > > > [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) > > > > > [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) > > > > > [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs > > > > > [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs > > > > > [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs > > > > > [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs > > > > > [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs > > > > > [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs > > > > > [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs > > > > > [ 206.520643] process_one_work (kernel/workqueue.c:2600) > > > > > [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) > > > > > [ 206.520672] kthread (kernel/kthread.c:389) > > > > > [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) > > > > > [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) > > > > > > > > > > [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: > > > > > [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) > > > > > [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) > > > > > [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) > > > > > [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) > > > > > [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs > > > > > [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs > > > > > [ 206.522495] vfs_fsync_range (fs/sync.c:188) > > > > > [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) > > > > > [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) > > > > > [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) > > > > > > > > > > [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 > > > > > > > > > > [ 206.522574] Reported by Kernel Concurrency Sanitizer on: > > > > > [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 > > > > > [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 > > > > > [ 206.522608] ================================================================== > > > > > > > > Thanks for working on this. I guess the full KCSAN warning isn't that > > > > useful in the changelog. Rather I'd spend more time explaining the real > > > > problem here ... > > > > > > > > > As Jan Kara explained, the problem is in the function xas_find_chuck(): > > > > > > > > > > /* Private */ > > > > > static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > > > > > xa_mark_t mark) > > > > > { > > > > > unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; > > > > > unsigned int offset = xas->xa_offset; > > > > > > > > > > if (advance) > > > > > offset++; > > > > > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > > > > > if (offset < XA_CHUNK_SIZE) { > > > > > → unsigned long data = *addr & (~0UL << offset); > > > > > if (data) > > > > > return __ffs(data); > > > > > > > > ... which is that xas_find_chunk() is called only under RCU protection and > > > > thus the two uses of 'data' in the above code can yield different results. > > > > > > > > > } > > > > > return XA_CHUNK_SIZE; > > > > > } > > > > > > > > > > return find_next_bit(addr, XA_CHUNK_SIZE, offset); > > > > > } > > > > > > > > > > In particular, the line > > > > > > > > > > unsigned long data = *addr & (~0UL << offset); > > > > > > > > > > contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN > > > > > data-race warning completely. > > > > > > > > Yes, this improves the situation for xarray use on 64-bit architectures but > > > > doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I > > > > mentioned in my previous reply, I'd rather: > > > > > > > > 1) Fix find_next_bit(), find_first_bit() and related functions in > > > > lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is > > > > quite some churn but I don't see how else to make these functions safe when > > > > the underlying contents can change. > > > > > > Thank you for your review. > > > > > > I assume you have the big picture, but just a stupid question: > > > > > > if (XA_CHUNK_SIZE == BITS_PER_LONG) { > > > if (offset < XA_CHUNK_SIZE) { > > > unsigned long data = READ_ONCE(*addr) & (~0UL << offset); > > > if (data) > > > return __ffs(data); > > > } > > > return XA_CHUNK_SIZE; > > > } > > > > > > I would hate to argue, but ... > > > > No problem, asking questions isn't argueing ;). > > > > > Wouldn't BITS_PER_LONG simply change to 32 on 32-bit architectures? > > > > Yes, they will. But XA_CHUNK_SIZE will still be 64 on 32-bit AFAICT so > > XA_CHUNK_SIZE != BITS_PER_LONG there. > > Ah, I see. This is definitely not good. But I managed to fix and test the find_next_bit() > family, but this seems that simply > > ------------------------------------------- > include/linux/xarray.h | 8 -------- > 1 file changed, 8 deletions(-) > > diff --git a/include/linux/xarray.h b/include/linux/xarray.h > index 1715fd322d62..89918b65b00d 100644 > --- a/include/linux/xarray.h > +++ b/include/linux/xarray.h > @@ -1718,14 +1718,6 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > if (advance) > offset++; > - if (XA_CHUNK_SIZE == BITS_PER_LONG) { > - if (offset < XA_CHUNK_SIZE) { > - unsigned long data = READ_ONCE(*addr) & (~0UL << offset); > - if (data) > - return __ffs(data); > - } > - return XA_CHUNK_SIZE; > - } > return find_next_bit(addr, XA_CHUNK_SIZE, offset); > } > > seems too good to be true. > > According to what you explained, the performance impact would be negligent or non-existing, > and the CONFIG_BASE_SMALL problem would disappear? If you fix find_next_bit(), then this will fix xas_find_chunk(), yes. > I did not even try to run that, as I am not 100% confident in the logic. > > Am I doing something very wrong? No, this is exactly what I think needs to happen. > > > Is there something I am missing? > > > > > > From include/asm-generic/bitsperlong.h: > > > ---------------------------------------- > > > #ifdef CONFIG_64BIT > > > #define BITS_PER_LONG 64 > > > #else > > > #define BITS_PER_LONG 32 > > > #endif /* CONFIG_64BIT */ > > > > > > About the CONFIG_BASE_SMALL I cannot tell: > > > ---------------------------------------- > > > #ifndef XA_CHUNK_SHIFT > > > #define XA_CHUNK_SHIFT (CONFIG_BASE_SMALL ? 4 : 6) > > > #endif > > > #define XA_CHUNK_SIZE (1UL << XA_CHUNK_SHIFT) > > > #define XA_CHUNK_MASK (XA_CHUNK_SIZE - 1) > > > #define XA_MAX_MARKS 3 > > > #define XA_MARK_LONGS DIV_ROUND_UP(XA_CHUNK_SIZE, BITS_PER_LONG) > > > ---------------------------------------- > > > > Again with CONFIG_BASE_SMALL we have XA_CHUNK_SIZE == 16 so it will not be > > equal to BITS_PER_LONG. > > > I see why you would want find_next_bit() and find_first_bit() fixed, but > > > I am not that deep into those bitops, so I guess I cannot make this in > > > one step ... Probably it would require a lot of homework. > > > > > > _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: > > > > > > 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) > > > 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) > > > 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) > > > 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) > > > 5 _find_next_zero_bit > > > > > > ... but I am simply not certain what is the right thing to do ATM about > > > those and whether they are false positives. > > > > Well, it would require some auditing to be sure but there is at least one > > user of these functions (xarray) where the problem is real so given the fix > > has no real runtime cost the fix looks justified. > > > > > AFAICS, READ_ONCE() here solves the case of 64 and 32 architectures which is > > > an incremental step, and it works ... I am just not ready for an > > > universal solution ATM. > > > > > > > 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the > > > > special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days > > > > because find_next_bit() is inline and does small_const_nbits(size) check. > > > > > > I see your point. A generalised solution would of course be better. But > > > from the report about data-races in those functions it seems that they > > > need a major rethink. It isn't that obvious to me what should be > > > READ_ONCE()-ed in a bit field ... > > > > Well, it's actually not that difficult. They all need a treatment like: > > > > unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, uns > > { > > - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); > > + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); > > } > > > > > > > Those functions are extensively used throughout the kernel and I get the > > > notion it is a job for someone with more experience ... > > > > Sure, if you don't feel like doing the general change, I can look into it > > myself. > > > > Honza > > Hi, > > I tried this patch and the > > > > 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) > > > 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) > > > 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) > > > 