[RESEND PATCH v1 0/2] Add UFS LINERESET handling

[RESEND PATCH v1 0/2] Add UFS LINERESET handling

[Can Guo]

PA Layer issues a LINERESET to the PHY at the recovery step in the Power
Mode change operation. If it happens during auto or mannual hibern8 enter,
even if hibern8 enter succeeds, UFS power mode shall be set to PWM-G1 mode
and kept in that mode after exit from hibern8, leading to bad performance.
Handle the LINERESET in the eh_work by restoring power mode to HS mode
after all pending reqs and tasks are cleared from doorbell.

Can Guo (2):
  scsi: ufs: Abort tasks before clear them from doorbell
  scsi: ufs: Handle LINERESET indication in err handler

 drivers/scsi/ufs/ufshcd.c | 254 ++++++++++++++++++++++++++++++++--------------
 drivers/scsi/ufs/ufshcd.h |   2 +
 drivers/scsi/ufs/ufshci.h |   1 +
 drivers/scsi/ufs/unipro.h |   3 +
 4 files changed, 184 insertions(+), 76 deletions(-)

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.

[PATCH v1 1/2] scsi: ufs: Abort tasks before clear them from doorbell

[Can Guo]

To recovery non-fatal errors, no full reset is required, err_handler only
clears those pending TRs/TMRs so that scsi layer can re-issue them. In
current err_handler, TRs are directly cleared from UFS host's doorbell but
not aborted from device side. However, according to the UFSHCI JEDEC spec,
the host software shall use UTP Transfer Request List CLear Register to
clear a task from UFS host's doorbell only when a UTP Transfer Request is
expected to not be completed, e.g. when the host software receives a
“FUNCTION COMPLETE” Task Management response which means a Transfer Request
was aborted. To follow the UFSHCI JEDEC spec, in err_handler, aborts one TR
before clearing it from doorbell.

Signed-off-by: Can Guo <cang@codeaurora.org>
---
 drivers/scsi/ufs/ufshcd.c | 143 ++++++++++++++++++++++++++--------------------
 1 file changed, 81 insertions(+), 62 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index b8441ad..000895f 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -235,6 +235,7 @@ static int ufshcd_setup_hba_vreg(struct ufs_hba *hba, bool on);
 static int ufshcd_setup_vreg(struct ufs_hba *hba, bool on);
 static inline int ufshcd_config_vreg_hpm(struct ufs_hba *hba,
 					 struct ufs_vreg *vreg);
+static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag);
 static int ufshcd_wb_buf_flush_enable(struct ufs_hba *hba);
 static int ufshcd_wb_buf_flush_disable(struct ufs_hba *hba);
 static int ufshcd_wb_ctrl(struct ufs_hba *hba, bool enable);
@@ -5657,8 +5658,8 @@ static void ufshcd_err_handler(struct work_struct *work)
 {
 	struct ufs_hba *hba;
 	unsigned long flags;
-	u32 err_xfer = 0;
-	u32 err_tm = 0;
+	bool err_xfer = false;
+	bool err_tm = false;
 	int err = 0;
 	int tag;
 	bool needs_reset = false;
@@ -5734,7 +5735,7 @@ static void ufshcd_err_handler(struct work_struct *work)
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	/* Clear pending transfer requests */
 	for_each_set_bit(tag, &hba->outstanding_reqs, hba->nutrs) {
-		if (ufshcd_clear_cmd(hba, tag)) {
+		if (ufshcd_try_to_abort_task(hba, tag)) {
 			err_xfer = true;
 			goto lock_skip_pending_xfer_clear;
 		}
@@ -6486,7 +6487,7 @@ static void ufshcd_set_req_abort_skip(struct ufs_hba *hba, unsigned long bitmap)
 }
 
 /**
- * ufshcd_abort - abort a specific command
+ * ufshcd_try_to_abort_task - abort a specific task
  * @cmd: SCSI command pointer
  *
  * Abort the pending command in device by sending UFS_ABORT_TASK task management
@@ -6495,6 +6496,80 @@ static void ufshcd_set_req_abort_skip(struct ufs_hba *hba, unsigned long bitmap)
  * issued. To avoid that, first issue UFS_QUERY_TASK to check if the command is
  * really issued and then try to abort it.
  *
+ * Returns zero on success, non-zero on failure
+ */
+static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag)
+{
+	struct ufshcd_lrb *lrbp = &hba->lrb[tag];
+	int err = 0;
+	int poll_cnt;
+	u8 resp = 0xF;
+	u32 reg;
+
+	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
+		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
+				UFS_QUERY_TASK, &resp);
+		if (!err && resp == UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
+			/* cmd pending in the device */
+			dev_err(hba->dev, "%s: cmd pending in the device. tag = %d\n",
+				__func__, tag);
+			break;
+		} else if (!err && resp == UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
+			/*
+			 * cmd not pending in the device, check if it is
+			 * in transition.
+			 */
+			dev_err(hba->dev, "%s: cmd at tag %d not pending in the device.\n",
+				__func__, tag);
+			reg = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
+			if (reg & (1 << tag)) {
+				/* sleep for max. 200us to stabilize */
+				usleep_range(100, 200);
+				continue;
+			}
+			/* command completed already */
+			dev_err(hba->dev, "%s: cmd at tag %d successfully cleared from DB.\n",
+				__func__, tag);
+			goto out;
+		} else {
+			dev_err(hba->dev,
+				"%s: no response from device. tag = %d, err %d\n",
+				__func__, tag, err);
+			if (!err)
+				err = resp; /* service response error */
+			goto out;
+		}
+	}
+
+	if (!poll_cnt) {
+		err = -EBUSY;
+		goto out;
+	}
+
+	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
+			UFS_ABORT_TASK, &resp);
+	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
+		if (!err) {
+			err = resp; /* service response error */
+			dev_err(hba->dev, "%s: issued. tag = %d, err %d\n",
+				__func__, tag, err);
+		}
+		goto out;
+	}
+
+	err = ufshcd_clear_cmd(hba, tag);
+	if (err)
+		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d, err %d\n",
+			__func__, tag, err);
+
+out:
+	return err;
+}
+
+/**
+ * ufshcd_abort - scsi host template eh_abort_handler callback
+ * @cmd: SCSI command pointer
+ *
  * Returns SUCCESS/FAILED
  */
 static int ufshcd_abort(struct scsi_cmnd *cmd)
@@ -6504,8 +6579,6 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 	unsigned long flags;
 	unsigned int tag;
 	int err = 0;
-	int poll_cnt;
-	u8 resp = 0xF;
 	struct ufshcd_lrb *lrbp;
 	u32 reg;
 
