KCSAN has reported a potential data race in blk_mq subsystem where
reading the rq->flag.
BUG: KCSAN: data-race in __blk_mq_end_request / blk_mq_check_inflight
read-write to 0xffff888120514d1c of 4 bytes by interrupt on cpu 6:
__blk_mq_end_request (block/blk-mq.c:700 block/blk-mq.c:1040)
scsi_end_request (drivers/scsi/scsi_lib.c:667)
scsi_io_completion (drivers/scsi/scsi_lib.c:1068)
scsi_finish_command (drivers/scsi/scsi.c:199)
scsi_complete (drivers/scsi/scsi_lib.c:?)
blk_done_softirq (block/blk-mq.c:? block/blk-mq.c:1134)
handle_softirqs (./arch/x86/include/asm/jump_label.h:27
./include/linux/jump_label.h:207
./include/trace/events/irq.h:142 kernel/softirq.c:555)
__irq_exit_rcu (kernel/softirq.c:617 kernel/softirq.c:639)
irq_exit_rcu (kernel/softirq.c:651)
common_interrupt (arch/x86/kernel/irq.c:247)
asm_common_interrupt (./arch/x86/include/asm/idtentry.h:693)
cpuidle_enter_state (drivers/cpuidle/cpuidle.c:291)
cpuidle_enter (drivers/cpuidle/cpuidle.c:388)
do_idle (kernel/sched/idle.c:155 kernel/sched/idle.c:236
kernel/sched/idle.c:332)
cpu_startup_entry (kernel/sched/idle.c:429)
start_secondary (arch/x86/kernel/smpboot.c:313)
common_startup_64 (arch/x86/kernel/head_64.S:421)
read to 0xffff888120514d1c of 4 bytes by task 9106 on cpu 51:
blk_mq_check_inflight (block/blk.h:356 block/blk-mq.c:94)
14:06:18 bt_iter (block/blk-mq-tag.c:292)
sbitmap_for_each_set (./include/linux/sbitmap.h:284
./include/linux/sbitmap.h:302)
blk_mq_queue_tag_busy_iter (block/blk-mq-tag.c:? block/blk-mq-tag.c:533)
blk_mq_in_flight (block/blk-mq.c:109)
diskstats_show (block/genhd.c:?)
seq_read_iter (fs/seq_file.c:?)
proc_reg_read_iter (fs/proc/inode.c:299)
vfs_read (fs/read_write.c:396 fs/read_write.c:476)
ksys_read (fs/read_write.c:619)
__x64_sys_read (fs/read_write.c:627)
x64_sys_call (arch/x86/entry/syscall_64.c:33)
do_syscall_64 (arch/x86/entry/common.c:?)
entry_SYSCALL_64_after_hwframe (arch/x86/entry/entry_64.S:130)
value changed: 0x00022382 -> 0x00022182
Discussing it with Jens Axboe and Pavel Begunkov, they suggested we just
want to annotated this with data_race(), since disk statistic reading
isn't critical, and it will not be a big deal if this bit is not stable.
Suggested-by: Jens Axboe <axboe@xxxxxxxxx>
Signed-off-by: Breno Leitao <leitao@xxxxxxxxxx>
---
block/blk.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/block/blk.h b/block/blk.h
index d9f584984bc4..57a1d73a0718 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -353,7 +353,8 @@ int blk_dev_init(void);
*/
static inline bool blk_do_io_stat(struct request *rq)
{
- return (rq->rq_flags & RQF_IO_STAT) && !blk_rq_is_passthrough(rq);
+ /* Disk stats reading isn’t critical, let it race */
+ return (data_race(rq->rq_flags) & RQF_IO_STAT) && !blk_rq_is_passthrough(rq);
}
void update_io_ticks(struct block_device *part, unsigned long now, bool end);
--
2.43.0
