[PATCH v2 0/4] scsi: ufs: Provide fatal and auto-hibern8 error history

[PATCH v2 0/4] scsi: ufs: Provide fatal and auto-hibern8 error history

[Stanley Chu]

This patchset provides more information of fatal errros and auto-hibern8 errors
to improve debugging by keeping their error history as completed as possible.

Thanks Avri so much for prompt reviewing patchset v1.

I would like to post v2 to add one more patch "scsi: ufs: Add history of fatal events"
to add history for "non-interrupt-based" errors as well, for example,

- Link startup fail
- Suspend fail
- Resume fail
- Task or request abort event

Changes in v2:
- Add new patch "scsi: ufs: Add history of fatal events".

Stanley Chu (4):
  scsi: ufs: Change names related to error history
  scsi: ufs: Add fatal and auto-hibern8 error history
  scsi: ufs: Do not reset error history during host reset
  scsi: ufs: Add history of fatal events

 drivers/scsi/ufs/ufshcd.c | 87 +++++++++++++++++++++++----------------
 drivers/scsi/ufs/ufshcd.h | 38 ++++++++++++-----
 2 files changed, 80 insertions(+), 45 deletions(-)

-- 
2.18.0

[PATCH v2 0/4] scsi: ufs: Provide fatal and auto-hibern8 error history

[Stanley Chu]

This patchset provides more information of fatal errros and auto-hibern8 errors
to improve debugging by keeping their error history as completed as possible.

Thanks Avri so much for prompt reviewing patchset v1.

I would like to post v2 to add one more patch "scsi: ufs: Add history of fatal events"
to add history for "non-interrupt-based" errors as well, for example,

- Link startup fail
- Suspend fail
- Resume fail
- Task or request abort event

Changes in v2:
- Add new patch "scsi: ufs: Add history of fatal events".

Stanley Chu (4):
  scsi: ufs: Change names related to error history
  scsi: ufs: Add fatal and auto-hibern8 error history
  scsi: ufs: Do not reset error history during host reset
  scsi: ufs: Add history of fatal events

 drivers/scsi/ufs/ufshcd.c | 87 +++++++++++++++++++++++----------------
 drivers/scsi/ufs/ufshcd.h | 38 ++++++++++++-----
 2 files changed, 80 insertions(+), 45 deletions(-)

-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 1/4] scsi: ufs: Change names related to error history

[Stanley Chu]

Remove "uic" term in below error history functions and structures
for more general usages,

struct ufs_uic_err_reg_hist;
void ufshcd_update_uic_reg_hist(struct ufs_uic_err_reg_hist *reg_hist,
	u32 reg);
void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
	struct ufs_uic)err_reg_hist *err_hist, char *err_name);

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 39 ++++++++++++++++++++-------------------
 drivers/scsi/ufs/ufshcd.h | 20 ++++++++++----------
 2 files changed, 30 insertions(+), 29 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index a208589426b1..eb062aba0d21 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -390,14 +390,15 @@ static void ufshcd_print_clk_freqs(struct ufs_hba *hba)
 	}
 }
 
-static void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
-		struct ufs_uic_err_reg_hist *err_hist, char *err_name)
+static void ufshcd_print_err_hist(struct ufs_hba *hba,
+				  struct ufs_err_reg_hist *err_hist,
+				  char *err_name)
 {
 	int i;
 	bool found = false;
 
-	for (i = 0; i < UIC_ERR_REG_HIST_LENGTH; i++) {
-		int p = (i + err_hist->pos) % UIC_ERR_REG_HIST_LENGTH;
+	for (i = 0; i < UFS_ERR_REG_HIST_LENGTH; i++) {
+		int p = (i + err_hist->pos) % UFS_ERR_REG_HIST_LENGTH;
 
 		if (err_hist->reg[p] == 0)
 			continue;
@@ -407,7 +408,7 @@ static void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
 	}
 
 	if (!found)
-		dev_err(hba->dev, "No record of %s uic errors\n", err_name);
+		dev_err(hba->dev, "No record of %s errors\n", err_name);
 }
 
 static void ufshcd_print_host_regs(struct ufs_hba *hba)
@@ -423,11 +424,11 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 		ktime_to_us(hba->ufs_stats.last_hibern8_exit_tstamp),
 		hba->ufs_stats.hibern8_exit_cnt);
 
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.pa_err, "pa_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.dl_err, "dl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.pa_err, "pa_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.dl_err, "dl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -5346,12 +5347,12 @@ static void ufshcd_err_handler(struct work_struct *work)
 	pm_runtime_put_sync(hba->dev);
 }
 
-static void ufshcd_update_uic_reg_hist(struct ufs_uic_err_reg_hist *reg_hist,
-		u32 reg)
+static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
+				   u32 reg)
 {
 	reg_hist->reg[reg_hist->pos] = reg;
 	reg_hist->tstamp[reg_hist->pos] = ktime_get();
-	reg_hist->pos = (reg_hist->pos + 1) % UIC_ERR_REG_HIST_LENGTH;
+	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
 }
 
 /**
@@ -5372,13 +5373,13 @@ static void ufshcd_update_uic_error(struct ufs_hba *hba)
 		 * must be checked but this error is handled separately.
 		 */
 		dev_dbg(hba->dev, "%s: UIC Lane error reported\n", __func__);
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.pa_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.pa_err, reg);
 	}
 
 	/* PA_INIT_ERROR is fatal and needs UIC reset */
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_DATA_LINK_LAYER);
 	if (reg)
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.dl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.dl_err, reg);
 
 	if (reg & UIC_DATA_LINK_LAYER_ERROR_PA_INIT)
 		hba->uic_error |= UFSHCD_UIC_DL_PA_INIT_ERROR;
@@ -5394,19 +5395,19 @@ static void ufshcd_update_uic_error(struct ufs_hba *hba)
 	/* UIC NL/TL/DME errors needs software retry */
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_NETWORK_LAYER);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.nl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.nl_err, reg);
 		hba->uic_error |= UFSHCD_UIC_NL_ERROR;
 	}
 
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_TRANSPORT_LAYER);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.tl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.tl_err, reg);
 		hba->uic_error |= UFSHCD_UIC_TL_ERROR;
 	}
 