@@ -6574,63 +6647,9 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 		goto out;
 	}
 
-	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
-		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
-				UFS_QUERY_TASK, &resp);
-		if (!err && resp == UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
-			/* cmd pending in the device */
-			dev_err(hba->dev, "%s: cmd pending in the device. tag = %d\n",
-				__func__, tag);
-			break;
-		} else if (!err && resp == UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
-			/*
-			 * cmd not pending in the device, check if it is
-			 * in transition.
-			 */
-			dev_err(hba->dev, "%s: cmd at tag %d not pending in the device.\n",
-				__func__, tag);
-			reg = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
-			if (reg & (1 << tag)) {
-				/* sleep for max. 200us to stabilize */
-				usleep_range(100, 200);
-				continue;
-			}
-			/* command completed already */
-			dev_err(hba->dev, "%s: cmd at tag %d successfully cleared from DB.\n",
-				__func__, tag);
-			goto out;
-		} else {
-			dev_err(hba->dev,
-				"%s: no response from device. tag = %d, err %d\n",
-				__func__, tag, err);
-			if (!err)
-				err = resp; /* service response error */
-			goto out;
-		}
-	}
-
-	if (!poll_cnt) {
-		err = -EBUSY;
-		goto out;
-	}
-
-	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
-			UFS_ABORT_TASK, &resp);
-	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
-		if (!err) {
-			err = resp; /* service response error */
-			dev_err(hba->dev, "%s: issued. tag = %d, err %d\n",
-				__func__, tag, err);
-		}
-		goto out;
-	}
-
-	err = ufshcd_clear_cmd(hba, tag);
-	if (err) {
-		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d, err %d\n",
-			__func__, tag, err);
+	err = ufshcd_try_to_abort_task(hba, tag);
+	if (err)
 		goto out;
-	}
 
 	spin_lock_irqsave(host->host_lock, flags);
 	__ufshcd_transfer_req_compl(hba, (1UL << tag));
-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.

[PATCH v1 2/2] scsi: ufs: Handle LINERESET indication in err handler

[Can Guo]

PA Layer issues a LINERESET to the PHY at the recovery step in the Power
Mode change operation. If it happens during auto or mannual hibern8 enter,
even if hibern8 enter succeeds, UFS power mode shall be set to PWM-G1 mode
and kept in that mode after exit from hibern8, leading to bad performance.
Handle the LINERESET in the eh_work by restoring power mode to HS mode
after all pending reqs and tasks are cleared from doorbell.

Signed-off-by: Can Guo <cang@codeaurora.org>
---
 drivers/scsi/ufs/ufshcd.c | 111 ++++++++++++++++++++++++++++++++++++++++------
 drivers/scsi/ufs/ufshcd.h |   2 +
 drivers/scsi/ufs/ufshci.h |   1 +
 drivers/scsi/ufs/unipro.h |   3 ++
 4 files changed, 103 insertions(+), 14 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 000895f..8cc127d 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -143,6 +143,7 @@ enum {
 	UFSHCD_UIC_NL_ERROR = (1 << 3), /* Network layer error */
 	UFSHCD_UIC_TL_ERROR = (1 << 4), /* Transport Layer error */
 	UFSHCD_UIC_DME_ERROR = (1 << 5), /* DME error */
+	UFSHCD_UIC_PA_GENERIC_ERROR = (1 << 6), /* Generic PA error */
 };
 
 #define ufshcd_set_eh_in_progress(h) \
@@ -4066,7 +4067,8 @@ static int ufshcd_change_power_mode(struct ufs_hba *hba,
 	int ret;
 
 	/* if already configured to the requested pwr_mode */
-	if (pwr_mode->gear_rx == hba->pwr_info.gear_rx &&
+	if (!hba->force_pmc &&
+	    pwr_mode->gear_rx == hba->pwr_info.gear_rx &&
 	    pwr_mode->gear_tx == hba->pwr_info.gear_tx &&
 	    pwr_mode->lane_rx == hba->pwr_info.lane_rx &&
 	    pwr_mode->lane_tx == hba->pwr_info.lane_tx &&
@@ -4494,6 +4496,8 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
 	if (ret)
 		goto out;
 