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) > > > 5 _find_next_zero_bit > > data-races do not seem to appear any longer. Yup. You've just missed one case in _find_last_bit() and then all the functions in include/linux/find.h need a similar treatment... Honza > > -------------------------------------------------------- > lib/find_bit.c | 33 +++++++++++++++++---------------- > 1 file changed, 17 insertions(+), 16 deletions(-) > > diff --git a/lib/find_bit.c b/lib/find_bit.c > index 32f99e9a670e..56244e4f744e 100644 > --- a/lib/find_bit.c > +++ b/lib/find_bit.c > @@ -18,6 +18,7 @@ > #include <linux/math.h> > #include <linux/minmax.h> > #include <linux/swab.h> > +#include <asm/rwonce.h> > /* > * Common helper for find_bit() function family > @@ -98,7 +99,7 @@ out: \ > */ > unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) > { > - return FIND_FIRST_BIT(addr[idx], /* nop */, size); > + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); > } > EXPORT_SYMBOL(_find_first_bit); > #endif > @@ -111,7 +112,7 @@ unsigned long _find_first_and_bit(const unsigned long *addr1, > const unsigned long *addr2, > unsigned long size) > { > - return FIND_FIRST_BIT(addr1[idx] & addr2[idx], /* nop */, size); > + return FIND_FIRST_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, size); > } > EXPORT_SYMBOL(_find_first_and_bit); > #endif > @@ -122,7 +123,7 @@ EXPORT_SYMBOL(_find_first_and_bit); > */ > unsigned long _find_first_zero_bit(const unsigned long *addr, unsigned long size) > { > - return FIND_FIRST_BIT(~addr[idx], /* nop */, size); > + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), /* nop */, size); > } > EXPORT_SYMBOL(_find_first_zero_bit); > #endif > @@ -130,28 +131,28 @@ EXPORT_SYMBOL(_find_first_zero_bit); > #ifndef find_next_bit > unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, unsigned long start) > { > - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); > } > EXPORT_SYMBOL(_find_next_bit); > #endif > unsigned long __find_nth_bit(const unsigned long *addr, unsigned long size, unsigned long n) > { > - return FIND_NTH_BIT(addr[idx], size, n); > + return FIND_NTH_BIT(READ_ONCE(addr[idx]), size, n); > } > EXPORT_SYMBOL(__find_nth_bit); > unsigned long __find_nth_and_bit(const unsigned long *addr1, const unsigned long *addr2, > unsigned long size, unsigned long n) > { > - return FIND_NTH_BIT(addr1[idx] & addr2[idx], size, n); > + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), size, n); > } > EXPORT_SYMBOL(__find_nth_and_bit); > unsigned long __find_nth_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, > unsigned long size, unsigned long n) > { > - return FIND_NTH_BIT(addr1[idx] & ~addr2[idx], size, n); > + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), size, n); > } > EXPORT_SYMBOL(__find_nth_andnot_bit); > @@ -160,7 +161,7 @@ unsigned long __find_nth_and_andnot_bit(const unsigned long *addr1, > const unsigned long *addr3, > unsigned long size, unsigned long n) > { > - return FIND_NTH_BIT(addr1[idx] & addr2[idx] & ~addr3[idx], size, n); > + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]) & ~READ_ONCE(addr3[idx]), size, n); > } > EXPORT_SYMBOL(__find_nth_and_andnot_bit); > @@ -168,7 +169,7 @@ EXPORT_SYMBOL(__find_nth_and_andnot_bit); > unsigned long _find_next_and_bit(const unsigned long *addr1, const unsigned long *addr2, > unsigned long nbits, unsigned long start) > { > - return FIND_NEXT_BIT(addr1[idx] & addr2[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, nbits, start); > } > EXPORT_SYMBOL(_find_next_and_bit); > #endif > @@ -177,7 +178,7 @@ EXPORT_SYMBOL(_find_next_and_bit); > unsigned long _find_next_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, > unsigned long nbits, unsigned long start) > { > - return FIND_NEXT_BIT(addr1[idx] & ~addr2[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), /* nop */, nbits, start); > } > EXPORT_SYMBOL(_find_next_andnot_bit); > #endif > @@ -186,7 +187,7 @@ EXPORT_SYMBOL(_find_next_andnot_bit); > unsigned long _find_next_or_bit(const unsigned long *addr1, const unsigned long *addr2, > unsigned long nbits, unsigned long start) > { > - return FIND_NEXT_BIT(addr1[idx] | addr2[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) | READ_ONCE(addr2[idx]), /* nop */, nbits, start); > } > EXPORT_SYMBOL(_find_next_or_bit); > #endif > @@ -195,7 +196,7 @@ EXPORT_SYMBOL(_find_next_or_bit); > unsigned long _find_next_zero_bit(const unsigned long *addr, unsigned long nbits, > unsigned long start) > { > - return FIND_NEXT_BIT(~addr[idx], /* nop */, nbits, start); > + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), /* nop */, nbits, start); > } > EXPORT_SYMBOL(_find_next_zero_bit); > #endif > @@ -208,7 +209,7 @@ unsigned long _find_last_bit(const unsigned long *addr, unsigned long size) > unsigned long idx = (size-1) / BITS_PER_LONG; > do { > - val &= addr[idx]; > + val &= READ_ONCE(addr[idx]); > if (val) > return idx * BITS_PER_LONG + __fls(val); > @@ -242,7 +243,7 @@ EXPORT_SYMBOL(find_next_clump8); > */ > unsigned long _find_first_zero_bit_le(const unsigned long *addr, unsigned long size) > { > - return FIND_FIRST_BIT(~addr[idx], swab, size); > + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), swab, size); > } > EXPORT_SYMBOL(_find_first_zero_bit_le); > @@ -252,7 +253,7 @@ EXPORT_SYMBOL(_find_first_zero_bit_le); > unsigned long _find_next_zero_bit_le(const unsigned long *addr, > unsigned long size, unsigned long offset) > { > - return FIND_NEXT_BIT(~addr[idx], swab, size, offset); > + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), swab, size, offset); > } > EXPORT_SYMBOL(_find_next_zero_bit_le); > #endif > @@ -261,7 +262,7 @@ EXPORT_SYMBOL(_find_next_zero_bit_le); > unsigned long _find_next_bit_le(const unsigned long *addr, > unsigned long size, unsigned long offset) > { > - return FIND_NEXT_BIT(addr[idx], swab, size, offset); > + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), swab, size, offset); > } > EXPORT_SYMBOL(_find_next_bit_le); > --
On 9/18/23 15:18, Jan Kara wrote: > On Mon 18-09-23 14:46:02, Mirsad Todorovac wrote: >> On 9/18/23 13:38, Jan Kara wrote: >>> On Mon 18-09-23 12:20:09, Mirsad Todorovac wrote: >>>> On 9/18/23 11:41, Jan Kara wrote: >>>>> On Mon 18-09-23 06:47:40, Mirsad Goran Todorovac wrote: >>>>>> KCSAN has discovered the following data-race: >>>>>> >>>>>> [ 206.510010] ================================================================== >>>>>> [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked >>>>>> >>>>>> [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: >>>>>> [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) >>>>>> [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) >>>>>> [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) >>>>>> [ 206.510128] folio_end_writeback (mm/filemap.c:1616) >>>>>> [ 206.510143] end_page_writeback (mm/folio-compat.c:28) >>>>>> [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs >>>>>> [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs >>>>>> [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs >>>>>> [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs >>>>>> [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs >>>>>> [ 206.514350] bio_endio (block/bio.c:1617) >>>>>> [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) >>>>>> [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme >>>>>> [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme >>>>>> [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) >>>>>> [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) >>>>>> [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) >>>>>> [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) >>>>>> [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) >>>>>> [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) >>>>>> [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) >>>>>> [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) >>>>>> [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs >>>>>> [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs >>>>>> [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs >>>>>> [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs >>>>>> [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs >>>>>> [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs >>>>>> [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs >>>>>> [ 206.520643] process_one_work (kernel/workqueue.c:2600) >>>>>> [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) >>>>>> [ 206.520672] kthread (kernel/kthread.c:389) >>>>>> [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) >>>>>> [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) >>>>>> >>>>>> [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: >>>>>> [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) >>>>>> [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) >>>>>> [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) >>>>>> [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) >>>>>> [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs >>>>>> [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs >>>>>> [ 206.522495] vfs_fsync_range (fs/sync.c:188) >>>>>> [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) >>>>>> [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) >>>>>> [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) >>>>>> >>>>>> [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 >>>>>> >>>>>> [ 206.522574] Reported by Kernel Concurrency Sanitizer on: >>>>>> [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 >>>>>> [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 >>>>>> [ 206.522608] ================================================================== >>>>> >>>>> Thanks for working on this. I guess the full KCSAN warning isn't that >>>>> useful in the changelog. Rather I'd spend more time explaining the real >>>>> problem here ... >>>>> >>>>>> As Jan Kara explained, the problem is in the function xas_find_chuck(): >>>>>> >>>>>> /* Private */ >>>>>> static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >>>>>> xa_mark_t mark) >>>>>> { >>>>>> unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; >>>>>> unsigned int offset = xas->xa_offset; >>>>>> >>>>>> if (advance) >>>>>> offset++; >>>>>> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >>>>>> if (offset < XA_CHUNK_SIZE) { >>>>>> → unsigned long data = *addr & (~0UL << offset); >>>>>> if (data) >>>>>> return __ffs(data); >>>>> >>>>> ... which is that xas_find_chunk() is called only under RCU protection and >>>>> thus the two uses of 'data' in the above code can yield different results. >>>>> >>>>>> } >>>>>> return XA_CHUNK_SIZE; >>>>>> } >>>>>> >>>>>> return find_next_bit(addr, XA_CHUNK_SIZE, offset); >>>>>> } >>>>>> >>>>>> In particular, the line >>>>>> >>>>>> unsigned long data = *addr & (~0UL << offset); >>>>>> >>>>>> contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN >>>>>> data-race warning completely. >>>>> >>>>> Yes, this improves the situation for xarray use on 64-bit architectures but >>>>> doesn't fix cases on 32-bit archs or if CONFIG_BASE_SMALL is set. As I >>>>> mentioned in my previous reply, I'd rather: >>>>> >>>>> 1) Fix find_next_bit(), find_first_bit() and related functions in >>>>> lib/find_bit.c to use READ_ONCE() - such as _find_first_bit() etc. It is >>>>> quite some churn but I don't see how else to make these functions safe when >>>>> the underlying contents can change. >>>> >>>> Thank you for your review. >>>> >>>> I assume you have the big picture, but just a stupid question: >>>> >>>> if (XA_CHUNK_SIZE == BITS_PER_LONG) { >>>> if (offset < XA_CHUNK_SIZE) { >>>> unsigned long data = READ_ONCE(*addr) & (~0UL << offset); >>>> if (data) >>>> return __ffs(data); >>>> } >>>> return XA_CHUNK_SIZE; >>>> } >>>> >>>> I would hate to argue, but ... >>> >>> No problem, asking questions isn't argueing ;). >>> >>>> Wouldn't BITS_PER_LONG simply change to 32 on 32-bit architectures? >>> >>> Yes, they will. But XA_CHUNK_SIZE will still be 64 on 32-bit AFAICT so >>> XA_CHUNK_SIZE != BITS_PER_LONG there. >> >> Ah, I see. This is definitely not good. But I managed to fix and test the find_next_bit() >> family, but this seems that simply >> >> ------------------------------------------- >> include/linux/xarray.h | 8 -------- >> 1 file changed, 8 deletions(-) >> >> diff --git a/include/linux/xarray.h b/include/linux/xarray.h >> index 1715fd322d62..89918b65b00d 100644 >> --- a/include/linux/xarray.h >> +++ b/include/linux/xarray.h >> @@ -1718,14 +1718,6 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >> if (advance) >> offset++; >> - if (XA_CHUNK_SIZE == BITS_PER_LONG) { >> - if (offset < XA_CHUNK_SIZE) { >> - unsigned long data = READ_ONCE(*addr) & (~0UL << offset); >> - if (data) >> - return __ffs(data); >> - } >> - return XA_CHUNK_SIZE; >> - } >> return find_next_bit(addr, XA_CHUNK_SIZE, offset); >> } >> >> seems too good to be true. >> >> According to what you explained, the performance impact would be negligent or non-existing, >> and the CONFIG_BASE_SMALL problem would disappear? > > If you fix find_next_bit(), then this will fix xas_find_chunk(), yes. Just testing that. >> I did not even try to run that, as I am not 100% confident in the logic. >> >> Am I doing something very wrong? > > No, this is exactly what I think needs to happen. Great then, I will apply this fix and try a test run. Of course, under your supervision ... >>>> Is there something I am missing? >>>> >>>> From include/asm-generic/bitsperlong.h: >>>> ---------------------------------------- >>>> #ifdef CONFIG_64BIT >>>> #define BITS_PER_LONG 64 >>>> #else >>>> #define BITS_PER_LONG 32 >>>> #endif /* CONFIG_64BIT */ >>>> >>>> About the CONFIG_BASE_SMALL I cannot tell: >>>> ---------------------------------------- >>>> #ifndef XA_CHUNK_SHIFT >>>> #define XA_CHUNK_SHIFT (CONFIG_BASE_SMALL ? 4 : 6) >>>> #endif >>>> #define XA_CHUNK_SIZE (1UL << XA_CHUNK_SHIFT) >>>> #define XA_CHUNK_MASK (XA_CHUNK_SIZE - 1) >>>> #define XA_MAX_MARKS 3 >>>> #define XA_MARK_LONGS DIV_ROUND_UP(XA_CHUNK_SIZE, BITS_PER_LONG) >>>> ---------------------------------------- >>> >>> Again with CONFIG_BASE_SMALL we have XA_CHUNK_SIZE == 16 so it will not be >>> equal to BITS_PER_LONG. >>>> I see why you would want find_next_bit() and find_first_bit() fixed, but >>>> I am not that deep into those bitops, so I guess I cannot make this in >>>> one step ... Probably it would require a lot of homework. >>>> >>>> _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: >>>> >>>> 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) >>>> 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) >>>> 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) >>>> 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) >>>> 5 _find_next_zero_bit >>>> >>>> ... but I am simply not certain what is the right thing to do ATM about >>>> those and whether they are false positives. >>> >>> Well, it would require some auditing to be sure but there is at least one >>> user of these functions (xarray) where the problem is real so given the fix >>> has no real runtime cost the fix looks justified. >>> >>>> AFAICS, READ_ONCE() here solves the case of 64 and 32 architectures which is >>>> an incremental step, and it works ... I am just not ready for an >>>> universal solution ATM. >>>> >>>>> 2) Change xas_find_chunk() to unconditionally use find_next_bit() as the >>>>> special case XA_CHUNK_SIZE == BITS_PER_LONG seems pointless these days >>>>> because find_next_bit() is inline and does small_const_nbits(size) check. >>>> >>>> I see your point. A generalised solution would of course be better. But >>>> from the report about data-races in those functions it seems that they >>>> need a major rethink. It isn't that obvious to me what should be >>>> READ_ONCE()-ed in a bit field ... >>> >>> Well, it's actually not that difficult. They all need a treatment like: >>> >>> unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, uns >>> { >>> - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); >>> + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); >>> } >>> >>> >>>> Those functions are extensively used throughout the kernel and I get the >>>> notion it is a job for someone with more experience ... >>> >>> Sure, if you don't feel like doing the general change, I can look into it >>> myself. >>> >>> Honza >> >> Hi, >> >> I tried this patch and the >> >>>> 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) >>>> 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) >>>> 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) >>>> 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) >>>> 5 _find_next_zero_bit >> >> data-races do not seem to appear any longer. > > Yup. You've just missed one case in _find_last_bit() and then all the > functions in include/linux/find.h need a similar treatment... I seem to have this: ----------------------------------------------------------------------------------- #ifndef find_last_bit unsigned long _find_last_bit(const unsigned long *addr, unsigned long size) { if (size) { unsigned long val = BITMAP_LAST_WORD_MASK(size); unsigned long idx = (size-1) / BITS_PER_LONG; do { val &= READ_ONCE(addr[idx]); if (val) return idx * BITS_PER_LONG + __fls(val); val = ~0ul; } while (idx--); } return size; } EXPORT_SYMBOL(_find_last_bit); #endif ----------------------------------------------------------------------------------- Is there something I did not notice? Just in case, I am adding the find_bit.diff because copy/paste might have been wrong ... Mirsad > > Honza > >> >> -------------------------------------------------------- >> lib/find_bit.c | 33 +++++++++++++++++---------------- >> 1 file changed, 17 insertions(+), 16 deletions(-) >> >> diff --git a/lib/find_bit.c b/lib/find_bit.c >> index 32f99e9a670e..56244e4f744e 100644 >> --- a/lib/find_bit.c >> +++ b/lib/find_bit.c >> @@ -18,6 +18,7 @@ >> #include <linux/math.h> >> #include <linux/minmax.h> >> #include <linux/swab.h> >> +#include <asm/rwonce.h> >> /* >> * Common helper for find_bit() function family >> @@ -98,7 +99,7 @@ out: \ >> */ >> unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) >> { >> - return FIND_FIRST_BIT(addr[idx], /* nop */, size); >> + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); >> } >> EXPORT_SYMBOL(_find_first_bit); >> #endif >> @@ -111,7 +112,7 @@ unsigned long _find_first_and_bit(const unsigned long *addr1, >> const unsigned long *addr2, >> unsigned long size) >> { >> - return FIND_FIRST_BIT(addr1[idx] & addr2[idx], /* nop */, size); >> + return FIND_FIRST_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, size); >> } >> EXPORT_SYMBOL(_find_first_and_bit); >> #endif >> @@ -122,7 +123,7 @@ EXPORT_SYMBOL(_find_first_and_bit); >> */ >> unsigned long _find_first_zero_bit(const unsigned long *addr, unsigned long size) >> { >> - return FIND_FIRST_BIT(~addr[idx], /* nop */, size); >> + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), /* nop */, size); >> } >> EXPORT_SYMBOL(_find_first_zero_bit); >> #endif >> @@ -130,28 +131,28 @@ EXPORT_SYMBOL(_find_first_zero_bit); >> #ifndef find_next_bit >> unsigned long _find_next_bit(const unsigned long *addr, unsigned long nbits, unsigned long start) >> { >> - return FIND_NEXT_BIT(addr[idx], /* nop */, nbits, start); >> + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), /* nop */, nbits, start); >> } >> EXPORT_SYMBOL(_find_next_bit); >> #endif >> unsigned long __find_nth_bit(const unsigned long *addr, unsigned long size, unsigned long n) >> { >> - return FIND_NTH_BIT(addr[idx], size, n); >> + return FIND_NTH_BIT(READ_ONCE(addr[idx]), size, n); >> } >> EXPORT_SYMBOL(__find_nth_bit); >> unsigned long __find_nth_and_bit(const unsigned long *addr1, const unsigned long *addr2, >> unsigned long size, unsigned long n) >> { >> - return FIND_NTH_BIT(addr1[idx] & addr2[idx], size, n); >> + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), size, n); >> } >> EXPORT_SYMBOL(__find_nth_and_bit); >> unsigned long __find_nth_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, >> unsigned long size, unsigned long n) >> { >> - return FIND_NTH_BIT(addr1[idx] & ~addr2[idx], size, n); >> + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), size, n); >> } >> EXPORT_SYMBOL(__find_nth_andnot_bit); >> @@ -160,7 +161,7 @@ unsigned long __find_nth_and_andnot_bit(const unsigned long *addr1, >> const unsigned long *addr3, >> unsigned long size, unsigned long n) >> { >> - return FIND_NTH_BIT(addr1[idx] & addr2[idx] & ~addr3[idx], size, n); >> + return FIND_NTH_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]) & ~READ_ONCE(addr3[idx]), size, n); >> } >> EXPORT_SYMBOL(__find_nth_and_andnot_bit); >> @@ -168,7 +169,7 @@ EXPORT_SYMBOL(__find_nth_and_andnot_bit); >> unsigned long _find_next_and_bit(const unsigned long *addr1, const unsigned long *addr2, >> unsigned long nbits, unsigned long start) >> { >> - return FIND_NEXT_BIT(addr1[idx] & addr2[idx], /* nop */, nbits, start); >> + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & READ_ONCE(addr2[idx]), /* nop */, nbits, start); >> } >> EXPORT_SYMBOL(_find_next_and_bit); >> #endif >> @@ -177,7 +178,7 @@ EXPORT_SYMBOL(_find_next_and_bit); >> unsigned long _find_next_andnot_bit(const unsigned long *addr1, const unsigned long *addr2, >> unsigned long nbits, unsigned long start) >> { >> - return FIND_NEXT_BIT(addr1[idx] & ~addr2[idx], /* nop */, nbits, start); >> + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) & ~READ_ONCE(addr2[idx]), /* nop */, nbits, start); >> } >> EXPORT_SYMBOL(_find_next_andnot_bit); >> #endif >> @@ -186,7 +187,7 @@ EXPORT_SYMBOL(_find_next_andnot_bit); >> unsigned long _find_next_or_bit(const unsigned long *addr1, const unsigned long *addr2, >> unsigned long nbits, unsigned long start) >> { >> - return FIND_NEXT_BIT(addr1[idx] | addr2[idx], /* nop */, nbits, start); >> + return FIND_NEXT_BIT(READ_ONCE(addr1[idx]) | READ_ONCE(addr2[idx]), /* nop */, nbits, start); >> } >> EXPORT_SYMBOL(_find_next_or_bit); >> #endif >> @@ -195,7 +196,7 @@ EXPORT_SYMBOL(_find_next_or_bit); >> unsigned long _find_next_zero_bit(const unsigned long *addr, unsigned long nbits, >> unsigned long start) >> { >> - return FIND_NEXT_BIT(~addr[idx], /* nop */, nbits, start); >> + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), /* nop */, nbits, start); >> } >> EXPORT_SYMBOL(_find_next_zero_bit); >> #endif >> @@ -208,7 +209,7 @@ unsigned long _find_last_bit(const unsigned long *addr, unsigned long size) >> unsigned long idx = (size-1) / BITS_PER_LONG; >> do { >> - val &= addr[idx]; >> + val &= READ_ONCE(addr[idx]); >> if (val) >> return idx * BITS_PER_LONG + __fls(val); >> @@ -242,7 +243,7 @@ EXPORT_SYMBOL(find_next_clump8); >> */ >> unsigned long _find_first_zero_bit_le(const unsigned long *addr, unsigned long size) >> { >> - return FIND_FIRST_BIT(~addr[idx], swab, size); >> + return FIND_FIRST_BIT(~READ_ONCE(addr[idx]), swab, size); >> } >> EXPORT_SYMBOL(_find_first_zero_bit_le); >> @@ -252,7 +253,7 @@ EXPORT_SYMBOL(_find_first_zero_bit_le); >> unsigned long _find_next_zero_bit_le(const unsigned long *addr, >> unsigned long size, unsigned long offset) >> { >> - return FIND_NEXT_BIT(~addr[idx], swab, size, offset); >> + return FIND_NEXT_BIT(~READ_ONCE(addr[idx]), swab, size, offset); >> } >> EXPORT_SYMBOL(_find_next_zero_bit_le); >> #endif >> @@ -261,7 +262,7 @@ EXPORT_SYMBOL(_find_next_zero_bit_le); >> unsigned long _find_next_bit_le(const unsigned long *addr, >> unsigned long size, unsigned long offset) >> { >> - return FIND_NEXT_BIT(addr[idx], swab, size, offset); >> + return FIND_NEXT_BIT(READ_ONCE(addr[idx]), swab, size, offset); >> } >> EXPORT_SYMBOL(_find_next_bit_le); >> --
On Mon 18-09-23 15:34:46, Mirsad Todorovac wrote: > On 9/18/23 15:18, Jan Kara wrote: > > On Mon 18-09-23 14:46:02, Mirsad Todorovac wrote: > > > Hi, > > > > > > I tried this patch and the > > > > > > > > 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) > > > > > 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) > > > > > 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) > > > > > 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) > > > > > 5 _find_next_zero_bit > > > > > > data-races do not seem to appear any longer. > > > > Yup. You've just missed one case in _find_last_bit() and then all the > > functions in include/linux/find.h need a similar treatment... > > I seem to have this: > > ----------------------------------------------------------------------------------- > #ifndef find_last_bit > unsigned long _find_last_bit(const unsigned long *addr, unsigned long size) > { > if (size) { > unsigned long val = BITMAP_LAST_WORD_MASK(size); > unsigned long idx = (size-1) / BITS_PER_LONG; > > do { > val &= READ_ONCE(addr[idx]); > if (val) > return idx * BITS_PER_LONG + __fls(val); > > val = ~0ul; > } while (idx--); > } > return size; > } > EXPORT_SYMBOL(_find_last_bit); > #endif > ----------------------------------------------------------------------------------- > > Is there something I did not notice? No, this looks correct. I just somehow didn't see this hunk in the diff you've posted. Honza
On Mon, Sep 18, 2023 at 02:46:02PM +0200, Mirsad Todorovac wrote: ... > Ah, I see. This is definitely not good. But I managed to fix and test the find_next_bit() > family, but this seems that simply > > ------------------------------------------- > include/linux/xarray.h | 8 -------- > 1 file changed, 8 deletions(-) > > diff --git a/include/linux/xarray.h b/include/linux/xarray.h > index 1715fd322d62..89918b65b00d 100644 > --- a/include/linux/xarray.h > +++ b/include/linux/xarray.h > @@ -1718,14 +1718,6 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, > if (advance) > offset++; > - if (XA_CHUNK_SIZE == BITS_PER_LONG) { > - if (offset < XA_CHUNK_SIZE) { > - unsigned long data = READ_ONCE(*addr) & (~0UL << offset); > - if (data) > - return __ffs(data); > - } > - return XA_CHUNK_SIZE; > - } > return find_next_bit(addr, XA_CHUNK_SIZE, offset); > } This looks correct. As per my understanding, the removed part is the 1-word bitmap optimization for find_next_bit. If so, it's not needed because find_next_bit() bears this optimization itself. ... > -------------------------------------------------------- > lib/find_bit.c | 33 +++++++++++++++++---------------- > 1 file changed, 17 insertions(+), 16 deletions(-) > > diff --git a/lib/find_bit.c b/lib/find_bit.c > index 32f99e9a670e..56244e4f744e 100644 > --- a/lib/find_bit.c > +++ b/lib/find_bit.c > @@ -18,6 +18,7 @@ > #include <linux/math.h> > #include <linux/minmax.h> > #include <linux/swab.h> > +#include <asm/rwonce.h> > /* > * Common helper for find_bit() function family > @@ -98,7 +99,7 @@ out: \ > */ > unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) > { > - return FIND_FIRST_BIT(addr[idx], /* nop */, size); > + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); > } > EXPORT_SYMBOL(_find_first_bit); > #endif ... That doesn't look correct. READ_ONCE() implies that there's another thread modifying the bitmap concurrently. This is not the true for vast majority of bitmap API users, and I expect that forcing READ_ONCE() would affect performance for them. Bitmap functions, with a few rare exceptions like set_bit(), are not thread-safe and require users to perform locking/synchronization where needed. If you really need READ_ONCE, I think it's better to implement a new flavor of the function(s) separately, like: find_first_bit_read_once() Thanks, Yury
