Skip to contentSkip to conversation

Fix a rare crash in the UFS driver

PATCH Superseded Bart Van Assche 4 participants 16 messages Apr 12, 2024

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Bart Van Assche

author reviewer
#0 lore raw

cover Fix a rare crash in the UFS driver

Hi Martin,

Sporadic crashes have been observed with the UFS kernel driver if a timeout occurs. This patch series fixes these crashes. Please consider this patch series for the next merge window.

Thanks,

Bart.

Bart Van Assche (4):
scsi: ufs: Declare ufshcd_mcq_poll_cqe_lock() once
scsi: ufs: Make ufshcd_poll() complain about unsupported arguments
scsi: ufs: Make the polling code report which command has been
completed
scsi: ufs: Check for completion from the timeout handler

 drivers/ufs/core/ufs-mcq.c     | 23 ++++++++-----
 drivers/ufs/core/ufshcd-priv.h |  6 ++--
 drivers/ufs/core/ufshcd.c      | 59 +++++++++++++++++++++++++++-------
 drivers/ufs/host/ufs-qcom.c    |  2 +-
 include/ufs/ufshcd.h           |  3 +-
 5 files changed, 67 insertions(+), 26 deletions(-)
author reviewer
#1 lore raw

PATCH 1/4 scsi: ufs: Declare ufshcd_mcq_poll_cqe_lock() once

ufshcd_mcq_poll_cqe_lock() is declared in include/ufs/ufshcd.h and also in drivers/ufs/core/ufshcd-priv.h. Remove the declaration from the latter file.

---
 drivers/ufs/core/ufshcd-priv.h | 2 --
 1 file changed, 2 deletions(-)
drivers/ufs/core/ufshcd-priv.h +0 -2
index a1add22205db..fb4457a84d11 100644
@@ -72,8 +72,6 @@ u32 ufshcd_mcq_read_cqis(struct ufs_hba *hba, int i);
 void ufshcd_mcq_write_cqis(struct ufs_hba *hba, u32 val, int i);
 struct ufs_hw_queue *ufshcd_mcq_req_to_hwq(struct ufs_hba *hba,
 					   struct request *req);
-unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
-				       struct ufs_hw_queue *hwq);
 void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba *hba,
 				    struct ufs_hw_queue *hwq);
 bool ufshcd_cmd_inflight(struct scsi_cmnd *cmd);
author reviewer
#2 lore raw

PATCH 2/4 scsi: ufs: Make ufshcd_poll() complain about unsupported arguments

In reply to Bart Van Assche

The ufshcd_poll() implementation does not support queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode. Hence complain if queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode.

---
 drivers/ufs/core/ufshcd.c | 1 +
 1 file changed, 1 insertion(+)
drivers/ufs/core/ufshcd.c +1 -0
index 62c8575f2c67..66198eee51b0 100644
@@ -5617,6 +5617,7 @@ static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
 	struct ufs_hw_queue *hwq;
 
 	if (is_mcq_enabled(hba)) {
+		WARN_ON_ONCE(queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT);
 		hwq = &hba->uhq[queue_num];
 
 		return ufshcd_mcq_poll_cqe_lock(hba, hwq);
reviewer
#3 lore raw

RE: [PATCH 2/4] scsi: ufs: Make ufshcd_poll() complain about unsupported arguments

Quoted 4 lines

The ufshcd_poll() implementation does not support queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode. Hence complain if queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode.

Fixes tag - is it ed975065c31c ?

Thanks,
Avri

Quoted 19 lines
---
 drivers/ufs/core/ufshcd.c | 1 +
 1 file changed, 1 insertion(+)
drivers/ufs/core/ufshcd.c +1 -0
index 62c8575f2c67..66198eee51b0 100644
@@ -5617,6 +5617,7 @@ static int ufshcd_poll(struct Scsi_Host *shost,
 unsigned int queue_num)
        struct ufs_hw_queue *hwq;
 
        if (is_mcq_enabled(hba)) {
+               WARN_ON_ONCE(queue_num ==
 UFSHCD_POLL_FROM_INTERRUPT_CONTEXT);
                hwq = &hba->uhq[queue_num];
                return ufshcd_mcq_poll_cqe_lock(hba, hwq);
author reviewer
#4 lore raw

Re: [PATCH 2/4] scsi: ufs: Make ufshcd_poll() complain about unsupported arguments

Quoted 5 lines · Avri Altman
Quoted 3 lines

The ufshcd_poll() implementation does not support queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode. Hence complain if queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT in MCQ mode.

Fixes tag - is it ed975065c31c ?

Isn't the Fixes tag for patches that fix a shortcoming? This patch does not change the behavior of the UFS driver so I don't see it as a fix.

Thanks,

Bart.

author reviewer
#5 lore raw

PATCH 3/4 scsi: ufs: Make the polling code report which command has been completed

In reply to Bart Van Assche

Prepare for introducing a new __ufshcd_poll() caller that will need to know whether or not a specific command has been completed.

---
 drivers/ufs/core/ufs-mcq.c     | 23 +++++++++++++-------
 drivers/ufs/core/ufshcd-priv.h |  4 ++--
 drivers/ufs/core/ufshcd.c      | 39 +++++++++++++++++++++++-----------
 drivers/ufs/host/ufs-qcom.c    |  2 +-
 include/ufs/ufshcd.h           |  3 ++-
 5 files changed, 47 insertions(+), 24 deletions(-)
drivers/ufs/core/ufs-mcq.c +15 -8
index 228975caf68e..5a02e1b3b3a5 100644
@@ -268,17 +268,20 @@ static int ufshcd_mcq_get_tag(struct ufs_hba *hba, struct cq_entry *cqe)
 	return div_u64(addr, ufshcd_get_ucd_size(hba));
 }
 