+	/* Clear UECPA once due to LINERESET has happened during LINK_STARTUP */
+	ufshcd_readl(hba, REG_UIC_ERROR_CODE_PHY_ADAPTER_LAYER);
 	ret = ufshcd_make_hba_operational(hba);
 out:
 	if (ret) {
@@ -5650,6 +5654,22 @@ static inline void ufshcd_recover_pm_error(struct ufs_hba *hba)
 }
 #endif
 
+static bool ufshcd_is_pwr_mode_restore_needed(struct ufs_hba *hba)
+{
+	struct ufs_pa_layer_attr *pwr_info = &hba->pwr_info;
+	u32 mode;
+
+	ufshcd_dme_get(hba, UIC_ARG_MIB(PA_PWRMODE), &mode);
+
+	if (pwr_info->pwr_rx != ((mode >> PWRMODE_RX_OFFSET) & PWRMODE_MASK))
+		return true;
+
+	if (pwr_info->pwr_tx != (mode & PWRMODE_MASK))
+		return true;
+
+	return false;
+}
+
 /**
  * ufshcd_err_handler - handle UFS errors that require s/w attention
  * @work: pointer to work structure
@@ -5660,9 +5680,9 @@ static void ufshcd_err_handler(struct work_struct *work)
 	unsigned long flags;
 	bool err_xfer = false;
 	bool err_tm = false;
-	int err = 0;
+	int err = 0, pmc_err;
 	int tag;
-	bool needs_reset = false;
+	bool needs_reset = false, needs_restore = false;
 
 	hba = container_of(work, struct ufs_hba, eh_work);
 
@@ -5710,8 +5730,9 @@ static void ufshcd_err_handler(struct work_struct *work)
 				    UFSHCD_UIC_DL_TCx_REPLAY_ERROR))))
 		needs_reset = true;
 
-	if (hba->saved_err & (INT_FATAL_ERRORS | UIC_ERROR |
-			      UFSHCD_UIC_HIBERN8_MASK)) {
+	if ((hba->saved_err & (INT_FATAL_ERRORS | UFSHCD_UIC_HIBERN8_MASK)) ||
+	    (hba->saved_uic_err &&
+	     (hba->saved_uic_err != UFSHCD_UIC_PA_GENERIC_ERROR))) {
 		bool pr_prdt = !!(hba->saved_err & SYSTEM_BUS_FATAL_ERROR);
 
 		spin_unlock_irqrestore(hba->host->host_lock, flags);
@@ -5729,8 +5750,25 @@ static void ufshcd_err_handler(struct work_struct *work)
 	 * host reset and restore
 	 */
 	if (needs_reset)
-		goto skip_pending_xfer_clear;
+		goto do_reset;
 
+	/*
+	 * If LINERESET was caught, UFS might have been put to PWM mode,
+	 * check if power mode restore is needed.
+	 */
+	if (hba->saved_uic_err & UFSHCD_UIC_PA_GENERIC_ERROR) {
+		hba->saved_uic_err &= ~UFSHCD_UIC_PA_GENERIC_ERROR;
+		if (!hba->saved_uic_err)
+			hba->saved_err &= ~UIC_ERROR;
+		spin_unlock_irqrestore(hba->host->host_lock, flags);
+		if (ufshcd_is_pwr_mode_restore_needed(hba))
+			needs_restore = true;
+		spin_lock_irqsave(hba->host->host_lock, flags);
+		if (!hba->saved_err && !needs_restore)
+			goto skip_err_handling;
+	}
+
+	hba->silence_err_logs = true;
 	/* release lock as clear command might sleep */
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	/* Clear pending transfer requests */
@@ -5754,11 +5792,38 @@ static void ufshcd_err_handler(struct work_struct *work)
 
 	/* Complete the requests that are cleared by s/w */
 	ufshcd_complete_requests(hba);
+	hba->silence_err_logs = false;
 
-	if (err_xfer || err_tm)
+	if (err_xfer || err_tm) {
 		needs_reset = true;
+		goto do_reset;
+	}
 
-skip_pending_xfer_clear:
+	/*
+	 * After all reqs and tasks are cleared from doorbell,
+	 * now it is safe to retore power mode.
+	 */
+	if (needs_restore) {
+		spin_unlock_irqrestore(hba->host->host_lock, flags);
+		/*
+		 * Hold the scaling lock just in case dev cmds
+		 * are sent via bsg and/or sysfs.
+		 */
+		down_write(&hba->clk_scaling_lock);
+		hba->force_pmc = true;
+		pmc_err = ufshcd_config_pwr_mode(hba, &(hba->pwr_info));
+		if (pmc_err) {
+			needs_reset = true;
+			dev_err(hba->dev, "%s: Failed to restore power mode, err = %d\n",
+					__func__, pmc_err);
+		}
+		hba->force_pmc = false;
+		ufshcd_print_pwr_info(hba);
+		up_write(&hba->clk_scaling_lock);
+		spin_lock_irqsave(hba->host->host_lock, flags);
+	}
+
+do_reset:
 	/* Fatal errors need reset */
 	if (needs_reset) {
 		unsigned long max_doorbells = (1UL << hba->nutrs) - 1;
@@ -5814,17 +5879,33 @@ static irqreturn_t ufshcd_update_uic_error(struct ufs_hba *hba)
 	u32 reg;
 	irqreturn_t retval = IRQ_NONE;
 