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_DME);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.dme_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.dme_err, reg);
 		hba->uic_error |= UFSHCD_UIC_DME_ERROR;
 	}
 
@@ -6682,7 +6683,7 @@ static void ufshcd_tune_unipro_params(struct ufs_hba *hba)
 
 static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 {
-	int err_reg_hist_size = sizeof(struct ufs_uic_err_reg_hist);
+	int err_reg_hist_size = sizeof(struct ufs_err_reg_hist);
 
 	hba->ufs_stats.hibern8_exit_cnt = 0;
 	hba->ufs_stats.last_hibern8_exit_tstamp = ktime_set(0, 0);
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index 994d73d03207..dcc61f857c38 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -412,17 +412,17 @@ struct ufs_init_prefetch {
 	u32 icc_level;
 };
 
-#define UIC_ERR_REG_HIST_LENGTH 8
+#define UFS_ERR_REG_HIST_LENGTH 8
 /**
- * struct ufs_uic_err_reg_hist - keeps history of uic errors
+ * struct ufs_err_reg_hist - keeps history of uic errors
  * @pos: index to indicate cyclic buffer position
  * @reg: cyclic buffer for registers value
  * @tstamp: cyclic buffer for time stamp
  */
-struct ufs_uic_err_reg_hist {
+struct ufs_err_reg_hist {
 	int pos;
-	u32 reg[UIC_ERR_REG_HIST_LENGTH];
-	ktime_t tstamp[UIC_ERR_REG_HIST_LENGTH];
+	u32 reg[UFS_ERR_REG_HIST_LENGTH];
+	ktime_t tstamp[UFS_ERR_REG_HIST_LENGTH];
 };
 
 /**
@@ -440,11 +440,11 @@ struct ufs_uic_err_reg_hist {
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
 	ktime_t last_hibern8_exit_tstamp;
-	struct ufs_uic_err_reg_hist pa_err;
-	struct ufs_uic_err_reg_hist dl_err;
-	struct ufs_uic_err_reg_hist nl_err;
-	struct ufs_uic_err_reg_hist tl_err;
-	struct ufs_uic_err_reg_hist dme_err;
+	struct ufs_err_reg_hist pa_err;
+	struct ufs_err_reg_hist dl_err;
+	struct ufs_err_reg_hist nl_err;
+	struct ufs_err_reg_hist tl_err;
+	struct ufs_err_reg_hist dme_err;
 };
 
 /**
-- 
2.18.0

[PATCH v2 1/4] scsi: ufs: Change names related to error history

[Stanley Chu]

Remove "uic" term in below error history functions and structures
for more general usages,

struct ufs_uic_err_reg_hist;
void ufshcd_update_uic_reg_hist(struct ufs_uic_err_reg_hist *reg_hist,
	u32 reg);
void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
	struct ufs_uic)err_reg_hist *err_hist, char *err_name);

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 39 ++++++++++++++++++++-------------------
 drivers/scsi/ufs/ufshcd.h | 20 ++++++++++----------
 2 files changed, 30 insertions(+), 29 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index a208589426b1..eb062aba0d21 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -390,14 +390,15 @@ static void ufshcd_print_clk_freqs(struct ufs_hba *hba)
 	}
 }
 
-static void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
-		struct ufs_uic_err_reg_hist *err_hist, char *err_name)
+static void ufshcd_print_err_hist(struct ufs_hba *hba,
+				  struct ufs_err_reg_hist *err_hist,
+				  char *err_name)
 {
 	int i;
 	bool found = false;
 
-	for (i = 0; i < UIC_ERR_REG_HIST_LENGTH; i++) {
-		int p = (i + err_hist->pos) % UIC_ERR_REG_HIST_LENGTH;
+	for (i = 0; i < UFS_ERR_REG_HIST_LENGTH; i++) {
+		int p = (i + err_hist->pos) % UFS_ERR_REG_HIST_LENGTH;
 
 		if (err_hist->reg[p] == 0)
 			continue;
@@ -407,7 +408,7 @@ static void ufshcd_print_uic_err_hist(struct ufs_hba *hba,
 	}
 
 	if (!found)
-		dev_err(hba->dev, "No record of %s uic errors\n", err_name);
+		dev_err(hba->dev, "No record of %s errors\n", err_name);
 }
 
 static void ufshcd_print_host_regs(struct ufs_hba *hba)
@@ -423,11 +424,11 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 		ktime_to_us(hba->ufs_stats.last_hibern8_exit_tstamp),
 		hba->ufs_stats.hibern8_exit_cnt);
 
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.pa_err, "pa_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.dl_err, "dl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
-	ufshcd_print_uic_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.pa_err, "pa_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.dl_err, "dl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -5346,12 +5347,12 @@ static void ufshcd_err_handler(struct work_struct *work)
 	pm_runtime_put_sync(hba->dev);
 }
 
-static void ufshcd_update_uic_reg_hist(struct ufs_uic_err_reg_hist *reg_hist,
-		u32 reg)
+static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
+				   u32 reg)
 {
 	reg_hist->reg[reg_hist->pos] = reg;
 	reg_hist->tstamp[reg_hist->pos] = ktime_get();
-	reg_hist->pos = (reg_hist->pos + 1) % UIC_ERR_REG_HIST_LENGTH;
+	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
 }
 
 /**
@@ -5372,13 +5373,13 @@ static void ufshcd_update_uic_error(struct ufs_hba *hba)
 		 * must be checked but this error is handled separately.
 		 */
 		dev_dbg(hba->dev, "%s: UIC Lane error reported\n", __func__);
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.pa_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.pa_err, reg);
 	}
 
 	/* PA_INIT_ERROR is fatal and needs UIC reset */
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_DATA_LINK_LAYER);
 	if (reg)
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.dl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.dl_err, reg);
 
 	if (reg & UIC_DATA_LINK_LAYER_ERROR_PA_INIT)
 		hba->uic_error |= UFSHCD_UIC_DL_PA_INIT_ERROR;
@@ -5394,19 +5395,19 @@ static void ufshcd_update_uic_error(struct ufs_hba *hba)
 	/* UIC NL/TL/DME errors needs software retry */
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_NETWORK_LAYER);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.nl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.nl_err, reg);
 		hba->uic_error |= UFSHCD_UIC_NL_ERROR;
 	}
 