-static void ufshcd_mcq_process_cqe(struct ufs_hba *hba,
-				   struct ufs_hw_queue *hwq)
+/* Returns true if and only if @compl_cmd has been completed. */
+static bool ufshcd_mcq_process_cqe(struct ufs_hba *hba,
+				   struct ufs_hw_queue *hwq,
+				   struct scsi_cmnd *compl_cmd)
 {
 	struct cq_entry *cqe = ufshcd_mcq_cur_cqe(hwq);
-	int tag = ufshcd_mcq_get_tag(hba, cqe);
 
 	if (cqe->command_desc_base_addr) {
-		ufshcd_compl_one_cqe(hba, tag, cqe);
-		/* After processed the cqe, mark it empty (invalid) entry */
+		const int tag = ufshcd_mcq_get_tag(hba, cqe);
+		/* Mark the CQE as invalid. */
 		cqe->command_desc_base_addr = 0;
+		return ufshcd_compl_one_cqe(hba, tag, cqe, compl_cmd);
 	}
+	return false;
 }
 
 void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba *hba,
@@ -289,7 +292,7 @@ void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba *hba,
 
 	spin_lock_irqsave(&hwq->cq_lock, flags);
 	while (entries > 0) {
-		ufshcd_mcq_process_cqe(hba, hwq);
+		ufshcd_mcq_process_cqe(hba, hwq, NULL);
 		ufshcd_mcq_inc_cq_head_slot(hwq);
 		entries--;
 	}
@@ -299,8 +302,10 @@ void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba *hba,
 	spin_unlock_irqrestore(&hwq->cq_lock, flags);
 }
 
+/* Clears *@compl_cmd if and only if *@compl_cmd has been completed. */
 unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
-				       struct ufs_hw_queue *hwq)
+				       struct ufs_hw_queue *hwq,
+				       struct scsi_cmnd **compl_cmd)
 {
 	unsigned long completed_reqs = 0;
 	unsigned long flags;
@@ -308,7 +313,9 @@ unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
 	spin_lock_irqsave(&hwq->cq_lock, flags);
 	ufshcd_mcq_update_cq_tail_slot(hwq);
 	while (!ufshcd_mcq_is_cq_empty(hwq)) {
-		ufshcd_mcq_process_cqe(hba, hwq);
+		if (ufshcd_mcq_process_cqe(hba, hwq,
+					   compl_cmd ? *compl_cmd : NULL))
+			*compl_cmd = NULL;
 		ufshcd_mcq_inc_cq_head_slot(hwq);
 		completed_reqs++;
 	}
drivers/ufs/core/ufshcd-priv.h +2 -2
index fb4457a84d11..42802fd689fb 100644
@@ -61,8 +61,8 @@ int ufshcd_query_attr(struct ufs_hba *hba, enum query_opcode opcode,
 int ufshcd_query_flag(struct ufs_hba *hba, enum query_opcode opcode,
 	enum flag_idn idn, u8 index, bool *flag_res);
 void ufshcd_auto_hibern8_update(struct ufs_hba *hba, u32 ahit);
-void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
-			  struct cq_entry *cqe);
+bool ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
+			  struct cq_entry *cqe, struct scsi_cmnd *compl_cmd);
 int ufshcd_mcq_init(struct ufs_hba *hba);
 int ufshcd_mcq_decide_queue_depth(struct ufs_hba *hba);
 int ufshcd_mcq_memory_alloc(struct ufs_hba *hba);
drivers/ufs/core/ufshcd.c +27 -12
index 66198eee51b0..08abdd763c51 100644
@@ -5540,9 +5540,12 @@ void ufshcd_release_scsi_cmd(struct ufs_hba *hba,
  * @hba: per adapter instance
  * @task_tag: the task tag of the request to be completed
  * @cqe: pointer to the completion queue entry
+ * @compl_cmd: if not NULL, check whether this command has been completed
+ *
+ * Returns: true if and only if @compl_cmd has been completed.
  */
-void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
-			  struct cq_entry *cqe)
+bool ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
+			  struct cq_entry *cqe, struct scsi_cmnd *compl_cmd)
 {
 	struct ufshcd_lrb *lrbp;
 	struct scsi_cmnd *cmd;
@@ -5559,6 +5562,7 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
 		ufshcd_release_scsi_cmd(hba, lrbp);
 		/* Do not touch lrbp after scsi done */
 		scsi_done(cmd);
+		return cmd == compl_cmd;
 	} else if (lrbp->command_type == UTP_CMD_TYPE_DEV_MANAGE ||
 		   lrbp->command_type == UTP_CMD_TYPE_UFS_STORAGE) {
 		if (hba->dev_cmd.complete) {
@@ -5569,6 +5573,7 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
 			complete(hba->dev_cmd.complete);
 		}
 	}
+	return false;
 }
 
 /**
@@ -5577,12 +5582,15 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
  * @completed_reqs: bitmask that indicates which requests to complete
  */
 static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
-					unsigned long completed_reqs)
+					unsigned long completed_reqs,
+					struct scsi_cmnd **compl_cmd)
 {
 	int tag;
 
 	for_each_set_bit(tag, &completed_reqs, hba->nutrs)
-		ufshcd_compl_one_cqe(hba, tag, NULL);
+		if (ufshcd_compl_one_cqe(hba, tag, NULL,
+					 compl_cmd ? *compl_cmd : NULL))
+			*compl_cmd = NULL;
 }
 
 /* Any value that is not an existing queue number is fine for this constant. */
@@ -5609,7 +5617,8 @@ static void ufshcd_clear_polled(struct ufs_hba *hba,
  * Return: > 0 if one or more commands have been completed or 0 if no
  * requests have been completed.
  */
-static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
+static int __ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num,
+			 struct scsi_cmnd **compl_cmd)
 {
 	struct ufs_hba *hba = shost_priv(shost);
 	unsigned long completed_reqs, flags;
@@ -5620,7 +5629,7 @@ static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
 		WARN_ON_ONCE(queue_num == UFSHCD_POLL_FROM_INTERRUPT_CONTEXT);
 		hwq = &hba->uhq[queue_num];
 
-		return ufshcd_mcq_poll_cqe_lock(hba, hwq);
+		return ufshcd_mcq_poll_cqe_lock(hba, hwq, compl_cmd);
 	}
 
 	spin_lock_irqsave(&hba->outstanding_lock, flags);
@@ -5637,11 +5646,16 @@ static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
 	spin_unlock_irqrestore(&hba->outstanding_lock, flags);
 
 	if (completed_reqs)
-		__ufshcd_transfer_req_compl(hba, completed_reqs);
+		__ufshcd_transfer_req_compl(hba, completed_reqs, compl_cmd);
 
 	return completed_reqs != 0;
 }
 
+static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
+{
+	return __ufshcd_poll(shost, queue_num, NULL);
+}
+
 /**
  * ufshcd_mcq_compl_pending_transfer - MCQ mode function. It is
  * invoked from the error handler context or ufshcd_host_reset_and_restore()
@@ -5685,7 +5699,7 @@ static void ufshcd_mcq_compl_pending_transfer(struct ufs_hba *hba,
 			}
 			spin_unlock_irqrestore(&hwq->cq_lock, flags);
 		} else {
-			ufshcd_mcq_poll_cqe_lock(hba, hwq);
+			ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
 		}
 	}
 }
@@ -6960,7 +6974,7 @@ static irqreturn_t ufshcd_handle_mcq_cq_events(struct ufs_hba *hba)
 			ufshcd_mcq_write_cqis(hba, events, i);
 