-	/* PHY layer lane error */
+	/* PHY layer error */
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_PHY_ADAPTER_LAYER);
-	/* Ignore LINERESET indication, as this is not an error */
 	if ((reg & UIC_PHY_ADAPTER_LAYER_ERROR) &&
-	    (reg & UIC_PHY_ADAPTER_LAYER_LANE_ERR_MASK)) {
+	    (reg & UIC_PHY_ADAPTER_LAYER_ERROR_CODE_MASK)) {
+		ufshcd_update_reg_hist(&hba->ufs_stats.pa_err, reg);
 		/*
 		 * To know whether this error is fatal or not, DB timeout
 		 * must be checked but this error is handled separately.
 		 */
-		dev_dbg(hba->dev, "%s: UIC Lane error reported\n", __func__);
-		ufshcd_update_reg_hist(&hba->ufs_stats.pa_err, reg);
+		if (reg & UIC_PHY_ADAPTER_LAYER_LANE_ERR_MASK)
+			dev_dbg(hba->dev, "%s: UIC Lane error reported\n",
+					__func__);
+
+		/* Got a LINERESET indication. */
+		if (reg & UIC_PHY_ADAPTER_LAYER_GENERIC_ERROR) {
+			struct uic_command *cmd = NULL;
+
+			hba->uic_error |= UFSHCD_UIC_PA_GENERIC_ERROR;
+			if (hba->uic_async_done && hba->active_uic_cmd)
+				cmd = hba->active_uic_cmd;
+			/*
+			 * Ignore the LINERESET during power mode change
+			 * operation via DME_SET command.
+			 */
+			if (cmd && (cmd->command == UIC_CMD_DME_SET))
+				hba->uic_error &= ~UFSHCD_UIC_PA_GENERIC_ERROR;
+		}
 		retval |= IRQ_HANDLED;
 	}
 
@@ -5941,7 +6022,9 @@ static irqreturn_t ufshcd_check_errors(struct ufs_hba *hba)
 		hba->saved_uic_err |= hba->uic_error;
 
 		/* dump controller state before resetting */
-		if (hba->saved_err & (INT_FATAL_ERRORS | UIC_ERROR)) {
+		if ((hba->saved_err & (INT_FATAL_ERRORS)) ||
+		    (hba->saved_uic_err &&
+		     (hba->saved_uic_err != UFSHCD_UIC_PA_GENERIC_ERROR))) {
 			dev_err(hba->dev, "%s: saved_err 0x%x saved_uic_err 0x%x\n",
 					__func__, hba->saved_err,
 					hba->saved_uic_err);
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index 618b253..8817103 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -629,6 +629,7 @@ struct ufs_hba_variant_params {
  * @saved_err: sticky error mask
  * @saved_uic_err: sticky UIC error mask
  * @force_reset: flag to force eh_work perform a full reset
+ * @force_pmc: flag to force a power mode change
  * @silence_err_logs: flag to silence error logs
  * @dev_cmd: ufs device management command information
  * @last_dme_cmd_tstamp: time stamp of the last completed DME command
@@ -728,6 +729,7 @@ struct ufs_hba {
 	u32 saved_uic_err;
 	struct ufs_stats ufs_stats;
 	bool force_reset;
+	bool force_pmc;
 	bool silence_err_logs;
 
 	/* Device management request data */
diff --git a/drivers/scsi/ufs/ufshci.h b/drivers/scsi/ufs/ufshci.h
index ba31b09..6795e1f 100644
--- a/drivers/scsi/ufs/ufshci.h
+++ b/drivers/scsi/ufs/ufshci.h
@@ -171,6 +171,7 @@ enum {
 #define UIC_PHY_ADAPTER_LAYER_ERROR			0x80000000
 #define UIC_PHY_ADAPTER_LAYER_ERROR_CODE_MASK		0x1F
 #define UIC_PHY_ADAPTER_LAYER_LANE_ERR_MASK		0xF
+#define UIC_PHY_ADAPTER_LAYER_GENERIC_ERROR		0x10
 
 /* UECDL - Host UIC Error Code Data Link Layer 3Ch */
 #define UIC_DATA_LINK_LAYER_ERROR		0x80000000
diff --git a/drivers/scsi/ufs/unipro.h b/drivers/scsi/ufs/unipro.h
index 4ee6478..f6b52ce 100644
--- a/drivers/scsi/ufs/unipro.h
+++ b/drivers/scsi/ufs/unipro.h
@@ -205,6 +205,9 @@ enum {
 	UNCHANGED	= 7,
 };
 
+#define PWRMODE_MASK		0xF
+#define PWRMODE_RX_OFFSET	4
+
 /* PA TX/RX Frequency Series */
 enum {
 	PA_HS_MODE_A	= 1,
-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.

Re: [PATCH v1 1/2] scsi: ufs: Abort tasks before clear them from doorbell

[Stanley Chu]

On Mon, 2020-08-24 at 19:07 -0700, Can Guo wrote:
> To recovery non-fatal errors, no full reset is required, err_handler only
> clears those pending TRs/TMRs so that scsi layer can re-issue them. In
> current err_handler, TRs are directly cleared from UFS host's doorbell but
> not aborted from device side. However, according to the UFSHCI JEDEC spec,
> the host software shall use UTP Transfer Request List CLear Register to
> clear a task from UFS host's doorbell only when a UTP Transfer Request is
> expected to not be completed, e.g. when the host software receives a
> “FUNCTION COMPLETE” Task Management response which means a Transfer Request
> was aborted. To follow the UFSHCI JEDEC spec, in err_handler, aborts one TR
> before clearing it from doorbell.
> 
> Signed-off-by: Can Guo <cang@codeaurora.org>

Acked-by: Stanley Chu <stanley.chu@mediatek.com>

Re: [PATCH v1 1/2] scsi: ufs: Abort tasks before clear them from doorbell

[Bean Huo]

Hi Can
This patch conflicts and be not in line with this series patches : h
ttps://patchwork.kernel.org/cover/11709279/, which has been applied
into 5.9/scsi-fixes. But they are not apppiled in the 5.9/scsi-queue.