On 9/18/23 16:59, Yury Norov wrote: > On Mon, Sep 18, 2023 at 02:46:02PM +0200, Mirsad Todorovac wrote: > > ... > >> Ah, I see. This is definitely not good. But I managed to fix and test the find_next_bit() >> family, but this seems that simply >> >> ------------------------------------------- >> include/linux/xarray.h | 8 -------- >> 1 file changed, 8 deletions(-) >> >> diff --git a/include/linux/xarray.h b/include/linux/xarray.h >> index 1715fd322d62..89918b65b00d 100644 >> --- a/include/linux/xarray.h >> +++ b/include/linux/xarray.h >> @@ -1718,14 +1718,6 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, >> if (advance) >> offset++; >> - if (XA_CHUNK_SIZE == BITS_PER_LONG) { >> - if (offset < XA_CHUNK_SIZE) { >> - unsigned long data = READ_ONCE(*addr) & (~0UL << offset); >> - if (data) >> - return __ffs(data); >> - } >> - return XA_CHUNK_SIZE; >> - } >> return find_next_bit(addr, XA_CHUNK_SIZE, offset); >> } > > This looks correct. As per my understanding, the removed part is the > 1-word bitmap optimization for find_next_bit. If so, it's not needed > because find_next_bit() bears this optimization itself. > > ... > >> -------------------------------------------------------- >> lib/find_bit.c | 33 +++++++++++++++++---------------- >> 1 file changed, 17 insertions(+), 16 deletions(-) >> >> diff --git a/lib/find_bit.c b/lib/find_bit.c >> index 32f99e9a670e..56244e4f744e 100644 >> --- a/lib/find_bit.c >> +++ b/lib/find_bit.c >> @@ -18,6 +18,7 @@ >> #include <linux/math.h> >> #include <linux/minmax.h> >> #include <linux/swab.h> >> +#include <asm/rwonce.h> >> /* >> * Common helper for find_bit() function family >> @@ -98,7 +99,7 @@ out: \ >> */ >> unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) >> { >> - return FIND_FIRST_BIT(addr[idx], /* nop */, size); >> + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); >> } >> EXPORT_SYMBOL(_find_first_bit); >> #endif > > ... > > That doesn't look correct. READ_ONCE() implies that there's another > thread modifying the bitmap concurrently. This is not the true for > vast majority of bitmap API users, and I expect that forcing > READ_ONCE() would affect performance for them. > > Bitmap functions, with a few rare exceptions like set_bit(), are not > thread-safe and require users to perform locking/synchronization where > needed. > > If you really need READ_ONCE, I think it's better to implement a new > flavor of the function(s) separately, like: > find_first_bit_read_once() > > Thanks, > Yury Hi, I see the quirk. AFAICS, correct locking would be more expensive than READ_ONCE() flavour of functions. Only one has to inspect every line where they are used and see if there is the need for the *_read_once() version. I suppose people will not be happy because of the duplication of code. :-( It is not a lot of work, I will do a PoC code and see if KCSAN still complains. (Which was the basic reason in the first place for all this, because something changed data from underneath our fingers ...). Best regards, Mirsad
On Mon 18-09-23 07:59:03, Yury Norov wrote: > On Mon, Sep 18, 2023 at 02:46:02PM +0200, Mirsad Todorovac wrote: > > -------------------------------------------------------- > > lib/find_bit.c | 33 +++++++++++++++++---------------- > > 1 file changed, 17 insertions(+), 16 deletions(-) > > > > diff --git a/lib/find_bit.c b/lib/find_bit.c > > index 32f99e9a670e..56244e4f744e 100644 > > --- a/lib/find_bit.c > > +++ b/lib/find_bit.c > > @@ -18,6 +18,7 @@ > > #include <linux/math.h> > > #include <linux/minmax.h> > > #include <linux/swab.h> > > +#include <asm/rwonce.h> > > /* > > * Common helper for find_bit() function family > > @@ -98,7 +99,7 @@ out: \ > > */ > > unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) > > { > > - return FIND_FIRST_BIT(addr[idx], /* nop */, size); > > + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); > > } > > EXPORT_SYMBOL(_find_first_bit); > > #endif > > ... > > That doesn't look correct. READ_ONCE() implies that there's another > thread modifying the bitmap concurrently. This is not the true for > vast majority of bitmap API users, and I expect that forcing > READ_ONCE() would affect performance for them. > > Bitmap functions, with a few rare exceptions like set_bit(), are not > thread-safe and require users to perform locking/synchronization where > needed. Well, for xarray the write side is synchronized with a spinlock but the read side is not (only RCU protected). > If you really need READ_ONCE, I think it's better to implement a new > flavor of the function(s) separately, like: > find_first_bit_read_once() So yes, xarray really needs READ_ONCE(). And I don't think READ_ONCE() imposes any real perfomance overhead in this particular case because for any sane compiler the generated assembly with & without READ_ONCE() will be exactly the same. For example I've checked disassembly of _find_next_bit() using READ_ONCE(). The main loop is: 0xffffffff815a2b6d <+77>: inc %r8 0xffffffff815a2b70 <+80>: add $0x8,%rdx 0xffffffff815a2b74 <+84>: mov %r8,%rcx 0xffffffff815a2b77 <+87>: shl $0x6,%rcx 0xffffffff815a2b7b <+91>: cmp %rcx,%rax 0xffffffff815a2b7e <+94>: jbe 0xffffffff815a2b9b <_find_next_bit+123> 0xffffffff815a2b80 <+96>: mov (%rdx),%rcx 0xffffffff815a2b83 <+99>: test %rcx,%rcx 0xffffffff815a2b86 <+102>: je 0xffffffff815a2b6d <_find_next_bit+77> 0xffffffff815a2b88 <+104>: shl $0x6,%r8 0xffffffff815a2b8c <+108>: tzcnt %rcx,%rcx So you can see the value we work with is copied from the address (rdx) into a register (rcx) and the test and __ffs() happens on a register value and thus READ_ONCE() has no practical effect. It just prevents the compiler from doing some stupid de-optimization. Honza
On 9/18/23 17:54, Jan Kara wrote: > On Mon 18-09-23 07:59:03, Yury Norov wrote: >> On Mon, Sep 18, 2023 at 02:46:02PM +0200, Mirsad Todorovac wrote: >>> -------------------------------------------------------- >>> lib/find_bit.c | 33 +++++++++++++++++---------------- >>> 1 file changed, 17 insertions(+), 16 deletions(-) >>> >>> diff --git a/lib/find_bit.c b/lib/find_bit.c >>> index 32f99e9a670e..56244e4f744e 100644 >>> --- a/lib/find_bit.c >>> +++ b/lib/find_bit.c >>> @@ -18,6 +18,7 @@ >>> #include <linux/math.h> >>> #include <linux/minmax.h> >>> #include <linux/swab.h> >>> +#include <asm/rwonce.h> >>> /* >>> * Common helper for find_bit() function family >>> @@ -98,7 +99,7 @@ out: \ >>> */ >>> unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) >>> { >>> - return FIND_FIRST_BIT(addr[idx], /* nop */, size); >>> + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); >>> } >>> EXPORT_SYMBOL(_find_first_bit); >>> #endif >> >> ... >> >> That doesn't look correct. READ_ONCE() implies that there's another >> thread modifying the bitmap concurrently. This is not the true for >> vast majority of bitmap API users, and I expect that forcing >> READ_ONCE() would affect performance for them. >> >> Bitmap functions, with a few rare exceptions like set_bit(), are not >> thread-safe and require users to perform locking/synchronization where >> needed. > > Well, for xarray the write side is synchronized with a spinlock but the read > side is not (only RCU protected). > >> If you really need READ_ONCE, I think it's better to implement a new >> flavor of the function(s) separately, like: >> find_first_bit_read_once() > > So yes, xarray really needs READ_ONCE(). And I don't think READ_ONCE() > imposes any real perfomance overhead in this particular case because for > any sane compiler the generated assembly with & without READ_ONCE() will be > exactly the same. For example I've checked disassembly of _find_next_bit() > using READ_ONCE(). The main loop is: > > 0xffffffff815a2b6d <+77>: inc %r8 > 0xffffffff815a2b70 <+80>: add $0x8,%rdx > 0xffffffff815a2b74 <+84>: mov %r8,%rcx > 0xffffffff815a2b77 <+87>: shl $0x6,%rcx > 0xffffffff815a2b7b <+91>: cmp %rcx,%rax > 0xffffffff815a2b7e <+94>: jbe 0xffffffff815a2b9b <_find_next_bit+123> > 0xffffffff815a2b80 <+96>: mov (%rdx),%rcx > 0xffffffff815a2b83 <+99>: test %rcx,%rcx > 0xffffffff815a2b86 <+102>: je 0xffffffff815a2b6d <_find_next_bit+77> > 0xffffffff815a2b88 <+104>: shl $0x6,%r8 > 0xffffffff815a2b8c <+108>: tzcnt %rcx,%rcx > > So you can see the value we work with is copied from the address (rdx) into > a register (rcx) and the test and __ffs() happens on a register value and > thus READ_ONCE() has no practical effect. It just prevents the compiler > from doing some stupid de-optimization. > > Honza If I may also add, centralised READ_ONCE() version had fixed a couple of hundred of the instances of KCSAN data-races in dmesg. _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) 5 _find_next_zero_bit Finding each one find_bit_*() function and replacing it with find_bit_*_read_once() could be time-consuming and challenging. However, I will do both versions so you could compare, if you'd like. Note, in the PoC version I have only implemented find_next_bit_read_once() ATM to see if this works. Regards, Mirsad