 		if (events & UFSHCD_MCQ_CQIS_TAIL_ENT_PUSH_STS)
-			ufshcd_mcq_poll_cqe_lock(hba, hwq);
+			ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
 	}
 
 	return IRQ_HANDLED;
@@ -7453,7 +7467,7 @@ static int ufshcd_eh_device_reset_handler(struct scsi_cmnd *cmd)
 			    lrbp->lun == lun) {
 				ufshcd_clear_cmd(hba, pos);
 				hwq = ufshcd_mcq_req_to_hwq(hba, scsi_cmd_to_rq(lrbp->cmd));
-				ufshcd_mcq_poll_cqe_lock(hba, hwq);
+				ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
 			}
 		}
 		err = 0;
@@ -7481,7 +7495,8 @@ static int ufshcd_eh_device_reset_handler(struct scsi_cmnd *cmd)
 				__func__, pos);
 		}
 	}
-	__ufshcd_transfer_req_compl(hba, pending_reqs & ~not_cleared_mask);
+	__ufshcd_transfer_req_compl(hba, pending_reqs & ~not_cleared_mask,
+				    NULL);
 
 out:
 	hba->req_abort_count = 0;
@@ -7658,7 +7673,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 		dev_err(hba->dev,
 		"%s: cmd was completed, but without a notifying intr, tag = %d",
 		__func__, tag);
-		__ufshcd_transfer_req_compl(hba, 1UL << tag);
+		__ufshcd_transfer_req_compl(hba, 1UL << tag, NULL);
 		goto release;
 	}
 
drivers/ufs/host/ufs-qcom.c +1 -1
index b3ca9b3bf94b..0598cd8f53dd 100644
@@ -1704,7 +1704,7 @@ static irqreturn_t ufs_qcom_mcq_esi_handler(int irq, void *data)
 	struct ufs_hw_queue *hwq = &hba->uhq[id];
 
 	ufshcd_mcq_write_cqis(hba, 0x1, id);
-	ufshcd_mcq_poll_cqe_lock(hba, hwq);
+	ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
 
 	return IRQ_HANDLED;
 }
include/ufs/ufshcd.h +2 -1
index 4ba826fe7b62..8fac282e0476 100644
@@ -1263,7 +1263,8 @@ void ufshcd_mcq_config_mac(struct ufs_hba *hba, u32 max_active_cmds);
 u32 ufshcd_mcq_read_cqis(struct ufs_hba *hba, int i);
 void ufshcd_mcq_write_cqis(struct ufs_hba *hba, u32 val, int i);
 unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
-					 struct ufs_hw_queue *hwq);
+				       struct ufs_hw_queue *hwq,
+				       struct scsi_cmnd **compl_cmd);
 void ufshcd_mcq_make_queues_operational(struct ufs_hba *hba);
 void ufshcd_mcq_enable_esi(struct ufs_hba *hba);
 void ufshcd_mcq_enable(struct ufs_hba *hba);
reviewer
#6 lore raw

RE: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

Quoted 31 lines

Prepare for introducing a new __ufshcd_poll() caller that will need to know whether or not a specific command has been completed.

---
 drivers/ufs/core/ufs-mcq.c     | 23 +++++++++++++-------
 drivers/ufs/core/ufshcd-priv.h |  4 ++--
 drivers/ufs/core/ufshcd.c      | 39 +++++++++++++++++++++++-----------
 drivers/ufs/host/ufs-qcom.c    |  2 +-
 include/ufs/ufshcd.h           |  3 ++-
 5 files changed, 47 insertions(+), 24 deletions(-)
drivers/ufs/core/ufs-mcq.c +4 -3
index 228975caf68e..5a02e1b3b3a5 100644
@@ -268,17 +268,20 @@ static int ufshcd_mcq_get_tag(struct ufs_hba *hba,
 struct cq_entry *cqe)
        return div_u64(addr, ufshcd_get_ucd_size(hba));
 }
 
-static void ufshcd_mcq_process_cqe(struct ufs_hba *hba,
-                                  struct ufs_hw_queue *hwq)
+/* Returns true if and only if @compl_cmd has been completed. */
+static bool ufshcd_mcq_process_cqe(struct ufs_hba *hba,
+                                  struct ufs_hw_queue *hwq,
+                                  struct scsi_cmnd *compl_cmd)
 {
        struct cq_entry *cqe = ufshcd_mcq_cur_cqe(hwq);
-       int tag = ufshcd_mcq_get_tag(hba, cqe);
 

drivers/ufs/core/ufs-mcq.c static int ufshcd_mcq_get_tag(struct ufs_hba *hba, · line 286

Quoted 10 lines
drivers/ufs/core/ufs-mcq.c +4 -2
        if (cqe->command_desc_base_addr) {
-               ufshcd_compl_one_cqe(hba, tag, cqe);
-               /* After processed the cqe, mark it empty (invalid) entry */
+               const int tag = ufshcd_mcq_get_tag(hba, cqe);
+               /* Mark the CQE as invalid. */
                cqe->command_desc_base_addr = 0;
+               return ufshcd_compl_one_cqe(hba, tag, cqe, compl_cmd);
        }
+       return false;
 }

Maybe elaborate on explaining why the tag isn't enough to designate the completing command?