Maybe you should rebase your patch from scsi-fixes branch. or do you
have another better option?

Thanks,
Bean

On Mon, 2020-08-24 at 19:07 -0700, Can Guo wrote:
> To recovery non-fatal errors, no full reset is required, err_handler
> only
> clears those pending TRs/TMRs so that scsi layer can re-issue them.
> In
> current err_handler, TRs are directly cleared from UFS host's
> doorbell but
> not aborted from device side. However, according to the UFSHCI JEDEC
> spec,
> the host software shall use UTP Transfer Request List CLear Register
> to
> clear a task from UFS host's doorbell only when a UTP Transfer
> Request is
> expected to not be completed, e.g. when the host software receives a
> “FUNCTION COMPLETE” Task Management response which means a Transfer
> Request
> was aborted. To follow the UFSHCI JEDEC spec, in err_handler, aborts
> one TR
> before clearing it from doorbell.
> 
> Signed-off-by: Can Guo <cang@codeaurora.org>
> ---
>  drivers/scsi/ufs/ufshcd.c | 143 ++++++++++++++++++++++++++--------
> ------------
>  1 file changed, 81 insertions(+), 62 deletions(-)
> 
> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index b8441ad..000895f 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -235,6 +235,7 @@ static int ufshcd_setup_hba_vreg(struct ufs_hba
> *hba, bool on);
>  static int ufshcd_setup_vreg(struct ufs_hba *hba, bool on);
>  static inline int ufshcd_config_vreg_hpm(struct ufs_hba *hba,
>  					 struct ufs_vreg *vreg);
> +static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag);
>  static int ufshcd_wb_buf_flush_enable(struct ufs_hba *hba);
>  static int ufshcd_wb_buf_flush_disable(struct ufs_hba *hba);
>  static int ufshcd_wb_ctrl(struct ufs_hba *hba, bool enable);
> @@ -5657,8 +5658,8 @@ static void ufshcd_err_handler(struct
> work_struct *work)
>  {
>  	struct ufs_hba *hba;
>  	unsigned long flags;
> -	u32 err_xfer = 0;
> -	u32 err_tm = 0;
> +	bool err_xfer = false;
> +	bool err_tm = false;
>  	int err = 0;
>  	int tag;
>  	bool needs_reset = false;
> @@ -5734,7 +5735,7 @@ static void ufshcd_err_handler(struct
> work_struct *work)
>  	spin_unlock_irqrestore(hba->host->host_lock, flags);
>  	/* Clear pending transfer requests */
>  	for_each_set_bit(tag, &hba->outstanding_reqs, hba->nutrs) {
> -		if (ufshcd_clear_cmd(hba, tag)) {
> +		if (ufshcd_try_to_abort_task(hba, tag)) {
>  			err_xfer = true;
>  			goto lock_skip_pending_xfer_clear;
>  		}
> @@ -6486,7 +6487,7 @@ static void ufshcd_set_req_abort_skip(struct
> ufs_hba *hba, unsigned long bitmap)
>  }
>  
>  /**
> - * ufshcd_abort - abort a specific command
> + * ufshcd_try_to_abort_task - abort a specific task
>   * @cmd: SCSI command pointer
>   *
>   * Abort the pending command in device by sending UFS_ABORT_TASK
> task management
> @@ -6495,6 +6496,80 @@ static void ufshcd_set_req_abort_skip(struct
> ufs_hba *hba, unsigned long bitmap)
>   * issued. To avoid that, first issue UFS_QUERY_TASK to check if the
> command is
>   * really issued and then try to abort it.
>   *
> + * Returns zero on success, non-zero on failure
> + */
> +static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag)
> +{
> +	struct ufshcd_lrb *lrbp = &hba->lrb[tag];
> +	int err = 0;
> +	int poll_cnt;
> +	u8 resp = 0xF;
> +	u32 reg;
> +
> +	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
> +		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp-
> >task_tag,
> +				UFS_QUERY_TASK, &resp);
> +		if (!err && resp ==
> UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
> +			/* cmd pending in the device */
> +			dev_err(hba->dev, "%s: cmd pending in the
> device. tag = %d\n",
> +				__func__, tag);
> +			break;
> +		} else if (!err && resp ==
> UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
> +			/*
> +			 * cmd not pending in the device, check if it
> is
> +			 * in transition.
> +			 */
> +			dev_err(hba->dev, "%s: cmd at tag %d not
> pending in the device.\n",
> +				__func__, tag);
> +			reg = ufshcd_readl(hba,
> REG_UTP_TRANSFER_REQ_DOOR_BELL);
> +			if (reg & (1 << tag)) {
> +				/* sleep for max. 200us to stabilize */
> +				usleep_range(100, 200);
> +				continue;
> +			}
> +			/* command completed already */
> +			dev_err(hba->dev, "%s: cmd at tag %d
> successfully cleared from DB.\n",
> +				__func__, tag);
> +			goto out;
> +		} else {
> +			dev_err(hba->dev,
> +				"%s: no response from device. tag = %d,
> err %d\n",
> +				__func__, tag, err);
> +			if (!err)
> +				err = resp; /* service response error
> */
> +			goto out;
> +		}
> +	}
> +
> +	if (!poll_cnt) {
> +		err = -EBUSY;
> +		goto out;
> +	}
> +
> +	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
> +			UFS_ABORT_TASK, &resp);
> +	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
> +		if (!err) {
> +			err = resp; /* service response error */
> +			dev_err(hba->dev, "%s: issued. tag = %d, err
> %d\n",
> +				__func__, tag, err);
> +		}
> +		goto out;
> +	}
> +
> +	err = ufshcd_clear_cmd(hba, tag);
> +	if (err)
> +		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d,
> err %d\n",
> +			__func__, tag, err);
> +
> +out:
> +	return err;
> +}
> +
> +/**
> + * ufshcd_abort - scsi host template eh_abort_handler callback
> + * @cmd: SCSI command pointer
> + *
>   * Returns SUCCESS/FAILED
>   */
>  static int ufshcd_abort(struct scsi_cmnd *cmd)
> @@ -6504,8 +6579,6 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>  	unsigned long flags;
>  	unsigned int tag;
>  	int err = 0;
> -	int poll_cnt;
> -	u8 resp = 0xF;
>  	struct ufshcd_lrb *lrbp;
>  	u32 reg;
>  
> @@ -6574,63 +6647,9 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>  		goto out;
>  	}
>  
> -	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
> -		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp-
> >task_tag,
> -				UFS_QUERY_TASK, &resp);
> -		if (!err && resp ==
> UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
> -			/* cmd pending in the device */
> -			dev_err(hba->dev, "%s: cmd pending in the
> device. tag = %d\n",
> -				__func__, tag);
> -			break;
> -		} else if (!err && resp ==
> UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
> -			/*
> -			 * cmd not pending in the device, check if it
> is
> -			 * in transition.
> -			 */
> -			dev_err(hba->dev, "%s: cmd at tag %d not
> pending in the device.\n",
> -				__func__, tag);
> -			reg = ufshcd_readl(hba,
> REG_UTP_TRANSFER_REQ_DOOR_BELL);
> -			if (reg & (1 << tag)) {
> -				/* sleep for max. 200us to stabilize */
> -				usleep_range(100, 200);
> -				continue;
> -			}
> -			/* command completed already */
> -			dev_err(hba->dev, "%s: cmd at tag %d
> successfully cleared from DB.\n",
> -				__func__, tag);
> -			goto out;
> -		} else {
> -			dev_err(hba->dev,
> -				"%s: no response from device. tag = %d,
> err %d\n",
> -				__func__, tag, err);
> -			if (!err)
> -				err = resp; /* service response error
> */
> -			goto out;
> -		}
> -	}
> -
> -	if (!poll_cnt) {
> -		err = -EBUSY;
> -		goto out;
> -	}
> -
> -	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
> -			UFS_ABORT_TASK, &resp);
> -	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
> -		if (!err) {
> -			err = resp; /* service response error */
> -			dev_err(hba->dev, "%s: issued. tag = %d, err
> %d\n",
> -				__func__, tag, err);
> -		}
> -		goto out;
> -	}
> -
> -	err = ufshcd_clear_cmd(hba, tag);
> -	if (err) {
> -		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d,
> err %d\n",
> -			__func__, tag, err);
> +	err = ufshcd_try_to_abort_task(hba, tag);
> +	if (err)
>  		goto out;
> -	}
>  
>  	spin_lock_irqsave(host->host_lock, flags);
>  	__ufshcd_transfer_req_compl(hba, (1UL << tag));