On Mon, Sep 18, 2023 at 06:28:07PM +0200, Mirsad Todorovac wrote: > > > On 9/18/23 17:54, Jan Kara wrote: > > On Mon 18-09-23 07:59:03, Yury Norov wrote: > > > On Mon, Sep 18, 2023 at 02:46:02PM +0200, Mirsad Todorovac wrote: > > > > -------------------------------------------------------- > > > > lib/find_bit.c | 33 +++++++++++++++++---------------- > > > > 1 file changed, 17 insertions(+), 16 deletions(-) > > > > > > > > diff --git a/lib/find_bit.c b/lib/find_bit.c > > > > index 32f99e9a670e..56244e4f744e 100644 > > > > --- a/lib/find_bit.c > > > > +++ b/lib/find_bit.c > > > > @@ -18,6 +18,7 @@ > > > > #include <linux/math.h> > > > > #include <linux/minmax.h> > > > > #include <linux/swab.h> > > > > +#include <asm/rwonce.h> > > > > /* > > > > * Common helper for find_bit() function family > > > > @@ -98,7 +99,7 @@ out: \ > > > > */ > > > > unsigned long _find_first_bit(const unsigned long *addr, unsigned long size) > > > > { > > > > - return FIND_FIRST_BIT(addr[idx], /* nop */, size); > > > > + return FIND_FIRST_BIT(READ_ONCE(addr[idx]), /* nop */, size); > > > > } > > > > EXPORT_SYMBOL(_find_first_bit); > > > > #endif > > > > > > ... > > > > > > That doesn't look correct. READ_ONCE() implies that there's another > > > thread modifying the bitmap concurrently. This is not the true for > > > vast majority of bitmap API users, and I expect that forcing > > > READ_ONCE() would affect performance for them. > > > > > > Bitmap functions, with a few rare exceptions like set_bit(), are not > > > thread-safe and require users to perform locking/synchronization where > > > needed. > > > > Well, for xarray the write side is synchronized with a spinlock but the read > > side is not (only RCU protected). > > > > > If you really need READ_ONCE, I think it's better to implement a new > > > flavor of the function(s) separately, like: > > > find_first_bit_read_once() > > > > So yes, xarray really needs READ_ONCE(). And I don't think READ_ONCE() > > imposes any real perfomance overhead in this particular case because for > > any sane compiler the generated assembly with & without READ_ONCE() will be > > exactly the same. For example I've checked disassembly of _find_next_bit() > > using READ_ONCE(). The main loop is: > > > > 0xffffffff815a2b6d <+77>: inc %r8 > > 0xffffffff815a2b70 <+80>: add $0x8,%rdx > > 0xffffffff815a2b74 <+84>: mov %r8,%rcx > > 0xffffffff815a2b77 <+87>: shl $0x6,%rcx > > 0xffffffff815a2b7b <+91>: cmp %rcx,%rax > > 0xffffffff815a2b7e <+94>: jbe 0xffffffff815a2b9b <_find_next_bit+123> > > 0xffffffff815a2b80 <+96>: mov (%rdx),%rcx > > 0xffffffff815a2b83 <+99>: test %rcx,%rcx > > 0xffffffff815a2b86 <+102>: je 0xffffffff815a2b6d <_find_next_bit+77> > > 0xffffffff815a2b88 <+104>: shl $0x6,%r8 > > 0xffffffff815a2b8c <+108>: tzcnt %rcx,%rcx > > > > So you can see the value we work with is copied from the address (rdx) into > > a register (rcx) and the test and __ffs() happens on a register value and > > thus READ_ONCE() has no practical effect. It just prevents the compiler > > from doing some stupid de-optimization. > > > > Honza > > If I may also add, centralised READ_ONCE() version had fixed a couple of hundred of > the instances of KCSAN data-races in dmesg. > > _find_*_bit() functions and/or macros cause quite a number of KCSAN BUG warnings: > > 95 _find_first_and_bit (lib/find_bit.c:114 (discriminator 10)) > 31 _find_first_zero_bit (lib/find_bit.c:125 (discriminator 10)) > 173 _find_next_and_bit (lib/find_bit.c:171 (discriminator 2)) > 655 _find_next_bit (lib/find_bit.c:133 (discriminator 2)) > 5 _find_next_zero_bit > > Finding each one find_bit_*() function and replacing it with find_bit_*_read_once() > could be time-consuming and challenging. > > However, I will do both versions so you could compare, if you'd like. > > Note, in the PoC version I have only implemented find_next_bit_read_once() ATM to see if > this works. > > Regards, > Mirsad Guys, I lost the track of the conversation. In the other email Mirsad said: Which was the basic reason in the first place for all this, because something changed data from underneath our fingers .. It sounds clearly to me that this is a bug in xarray, *revealed* by find_next_bit() function. But later in discussion you're trying to 'fix' find_*_bit(), like if find_bit() corrupted the bitmap, but it's not. In previous email Jan said: for any sane compiler the generated assembly with & without READ_ONCE() will be exactly the same. If the code generated with and without READ_ONCE() is the same, the behavior would be the same, right? If you see the difference, the code should differ. You say that READ_ONCE() in find_bit() 'fixes' 200 KCSAN BUG warnings. To me it sounds like hiding the problems instead of fixing. If there's a race between writing and reading bitmaps, it should be fixed properly by adding an appropriate serialization mechanism. Shutting Kcsan up with READ_ONCE() here and there is exactly the opposite path to the right direction. Every READ_ONCE must be paired with WRITE_ONCE, just like atomic reads/writes or spin locks/unlocks. Having that in mind, adding READ_ONCE() in find_bit() requires adding it to every bitmap function out there. And this is, as I said before, would be an overhead for most users.
On Mon, Sep 18, 2023 at 11:56:36AM -0700, Yury Norov wrote: > Guys, I lost the track of the conversation. In the other email Mirsad > said: > Which was the basic reason in the first place for all this, because something changed > data from underneath our fingers .. > > It sounds clearly to me that this is a bug in xarray, *revealed* by > find_next_bit() function. But later in discussion you're trying to 'fix' > find_*_bit(), like if find_bit() corrupted the bitmap, but it's not. No, you're really confused. That happens. KCSAN is looking for concurrency bugs. That is, does another thread mutate the data "while" we're reading it. It does that by reading the data, delaying for a few instructions and reading it again. If it changed, clearly there's a race. That does not mean there's a bug! Some races are innocuous. Many races are innocuous! The problem is that compilers sometimes get overly clever and don't do the obvious thing you ask them to do. READ_ONCE() serves two functions here; one is that it tells the compiler not to try anything fancy, and the other is that it tells KCSAN to not bother instrumenting this load; no load-delay-reload. > In previous email Jan said: > for any sane compiler the generated assembly with & without READ_ONCE() > will be exactly the same. > > If the code generated with and without READ_ONCE() is the same, the > behavior would be the same, right? If you see the difference, the code > should differ. Hopefully now you understand why this argument is wrong ... > You say that READ_ONCE() in find_bit() 'fixes' 200 KCSAN BUG warnings. To > me it sounds like hiding the problems instead of fixing. If there's a race > between writing and reading bitmaps, it should be fixed properly by > adding an appropriate serialization mechanism. Shutting Kcsan up with > READ_ONCE() here and there is exactly the opposite path to the right direction. Counterpoint: generally bitmaps are modified with set_bit() which actually is atomic. We define so many bitmap things as being atomic already, it doesn't feel like making find_bit() "must be protected" as a useful use of time. But hey, maybe I'm wrong. Mirsad, can you send Yury the bug reports for find_bit and friends, and Yury can take the time to dig through them and see if there are any real races in that mess? > Every READ_ONCE must be paired with WRITE_ONCE, just like atomic > reads/writes or spin locks/unlocks. Having that in mind, adding > READ_ONCE() in find_bit() requires adding it to every bitmap function > out there. And this is, as I said before, would be an overhead for > most users. I don't believe you. Telling the compiler to stop trying to be clever rarely results in a performance loss.