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_TRANSPORT_LAYER);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.tl_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.tl_err, reg);
 		hba->uic_error |= UFSHCD_UIC_TL_ERROR;
 	}
 
 	reg = ufshcd_readl(hba, REG_UIC_ERROR_CODE_DME);
 	if (reg) {
-		ufshcd_update_uic_reg_hist(&hba->ufs_stats.dme_err, reg);
+		ufshcd_update_reg_hist(&hba->ufs_stats.dme_err, reg);
 		hba->uic_error |= UFSHCD_UIC_DME_ERROR;
 	}
 
@@ -6682,7 +6683,7 @@ static void ufshcd_tune_unipro_params(struct ufs_hba *hba)
 
 static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 {
-	int err_reg_hist_size = sizeof(struct ufs_uic_err_reg_hist);
+	int err_reg_hist_size = sizeof(struct ufs_err_reg_hist);
 
 	hba->ufs_stats.hibern8_exit_cnt = 0;
 	hba->ufs_stats.last_hibern8_exit_tstamp = ktime_set(0, 0);
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index 994d73d03207..dcc61f857c38 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -412,17 +412,17 @@ struct ufs_init_prefetch {
 	u32 icc_level;
 };
 
-#define UIC_ERR_REG_HIST_LENGTH 8
+#define UFS_ERR_REG_HIST_LENGTH 8
 /**
- * struct ufs_uic_err_reg_hist - keeps history of uic errors
+ * struct ufs_err_reg_hist - keeps history of uic errors
  * @pos: index to indicate cyclic buffer position
  * @reg: cyclic buffer for registers value
  * @tstamp: cyclic buffer for time stamp
  */
-struct ufs_uic_err_reg_hist {
+struct ufs_err_reg_hist {
 	int pos;
-	u32 reg[UIC_ERR_REG_HIST_LENGTH];
-	ktime_t tstamp[UIC_ERR_REG_HIST_LENGTH];
+	u32 reg[UFS_ERR_REG_HIST_LENGTH];
+	ktime_t tstamp[UFS_ERR_REG_HIST_LENGTH];
 };
 
 /**
@@ -440,11 +440,11 @@ struct ufs_uic_err_reg_hist {
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
 	ktime_t last_hibern8_exit_tstamp;
-	struct ufs_uic_err_reg_hist pa_err;
-	struct ufs_uic_err_reg_hist dl_err;
-	struct ufs_uic_err_reg_hist nl_err;
-	struct ufs_uic_err_reg_hist tl_err;
-	struct ufs_uic_err_reg_hist dme_err;
+	struct ufs_err_reg_hist pa_err;
+	struct ufs_err_reg_hist dl_err;
+	struct ufs_err_reg_hist nl_err;
+	struct ufs_err_reg_hist tl_err;
+	struct ufs_err_reg_hist dme_err;
 };
 
 /**
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 2/4] scsi: ufs: Add fatal and auto-hibern8 error history

[Stanley Chu]

Provide more information of fatal errros and auto-hibern8 errors
to improve debugging by extending extend existed UFS error history
framework.

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 11 ++++++++++-
 drivers/scsi/ufs/ufshcd.h | 10 +++++++++-
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index eb062aba0d21..b8b874311509 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -429,6 +429,9 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
+			      "auto_hibern8_err");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -5440,8 +5443,10 @@ static void ufshcd_check_errors(struct ufs_hba *hba)
 {
 	bool queue_eh_work = false;
 
-	if (hba->errors & INT_FATAL_ERRORS)
+	if (hba->errors & INT_FATAL_ERRORS) {
+		ufshcd_update_reg_hist(&hba->ufs_stats.fatal_err, hba->errors);
 		queue_eh_work = true;
+	}
 
 	if (hba->errors & UIC_ERROR) {
 		hba->uic_error = 0;
@@ -5456,6 +5461,8 @@ static void ufshcd_check_errors(struct ufs_hba *hba)
 			__func__, (hba->errors & UIC_HIBERNATE_ENTER) ?
 			"Enter" : "Exit",
 			hba->errors, ufshcd_get_upmcrs(hba));
+		ufshcd_update_reg_hist(&hba->ufs_stats.auto_hibern8_err,
+				       hba->errors);
 		queue_eh_work = true;
 	}
 
@@ -6693,6 +6700,8 @@ static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 	memset(&hba->ufs_stats.nl_err, 0, err_reg_hist_size);
 	memset(&hba->ufs_stats.tl_err, 0, err_reg_hist_size);
 	memset(&hba->ufs_stats.dme_err, 0, err_reg_hist_size);
+	memset(&hba->ufs_stats.fatal_err, 0, err_reg_hist_size);
+	memset(&hba->ufs_stats.auto_hibern8_err, 0, err_reg_hist_size);
 
 	hba->req_abort_count = 0;
 }
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index dcc61f857c38..c6ec5c749ceb 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -414,7 +414,7 @@ struct ufs_init_prefetch {
 
 #define UFS_ERR_REG_HIST_LENGTH 8
 /**
- * struct ufs_err_reg_hist - keeps history of uic errors
+ * struct ufs_err_reg_hist - keeps history of errors
  * @pos: index to indicate cyclic buffer position
  * @reg: cyclic buffer for registers value
  * @tstamp: cyclic buffer for time stamp
@@ -436,15 +436,23 @@ struct ufs_err_reg_hist {
  * @nl_err: tracks nl-uic errors
  * @tl_err: tracks tl-uic errors
  * @dme_err: tracks dme errors
+ * @fatal_err: tracks fatal errors
+ * @auto_hibern8_err: tracks auto-hibernate errors
  */
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
 	ktime_t last_hibern8_exit_tstamp;
+
+	/* uic specific errors */
 	struct ufs_err_reg_hist pa_err;
 	struct ufs_err_reg_hist dl_err;
 	struct ufs_err_reg_hist nl_err;
 	struct ufs_err_reg_hist tl_err;
 	struct ufs_err_reg_hist dme_err;
+
+	/* fatal errors */
+	struct ufs_err_reg_hist fatal_err;
+	struct ufs_err_reg_hist auto_hibern8_err;
 };
 
 /**
-- 
2.18.0

[PATCH v2 2/4] scsi: ufs: Add fatal and auto-hibern8 error history

[Stanley Chu]

Provide more information of fatal errros and auto-hibern8 errors
to improve debugging by extending extend existed UFS error history
framework.

