24e44cc22a: BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc
Subsystems (3)
kernel test robot
build botHello,
kernel test robot noticed "BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc" on:
commit: 24e44cc22aa3112082f2ee23137d048c73ca96d5 ("mm: percpu: enable per-cpu allocation tagging")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
[test failed on linus/mater 68b59730459e5d1fe4e0bbeb04ceb9df0f002270]
[test failed on linux-next/master 73399b58e5e5a1b28a04baf42e321cfcfc663c2f]
in testcase: trinity
version: trinity-i386-abe9de86-1_20230429
with following parameters:
runtime: 300s group: group-04 nr_groups: 5
compiler: gcc-13
test machine: qemu-system-x86_64 -enable-kvm -cpu SandyBridge -smp 2 -m 16G
(please refer to attached dmesg/kmsg for entire log/backtrace)
as we understand, this commit is not the root-cause of KCSAN issue, just make the issue change from form (1) to (2). furthermore, we don't understand why the issue happens randomly. but we failed to bisect the issue (1). so we just make out this report FYI, not sure if it could supply some hints to any real issues.
60fa4a9e23231721 24e44cc22aa3112082f2ee23137
---------------- ---------------------------
fail:runs %reproduction fail:runs
| | |
281:999 -28% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_alloc <--- (1)
296:999 -30% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_free
25:999 -3% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_chunk_populated
:999 29% 292:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc <---(2)
:999 27% 269:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_free
:999 4% 44:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_chunk_populated
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-lkp/[email protected]
[ 117.357897][ T3648] ================================================================== [ 117.358568][ T3648] BUG: KCSAN: data-race in pcpu_alloc_noprof / pcpu_block_update_hint_alloc [ 117.359222][ T3648] [ 117.359424][ T3648] write to 0xffffffff8559e600 of 4 bytes by task 3653 on cpu 1: [ 117.360048][ T3648] pcpu_block_update_hint_alloc (mm/percpu.c:602 mm/percpu.c:598 mm/percpu.c:923) [ 117.360554][ T3648] pcpu_alloc_area (mm/percpu.c:1260) [ 117.360960][ T3648] pcpu_alloc_noprof (mm/percpu.c:1835) [ 117.361393][ T3648] bpf_map_alloc_percpu (kernel/bpf/syscall.c:466) [ 117.361825][ T3648] prealloc_init (kernel/bpf/hashtab.c:338) [ 117.362211][ T3648] htab_map_alloc (kernel/bpf/hashtab.c:573) [ 117.362618][ T3648] map_create (kernel/bpf/syscall.c:1320) [ 117.362987][ T3648] __sys_bpf (kernel/bpf/syscall.c:5642) [ 117.363348][ T3648] __ia32_sys_bpf (kernel/bpf/syscall.c:5765) [ 117.363738][ T3648] ia32_sys_call (kbuild/obj/consumer/x86_64-randconfig-013-20240713/./arch/x86/include/generated/asm/syscalls_32.h:358) [ 117.364151][ T3648] do_int80_emulation (arch/x86/entry/common.c:165 (discriminator 1) arch/x86/entry/common.c:253 (discriminator 1)) [ 117.364578][ T3648] asm_int80_emulation (arch/x86/include/asm/idtentry.h:626) [ 117.365004][ T3648] [ 117.365203][ T3648] read to 0xffffffff8559e600 of 4 bytes by task 3648 on cpu 0: [ 117.365797][ T3648] pcpu_alloc_noprof (mm/percpu.c:1894) [ 117.366210][ T3648] bpf_map_alloc_percpu (kernel/bpf/syscall.c:466) [ 117.366627][ T3648] prealloc_init (kernel/bpf/hashtab.c:338) [ 117.367000][ T3648] htab_map_alloc (kernel/bpf/hashtab.c:573) [ 117.367400][ T3648] map_create (kernel/bpf/syscall.c:1320) [ 117.367766][ T3648] __sys_bpf (kernel/bpf/syscall.c:5642) [ 117.368120][ T3648] __ia32_sys_bpf (kernel/bpf/syscall.c:5765) [ 117.368503][ T3648] ia32_sys_call (kbuild/obj/consumer/x86_64-randconfig-013-20240713/./arch/x86/include/generated/asm/syscalls_32.h:358) [ 117.368907][ T3648] do_int80_emulation (arch/x86/entry/common.c:165 (discriminator 1) arch/x86/entry/common.c:253 (discriminator 1)) [ 117.369341][ T3648] asm_int80_emulation (arch/x86/include/asm/idtentry.h:626) [ 117.369761][ T3648] [ 117.369968][ T3648] value changed: 0x00000003 -> 0x00000000 [ 117.370454][ T3648] [ 117.370662][ T3648] Reported by Kernel Concurrency Sanitizer on: [ 117.371179][ T3648] CPU: 0 PID: 3648 Comm: trinity-c2 Not tainted 6.9.0-rc4-00089-g24e44cc22aa3 #1 [ 117.371943][ T3648] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 117.372811][ T3648] ==================================================================