On 9/19/2023 6:20 AM, Matthew Wilcox wrote: > On Mon, Sep 18, 2023 at 11:56:36AM -0700, Yury Norov wrote: >> Guys, I lost the track of the conversation. In the other email Mirsad >> said: >> Which was the basic reason in the first place for all this, because something changed >> data from underneath our fingers .. >> >> It sounds clearly to me that this is a bug in xarray, *revealed* by >> find_next_bit() function. But later in discussion you're trying to 'fix' >> find_*_bit(), like if find_bit() corrupted the bitmap, but it's not. > > No, you're really confused. That happens. > > KCSAN is looking for concurrency bugs. That is, does another thread > mutate the data "while" we're reading it. It does that by reading > the data, delaying for a few instructions and reading it again. If it > changed, clearly there's a race. That does not mean there's a bug! > > Some races are innocuous. Many races are innocuous! The problem is > that compilers sometimes get overly clever and don't do the obvious > thing you ask them to do. READ_ONCE() serves two functions here; > one is that it tells the compiler not to try anything fancy, and > the other is that it tells KCSAN to not bother instrumenting this > load; no load-delay-reload. > >> In previous email Jan said: >> for any sane compiler the generated assembly with & without READ_ONCE() >> will be exactly the same. >> >> If the code generated with and without READ_ONCE() is the same, the >> behavior would be the same, right? If you see the difference, the code >> should differ. > > Hopefully now you understand why this argument is wrong ... > >> You say that READ_ONCE() in find_bit() 'fixes' 200 KCSAN BUG warnings. To >> me it sounds like hiding the problems instead of fixing. If there's a race >> between writing and reading bitmaps, it should be fixed properly by >> adding an appropriate serialization mechanism. Shutting Kcsan up with >> READ_ONCE() here and there is exactly the opposite path to the right direction. > > Counterpoint: generally bitmaps are modified with set_bit() which > actually is atomic. We define so many bitmap things as being atomic > already, it doesn't feel like making find_bit() "must be protected" > as a useful use of time. > > But hey, maybe I'm wrong. Mirsad, can you send Yury the bug reports > for find_bit and friends, and Yury can take the time to dig through them > and see if there are any real races in that mess? > >> Every READ_ONCE must be paired with WRITE_ONCE, just like atomic >> reads/writes or spin locks/unlocks. Having that in mind, adding >> READ_ONCE() in find_bit() requires adding it to every bitmap function >> out there. And this is, as I said before, would be an overhead for >> most users. > > I don't believe you. Telling the compiler to stop trying to be clever > rarely results in a performance loss. Hi Mr. Wilcox, Do you think we should submit a formal patch for this data-race? Thank you. Best regards, Mirsad Todorovac
On Fri 06-10-23 16:39:54, Mirsad Todorovac wrote: > On 9/19/2023 6:20 AM, Matthew Wilcox wrote: > > On Mon, Sep 18, 2023 at 11:56:36AM -0700, Yury Norov wrote: > > > Guys, I lost the track of the conversation. In the other email Mirsad > > > said: > > > Which was the basic reason in the first place for all this, because something changed > > > data from underneath our fingers .. > > > > > > It sounds clearly to me that this is a bug in xarray, *revealed* by > > > find_next_bit() function. But later in discussion you're trying to 'fix' > > > find_*_bit(), like if find_bit() corrupted the bitmap, but it's not. > > > > No, you're really confused. That happens. > > > > KCSAN is looking for concurrency bugs. That is, does another thread > > mutate the data "while" we're reading it. It does that by reading > > the data, delaying for a few instructions and reading it again. If it > > changed, clearly there's a race. That does not mean there's a bug! > > > > Some races are innocuous. Many races are innocuous! The problem is > > that compilers sometimes get overly clever and don't do the obvious > > thing you ask them to do. READ_ONCE() serves two functions here; > > one is that it tells the compiler not to try anything fancy, and > > the other is that it tells KCSAN to not bother instrumenting this > > load; no load-delay-reload. > > > > > In previous email Jan said: > > > for any sane compiler the generated assembly with & without READ_ONCE() > > > will be exactly the same. > > > > > > If the code generated with and without READ_ONCE() is the same, the > > > behavior would be the same, right? If you see the difference, the code > > > should differ. > > > > Hopefully now you understand why this argument is wrong ... > > > > > You say that READ_ONCE() in find_bit() 'fixes' 200 KCSAN BUG warnings. To > > > me it sounds like hiding the problems instead of fixing. If there's a race > > > between writing and reading bitmaps, it should be fixed properly by > > > adding an appropriate serialization mechanism. Shutting Kcsan up with > > > READ_ONCE() here and there is exactly the opposite path to the right direction. > > > > Counterpoint: generally bitmaps are modified with set_bit() which > > actually is atomic. We define so many bitmap things as being atomic > > already, it doesn't feel like making find_bit() "must be protected" > > as a useful use of time. > > > > But hey, maybe I'm wrong. Mirsad, can you send Yury the bug reports > > for find_bit and friends, and Yury can take the time to dig through them > > and see if there are any real races in that mess? > > > > > Every READ_ONCE must be paired with WRITE_ONCE, just like atomic > > > reads/writes or spin locks/unlocks. Having that in mind, adding > > > READ_ONCE() in find_bit() requires adding it to every bitmap function > > > out there. And this is, as I said before, would be an overhead for > > > most users. > > > > I don't believe you. Telling the compiler to stop trying to be clever > > rarely results in a performance loss. > > Hi Mr. Wilcox, > > Do you think we should submit a formal patch for this data-race? So I did some benchmarking with various GCC versions and the truth is that READ_ONCE() does affect code generation a bit (although the original code does not refetch the value from memory). As a result my benchmarks show the bit searching functions are about 2% slower. This is not much but it is stupid to cause a performance regression due to non-issue. I'm trying to get some compiler guys look into this whether we can improve it somehow... Honza
On 10/9/23 12:15, Jan Kara wrote: > On Fri 06-10-23 16:39:54, Mirsad Todorovac wrote: >> On 9/19/2023 6:20 AM, Matthew Wilcox wrote: >>> On Mon, Sep 18, 2023 at 11:56:36AM -0700, Yury Norov wrote: >>>> Guys, I lost the track of the conversation. In the other email Mirsad >>>> said: >>>> Which was the basic reason in the first place for all this, because something changed >>>> data from underneath our fingers .. >>>> >>>> It sounds clearly to me that this is a bug in xarray, *revealed* by >>>> find_next_bit() function. But later in discussion you're trying to 'fix' >>>> find_*_bit(), like if find_bit() corrupted the bitmap, but it's not. >>> >>> No, you're really confused. That happens. >>> >>> KCSAN is looking for concurrency bugs. That is, does another thread >>> mutate the data "while" we're reading it. It does that by reading >>> the data, delaying for a few instructions and reading it again. If it >>> changed, clearly there's a race. That does not mean there's a bug! >>> >>> Some races are innocuous. Many races are innocuous! The problem is >>> that compilers sometimes get overly clever and don't do the obvious >>> thing you ask them to do. READ_ONCE() serves two functions here; >>> one is that it tells the compiler not to try anything fancy, and >>> the other is that it tells KCSAN to not bother instrumenting this >>> load; no load-delay-reload. >>> >>>> In previous email Jan said: >>>> for any sane compiler the generated assembly with & without READ_ONCE() >>>> will be exactly the same. >>>> >>>> If the code generated with and without READ_ONCE() is the same, the >>>> behavior would be the same, right? If you see the difference, the code >>>> should differ. >>> >>> Hopefully now you understand why this argument is wrong ... >>> >>>> You say that READ_ONCE() in find_bit() 'fixes' 200 KCSAN BUG warnings. To >>>> me it sounds like hiding the problems instead of fixing. If