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 11 ++++++++++-
 drivers/scsi/ufs/ufshcd.h | 10 +++++++++-
 2 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index eb062aba0d21..b8b874311509 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -429,6 +429,9 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.nl_err, "nl_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.tl_err, "tl_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.dme_err, "dme_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
+			      "auto_hibern8_err");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -5440,8 +5443,10 @@ static void ufshcd_check_errors(struct ufs_hba *hba)
 {
 	bool queue_eh_work = false;
 
-	if (hba->errors & INT_FATAL_ERRORS)
+	if (hba->errors & INT_FATAL_ERRORS) {
+		ufshcd_update_reg_hist(&hba->ufs_stats.fatal_err, hba->errors);
 		queue_eh_work = true;
+	}
 
 	if (hba->errors & UIC_ERROR) {
 		hba->uic_error = 0;
@@ -5456,6 +5461,8 @@ static void ufshcd_check_errors(struct ufs_hba *hba)
 			__func__, (hba->errors & UIC_HIBERNATE_ENTER) ?
 			"Enter" : "Exit",
 			hba->errors, ufshcd_get_upmcrs(hba));
+		ufshcd_update_reg_hist(&hba->ufs_stats.auto_hibern8_err,
+				       hba->errors);
 		queue_eh_work = true;
 	}
 
@@ -6693,6 +6700,8 @@ static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 	memset(&hba->ufs_stats.nl_err, 0, err_reg_hist_size);
 	memset(&hba->ufs_stats.tl_err, 0, err_reg_hist_size);
 	memset(&hba->ufs_stats.dme_err, 0, err_reg_hist_size);
+	memset(&hba->ufs_stats.fatal_err, 0, err_reg_hist_size);
+	memset(&hba->ufs_stats.auto_hibern8_err, 0, err_reg_hist_size);
 
 	hba->req_abort_count = 0;
 }
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index dcc61f857c38..c6ec5c749ceb 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -414,7 +414,7 @@ struct ufs_init_prefetch {
 
 #define UFS_ERR_REG_HIST_LENGTH 8
 /**
- * struct ufs_err_reg_hist - keeps history of uic errors
+ * struct ufs_err_reg_hist - keeps history of errors
  * @pos: index to indicate cyclic buffer position
  * @reg: cyclic buffer for registers value
  * @tstamp: cyclic buffer for time stamp
@@ -436,15 +436,23 @@ struct ufs_err_reg_hist {
  * @nl_err: tracks nl-uic errors
  * @tl_err: tracks tl-uic errors
  * @dme_err: tracks dme errors
+ * @fatal_err: tracks fatal errors
+ * @auto_hibern8_err: tracks auto-hibernate errors
  */
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
 	ktime_t last_hibern8_exit_tstamp;
+
+	/* uic specific errors */
 	struct ufs_err_reg_hist pa_err;
 	struct ufs_err_reg_hist dl_err;
 	struct ufs_err_reg_hist nl_err;
 	struct ufs_err_reg_hist tl_err;
 	struct ufs_err_reg_hist dme_err;
+
+	/* fatal errors */
+	struct ufs_err_reg_hist fatal_err;
+	struct ufs_err_reg_hist auto_hibern8_err;
 };
 
 /**
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 3/4] scsi: ufs: Do not reset error history during host reset

[Stanley Chu]

Currently UFS error history will be reset and lost during host reset
flow by ufschd_probe_hba().

We shall not reset it and then error history can be kept as completed
as possible to improve debugging.

In addition, fix a minor display error in ufshcd_print_err_hist().

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index b8b874311509..a46c3d2b2ea3 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -402,7 +402,7 @@ static void ufshcd_print_err_hist(struct ufs_hba *hba,
 
 		if (err_hist->reg[p] == 0)
 			continue;
-		dev_err(hba->dev, "%s[%d] = 0x%x at %lld us\n", err_name, i,
+		dev_err(hba->dev, "%s[%d] = 0x%x at %lld us\n", err_name, p,
 			err_hist->reg[p], ktime_to_us(err_hist->tstamp[p]));
 		found = true;
 	}
@@ -6690,19 +6690,8 @@ static void ufshcd_tune_unipro_params(struct ufs_hba *hba)
 
 static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 {
-	int err_reg_hist_size = sizeof(struct ufs_err_reg_hist);
-
 	hba->ufs_stats.hibern8_exit_cnt = 0;
 	hba->ufs_stats.last_hibern8_exit_tstamp = ktime_set(0, 0);
-
-	memset(&hba->ufs_stats.pa_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.dl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.nl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.tl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.dme_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.fatal_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.auto_hibern8_err, 0, err_reg_hist_size);
-
 	hba->req_abort_count = 0;
 }
 
-- 
2.18.0

[PATCH v2 3/4] scsi: ufs: Do not reset error history during host reset

[Stanley Chu]

Currently UFS error history will be reset and lost during host reset
flow by ufschd_probe_hba().

We shall not reset it and then error history can be kept as completed
as possible to improve debugging.

In addition, fix a minor display error in ufshcd_print_err_hist().