The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20240719/[email protected]
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
Quoted 81 lines · kernel test robot
Hello,
kernel test robot noticed "BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc" on:
commit:
24e44cc22aa3112082f2ee23137d048c73ca96d5("mm: percpu: enable per-cpu allocation tagging")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master[test failed on linus/mater
68b59730459e5d1fe4e0bbeb04ceb9df0f002270]
[test failed on linux-next/master73399b58e5e5a1b28a04baf42e321cfcfc663c2f]in testcase: trinity
version: trinity-i386-abe9de86-1_20230429
with following parameters:runtime: 300s group: group-04 nr_groups: 5compiler: gcc-13
test machine: qemu-system-x86_64 -enable-kvm -cpu SandyBridge -smp 2 -m 16G(please refer to attached dmesg/kmsg for entire log/backtrace)
as we understand, this commit is not the root-cause of KCSAN issue, just make the issue change from form (1) to (2). furthermore, we don't understand why the issue happens randomly. but we failed to bisect the issue (1). so we just make out this report FYI, not sure if it could supply some hints to any real issues.
60fa4a9e2323172124e44cc22aa3112082f2ee23137
---------------- ---------------------------
fail:runs %reproduction fail:runs
| | |
281:999 -28% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_alloc <--- (1)
296:999 -30% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_free
25:999 -3% :998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_chunk_populated
:999 29% 292:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc <---(2)
:999 27% 269:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_free
:999 4% 44:998 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_chunk_populatedIf you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-lkp/[email protected][ 117.357897][ T3648] ================================================================== [ 117.358568][ T3648] BUG: KCSAN: data-race in pcpu_alloc_noprof / pcpu_block_update_hint_alloc [ 117.359222][ T3648] [ 117.359424][ T3648] write to 0xffffffff8559e600 of 4 bytes by task 3653 on cpu 1: [ 117.360048][ T3648] pcpu_block_update_hint_alloc (mm/percpu.c:602 mm/percpu.c:598 mm/percpu.c:923) [ 117.360554][ T3648] pcpu_alloc_area (mm/percpu.c:1260) [ 117.360960][ T3648] pcpu_alloc_noprof (mm/percpu.c:1835) [ 117.361393][ T3648] bpf_map_alloc_percpu (kernel/bpf/syscall.c:466) [ 117.361825][ T3648] prealloc_init (kernel/bpf/hashtab.c:338) [ 117.362211][ T3648] htab_map_alloc (kernel/bpf/hashtab.c:573) [ 117.362618][ T3648] map_create (kernel/bpf/syscall.c:1320) [ 117.362987][ T3648] __sys_bpf (kernel/bpf/syscall.c:5642) [ 117.363348][ T3648] __ia32_sys_bpf (kernel/bpf/syscall.c:5765) [ 117.363738][ T3648] ia32_sys_call (kbuild/obj/consumer/x86_64-randconfig-013-20240713/./arch/x86/include/generated/asm/syscalls_32.h:358) [ 117.364151][ T3648] do_int80_emulation (arch/x86/entry/common.c:165 (discriminator 1) arch/x86/entry/common.c:253 (discriminator 1)) [ 117.364578][ T3648] asm_int80_emulation (arch/x86/include/asm/idtentry.h:626) [ 117.365004][ T3648] [ 117.365203][ T3648] read to 0xffffffff8559e600 of 4 bytes by task 3648 on cpu 0: [ 117.365797][ T3648] pcpu_alloc_noprof (mm/percpu.c:1894) [ 117.366210][ T3648] bpf_map_alloc_percpu (kernel/bpf/syscall.c:466) [ 117.366627][ T3648] prealloc_init (kernel/bpf/hashtab.c:338) [ 117.367000][ T3648] htab_map_alloc (kernel/bpf/hashtab.c:573) [ 117.367400][ T3648] map_create (kernel/bpf/syscall.c:1320) [ 117.367766][ T3648] __sys_bpf (kernel/bpf/syscall.c:5642) [ 117.368120][ T3648] __ia32_sys_bpf (kernel/bpf/syscall.c:5765) [ 117.368503][ T3648] ia32_sys_call (kbuild/obj/consumer/x86_64-randconfig-013-20240713/./arch/x86/include/generated/asm/syscalls_32.h:358) [ 117.368907][ T3648] do_int80_emulation (arch/x86/entry/common.c:165 (discriminator 1) arch/x86/entry/common.c:253 (discriminator 1))
