scsi: ufs: core: Fix another task management completion race
authorAdrian Hunter <adrian.hunter@intel.com>
Mon, 8 Nov 2021 06:48:15 +0000 (08:48 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 25 Nov 2021 08:48:38 +0000 (09:48 +0100)
[ Upstream commit 5cb37a26355d79ab290220677b1b57d28e99a895 ]

hba->outstanding_tasks, which is read under host_lock spinlock, tells the
interrupt handler what task management tags are in use by the driver.  The
doorbell register bits indicate which tags are in use by the hardware.  A
doorbell bit that is 0 is because the bit has yet to be set by the driver,
or because the task is complete. It is only possible to disambiguate the 2
cases, if reading/writing the doorbell register is synchronized with
reading/writing hba->outstanding_tasks.

For that reason, reading REG_UTP_TASK_REQ_DOOR_BELL must be done under
spinlock.

Link: https://lore.kernel.org/r/20211108064815.569494-3-adrian.hunter@intel.com
Fixes: f5ef336fd2e4 ("scsi: ufs: core: Fix task management completion")
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/scsi/ufs/ufshcd.c

index 3d0da8b3fed8a8299686db590cf8107f5ea403eb..55f2e4d6f10b7ccf29a15bdb41ecc402b7ff2d42 100644 (file)
@@ -6382,9 +6382,8 @@ static irqreturn_t ufshcd_tmc_handler(struct ufs_hba *hba)
        irqreturn_t ret = IRQ_NONE;
        int tag;
 
-       pending = ufshcd_readl(hba, REG_UTP_TASK_REQ_DOOR_BELL);
-
        spin_lock_irqsave(hba->host->host_lock, flags);
+       pending = ufshcd_readl(hba, REG_UTP_TASK_REQ_DOOR_BELL);
        issued = hba->outstanding_tasks & ~pending;
        for_each_set_bit(tag, &issued, hba->nutmrs) {
                struct request *req = hba->tmf_rqs[tag];