scsi: ufs: core: Fix the polling implementation
authorBart Van Assche <bvanassche@acm.org>
Fri, 18 Nov 2022 23:37:03 +0000 (15:37 -0800)
committerMartin K. Petersen <martin.petersen@oracle.com>
Sat, 26 Nov 2022 02:31:10 +0000 (02:31 +0000)
Fix the following issues in ufshcd_poll():

 - If polling succeeds, return a positive value.

 - Do not complete polling requests from interrupt context because the
   block layer expects these requests to be completed from thread
   context. From block/bio.c:

     If REQ_ALLOC_CACHE is set, the final put of the bio MUST be done from
     process context, not hard/soft IRQ.

Fixes: eaab9b573054 ("scsi: ufs: Implement polling support")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20221118233717.441298-1-bvanassche@acm.org
Reviewed-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/ufs/core/ufshcd.c

index 0e88508..2dbe249 100644 (file)
@@ -5344,6 +5344,26 @@ static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
        }
 }
 
+/* Any value that is not an existing queue number is fine for this constant. */
+enum {
+       UFSHCD_POLL_FROM_INTERRUPT_CONTEXT = -1
+};
+
+static void ufshcd_clear_polled(struct ufs_hba *hba,
+                               unsigned long *completed_reqs)
+{
+       int tag;
+
+       for_each_set_bit(tag, completed_reqs, hba->nutrs) {
+               struct scsi_cmnd *cmd = hba->lrb[tag].cmd;
+
+               if (!cmd)
+                       continue;
+               if (scsi_cmd_to_rq(cmd)->cmd_flags & REQ_POLLED)
+                       __clear_bit(tag, completed_reqs);
+       }
+}
+
 /*
  * Returns > 0 if one or more commands have been completed or 0 if no
  * requests have been completed.
@@ -5360,13 +5380,17 @@ static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
        WARN_ONCE(completed_reqs & ~hba->outstanding_reqs,
                  "completed: %#lx; outstanding: %#lx\n", completed_reqs,
                  hba->outstanding_reqs);
+       if (queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT) {
+               /* Do not complete polled requests from interrupt context. */
+               ufshcd_clear_polled(hba, &completed_reqs);
+       }
        hba->outstanding_reqs &= ~completed_reqs;
        spin_unlock_irqrestore(&hba->outstanding_lock, flags);
 
        if (completed_reqs)
                __ufshcd_transfer_req_compl(hba, completed_reqs);
 
-       return completed_reqs;
+       return completed_reqs != 0;
 }
 
 /**
@@ -5397,7 +5421,7 @@ static irqreturn_t ufshcd_transfer_req_compl(struct ufs_hba *hba)
         * Ignore the ufshcd_poll() return value and return IRQ_HANDLED since we
         * do not want polling to trigger spurious interrupt complaints.
         */
-       ufshcd_poll(hba->host, 0);
+       ufshcd_poll(hba->host, UFSHCD_POLL_FROM_INTERRUPT_CONTEXT);
 
        return IRQ_HANDLED;
 }