Quoted 9 lines
[ 117.369341][ T3648] asm_int80_emulation (arch/x86/include/asm/idtentry.h:626) [ 117.369761][ T3648] [ 117.369968][ T3648] value changed: 0x00000003 -> 0x00000000 [ 117.370454][ T3648] [ 117.370662][ T3648] Reported by Kernel Concurrency Sanitizer on: [ 117.371179][ T3648] CPU: 0 PID: 3648 Comm: trinity-c2 Not tainted 6.9.0-rc4-00089-g24e44cc22aa3 #1 [ 117.371943][ T3648] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 117.372811][ T3648] ==================================================================
This looks like a data race because we read pcpu_nr_empty_pop_pages out of the lock for a best effort checking, @Tejun, maybe you could confirm on this?
If so, a fix could be as the follow, i.e. telling KCSAN that the data race is expected. Another fix is making all accesses to pcpu_nr_empty_pop_pages atomics (via WRITE_ONCE() and READ_ONCE()).
Regards,
Boqun
(is there a way to tell the bot to issue a rerun with a diff? ;-))
------------------------>8
mm/percpu.c +6 -1
@@ -1891,7 +1891,12 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +/* +* Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may +* occur but this is just a best-effort checking, everything is synced +* in pcpu_balance_work. +*/ +if (data_race (pcpu_nr_empty_pop_pages) <PCPU_EMPTY_POP_PAGES_LOW )pcpu_schedule_balance_work ();/* clear the areas and return address relative to base address */
Tejun Heo
maintainerQuoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?
That does sound plausible.
mm/percpu.c void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, · line 1900
Quoted 8 lines
mm/percpu.c +6
-1-if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +/* +* Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may +* occur but this is just a best-effort checking, everything is synced +* in pcpu_balance_work. +*/ +if (data_race (pcpu_nr_empty_pop_pages) <PCPU_EMPTY_POP_PAGES_LOW )pcpu_schedule_balance_work ();
Would it be better to use READ/WRITE_ONCE() for the variable?
Thanks.
Signature · 1 line
tejun
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();
Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.
Regards,
Boqun
----->8
mm/percpu.c +8 -5
@@ -570,7 +570,8 @@ static void pcpu_isolate_chunk(struct pcpu_chunk *chunk)if (!chunk->isolated) { chunk->isolated = true; - pcpu_nr_empty_pop_pages -= chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages - chunk->nr_empty_pop_pages); }list_move (&chunk->list, &pcpu_chunk_lists[pcpu_to_depopulate_slot]); }
@@ -581,7 +582,8 @@ static void pcpu_reintegrate_chunk(struct pcpu_chunk *chunk)if (chunk->isolated) { chunk->isolated = false; - pcpu_nr_empty_pop_pages += chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + chunk->nr_empty_pop_pages);pcpu_chunk_relocate (chunk,-1 ); } }
@@ -599,7 +601,8 @@ static inline void pcpu_update_empty_pages(struct pcpu_chunk *chunk, int nr) { chunk->nr_empty_pop_pages += nr;if (chunk != pcpu_reserved_chunk && !chunk->isolated) - pcpu_nr_empty_pop_pages += nr; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + nr); }/*
@@ -1891,7 +1894,7 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +if (READ_ONCE (pcpu_nr_empty_pop_pages) <PCPU_EMPTY_POP_PAGES_LOW )pcpu_schedule_balance_work ();/* clear the areas and return address relative to base address */
@@ -2754,7 +2757,7 @@ void __init pcpu_setup_first_chunk(const struct pcpu_alloc_info *ai, tmp_addr = (unsigned long )base_addr + static_size + ai->reserved_size; pcpu_first_chunk =pcpu_alloc_first_chunk (tmp_addr, dyn_size); - pcpu_nr_empty_pop_pages = pcpu_first_chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, pcpu_first_chunk->nr_empty_pop_pages);pcpu_chunk_relocate (pcpu_first_chunk,-1 );/* include all regions of the first chunk */
Tejun Heo
maintainerHello,
On Mon, Jul 22, 2024 at 11:03:00AM -0700, Boqun Feng wrote:
...