Thanks,
Avri

Quoted 230 lines
drivers/ufs/core/ufs-mcq.c +1 -1
 void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba *hba,
@@ -289,7 +292,7 @@ void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba
 *hba,
 
        spin_lock_irqsave(&hwq->cq_lock, flags);
        while (entries > 0) {
-               ufshcd_mcq_process_cqe(hba, hwq);
+               ufshcd_mcq_process_cqe(hba, hwq, NULL);
                ufshcd_mcq_inc_cq_head_slot(hwq);
                entries--;
        }
drivers/ufs/core/ufs-mcq.c +3 -1
@@ -299,8 +302,10 @@ void ufshcd_mcq_compl_all_cqes_lock(struct ufs_hba
 *hba,
        spin_unlock_irqrestore(&hwq->cq_lock, flags);
 }
 
+/* Clears *@compl_cmd if and only if *@compl_cmd has been completed. */
 unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
-                                      struct ufs_hw_queue *hwq)
+                                      struct ufs_hw_queue *hwq,
+                                      struct scsi_cmnd **compl_cmd)
 {
        unsigned long completed_reqs = 0;
        unsigned long flags;
drivers/ufs/core/ufs-mcq.c +3 -1
@@ -308,7 +313,9 @@ unsigned long ufshcd_mcq_poll_cqe_lock(struct
 ufs_hba *hba,
        spin_lock_irqsave(&hwq->cq_lock, flags);
        ufshcd_mcq_update_cq_tail_slot(hwq);
        while (!ufshcd_mcq_is_cq_empty(hwq)) {
-               ufshcd_mcq_process_cqe(hba, hwq);
+               if (ufshcd_mcq_process_cqe(hba, hwq,
+                                          compl_cmd ? *compl_cmd : NULL))
+                       *compl_cmd = NULL;
                ufshcd_mcq_inc_cq_head_slot(hwq);
                completed_reqs++;
        }
drivers/ufs/core/ufshcd-priv.h +2 -2
index fb4457a84d11..42802fd689fb 100644
@@ -61,8 +61,8 @@ int ufshcd_query_attr(struct ufs_hba *hba, enum
 query_opcode opcode,
 int ufshcd_query_flag(struct ufs_hba *hba, enum query_opcode opcode,
        enum flag_idn idn, u8 index, bool *flag_res);
 void ufshcd_auto_hibern8_update(struct ufs_hba *hba, u32 ahit);
-void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
-                         struct cq_entry *cqe);
+bool ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
+                         struct cq_entry *cqe, struct scsi_cmnd *compl_cmd);
 int ufshcd_mcq_init(struct ufs_hba *hba);
 int ufshcd_mcq_decide_queue_depth(struct ufs_hba *hba);

int ufshcd_mcq_memory_alloc(struct ufs_hba *hba);

drivers/ufs/core/ufshcd.c +1 -0
index 66198eee51b0..08abdd763c51 100644
@@ -5540,9 +5540,12 @@ void ufshcd_release_scsi_cmd(struct ufs_hba *hba,
  * @hba: per adapter instance
  * @task_tag: the task tag of the request to be completed
  * @cqe: pointer to the completion queue entry
+ * @compl_cmd: if not NULL, check whether this command has been

completed
+ *
+ * Returns: true if and only if @compl_cmd has been completed. */

drivers/ufs/core/ufshcd.c +3 -2
-void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
-                         struct cq_entry *cqe)
+bool ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag,
+                         struct cq_entry *cqe, struct scsi_cmnd *compl_cmd)
 {
        struct ufshcd_lrb *lrbp;
        struct scsi_cmnd *cmd;
@@ -5559,6 +5562,7 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba, int
 task_tag,
                ufshcd_release_scsi_cmd(hba, lrbp);
                /* Do not touch lrbp after scsi done */
                scsi_done(cmd);
+               return cmd == compl_cmd;
        } else if (lrbp->command_type == UTP_CMD_TYPE_DEV_MANAGE ||
                   lrbp->command_type == UTP_CMD_TYPE_UFS_STORAGE) {
                if (hba->dev_cmd.complete) {
drivers/ufs/core/ufshcd.c +1 -0
@@ -5569,6 +5573,7 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba, int
 task_tag,
                        complete(hba->dev_cmd.complete);
                }
        }
+       return false;
 }
 

/**

drivers/ufs/core/ufshcd.c +5 -2
@@ -5577,12 +5582,15 @@ void ufshcd_compl_one_cqe(struct ufs_hba *hba,
 int task_tag,
  * @completed_reqs: bitmask that indicates which requests to complete
  */
 static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
-                                       unsigned long completed_reqs)
+                                       unsigned long completed_reqs,
+                                       struct scsi_cmnd **compl_cmd)
 {
        int tag;
 
        for_each_set_bit(tag, &completed_reqs, hba->nutrs)
-               ufshcd_compl_one_cqe(hba, tag, NULL);
+               if (ufshcd_compl_one_cqe(hba, tag, NULL,
+                                        compl_cmd ? *compl_cmd : NULL))
+                       *compl_cmd = NULL;
 }
 

/* Any value that is not an existing queue number is fine for this constant. */

drivers/ufs/core/ufshcd.c +3 -2
@@ -5609,7 +5617,8 @@ static void ufshcd_clear_polled(struct ufs_hba *hba,
  * Return: > 0 if one or more commands have been completed or 0 if no
  * requests have been completed.
  */
-static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
+static int __ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num,
+                        struct scsi_cmnd **compl_cmd)
 {
        struct ufs_hba *hba = shost_priv(shost);
        unsigned long completed_reqs, flags;
@@ -5620,7 +5629,7 @@ static int ufshcd_poll(struct Scsi_Host *shost,
 unsigned int queue_num)
                WARN_ON_ONCE(queue_num ==
 UFSHCD_POLL_FROM_INTERRUPT_CONTEXT);
                hwq = &hba->uhq[queue_num];
 
-               return ufshcd_mcq_poll_cqe_lock(hba, hwq);
+               return ufshcd_mcq_poll_cqe_lock(hba, hwq, compl_cmd);
        }
        spin_lock_irqsave(&hba->outstanding_lock, flags);
drivers/ufs/core/ufshcd.c +6 -1
@@ -5637,11 +5646,16 @@ static int ufshcd_poll(struct Scsi_Host *shost,
 unsigned int queue_num)
        spin_unlock_irqrestore(&hba->outstanding_lock, flags);
 
        if (completed_reqs)
-               __ufshcd_transfer_req_compl(hba, completed_reqs);
+               __ufshcd_transfer_req_compl(hba, completed_reqs, compl_cmd);
 
        return completed_reqs != 0;
 }
 
+static int ufshcd_poll(struct Scsi_Host *shost, unsigned int queue_num)
+{
+       return __ufshcd_poll(shost, queue_num, NULL);
+}
+
 /**
  * ufshcd_mcq_compl_pending_transfer - MCQ mode function. It is

* invoked from the error handler context or ufshcd_host_reset_and_restore()

drivers/ufs/core/ufshcd.c +1 -1
@@ -5685,7 +5699,7 @@ static void
 ufshcd_mcq_compl_pending_transfer(struct ufs_hba *hba,
                        }
                        spin_unlock_irqrestore(&hwq->cq_lock, flags);
                } else {
-                       ufshcd_mcq_poll_cqe_lock(hba, hwq);
+                       ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
                }
        }