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Reviewed-by: Avri Altman <avri.altman@wdc.com>
---
 drivers/scsi/ufs/ufshcd.c | 13 +------------
 1 file changed, 1 insertion(+), 12 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index b8b874311509..a46c3d2b2ea3 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -402,7 +402,7 @@ static void ufshcd_print_err_hist(struct ufs_hba *hba,
 
 		if (err_hist->reg[p] == 0)
 			continue;
-		dev_err(hba->dev, "%s[%d] = 0x%x at %lld us\n", err_name, i,
+		dev_err(hba->dev, "%s[%d] = 0x%x at %lld us\n", err_name, p,
 			err_hist->reg[p], ktime_to_us(err_hist->tstamp[p]));
 		found = true;
 	}
@@ -6690,19 +6690,8 @@ static void ufshcd_tune_unipro_params(struct ufs_hba *hba)
 
 static void ufshcd_clear_dbg_ufs_stats(struct ufs_hba *hba)
 {
-	int err_reg_hist_size = sizeof(struct ufs_err_reg_hist);
-
 	hba->ufs_stats.hibern8_exit_cnt = 0;
 	hba->ufs_stats.last_hibern8_exit_tstamp = ktime_set(0, 0);
-
-	memset(&hba->ufs_stats.pa_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.dl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.nl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.tl_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.dme_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.fatal_err, 0, err_reg_hist_size);
-	memset(&hba->ufs_stats.auto_hibern8_err, 0, err_reg_hist_size);
-
 	hba->req_abort_count = 0;
 }
 
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Stanley Chu]

Currently only "interrupt-based" errors have their own history,
however there are "non-interrupt-based" errors which may be
fatal also needing history to improve debugging or help know
the health status of UFS devices.

For example,
- Link startup fail
- Suspend fail
- Resume fail
- Task or request abort event

This patch tries to add those failed events by existed UFS error
history mechanism.

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
---
 drivers/scsi/ufs/ufshcd.c | 36 +++++++++++++++++++++++++++---------
 drivers/scsi/ufs/ufshcd.h | 10 ++++++++++
 2 files changed, 37 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index a46c3d2b2ea3..969128a731e1 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -432,6 +432,14 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
 			      "auto_hibern8_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.task_abort_err,
+			      "task_abort");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.link_startup_err,
+			      "link_startup_fail");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.suspend_err,
+			      "suspend_fail");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.resume_err,
+			      "resume_fail");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -4329,6 +4337,14 @@ static inline int ufshcd_disable_device_tx_lcc(struct ufs_hba *hba)
 	return ufshcd_disable_tx_lcc(hba, true);
 }
 
+static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
+				   u32 reg)
+{
+	reg_hist->reg[reg_hist->pos] = reg;
+	reg_hist->tstamp[reg_hist->pos] = ktime_get();
+	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
+}
+
 /**
  * ufshcd_link_startup - Initialize unipro link startup
  * @hba: per adapter instance
@@ -4356,6 +4372,8 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
 
 		/* check if device is detected by inter-connect layer */
 		if (!ret && !ufshcd_is_device_present(hba)) {
+			ufshcd_update_reg_hist(&hba->ufs_stats.link_startup_err,
+					       0);
 			dev_err(hba->dev, "%s: Device not present\n", __func__);
 			ret = -ENXIO;
 			goto out;
@@ -4366,8 +4384,11 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
 		 * but we can't be sure if the link is up until link startup
 		 * succeeds. So reset the local Uni-Pro and try again.
 		 */
-		if (ret && ufshcd_hba_enable(hba))
+		if (ret && ufshcd_hba_enable(hba)) {
+			ufshcd_update_reg_hist(&hba->ufs_stats.link_startup_err,
+					       (u32)ret);
 			goto out;
+		}
 	} while (ret && retries--);
 
 	if (ret)
@@ -5350,14 +5371,6 @@ static void ufshcd_err_handler(struct work_struct *work)
 	pm_runtime_put_sync(hba->dev);
 }
 
-static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
-				   u32 reg)
-{
-	reg_hist->reg[reg_hist->pos] = reg;
-	reg_hist->tstamp[reg_hist->pos] = ktime_get();
-	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
-}
-
 /**
  * ufshcd_update_uic_error - check and set fatal UIC error flags.
  * @hba: per-adapter instance
@@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 	 */
 	scsi_print_command(hba->lrb[tag].cmd);
 	if (!hba->req_abort_count) {
+		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err, 0);
 		ufshcd_print_host_regs(hba);
 		ufshcd_print_host_state(hba);
 		ufshcd_print_pwr_info(hba);
@@ -7819,6 +7833,8 @@ static int ufshcd_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	ufshcd_release(hba);
 out:
 	hba->pm_op_in_progress = 0;
+	if (ret)
+		ufshcd_update_reg_hist(&hba->ufs_stats.suspend_err, (u32)ret);
 	return ret;
 }
 
@@ -7921,6 +7937,8 @@ static int ufshcd_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	ufshcd_setup_clocks(hba, false);
 out:
 	hba->pm_op_in_progress = 0;
+	if (ret)
+		ufshcd_update_reg_hist(&hba->ufs_stats.resume_err, (u32)ret);
 	return ret;
 }
 
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index c6ec5c749ceb..f9f109da7f18 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
  * @dme_err: tracks dme errors
  * @fatal_err: tracks fatal errors
  * @auto_hibern8_err: tracks auto-hibernate errors
+ * @tsk_abort_err: tracks task abort events
+ * @linkup_err: tracks link-startup fail events
+ * @suspend_err: tracks suspend fail events
+ * @resume_err: tracks resume fail events
  */
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
@@ -453,6 +457,12 @@ struct ufs_stats {
 	/* fatal errors */
 	struct ufs_err_reg_hist fatal_err;
 	struct ufs_err_reg_hist auto_hibern8_err;
+
+	/* fatal events */
+	struct ufs_err_reg_hist task_abort_err;
+	struct ufs_err_reg_hist link_startup_err;
+	struct ufs_err_reg_hist suspend_err;
+	struct ufs_err_reg_hist resume_err;
 };
 
 /**
-- 
2.18.0

[PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Stanley Chu]

Currently only "interrupt-based" errors have their own history,
however there are "non-interrupt-based" errors which may be
fatal also needing history to improve debugging or help know
the health status of UFS devices.

For example,
- Link startup fail
- Suspend fail
- Resume fail
- Task or request abort event

This patch tries to add those failed events by existed UFS error
history mechanism.