Quoted 4 lines
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.
The patch looks great and I doubt this would impact performance in any noticeable way.
Thanks.
Signature · 1 line
tejun
Dennis Zhou
maintainerHello,
Quoted 17 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();
Quoted 8 lines
Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.
I think a better alternative is we can move it up into the lock under area_found. The value gets updated as part of pcpu_alloc_area() as the code above populates percpu memory that is already allocated.
We should probably annotate pcpu_update_empty_pages() with:
lockdep_assert_held(&pcpu_lock);
Thanks,
Dennis
Quoted 57 lines
Regards,
Boqun----->8
mm/percpu.c +8
-5@@ -570,7 +570,8 @@ static void pcpu_isolate_chunk(struct pcpu_chunk *chunk)if (!chunk->isolated) { chunk->isolated = true; - pcpu_nr_empty_pop_pages -= chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages - chunk->nr_empty_pop_pages); }list_move (&chunk->list, &pcpu_chunk_lists[pcpu_to_depopulate_slot]); }@@ -581,7 +582,8 @@ static void pcpu_reintegrate_chunk(struct pcpu_chunk *chunk)if (chunk->isolated) { chunk->isolated = false; - pcpu_nr_empty_pop_pages += chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + chunk->nr_empty_pop_pages);pcpu_chunk_relocate (chunk,-1 ); } }@@ -599,7 +601,8 @@ static inline void pcpu_update_empty_pages(struct pcpu_chunk *chunk, int nr) { chunk->nr_empty_pop_pages += nr;if (chunk != pcpu_reserved_chunk && !chunk->isolated) - pcpu_nr_empty_pop_pages += nr; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + nr); }/* @@ -1891,7 +1894,7 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +if (READ_ONCE (pcpu_nr_empty_pop_pages) <PCPU_EMPTY_POP_PAGES_LOW )pcpu_schedule_balance_work ();/* clear the areas and return address relative to base address */ @@ -2754,7 +2757,7 @@ void __init pcpu_setup_first_chunk(const struct pcpu_alloc_info *ai, tmp_addr = (unsigned long )base_addr + static_size + ai->reserved_size; pcpu_first_chunk =pcpu_alloc_first_chunk (tmp_addr, dyn_size); - pcpu_nr_empty_pop_pages = pcpu_first_chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, pcpu_first_chunk->nr_empty_pop_pages);pcpu_chunk_relocate (pcpu_first_chunk,-1 );/* include all regions of the first chunk */
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 17 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 8 lines
Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.
Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
mm/percpu.c +4 -2
@@ -1891,8 +1891,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); + scoped_guard(spinlock_irqsave, &pcpu_lock) { +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); + }/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)
I.e. just locking while checking.
Regards,
Boqun
Quoted 63 lines
We should probably annotate
pcpu_update_empty_pages()with:lockdep_assert_held(&pcpu_lock);Thanks,
DennisQuoted 57 lines
Regards,
Boqun----->8
mm/percpu.c +8
-5@@ -570,7 +570,8 @@ static void pcpu_isolate_chunk(struct pcpu_chunk *chunk)if (!chunk->isolated) { chunk->isolated = true; - pcpu_nr_empty_pop_pages -= chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages - chunk->nr_empty_pop_pages); }list_move (&chunk->list, &pcpu_chunk_lists[pcpu_to_depopulate_slot]); }@@ -581,7 +582,8 @@ static void pcpu_reintegrate_chunk(struct pcpu_chunk *chunk)if (chunk->isolated) { chunk->isolated = false; - pcpu_nr_empty_pop_pages += chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + chunk->nr_empty_pop_pages);pcpu_chunk_relocate (chunk,-1 ); } }@@ -599,7 +601,8 @@ static inline void pcpu_update_empty_pages(struct pcpu_chunk *chunk, int nr) { chunk->nr_empty_pop_pages += nr;if (chunk != pcpu_reserved_chunk && !chunk->isolated) - pcpu_nr_empty_pop_pages += nr; +WRITE_ONCE (pcpu_nr_empty_pop_pages, + pcpu_nr_empty_pop_pages + nr); }/* @@ -1891,7 +1894,7 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +if (READ_ONCE (pcpu_nr_empty_pop_pages) <PCPU_EMPTY_POP_PAGES_LOW )pcpu_schedule_balance_work ();/* clear the areas and return address relative to base address */ @@ -2754,7 +2757,7 @@ void __init pcpu_setup_first_chunk(const struct pcpu_alloc_info *ai, tmp_addr = (unsigned long )base_addr + static_size + ai->reserved_size; pcpu_first_chunk =pcpu_alloc_first_chunk (tmp_addr, dyn_size); - pcpu_nr_empty_pop_pages = pcpu_first_chunk->nr_empty_pop_pages; +WRITE_ONCE (pcpu_nr_empty_pop_pages, pcpu_first_chunk->nr_empty_pop_pages);pcpu_chunk_relocate (pcpu_first_chunk,-1 );/* include all regions of the first chunk */
Dennis Zhou
maintainerQuoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.
Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
I believe that's the only naked access of pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.
I also don't know how to rerun this CI tho..
---
mm/percpu.c +4 -3
@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */
@@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);
hi, Dennis Zhou,
Quoted 41 lines · Dennis Zhou
Quoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
Quoted 4 lines
I believe that's the only naked access of
pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.I also don't know how to rerun this CI tho..
we could test this patch. what's the base? could we apply it directly upon 24e44cc22a?
BTW, our bot is not so clever so far to auto test fix patches, so this is kind of manual effort. due to resource constraint, it will be hard for us to test each patch (we saw several patches in this thread already) or test very fast.
Quoted 27 lines
---
mm/percpu.c +4
-3@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */ @@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);
Dennis Zhou
maintainerHi Oliver,
Quoted 47 lines · Oliver Sang
hi, Dennis Zhou,
Quoted 44 lines · Dennis Zhou
Quoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
I believe that's the only naked access of
pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.
Quoted 9 lines
Quoted 1 line
I also don't know how to rerun this CI tho..
we could test this patch. what's the base? could we apply it directly upon
24e44cc22a?BTW, our bot is not so clever so far to auto test fix patches, so this is kind of manual effort. due to resource constraint, it will be hard for us to test each patch (we saw several patches in this thread already) or test very fast.
Ah yeah that makes sense. If you don't mind testing the last one I sent, the one below, that applies cleanly to 24e44cc22a.
Thanks,
Dennis
Quoted 27 lines
Quoted 27 lines
---
mm/percpu.c +4
-3@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */ @@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);
hi, Dannis,
Quoted 55 lines · Dennis Zhou
Hi Oliver,
Quoted 52 lines · Oliver Sang
hi, Dennis Zhou,
Quoted 41 lines · Dennis Zhou
Quoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
Quoted 4 lines
I believe that's the only naked access of
pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.I also don't know how to rerun this CI tho..
we could test this patch. what's the base? could we apply it directly upon
24e44cc22a?
Quoted 7 lines
Quoted 4 lines
BTW, our bot is not so clever so far to auto test fix patches, so this is kind of manual effort. due to resource constraint, it will be hard for us to test each patch (we saw several patches in this thread already) or test very fast.
Ah yeah that makes sense. If you don't mind testing the last one I sent, the one below, that applies cleanly to
24e44cc22a.
in our tests, you patch could solve the KCSAN issues. thanks
=========================================================================================
compiler/group/kconfig/nr_groups/rootfs/runtime/tbox_group/testcase:
gcc-13/group-04/x86_64-randconfig-013-20240713/5/debian-11.1-i386-20220923.cgz/300s/vm-snb/trinity
commit:
60fa4a9e23231 ("mm: percpu: add codetag reference into pcpuobj_ext") 24e44cc22aa31 ("mm: percpu: enable per-cpu allocation tagging") dcfbb68202759 <--- your patch
60fa4a9e23231721 24e44cc22aa3112082f2ee23137 dcfbb6820275994e92a9dcf309e ---------------- --------------------------- ---------------------------
fail:runs %reproduction fail:runs %reproduction fail:runs
| | | | |
281:999 -28% :998 -28% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_alloc
296:999 -30% :998 -30% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_block_update_hint_free
25:999 -3% :998 -3% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc/pcpu_chunk_populated
:999 29% 292:998 0% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_alloc
:999 27% 269:998 0% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_block_update_hint_free
:999 4% 44:998 0% :999 dmesg.BUG:KCSAN:data-race_in_pcpu_alloc_noprof/pcpu_chunk_populated
Quoted 31 lines
Thanks,
DennisQuoted 27 lines
Quoted 27 lines
---
mm/percpu.c +4
-3@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */ @@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);
Quoted 52 lines · Dennis Zhou
Quoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
I believe that's the only naked access of
pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.I also don't know how to rerun this CI tho..