}

drivers/ufs/core/ufshcd.c +1 -1
@@ -6960,7 +6974,7 @@ static irqreturn_t
 ufshcd_handle_mcq_cq_events(struct ufs_hba *hba)
                        ufshcd_mcq_write_cqis(hba, events, i);
 
                if (events & UFSHCD_MCQ_CQIS_TAIL_ENT_PUSH_STS)
-                       ufshcd_mcq_poll_cqe_lock(hba, hwq);
+                       ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
        }
 
        return IRQ_HANDLED;
drivers/ufs/core/ufshcd.c +1 -1
@@ -7453,7 +7467,7 @@ static int ufshcd_eh_device_reset_handler(struct
 scsi_cmnd *cmd)
                            lrbp->lun == lun) {
                                ufshcd_clear_cmd(hba, pos);
                                hwq = ufshcd_mcq_req_to_hwq(hba, scsi_cmd_to_rq(lrbp-
 >cmd));
-                               ufshcd_mcq_poll_cqe_lock(hba, hwq);
+                               ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
                        }
                }
                err = 0;
drivers/ufs/core/ufshcd.c +2 -1
@@ -7481,7 +7495,8 @@ static int ufshcd_eh_device_reset_handler(struct
 scsi_cmnd *cmd)
                                __func__, pos);
                }
        }
-       __ufshcd_transfer_req_compl(hba, pending_reqs & ~not_cleared_mask);
+       __ufshcd_transfer_req_compl(hba, pending_reqs & ~not_cleared_mask,
+                                   NULL);
 
 out:
        hba->req_abort_count = 0;
drivers/ufs/core/ufshcd.c +1 -1
@@ -7658,7 +7673,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
                dev_err(hba->dev,
                "%s: cmd was completed, but without a notifying intr, tag = %d",
                __func__, tag);
-               __ufshcd_transfer_req_compl(hba, 1UL << tag);
+               __ufshcd_transfer_req_compl(hba, 1UL << tag, NULL);
                goto release;
        }
 
drivers/ufs/host/ufs-qcom.c +1 -1
index b3ca9b3bf94b..0598cd8f53dd 100644
@@ -1704,7 +1704,7 @@ static irqreturn_t ufs_qcom_mcq_esi_handler(int irq,
 void *data)
        struct ufs_hw_queue *hwq = &hba->uhq[id];
 
        ufshcd_mcq_write_cqis(hba, 0x1, id);
-       ufshcd_mcq_poll_cqe_lock(hba, hwq);
+       ufshcd_mcq_poll_cqe_lock(hba, hwq, NULL);
 
        return IRQ_HANDLED;

}

include/ufs/ufshcd.h +2 -1
index 4ba826fe7b62..8fac282e0476 100644
@@ -1263,7 +1263,8 @@ void ufshcd_mcq_config_mac(struct ufs_hba *hba,
 u32 max_active_cmds);
 u32 ufshcd_mcq_read_cqis(struct ufs_hba *hba, int i);
 void ufshcd_mcq_write_cqis(struct ufs_hba *hba, u32 val, int i);
 unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
-                                        struct ufs_hw_queue *hwq);
+                                      struct ufs_hw_queue *hwq,
+                                      struct scsi_cmnd **compl_cmd);
 void ufshcd_mcq_make_queues_operational(struct ufs_hba *hba);
 void ufshcd_mcq_enable_esi(struct ufs_hba *hba);

void ufshcd_mcq_enable(struct ufs_hba *hba);

author reviewer
#7 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

Quoted 10 lines · Avri Altman
Quoted 10 lines
drivers/ufs/core/ufs-mcq.c +4 -3
-static void ufshcd_mcq_process_cqe(struct ufs_hba *hba,
-                                  struct ufs_hw_queue *hwq)
+/* Returns true if and only if @compl_cmd has been completed. */
+static bool ufshcd_mcq_process_cqe(struct ufs_hba *hba,
+                                  struct ufs_hw_queue *hwq,
+                                  struct scsi_cmnd *compl_cmd)
  {
         struct cq_entry *cqe = ufshcd_mcq_cur_cqe(hwq);
-       int tag = ufshcd_mcq_get_tag(hba, cqe);
Quoted 12 lines

drivers/ufs/core/ufs-mcq.c static int ufshcd_mcq_get_tag(struct ufs_hba *hba, struct cq_entry *cqe) · line 285

Quoted 10 lines
         if (cqe->command_desc_base_addr) {
drivers/ufs/core/ufs-mcq.c +4 -2
-               ufshcd_compl_one_cqe(hba, tag, cqe);
-               /* After processed the cqe, mark it empty (invalid) entry */
+               const int tag = ufshcd_mcq_get_tag(hba, cqe);
+               /* Mark the CQE as invalid. */
                 cqe->command_desc_base_addr = 0;
+               return ufshcd_compl_one_cqe(hba, tag, cqe, compl_cmd);
         }
+       return false;
  }

Maybe elaborate on explaining why the tag isn't enough to designate the completing command?

Hi Avri,

Anything that uniquely identifies a SCSI command would work. I think it's more a matter of making a choice rather than arguing why not using the blk_mq_unique_tag_to_hwq() return value?

Thanks,

Bart.

#8 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

In reply to Bart Van Assche
Quoted 9 lines · Bart Van Assche

External email : Please do not click links or open attachments until you have verified the sender or the content.
Prepare for introducing a new __ufshcd_poll() caller that will need
to
know whether or not a specific command has been completed.

Hi Bart,

Please also change ufs-medatek.c ufshcd_mcq_poll_cqe_lock usage.

Thanks.
Peter

author reviewer
#9 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

Quoted 1 line · Peter Wang (王信友)

Please also change ufs-medatek.c ufshcd_mcq_poll_cqe_lock usage.

Will do. Thanks for having reported this.

Bart.