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
---
 drivers/scsi/ufs/ufshcd.c | 36 +++++++++++++++++++++++++++---------
 drivers/scsi/ufs/ufshcd.h | 10 ++++++++++
 2 files changed, 37 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index a46c3d2b2ea3..969128a731e1 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -432,6 +432,14 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
 			      "auto_hibern8_err");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.task_abort_err,
+			      "task_abort");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.link_startup_err,
+			      "link_startup_fail");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.suspend_err,
+			      "suspend_fail");
+	ufshcd_print_err_hist(hba, &hba->ufs_stats.resume_err,
+			      "resume_fail");
 
 	ufshcd_print_clk_freqs(hba);
 
@@ -4329,6 +4337,14 @@ static inline int ufshcd_disable_device_tx_lcc(struct ufs_hba *hba)
 	return ufshcd_disable_tx_lcc(hba, true);
 }
 
+static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
+				   u32 reg)
+{
+	reg_hist->reg[reg_hist->pos] = reg;
+	reg_hist->tstamp[reg_hist->pos] = ktime_get();
+	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
+}
+
 /**
  * ufshcd_link_startup - Initialize unipro link startup
  * @hba: per adapter instance
@@ -4356,6 +4372,8 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
 
 		/* check if device is detected by inter-connect layer */
 		if (!ret && !ufshcd_is_device_present(hba)) {
+			ufshcd_update_reg_hist(&hba->ufs_stats.link_startup_err,
+					       0);
 			dev_err(hba->dev, "%s: Device not present\n", __func__);
 			ret = -ENXIO;
 			goto out;
@@ -4366,8 +4384,11 @@ static int ufshcd_link_startup(struct ufs_hba *hba)
 		 * but we can't be sure if the link is up until link startup
 		 * succeeds. So reset the local Uni-Pro and try again.
 		 */
-		if (ret && ufshcd_hba_enable(hba))
+		if (ret && ufshcd_hba_enable(hba)) {
+			ufshcd_update_reg_hist(&hba->ufs_stats.link_startup_err,
+					       (u32)ret);
 			goto out;
+		}
 	} while (ret && retries--);
 
 	if (ret)
@@ -5350,14 +5371,6 @@ static void ufshcd_err_handler(struct work_struct *work)
 	pm_runtime_put_sync(hba->dev);
 }
 
-static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
-				   u32 reg)
-{
-	reg_hist->reg[reg_hist->pos] = reg;
-	reg_hist->tstamp[reg_hist->pos] = ktime_get();
-	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
-}
-
 /**
  * ufshcd_update_uic_error - check and set fatal UIC error flags.
  * @hba: per-adapter instance
@@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 	 */
 	scsi_print_command(hba->lrb[tag].cmd);
 	if (!hba->req_abort_count) {
+		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err, 0);
 		ufshcd_print_host_regs(hba);
 		ufshcd_print_host_state(hba);
 		ufshcd_print_pwr_info(hba);
@@ -7819,6 +7833,8 @@ static int ufshcd_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	ufshcd_release(hba);
 out:
 	hba->pm_op_in_progress = 0;
+	if (ret)
+		ufshcd_update_reg_hist(&hba->ufs_stats.suspend_err, (u32)ret);
 	return ret;
 }
 
@@ -7921,6 +7937,8 @@ static int ufshcd_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	ufshcd_setup_clocks(hba, false);
 out:
 	hba->pm_op_in_progress = 0;
+	if (ret)
+		ufshcd_update_reg_hist(&hba->ufs_stats.resume_err, (u32)ret);
 	return ret;
 }
 
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index c6ec5c749ceb..f9f109da7f18 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
  * @dme_err: tracks dme errors
  * @fatal_err: tracks fatal errors
  * @auto_hibern8_err: tracks auto-hibernate errors
+ * @tsk_abort_err: tracks task abort events
+ * @linkup_err: tracks link-startup fail events
+ * @suspend_err: tracks suspend fail events
+ * @resume_err: tracks resume fail events
  */
 struct ufs_stats {
 	u32 hibern8_exit_cnt;
@@ -453,6 +457,12 @@ struct ufs_stats {
 	/* fatal errors */
 	struct ufs_err_reg_hist fatal_err;
 	struct ufs_err_reg_hist auto_hibern8_err;
+
+	/* fatal events */
+	struct ufs_err_reg_hist task_abort_err;
+	struct ufs_err_reg_hist link_startup_err;
+	struct ufs_err_reg_hist suspend_err;
+	struct ufs_err_reg_hist resume_err;
 };
 
 /**
-- 
2.18.0


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

RE: [PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Avri Altman]

Hi Stanley,

> 
> Currently only "interrupt-based" errors have their own history,
> however there are "non-interrupt-based" errors which may be
> fatal also needing history to improve debugging or help know
> the health status of UFS devices.
> 
> For example,
> - Link startup fail
> - Suspend fail
> - Resume fail
> - Task or request abort event
> 
> This patch tries to add those failed events by existed UFS error
> history mechanism.
> 
> Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
> ---
>  drivers/scsi/ufs/ufshcd.c | 36 +++++++++++++++++++++++++++---------
>  drivers/scsi/ufs/ufshcd.h | 10 ++++++++++
>  2 files changed, 37 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index a46c3d2b2ea3..969128a731e1 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -432,6 +432,14 @@ static void ufshcd_print_host_regs(struct ufs_hba
> *hba)
>  	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
>  	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
>  			      "auto_hibern8_err");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.task_abort_err,
> +			      "task_abort");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.link_startup_err,
> +			      "link_startup_fail");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.suspend_err,
> +			      "suspend_fail");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.resume_err,
> +			      "resume_fail");
> 
>  	ufshcd_print_clk_freqs(hba);
> 
> @@ -4329,6 +4337,14 @@ static inline int
> ufshcd_disable_device_tx_lcc(struct ufs_hba *hba)
>  	return ufshcd_disable_tx_lcc(hba, true);
>  }
> 
> +static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
> +				   u32 reg)
> +{
> +	reg_hist->reg[reg_hist->pos] = reg;
> +	reg_hist->tstamp[reg_hist->pos] = ktime_get();
> +	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
> +}
> +
>  /**
>   * ufshcd_link_startup - Initialize unipro link startup
>   * @hba: per adapter instance
> @@ -4356,6 +4372,8 @@ static int ufshcd_link_startup(struct ufs_hba
> *hba)
> 
>  		/* check if device is detected by inter-connect layer */
>  		if (!ret && !ufshcd_is_device_present(hba)) {
> +			ufshcd_update_reg_hist(&hba-
> >ufs_stats.link_startup_err,
> +					       0);
>  			dev_err(hba->dev, "%s: Device not present\n",
> __func__);
>  			ret = -ENXIO;
>  			goto out;
> @@ -4366,8 +4384,11 @@ static int ufshcd_link_startup(struct ufs_hba
> *hba)
>  		 * but we can't be sure if the link is up until link startup
>  		 * succeeds. So reset the local Uni-Pro and try again.
>  		 */
> -		if (ret && ufshcd_hba_enable(hba))
> +		if (ret && ufshcd_hba_enable(hba)) {
> +			ufshcd_update_reg_hist(&hba-
> >ufs_stats.link_startup_err,
> +					       (u32)ret);
>  			goto out;
> +		}
>  	} while (ret && retries--);
> 
>  	if (ret)
Here also link startup fails...