Re: [PATCH v1 1/2] scsi: ufs: Abort tasks before clear them from doorbell

[Can Guo]

Hi Bean,

On 2020-08-31 02:11, Bean Huo wrote:
> Hi Can
> This patch conflicts and be not in line with this series patches : h
> ttps://patchwork.kernel.org/cover/11709279/, which has been applied
> into 5.9/scsi-fixes. But they are not apppiled in the 5.9/scsi-queue.
> 
> Maybe you should rebase your patch from scsi-fixes branch. or do you
> have another better option?
> 
> Thanks,
> Bean
> 

I am pushing this change due to LINERSET handling needs it and LINERESET
handling is added based on my previous changes to err_handler, which are
picked by 5.10/scsi-queue. So the two are applied for 5.10/scsi-queue 
only.
Any conflicts you see on 5.10/scsi-queue?

Thanks,

Can Guo.

> On Mon, 2020-08-24 at 19:07 -0700, Can Guo wrote:
>> To recovery non-fatal errors, no full reset is required, err_handler
>> only
>> clears those pending TRs/TMRs so that scsi layer can re-issue them.
>> In
>> current err_handler, TRs are directly cleared from UFS host's
>> doorbell but
>> not aborted from device side. However, according to the UFSHCI JEDEC
>> spec,
>> the host software shall use UTP Transfer Request List CLear Register
>> to
>> clear a task from UFS host's doorbell only when a UTP Transfer
>> Request is
>> expected to not be completed, e.g. when the host software receives a
>> “FUNCTION COMPLETE” Task Management response which means a Transfer
>> Request
>> was aborted. To follow the UFSHCI JEDEC spec, in err_handler, aborts
>> one TR
>> before clearing it from doorbell.
>> 
>> Signed-off-by: Can Guo <cang@codeaurora.org>
>> ---
>>  drivers/scsi/ufs/ufshcd.c | 143 ++++++++++++++++++++++++++--------
>> ------------
>>  1 file changed, 81 insertions(+), 62 deletions(-)
>> 
>> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
>> index b8441ad..000895f 100644
>> --- a/drivers/scsi/ufs/ufshcd.c
>> +++ b/drivers/scsi/ufs/ufshcd.c
>> @@ -235,6 +235,7 @@ static int ufshcd_setup_hba_vreg(struct ufs_hba
>> *hba, bool on);
>>  static int ufshcd_setup_vreg(struct ufs_hba *hba, bool on);
>>  static inline int ufshcd_config_vreg_hpm(struct ufs_hba *hba,
>>  					 struct ufs_vreg *vreg);
>> +static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag);
>>  static int ufshcd_wb_buf_flush_enable(struct ufs_hba *hba);
>>  static int ufshcd_wb_buf_flush_disable(struct ufs_hba *hba);
>>  static int ufshcd_wb_ctrl(struct ufs_hba *hba, bool enable);
>> @@ -5657,8 +5658,8 @@ static void ufshcd_err_handler(struct
>> work_struct *work)
>>  {
>>  	struct ufs_hba *hba;
>>  	unsigned long flags;
>> -	u32 err_xfer = 0;
>> -	u32 err_tm = 0;
>> +	bool err_xfer = false;
>> +	bool err_tm = false;
>>  	int err = 0;
>>  	int tag;
>>  	bool needs_reset = false;
>> @@ -5734,7 +5735,7 @@ static void ufshcd_err_handler(struct
>> work_struct *work)
>>  	spin_unlock_irqrestore(hba->host->host_lock, flags);
>>  	/* Clear pending transfer requests */
>>  	for_each_set_bit(tag, &hba->outstanding_reqs, hba->nutrs) {
>> -		if (ufshcd_clear_cmd(hba, tag)) {
>> +		if (ufshcd_try_to_abort_task(hba, tag)) {
>>  			err_xfer = true;
>>  			goto lock_skip_pending_xfer_clear;
>>  		}
>> @@ -6486,7 +6487,7 @@ static void ufshcd_set_req_abort_skip(struct
>> ufs_hba *hba, unsigned long bitmap)
>>  }
>> 
>>  /**
>> - * ufshcd_abort - abort a specific command
>> + * ufshcd_try_to_abort_task - abort a specific task
>>   * @cmd: SCSI command pointer
>>   *
>>   * Abort the pending command in device by sending UFS_ABORT_TASK
>> task management
>> @@ -6495,6 +6496,80 @@ static void ufshcd_set_req_abort_skip(struct
>> ufs_hba *hba, unsigned long bitmap)
>>   * issued. To avoid that, first issue UFS_QUERY_TASK to check if the
>> command is
>>   * really issued and then try to abort it.
>>   *
>> + * Returns zero on success, non-zero on failure
>> + */
>> +static int ufshcd_try_to_abort_task(struct ufs_hba *hba, int tag)
>> +{
>> +	struct ufshcd_lrb *lrbp = &hba->lrb[tag];
>> +	int err = 0;
>> +	int poll_cnt;
>> +	u8 resp = 0xF;
>> +	u32 reg;
>> +
>> +	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
>> +		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp-
>> >task_tag,
>> +				UFS_QUERY_TASK, &resp);
>> +		if (!err && resp ==
>> UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