---
mm/percpu.c +0
-0
mm/percpu.c void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, · line 1870
Quoted 7 lines
mm/percpu.c +4
-0@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +
But the pcpu_chunk_populated() afterwards could modify the pcpu_nr_empty_pop_pages again, wouldn't this be a behavior changing?
Regards,
Boqun
Quoted 13 lines
mm/percpu.c +0
-3spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */ @@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);
Dennis Zhou
maintainerQuoted 54 lines · Boqun Feng
Quoted 53 lines · Dennis Zhou
Quoted 30 lines · Boqun Feng
Quoted 29 lines · Dennis Zhou
Hello,
Quoted 25 lines · Boqun Feng
Quoted 16 lines · Tejun Heo
Quoted 3 lines · Boqun Feng
This looks like a data race because we read
pcpu_nr_empty_pop_pagesout of the lock for a best effort checking, @Tejun, maybe you could confirm on this?That does sound plausible.
Review
Quoted 8 lines
(quoted diff) +6
-1- if (pcpu_nr_empty_pop_pages < PCPU_EMPTY_POP_PAGES_LOW) + /* + * Checks pcpu_nr_empty_pop_pages out of the pcpu_lock, data races may + * occur but this is just a best-effort checking, everything is synced + * in pcpu_balance_work. + */ + if (data_race(pcpu_nr_empty_pop_pages) < PCPU_EMPTY_POP_PAGES_LOW) pcpu_schedule_balance_work();Quoted 2 lines
Would it be better to use READ/WRITE_ONCE() for the variable?
For READ/WRITE_ONCE(), we will need to replace all write accesses and all out-of-lock read accesses to
pcpu_nr_empty_pop_pages, like below. It's better in the sense that it doesn't rely on compiler behaviors on data races, not sure about the performance impact though.Quoted 9 lines
Quoted 4 lines
I think a better alternative is we can move it up into the lock under
area_found. The value gets updated as part ofpcpu_alloc_area()as the code above populates percpu memory that is already allocated.Not sure I followed what exactly you suggested here because I'm not familiar with the logic, but a simpler version would be:
I believe that's the only naked access of
pcpu_nr_empty_pop_pages. So I was thinking this'll fix this problem.I also don't know how to rerun this CI tho..
---
mm/percpu.c +0
-0@@ -1864,6 +1864,10 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,
Quoted 10 lines
mm/percpu.c void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved, · line 1870
Quoted 6 lines
mm/percpu.c +4
-0area_found:pcpu_stats_area_alloc (chunk, size); + +if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) +pcpu_schedule_balance_work (); +But the
pcpu_chunk_populated()afterwards could modify thepcpu_nr_empty_pop_pagesagain, wouldn't this be a behavior changing?
It does, but really at this point it's a mixed bag because the lock isn't permanently held at all while we do all these operations. The value is read at best effort.
Ultimately the code below is populating backing pages for non-atomic allocations. At this point the ideal situation is we're using an already populated page. There are caveats but I can't say the prior is any better than this version.
The code you mentioned pairs with the comment on line 916 below.
/*
* If the allocation is not atomic, some blocks may not be
* populated with pages, while we account it here. The number
* of pages will be added back with pcpu_chunk_populated()
* when populating pages.
*/
Thanks,
Dennis
Quoted 16 lines
Regards,
BoqunQuoted 13 lines
mm/percpu.c +0
-3spin_unlock_irqrestore (&pcpu_lock, flags);/* populate if not all pages are already there */ @@ -1891,9 +1895,6 @@ void __percpu *pcpu_alloc_noprof(size_t size, size_t align, bool reserved,mutex_unlock (&pcpu_alloc_mutex); } -if (pcpu_nr_empty_pop_pages <PCPU_EMPTY_POP_PAGES_LOW ) -pcpu_schedule_balance_work (); -/* clear the areas and return address relative to base address */ for_each_possible_cpu (cpu)memset ((void *)pcpu_chunk_addr (chunk, cpu,0 ) + off,0 , size);