> @@ -5350,14 +5371,6 @@ static void ufshcd_err_handler(struct
> work_struct *work)
>  	pm_runtime_put_sync(hba->dev);
>  }
> 
> -static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
> -				   u32 reg)
> -{
> -	reg_hist->reg[reg_hist->pos] = reg;
> -	reg_hist->tstamp[reg_hist->pos] = ktime_get();
> -	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
> -}
> -
>  /**
>   * ufshcd_update_uic_error - check and set fatal UIC error flags.
>   * @hba: per-adapter instance
> @@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>  	 */
>  	scsi_print_command(hba->lrb[tag].cmd);
>  	if (!hba->req_abort_count) {
> +		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err,
> 0);
Here you are collecting abort events statistics, not abort errors.
If this is what you meant, then it's not task_abort_err, but task_abort.
And if indeed you are tracking task aborts, maybe add lun resets as well?


>  		ufshcd_print_host_regs(hba);
>  		ufshcd_print_host_state(hba);
>  		ufshcd_print_pwr_info(hba);
> @@ -7819,6 +7833,8 @@ static int ufshcd_suspend(struct ufs_hba *hba,
> enum ufs_pm_op pm_op)
>  	ufshcd_release(hba);
>  out:
>  	hba->pm_op_in_progress = 0;
> +	if (ret)
> +		ufshcd_update_reg_hist(&hba->ufs_stats.suspend_err,
> (u32)ret);
>  	return ret;
>  }
> 
> @@ -7921,6 +7937,8 @@ static int ufshcd_resume(struct ufs_hba *hba,
> enum ufs_pm_op pm_op)
>  	ufshcd_setup_clocks(hba, false);
>  out:
>  	hba->pm_op_in_progress = 0;
> +	if (ret)
> +		ufshcd_update_reg_hist(&hba->ufs_stats.resume_err,
> (u32)ret);
>  	return ret;
>  }
> 
> diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
> index c6ec5c749ceb..f9f109da7f18 100644
> --- a/drivers/scsi/ufs/ufshcd.h
> +++ b/drivers/scsi/ufs/ufshcd.h
> @@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
>   * @dme_err: tracks dme errors
>   * @fatal_err: tracks fatal errors
>   * @auto_hibern8_err: tracks auto-hibernate errors
> + * @tsk_abort_err: tracks task abort events
> + * @linkup_err: tracks link-startup fail events
> + * @suspend_err: tracks suspend fail events
> + * @resume_err: tracks resume fail events
>   */
>  struct ufs_stats {
>  	u32 hibern8_exit_cnt;
> @@ -453,6 +457,12 @@ struct ufs_stats {
>  	/* fatal errors */
>  	struct ufs_err_reg_hist fatal_err;
>  	struct ufs_err_reg_hist auto_hibern8_err;
> +
> +	/* fatal events */
Maybe move here fatal_err as well?

> +	struct ufs_err_reg_hist task_abort_err;
> +	struct ufs_err_reg_hist link_startup_err;
> +	struct ufs_err_reg_hist suspend_err;
> +	struct ufs_err_reg_hist resume_err;
>  };
> 
>  /**
> --
> 2.18.0


Thanks,
Avri

RE: [PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Avri Altman]

Hi Stanley,

> 
> Currently only "interrupt-based" errors have their own history,
> however there are "non-interrupt-based" errors which may be
> fatal also needing history to improve debugging or help know
> the health status of UFS devices.
> 
> For example,
> - Link startup fail
> - Suspend fail
> - Resume fail
> - Task or request abort event
> 
> This patch tries to add those failed events by existed UFS error
> history mechanism.
> 
> Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
> ---
>  drivers/scsi/ufs/ufshcd.c | 36 +++++++++++++++++++++++++++---------
>  drivers/scsi/ufs/ufshcd.h | 10 ++++++++++
>  2 files changed, 37 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index a46c3d2b2ea3..969128a731e1 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -432,6 +432,14 @@ static void ufshcd_print_host_regs(struct ufs_hba
> *hba)
>  	ufshcd_print_err_hist(hba, &hba->ufs_stats.fatal_err, "fatal_err");
>  	ufshcd_print_err_hist(hba, &hba->ufs_stats.auto_hibern8_err,
>  			      "auto_hibern8_err");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.task_abort_err,
> +			      "task_abort");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.link_startup_err,
> +			      "link_startup_fail");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.suspend_err,
> +			      "suspend_fail");
> +	ufshcd_print_err_hist(hba, &hba->ufs_stats.resume_err,
> +			      "resume_fail");
> 
>  	ufshcd_print_clk_freqs(hba);
> 
> @@ -4329,6 +4337,14 @@ static inline int
> ufshcd_disable_device_tx_lcc(struct ufs_hba *hba)
>  	return ufshcd_disable_tx_lcc(hba, true);
>  }
> 
> +static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
> +				   u32 reg)
> +{
> +	reg_hist->reg[reg_hist->pos] = reg;
> +	reg_hist->tstamp[reg_hist->pos] = ktime_get();
> +	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
> +}
> +
>  /**
>   * ufshcd_link_startup - Initialize unipro link startup
>   * @hba: per adapter instance
> @@ -4356,6 +4372,8 @@ static int ufshcd_link_startup(struct ufs_hba
> *hba)
> 
>  		/* check if device is detected by inter-connect layer */
>  		if (!ret && !ufshcd_is_device_present(hba)) {
> +			ufshcd_update_reg_hist(&hba-
> >ufs_stats.link_startup_err,
> +					       0);
>  			dev_err(hba->dev, "%s: Device not present\n",
> __func__);
>  			ret = -ENXIO;
>  			goto out;
> @@ -4366,8 +4384,11 @@ static int ufshcd_link_startup(struct ufs_hba
> *hba)
>  		 * but we can't be sure if the link is up until link startup
>  		 * succeeds. So reset the local Uni-Pro and try again.
>  		 */
> -		if (ret && ufshcd_hba_enable(hba))
> +		if (ret && ufshcd_hba_enable(hba)) {
> +			ufshcd_update_reg_hist(&hba-
> >ufs_stats.link_startup_err,
> +					       (u32)ret);
>  			goto out;
> +		}
>  	} while (ret && retries--);
> 
>  	if (ret)
Here also link startup fails...