#10 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

In reply to Bart Van Assche
kernel test robot noticed the following build errors:

Hi Bart,

kernel test robot noticed the following build errors:

[auto build test ERROR on mkp-scsi/for-next]
[also build test ERROR on jejb-scsi/for-next linus/master v6.9-rc3 next-20240412] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Bart-Van-Assche/scsi-ufs-Declare-ufshcd_mcq_poll_cqe_lock-once/20240412-080540
base:   https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git for-next
patch link:    https://lore.kernel.org/r/20240412000246.1167600-4-bvanassche%40acm.org
patch subject: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed
config: arm-allmodconfig (https://download.01.org/0day-ci/archive/20240412/[email protected]/config)
compiler: arm-linux-gnueabi-gcc (GCC) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240412/[email protected]/reproduce)

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-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

drivers/ufs/host/ufs-mediatek.c: In function 'ufs_mtk_mcq_intr':

>> drivers/ufs/host/ufs-mediatek.c:1717:17: error: too few arguments to function 'ufshcd_mcq_poll_cqe_lock'
    1717 |                 ufshcd_mcq_poll_cqe_lock(hba, hwq);
         |                 ^~~~~~~~~~~~~~~~~~~~~~~~

In file included from drivers/ufs/host/ufs-mediatek.c:23:
include/ufs/ufshcd.h:1267:15: note: declared here
1267 | unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
| ^~~~~~~~~~~~~~~~~~~~~~~~

vim +/ufshcd_mcq_poll_cqe_lock +1717 drivers/ufs/host/ufs-mediatek.c

e152a616c88653 Po-Wen Kao 2023-07-01  1701  
e152a616c88653 Po-Wen Kao 2023-07-01  1702  static irqreturn_t ufs_mtk_mcq_intr(int irq, void *__intr_info)
e152a616c88653 Po-Wen Kao 2023-07-01  1703  {
e152a616c88653 Po-Wen Kao 2023-07-01  1704  	struct ufs_mtk_mcq_intr_info *mcq_intr_info = __intr_info;
e152a616c88653 Po-Wen Kao 2023-07-01  1705  	struct ufs_hba *hba = mcq_intr_info->hba;
e152a616c88653 Po-Wen Kao 2023-07-01  1706  	struct ufs_hw_queue *hwq;
e152a616c88653 Po-Wen Kao 2023-07-01  1707  	u32 events;
e152a616c88653 Po-Wen Kao 2023-07-01  1708  	int qid = mcq_intr_info->qid;
e152a616c88653 Po-Wen Kao 2023-07-01  1709  
e152a616c88653 Po-Wen Kao 2023-07-01  1710  	hwq = &hba->uhq[qid];
e152a616c88653 Po-Wen Kao 2023-07-01  1711  
e152a616c88653 Po-Wen Kao 2023-07-01  1712  	events = ufshcd_mcq_read_cqis(hba, qid);
e152a616c88653 Po-Wen Kao 2023-07-01  1713  	if (events)
e152a616c88653 Po-Wen Kao 2023-07-01  1714  		ufshcd_mcq_write_cqis(hba, events, qid);
e152a616c88653 Po-Wen Kao 2023-07-01  1715  
e152a616c88653 Po-Wen Kao 2023-07-01  1716  	if (events & UFSHCD_MCQ_CQIS_TAIL_ENT_PUSH_STS)
e152a616c88653 Po-Wen Kao 2023-07-01 @1717  		ufshcd_mcq_poll_cqe_lock(hba, hwq);
e152a616c88653 Po-Wen Kao 2023-07-01  1718  
e152a616c88653 Po-Wen Kao 2023-07-01  1719  	return IRQ_HANDLED;
e152a616c88653 Po-Wen Kao 2023-07-01  1720  }
e152a616c88653 Po-Wen Kao 2023-07-01  1721  
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#11 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

In reply to Bart Van Assche
kernel test robot noticed the following build errors:

Hi Bart,

kernel test robot noticed the following build errors:

[auto build test ERROR on mkp-scsi/for-next]
[also build test ERROR on jejb-scsi/for-next linus/master v6.9-rc3 next-20240412] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Bart-Van-Assche/scsi-ufs-Declare-ufshcd_mcq_poll_cqe_lock-once/20240412-080540
base:   https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git for-next
patch link:    https://lore.kernel.org/r/20240412000246.1167600-4-bvanassche%40acm.org
patch subject: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed
config: arm64-allmodconfig (https://download.01.org/0day-ci/archive/20240412/[email protected]/config)
compiler: clang version 19.0.0git (https://github.com/llvm/llvm-project 8b3b4a92adee40483c27f26c478a384cd69c6f05)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240412/[email protected]/reproduce)

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-kbuild-all/[email protected]/

All errors (new ones prefixed by >>):

In file included from drivers/ufs/host/ufs-mediatek.c:18:
In file included from include/linux/phy/phy.h:17:
In file included from include/linux/regulator/consumer.h:35:
In file included from include/linux/suspend.h:5:
In file included from include/linux/swap.h:9:
In file included from include/linux/memcontrol.h:21:
In file included from include/linux/mm.h:2208:
include/linux/vmstat.h:508:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]

     508 |         return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~ ^
     509 |                            item];
         |                            ~~~~
   include/linux/vmstat.h:515:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
     515 |         return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~ ^
     516 |                            NR_VM_NUMA_EVENT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~~
   include/linux/vmstat.h:522:36: warning: arithmetic between different enumeration types ('enum node_stat_item' and 'enum lru_list') [-Wenum-enum-conversion]
     522 |         return node_stat_name(NR_LRU_BASE + lru) + 3; // skip "nr_"
         |                               ~~~~~~~~~~~ ^ ~~~
   include/linux/vmstat.h:527:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
     527 |         return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~ ^
     528 |                            NR_VM_NUMA_EVENT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~~
   include/linux/vmstat.h:536:43: warning: arithmetic between different enumeration types ('enum zone_stat_item' and 'enum numa_stat_item') [-Wenum-enum-conversion]
     536 |         return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~ ^
     537 |                            NR_VM_NUMA_EVENT_ITEMS +
         |                            ~~~~~~~~~~~~~~~~~~~~~~
>> drivers/ufs/host/ufs-mediatek.c:1717:36: error: too few arguments to function call, expected 3, have 2
    1717 |                 ufshcd_mcq_poll_cqe_lock(hba, hwq);
         |                 ~~~~~~~~~~~~~~~~~~~~~~~~         ^
   include/ufs/ufshcd.h:1267:15: note: 'ufshcd_mcq_poll_cqe_lock' declared here
    1267 | unsigned long ufshcd_mcq_poll_cqe_lock(struct ufs_hba *hba,
         |               ^                        ~~~~~~~~~~~~~~~~~~~~
    1268 |                                        struct ufs_hw_queue *hwq,
         |                                        ~~~~~~~~~~~~~~~~~~~~~~~~~
    1269 |                                        struct scsi_cmnd **compl_cmd);
         |                                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~

5 warnings and 1 error generated.