there's a race >>>> between writing and reading bitmaps, it should be fixed properly by >>>> adding an appropriate serialization mechanism. Shutting Kcsan up with >>>> READ_ONCE() here and there is exactly the opposite path to the right direction. >>> >>> Counterpoint: generally bitmaps are modified with set_bit() which >>> actually is atomic. We define so many bitmap things as being atomic >>> already, it doesn't feel like making find_bit() "must be protected" >>> as a useful use of time. >>> >>> But hey, maybe I'm wrong. Mirsad, can you send Yury the bug reports >>> for find_bit and friends, and Yury can take the time to dig through them >>> and see if there are any real races in that mess? >>> >>>> Every READ_ONCE must be paired with WRITE_ONCE, just like atomic >>>> reads/writes or spin locks/unlocks. Having that in mind, adding >>>> READ_ONCE() in find_bit() requires adding it to every bitmap function >>>> out there. And this is, as I said before, would be an overhead for >>>> most users. >>> >>> I don't believe you. Telling the compiler to stop trying to be clever >>> rarely results in a performance loss. >> >> Hi Mr. Wilcox, >> >> Do you think we should submit a formal patch for this data-race? > > So I did some benchmarking with various GCC versions and the truth is that > READ_ONCE() does affect code generation a bit (although the original code > does not refetch the value from memory). As a result my benchmarks show the > bit searching functions are about 2% slower. This is not much but it is > stupid to cause a performance regression due to non-issue. I'm trying to > get some compiler guys look into this whether we can improve it somehow... > > Honza Dear Jan, First, I am not an expert or an authority on the subject, this is only my opinion. IMHO, 2% slower code is acceptable if it gives us data integrity. If a 16-core system manages to break and tear loads without READ_ONCE(), 2% faster code gives us nothing if the other core changes half of the location in the midst of the load, just because the optimiser did some "funny stuff". If I had a pacemaker and it is running Linux kernel, I would probably choose 2% slower but race-free code. Please allow me to assert that this is not a spin lock, memory bus lock, or a memory barrier that would affect the other cores - it will only slightly prevent some read reordering/tearing. I think you are on the good track, and that this patch is a good thing. Low-level functions have to be first safe, then fast. A faster algorithm, like replacing spinlocks with RCU, can certainly more than make up for that ... Sorry for a digression. Best regards, Mirsad
diff --git a/include/linux/xarray.h b/include/linux/xarray.h index cb571dfcf4b1..1715fd322d62 100644 --- a/include/linux/xarray.h +++ b/include/linux/xarray.h @@ -1720,7 +1720,7 @@ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, offset++; if (XA_CHUNK_SIZE == BITS_PER_LONG) { if (offset < XA_CHUNK_SIZE) { - unsigned long data = *addr & (~0UL << offset); + unsigned long data = READ_ONCE(*addr) & (~0UL << offset); if (data) return __ffs(data); }
KCSAN has discovered the following data-race: [ 206.510010] ================================================================== [ 206.510035] BUG: KCSAN: data-race in xas_clear_mark / xas_find_marked [ 206.510067] write to 0xffff963df6a90fe0 of 8 bytes by interrupt on cpu 22: [ 206.510081] xas_clear_mark (./arch/x86/include/asm/bitops.h:178 ./include/asm-generic/bitops/instrumented-non-atomic.h:115 lib/xarray.c:102 lib/xarray.c:914) [ 206.510097] __xa_clear_mark (lib/xarray.c:1923) [ 206.510114] __folio_end_writeback (mm/page-writeback.c:2981) [ 206.510128] folio_end_writeback (mm/filemap.c:1616) [ 206.510143] end_page_writeback (mm/folio-compat.c:28) [ 206.510155] btrfs_page_clear_writeback (fs/btrfs/subpage.c:646) btrfs [ 206.510994] end_bio_extent_writepage (./include/linux/bio.h:84 fs/btrfs/extent_io.c:542) btrfs [ 206.511817] __btrfs_bio_end_io (fs/btrfs/bio.c:117 fs/btrfs/bio.c:112) btrfs [ 206.512640] btrfs_orig_bbio_end_io (fs/btrfs/bio.c:164) btrfs [ 206.513497] btrfs_simple_end_io (fs/btrfs/bio.c:380) btrfs [ 206.514350] bio_endio (block/bio.c:1617) [ 206.514362] blk_mq_end_request_batch (block/blk-mq.c:837 block/blk-mq.c:1073) [ 206.514377] nvme_pci_complete_batch (drivers/nvme/host/pci.c:986) nvme [ 206.514437] nvme_irq (drivers/nvme/host/pci.c:1086) nvme [ 206.514500] __handle_irq_event_percpu (kernel/irq/handle.c:158) [ 206.514517] handle_irq_event (kernel/irq/handle.c:195 kernel/irq/handle.c:210) [ 206.514533] handle_edge_irq (kernel/irq/chip.c:836) [ 206.514549] __common_interrupt (./include/linux/irqdesc.h:161 arch/x86/kernel/irq.c:238 arch/x86/kernel/irq.c:257) [ 206.514563] common_interrupt (arch/x86/kernel/irq.c:247 (discriminator 14)) [ 206.514583] asm_common_interrupt (./arch/x86/include/asm/idtentry.h:636) [ 206.514599] kcsan_setup_watchpoint (kernel/kcsan/core.c:705 (discriminator 1)) [ 206.514612] __tsan_read8 (kernel/kcsan/core.c:1025) [ 206.514626] steal_from_bitmap.part.0 (./include/linux/find.h:186 fs/btrfs/free-space-cache.c:2557 fs/btrfs/free-space-cache.c:2613) btrfs [ 206.515491] __btrfs_add_free_space (fs/btrfs/free-space-cache.c:2689 fs/btrfs/free-space-cache.c:2667) btrfs [ 206.516361] btrfs_add_free_space_async_trimmed (fs/btrfs/free-space-cache.c:2798) btrfs [ 206.517231] add_new_free_space (fs/btrfs/block-group.c:550) btrfs [ 206.518095] load_free_space_tree (fs/btrfs/free-space-tree.c:1595 fs/btrfs/free-space-tree.c:1658) btrfs [ 206.518953] caching_thread (fs/btrfs/block-group.c:873) btrfs [ 206.519800] btrfs_work_helper (fs/btrfs/async-thread.c:314) btrfs [ 206.520643] process_one_work (kernel/workqueue.c:2600) [ 206.520658] worker_thread (./include/linux/list.h:292 kernel/workqueue.c:2752) [ 206.520672] kthread (kernel/kthread.c:389) [ 206.520684] ret_from_fork (arch/x86/kernel/process.c:145) [ 206.520701] ret_from_fork_asm (arch/x86/entry/entry_64.S:312) [ 206.520722] read to 0xffff963df6a90fe0 of 8 bytes by task 2793 on cpu 6: [ 206.520735] xas_find_marked (./include/linux/xarray.h:1706 lib/xarray.c:1354) [ 206.520750] filemap_get_folios_tag (mm/filemap.c:1975 mm/filemap.c:2273) [ 206.520763] __filemap_fdatawait_range (mm/filemap.c:519) [ 206.520777] filemap_fdatawait_range (mm/filemap.c:556) [ 206.520790] btrfs_wait_ordered_range (fs/btrfs/ordered-data.c:839) btrfs [ 206.521641] btrfs_sync_file (fs/btrfs/file.c:1859) btrfs [ 206.522495] vfs_fsync_range (fs/sync.c:188) [ 206.522509] __x64_sys_fsync (./include/linux/file.h:45 fs/sync.c:213 fs/sync.c:220 fs/sync.c:218 fs/sync.c:218) [ 206.522522] do_syscall_64 (arch/x86/entry/common.c:50 arch/x86/entry/common.c:80) [ 206.522535] entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:120) [ 206.522557] value changed: 0xfffffffffff80000 -> 0xfffffffffff00000 [ 206.522574] Reported by Kernel Concurrency Sanitizer on: [ 206.522585] CPU: 6 PID: 2793 Comm: tracker-extract Tainted: G L 6.5.0-rc6+ #44 [ 206.522600] Hardware name: ASRock X670E PG Lightning/X670E PG Lightning, BIOS 1.21 04/26/2023 [ 206.522608] ================================================================== As Jan Kara explained, the problem is in the function xas_find_chuck(): /* Private */ static inline unsigned int xas_find_chunk(struct xa_state *xas, bool advance, xa_mark_t mark) { unsigned long *addr = xas->xa_node->marks[(__force unsigned)mark]; unsigned int offset = xas->xa_offset; if (advance) offset++; if (XA_CHUNK_SIZE == BITS_PER_LONG) { if (offset < XA_CHUNK_SIZE) { → unsigned long data = *addr & (~0UL << offset); if (data) return __ffs(data); } return XA_CHUNK_SIZE; } return find_next_bit(addr, XA_CHUNK_SIZE, offset); } In particular, the line unsigned long data = *addr & (~0UL << offset); contains a data race that is best avoided using READ_ONCE(), which eliminated the KCSAN data-race warning completely. Reported-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> Suggested-by: Jan Kara <jack@suse.cz> Fixes: b803b42823d0d ("xarray: Add XArray iterators") Matthew Wilcox <willy@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Josef Bacik <josef@toxicpanda.com> Cc: David Sterba <dsterba@suse.com> Cc: linux-btrfs@vger.kernel.org Cc: linux-fsdevel@vger.kernel.org Cc: linux-mm@kvack.org Signed-off-by: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr> --- v1: the proposed fix (RFC) include/linux/xarray.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)