>> +			/* cmd pending in the device */
>> +			dev_err(hba->dev, "%s: cmd pending in the
>> device. tag = %d\n",
>> +				__func__, tag);
>> +			break;
>> +		} else if (!err && resp ==
>> UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
>> +			/*
>> +			 * cmd not pending in the device, check if it
>> is
>> +			 * in transition.
>> +			 */
>> +			dev_err(hba->dev, "%s: cmd at tag %d not
>> pending in the device.\n",
>> +				__func__, tag);
>> +			reg = ufshcd_readl(hba,
>> REG_UTP_TRANSFER_REQ_DOOR_BELL);
>> +			if (reg & (1 << tag)) {
>> +				/* sleep for max. 200us to stabilize */
>> +				usleep_range(100, 200);
>> +				continue;
>> +			}
>> +			/* command completed already */
>> +			dev_err(hba->dev, "%s: cmd at tag %d
>> successfully cleared from DB.\n",
>> +				__func__, tag);
>> +			goto out;
>> +		} else {
>> +			dev_err(hba->dev,
>> +				"%s: no response from device. tag = %d,
>> err %d\n",
>> +				__func__, tag, err);
>> +			if (!err)
>> +				err = resp; /* service response error
>> */
>> +			goto out;
>> +		}
>> +	}
>> +
>> +	if (!poll_cnt) {
>> +		err = -EBUSY;
>> +		goto out;
>> +	}
>> +
>> +	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
>> +			UFS_ABORT_TASK, &resp);
>> +	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
>> +		if (!err) {
>> +			err = resp; /* service response error */
>> +			dev_err(hba->dev, "%s: issued. tag = %d, err
>> %d\n",
>> +				__func__, tag, err);
>> +		}
>> +		goto out;
>> +	}
>> +
>> +	err = ufshcd_clear_cmd(hba, tag);
>> +	if (err)
>> +		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d,
>> err %d\n",
>> +			__func__, tag, err);
>> +
>> +out:
>> +	return err;
>> +}
>> +
>> +/**
>> + * ufshcd_abort - scsi host template eh_abort_handler callback
>> + * @cmd: SCSI command pointer
>> + *
>>   * Returns SUCCESS/FAILED
>>   */
>>  static int ufshcd_abort(struct scsi_cmnd *cmd)
>> @@ -6504,8 +6579,6 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>>  	unsigned long flags;
>>  	unsigned int tag;
>>  	int err = 0;
>> -	int poll_cnt;
>> -	u8 resp = 0xF;
>>  	struct ufshcd_lrb *lrbp;
>>  	u32 reg;
>> 
>> @@ -6574,63 +6647,9 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>>  		goto out;
>>  	}
>> 
>> -	for (poll_cnt = 100; poll_cnt; poll_cnt--) {
>> -		err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp-
>> >task_tag,
>> -				UFS_QUERY_TASK, &resp);
>> -		if (!err && resp ==
>> UPIU_TASK_MANAGEMENT_FUNC_SUCCEEDED) {
>> -			/* cmd pending in the device */
>> -			dev_err(hba->dev, "%s: cmd pending in the
>> device. tag = %d\n",
>> -				__func__, tag);
>> -			break;
>> -		} else if (!err && resp ==
>> UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
>> -			/*
>> -			 * cmd not pending in the device, check if it
>> is
>> -			 * in transition.
>> -			 */
>> -			dev_err(hba->dev, "%s: cmd at tag %d not
>> pending in the device.\n",
>> -				__func__, tag);
>> -			reg = ufshcd_readl(hba,
>> REG_UTP_TRANSFER_REQ_DOOR_BELL);
>> -			if (reg & (1 << tag)) {
>> -				/* sleep for max. 200us to stabilize */
>> -				usleep_range(100, 200);
>> -				continue;
>> -			}
>> -			/* command completed already */
>> -			dev_err(hba->dev, "%s: cmd at tag %d
>> successfully cleared from DB.\n",
>> -				__func__, tag);
>> -			goto out;
>> -		} else {
>> -			dev_err(hba->dev,
>> -				"%s: no response from device. tag = %d,
>> err %d\n",
>> -				__func__, tag, err);
>> -			if (!err)
>> -				err = resp; /* service response error
>> */
>> -			goto out;
>> -		}
>> -	}
>> -
>> -	if (!poll_cnt) {
>> -		err = -EBUSY;
>> -		goto out;
>> -	}
>> -
>> -	err = ufshcd_issue_tm_cmd(hba, lrbp->lun, lrbp->task_tag,
>> -			UFS_ABORT_TASK, &resp);
>> -	if (err || resp != UPIU_TASK_MANAGEMENT_FUNC_COMPL) {
>> -		if (!err) {
>> -			err = resp; /* service response error */
>> -			dev_err(hba->dev, "%s: issued. tag = %d, err
>> %d\n",
>> -				__func__, tag, err);
>> -		}
>> -		goto out;
>> -	}
>> -
>> -	err = ufshcd_clear_cmd(hba, tag);
>> -	if (err) {
>> -		dev_err(hba->dev, "%s: Failed clearing cmd at tag %d,
>> err %d\n",
>> -			__func__, tag, err);
>> +	err = ufshcd_try_to_abort_task(hba, tag);
>> +	if (err)
>>  		goto out;
>> -	}
>> 
>>  	spin_lock_irqsave(host->host_lock, flags);
>>  	__ufshcd_transfer_req_compl(hba, (1UL << tag));