vim +1717 drivers/ufs/host/ufs-mediatek.c

e152a616c88653 Po-Wen Kao 2023-07-01  1701  
e152a616c88653 Po-Wen Kao 2023-07-01  1702  static irqreturn_t ufs_mtk_mcq_intr(int irq, void *__intr_info)
e152a616c88653 Po-Wen Kao 2023-07-01  1703  {
e152a616c88653 Po-Wen Kao 2023-07-01  1704  	struct ufs_mtk_mcq_intr_info *mcq_intr_info = __intr_info;
e152a616c88653 Po-Wen Kao 2023-07-01  1705  	struct ufs_hba *hba = mcq_intr_info->hba;
e152a616c88653 Po-Wen Kao 2023-07-01  1706  	struct ufs_hw_queue *hwq;
e152a616c88653 Po-Wen Kao 2023-07-01  1707  	u32 events;
e152a616c88653 Po-Wen Kao 2023-07-01  1708  	int qid = mcq_intr_info->qid;
e152a616c88653 Po-Wen Kao 2023-07-01  1709  
e152a616c88653 Po-Wen Kao 2023-07-01  1710  	hwq = &hba->uhq[qid];
e152a616c88653 Po-Wen Kao 2023-07-01  1711  
e152a616c88653 Po-Wen Kao 2023-07-01  1712  	events = ufshcd_mcq_read_cqis(hba, qid);
e152a616c88653 Po-Wen Kao 2023-07-01  1713  	if (events)
e152a616c88653 Po-Wen Kao 2023-07-01  1714  		ufshcd_mcq_write_cqis(hba, events, qid);
e152a616c88653 Po-Wen Kao 2023-07-01  1715  
e152a616c88653 Po-Wen Kao 2023-07-01  1716  	if (events & UFSHCD_MCQ_CQIS_TAIL_ENT_PUSH_STS)
e152a616c88653 Po-Wen Kao 2023-07-01 @1717  		ufshcd_mcq_poll_cqe_lock(hba, hwq);
e152a616c88653 Po-Wen Kao 2023-07-01  1718  
e152a616c88653 Po-Wen Kao 2023-07-01  1719  	return IRQ_HANDLED;
e152a616c88653 Po-Wen Kao 2023-07-01  1720  }
e152a616c88653 Po-Wen Kao 2023-07-01  1721  
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#12 lore raw

Re: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed

In reply to Bart Van Assche
kernel test robot noticed the following build warnings:

Hi Bart,

kernel test robot noticed the following build warnings:

[auto build test WARNING on mkp-scsi/for-next]
[also build test WARNING on jejb-scsi/for-next linus/master v6.9-rc3 next-20240412] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Bart-Van-Assche/scsi-ufs-Declare-ufshcd_mcq_poll_cqe_lock-once/20240412-080540
base:   https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git for-next
patch link:    https://lore.kernel.org/r/20240412000246.1167600-4-bvanassche%40acm.org
patch subject: [PATCH 3/4] scsi: ufs: Make the polling code report which command has been completed
config: i386-randconfig-061-20240412 (https://download.01.org/0day-ci/archive/20240412/[email protected]/config)
compiler: gcc-13 (Ubuntu 13.2.0-4ubuntu3) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240412/[email protected]/reproduce)

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-kbuild-all/[email protected]/

All warnings (new ones prefixed by >>):

>> drivers/ufs/core/ufshcd.c:5587: warning: Function parameter or struct member 'compl_cmd' not described in '__ufshcd_transfer_req_compl'

vim +5587 drivers/ufs/core/ufshcd.c

c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5578  
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5579  /**
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5580   * __ufshcd_transfer_req_compl - handle SCSI and query command completion
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5581   * @hba: per adapter instance
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5582   * @completed_reqs: bitmask that indicates which requests to complete
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5583   */
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5584  static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
56399aa6ca74cf drivers/ufs/core/ufshcd.c Bart Van Assche    2024-04-11  5585  					unsigned long completed_reqs,
56399aa6ca74cf drivers/ufs/core/ufshcd.c Bart Van Assche    2024-04-11  5586  					struct scsi_cmnd **compl_cmd)
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13 @5587  {
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5588  	int tag;
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5589  
c30d8d010b5efd drivers/ufs/core/ufshcd.c Asutosh Das        2023-01-13  5590  	for_each_set_bit(tag, &completed_reqs, hba->nutrs)
56399aa6ca74cf drivers/ufs/core/ufshcd.c Bart Van Assche    2024-04-11  5591  		if (ufshcd_compl_one_cqe(hba, tag, NULL,
56399aa6ca74cf drivers/ufs/core/ufshcd.c Bart Van Assche    2024-04-11  5592  					 compl_cmd ? *compl_cmd : NULL))
56399aa6ca74cf drivers/ufs/core/ufshcd.c Bart Van Assche    2024-04-11  5593  			*compl_cmd = NULL;
a45f937110fa6b drivers/scsi/ufs/ufshcd.c Can Guo            2021-05-24  5594  }

7a3e97b0dc4bba drivers/scsi/ufs/ufshcd.c Santosh Yaraganavi 2012-02-29 5595

Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
author reviewer
#13 lore raw

PATCH 4/4 scsi: ufs: Check for completion from the timeout handler

In reply to Bart Van Assche

If ufshcd_abort() returns SUCCESS for an already completed command then that command is completed twice. This results in a crash. Prevent this by checking whether a command has completed without completion interrupt from the timeout handler. This CL fixes the following kernel crash:

Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
Call trace:
 dma_direct_map_sg+0x70/0x274
 scsi_dma_map+0x84/0x124
 ufshcd_queuecommand+0x3fc/0x880
 scsi_queue_rq+0x7d0/0x111c
 blk_mq_dispatch_rq_list+0x440/0xebc
 blk_mq_do_dispatch_sched+0x5a4/0x6b8
 __blk_mq_sched_dispatch_requests+0x150/0x220
 __blk_mq_run_hw_queue+0xf0/0x218
 __blk_mq_delay_run_hw_queue+0x8c/0x18c
 blk_mq_run_hw_queue+0x1a4/0x360
 blk_mq_sched_insert_requests+0x130/0x334
 blk_mq_flush_plug_list+0x138/0x234
 blk_flush_plug_list+0x118/0x164
 blk_finish_plug()
 read_pages+0x38c/0x408
 page_cache_ra_unbounded+0x230/0x2f8
 do_sync_mmap_readahead+0x1a4/0x208
 filemap_fault+0x27c/0x8f4
 f2fs_filemap_fault+0x28/0xfc
 __do_fault+0xc4/0x208
 handle_pte_fault+0x290/0xe04
 do_handle_mm_fault+0x52c/0x858
 do_page_fault+0x5dc/0x798
 do_translation_fault+0x40/0x54
 do_mem_abort+0x60/0x134
 el0_da+0x40/0xb8
 el0t_64_sync_handler+0xc4/0xe4
 el0t_64_sync+0x1b4/0x1b8
---
 drivers/ufs/core/ufshcd.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
drivers/ufs/core/ufshcd.c +19 -0
index 08abdd763c51..98b14623317f 100644
@@ -9022,6 +9022,25 @@ static void ufshcd_async_scan(void *data, async_cookie_t cookie)
 static enum scsi_timeout_action ufshcd_eh_timed_out(struct scsi_cmnd *scmd)
 {
 	struct ufs_hba *hba = shost_priv(scmd->device->host);
+	struct scsi_cmnd *cmd2 = scmd;
+
+	WARN_ON_ONCE(!scmd);
+
+	if (is_mcq_enabled(hba)) {
+		struct ufs_hw_queue *hwq =
+			ufshcd_mcq_req_to_hwq(hba, scsi_cmd_to_rq(scmd));
+
+		ufshcd_mcq_poll_cqe_lock(hba, hwq, &cmd2);
+	} else {
+		__ufshcd_poll(hba->host, UFSHCD_POLL_FROM_INTERRUPT_CONTEXT,
+			      &cmd2);
+	}
+	if (cmd2 == NULL) {
+		sdev_printk(KERN_INFO, scmd->device,
+			    "%s: cmd with tag %#x has already been completed\n",
+			    __func__, blk_mq_unique_tag(scsi_cmd_to_rq(scmd)));
+		return SCSI_EH_DONE;
+	}
 
 	if (!hba->system_suspending) {
 		/* Activate the error handler in the SCSI core. */
#14 lore raw

Re: [PATCH 4/4] scsi: ufs: Check for completion from the timeout handler

Quoted 61 lines · Bart Van Assche

External email : Please do not click links or open attachments until you have verified the sender or the content.
If ufshcd_abort() returns SUCCESS for an already completed command
then
that command is completed twice. This results in a crash. Prevent this by
checking whether a command has completed without completion interrupt from
the timeout handler. This CL fixes the following kernel crash:

Unable to handle kernel NULL pointer dereference at virtual address
0000000000000000
Call trace:
 dma_direct_map_sg+0x70/0x274
 scsi_dma_map+0x84/0x124
 ufshcd_queuecommand+0x3fc/0x880
 scsi_queue_rq+0x7d0/0x111c
 blk_mq_dispatch_rq_list+0x440/0xebc
 blk_mq_do_dispatch_sched+0x5a4/0x6b8
 __blk_mq_sched_dispatch_requests+0x150/0x220
 __blk_mq_run_hw_queue+0xf0/0x218
 __blk_mq_delay_run_hw_queue+0x8c/0x18c
 blk_mq_run_hw_queue+0x1a4/0x360
 blk_mq_sched_insert_requests+0x130/0x334
 blk_mq_flush_plug_list+0x138/0x234
 blk_flush_plug_list+0x118/0x164
 blk_finish_plug()
 read_pages+0x38c/0x408
 page_cache_ra_unbounded+0x230/0x2f8
 do_sync_mmap_readahead+0x1a4/0x208
 filemap_fault+0x27c/0x8f4
 f2fs_filemap_fault+0x28/0xfc
 __do_fault+0xc4/0x208
 handle_pte_fault+0x290/0xe04
 do_handle_mm_fault+0x52c/0x858
 do_page_fault+0x5dc/0x798
 do_translation_fault+0x40/0x54
 do_mem_abort+0x60/0x134
 el0_da+0x40/0xb8
 el0t_64_sync_handler+0xc4/0xe4
 el0t_64_sync+0x1b4/0x1b8
---
 drivers/ufs/core/ufshcd.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
drivers/ufs/core/ufshcd.c +3 -0
index 08abdd763c51..98b14623317f 100644
@@ -9022,6 +9022,25 @@ static void ufshcd_async_scan(void *data,
 async_cookie_t cookie)
 static enum scsi_timeout_action ufshcd_eh_timed_out(struct scsi_cmnd
 *scmd)
 {
 struct ufs_hba *hba = shost_priv(scmd->device->host);
+struct scsi_cmnd *cmd2 = scmd;
+
+WARN_ON_ONCE(!scmd);

drivers/ufs/core/ufshcd.c static void ufshcd_async_scan(void *data, · line 9039

Quoted 10 lines
drivers/ufs/core/ufshcd.c +9 -0
+
+if (is_mcq_enabled(hba)) {
+struct ufs_hw_queue *hwq =
+ufshcd_mcq_req_to_hwq(hba, scsi_cmd_to_rq(scmd));
+
+ufshcd_mcq_poll_cqe_lock(hba, hwq, &cmd2);
+} else {
+__ufshcd_poll(hba->host, UFSHCD_POLL_FROM_INTERRUPT_CONTEXT,
+      &cmd2);
 

may need check __ufshcd_poll return value?

Quoted 10 lines
drivers/ufs/core/ufshcd.c +7 -0
+}
+if (cmd2 == NULL) {
+sdev_printk(KERN_INFO, scmd->device,
+    "%s: cmd with tag %#x has already been completed\n",
+    __func__, blk_mq_unique_tag(scsi_cmd_to_rq(scmd)));
+return SCSI_EH_DONE;
+}

if (!hba->system_suspending) {
/* Activate the error handler in the SCSI core. */

author reviewer
#15 lore raw

Re: [PATCH 4/4] scsi: ufs: Check for completion from the timeout handler

Quoted 10 lines · Peter Wang (王信友)

drivers/ufs/core/ufshcd.c static void ufshcd_async_scan(void *data, async_cookie_t cookie) · line 9036

Quoted 8 lines
drivers/ufs/core/ufshcd.c +8 -0
+if (is_mcq_enabled(hba)) {
+struct ufs_hw_queue *hwq =
+ufshcd_mcq_req_to_hwq(hba, scsi_cmd_to_rq(scmd));
+
+ufshcd_mcq_poll_cqe_lock(hba, hwq, &cmd2);
+} else {
+__ufshcd_poll(hba->host, UFSHCD_POLL_FROM_INTERRUPT_CONTEXT,
+      &cmd2);

may need check __ufshcd_poll return value?

I don't think we should do that. __ufshcd_poll() returns a value > 0 if it has completed any SCSI command and may complete commands other than @scmd. In this function we need to know whether or not @scmd has been completed.

Thanks,

Bart.