Re: [PATCH v1 1/2] scsi: ufs: Abort tasks before clear them from doorbell

[Bean Huo]

On Mon, 2020-08-31 at 09:20 +0800, Can Guo wrote:
> On 2020-08-31 02:11, Bean Huo wrote:
> > Hi Can
> > This patch conflicts and be not in line with this series patches :
> > h
> > ttps://patchwork.kernel.org/cover/11709279/, which has been applied
> > into 5.9/scsi-fixes. But they are not apppiled in the 5.9/scsi-
> > queue.
> > 
> > Maybe you should rebase your patch from scsi-fixes branch. or do
> > you
> > have another better option?
> > 
> > Thanks,
> > Bean
> > 
> 
> I am pushing this change due to LINERSET handling needs it and
> LINERESET
> handling is added based on my previous changes to err_handler, which
> are
> picked by 5.10/scsi-queue. So the two are applied for 5.10/scsi-
> queue 
> only.
> Any conflicts you see on 5.10/scsi-queue?

Hi Can
I meant scsi-fixes branch. no conflict with scsi-queue branch. If the
the changes in the scsi-fixes branch  will never be merged to scsi-
queue branch.It is fine.

Thanks,
Bean

> 
> Thanks,
> 
> Can Guo.

Re: [RESEND PATCH v1 0/2] Add UFS LINERESET handling

[Martin K. Petersen]

On Mon, 24 Aug 2020 19:07:04 -0700, Can Guo wrote:

> PA Layer issues a LINERESET to the PHY at the recovery step in the Power
> Mode change operation. If it happens during auto or mannual hibern8 enter,
> even if hibern8 enter succeeds, UFS power mode shall be set to PWM-G1 mode
> and kept in that mode after exit from hibern8, leading to bad performance.
> Handle the LINERESET in the eh_work by restoring power mode to HS mode
> after all pending reqs and tasks are cleared from doorbell.
> 
> [...]

Applied to 5.10/scsi-queue, thanks!

[1/2] scsi: ufs: Abort tasks before clearing them from doorbell
      https://git.kernel.org/mkp/scsi/c/307348f6ab14
[2/2] scsi: ufs: Handle LINERESET indication in err handler
      https://git.kernel.org/mkp/scsi/c/2355b66ed20c

-- 
Martin K. Petersen	Oracle Linux Engineering