> @@ -5350,14 +5371,6 @@ static void ufshcd_err_handler(struct
> work_struct *work)
>  	pm_runtime_put_sync(hba->dev);
>  }
> 
> -static void ufshcd_update_reg_hist(struct ufs_err_reg_hist *reg_hist,
> -				   u32 reg)
> -{
> -	reg_hist->reg[reg_hist->pos] = reg;
> -	reg_hist->tstamp[reg_hist->pos] = ktime_get();
> -	reg_hist->pos = (reg_hist->pos + 1) % UFS_ERR_REG_HIST_LENGTH;
> -}
> -
>  /**
>   * ufshcd_update_uic_error - check and set fatal UIC error flags.
>   * @hba: per-adapter instance
> @@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
>  	 */
>  	scsi_print_command(hba->lrb[tag].cmd);
>  	if (!hba->req_abort_count) {
> +		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err,
> 0);
Here you are collecting abort events statistics, not abort errors.
If this is what you meant, then it's not task_abort_err, but task_abort.
And if indeed you are tracking task aborts, maybe add lun resets as well?


>  		ufshcd_print_host_regs(hba);
>  		ufshcd_print_host_state(hba);
>  		ufshcd_print_pwr_info(hba);
> @@ -7819,6 +7833,8 @@ static int ufshcd_suspend(struct ufs_hba *hba,
> enum ufs_pm_op pm_op)
>  	ufshcd_release(hba);
>  out:
>  	hba->pm_op_in_progress = 0;
> +	if (ret)
> +		ufshcd_update_reg_hist(&hba->ufs_stats.suspend_err,
> (u32)ret);
>  	return ret;
>  }
> 
> @@ -7921,6 +7937,8 @@ static int ufshcd_resume(struct ufs_hba *hba,
> enum ufs_pm_op pm_op)
>  	ufshcd_setup_clocks(hba, false);
>  out:
>  	hba->pm_op_in_progress = 0;
> +	if (ret)
> +		ufshcd_update_reg_hist(&hba->ufs_stats.resume_err,
> (u32)ret);
>  	return ret;
>  }
> 
> diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
> index c6ec5c749ceb..f9f109da7f18 100644
> --- a/drivers/scsi/ufs/ufshcd.h
> +++ b/drivers/scsi/ufs/ufshcd.h
> @@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
>   * @dme_err: tracks dme errors
>   * @fatal_err: tracks fatal errors
>   * @auto_hibern8_err: tracks auto-hibernate errors
> + * @tsk_abort_err: tracks task abort events
> + * @linkup_err: tracks link-startup fail events
> + * @suspend_err: tracks suspend fail events
> + * @resume_err: tracks resume fail events
>   */
>  struct ufs_stats {
>  	u32 hibern8_exit_cnt;
> @@ -453,6 +457,12 @@ struct ufs_stats {
>  	/* fatal errors */
>  	struct ufs_err_reg_hist fatal_err;
>  	struct ufs_err_reg_hist auto_hibern8_err;
> +
> +	/* fatal events */
Maybe move here fatal_err as well?

> +	struct ufs_err_reg_hist task_abort_err;
> +	struct ufs_err_reg_hist link_startup_err;
> +	struct ufs_err_reg_hist suspend_err;
> +	struct ufs_err_reg_hist resume_err;
>  };
> 
>  /**
> --
> 2.18.0


Thanks,
Avri


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

RE: [PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Stanley Chu]

Hi Avri,

On Wed, 2019-07-10 at 08:04 +0000, Avri Altman wrote:
> Hi Stanley,
> 
> > +					       (u32)ret);
> >  			goto out;
> > +		}
> >  	} while (ret && retries--);
> > 
> >  	if (ret)
> Here also link startup fails...

Thanks! Will track this place as well in next version.

> >   * ufshcd_update_uic_error - check and set fatal UIC error flags.
> >   * @hba: per-adapter instance
> > @@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
> >  	 */
> >  	scsi_print_command(hba->lrb[tag].cmd);
> >  	if (!hba->req_abort_count) {
> > +		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err,
> > 0);
> Here you are collecting abort events statistics, not abort errors.
> If this is what you meant, then it's not task_abort_err, but task_abort.
> And if indeed you are tracking task aborts, maybe add lun resets as well?

Good suggestion! I would add history of lun reset and host reset as well
in next version.

> > diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
> > index c6ec5c749ceb..f9f109da7f18 100644
> > --- a/drivers/scsi/ufs/ufshcd.h
> > +++ b/drivers/scsi/ufs/ufshcd.h
> > @@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
> >   * @dme_err: tracks dme errors
> >   * @fatal_err: tracks fatal errors
> >   * @auto_hibern8_err: tracks auto-hibernate errors
> > + * @tsk_abort_err: tracks task abort events
> > + * @linkup_err: tracks link-startup fail events
> > + * @suspend_err: tracks suspend fail events
> > + * @resume_err: tracks resume fail events
> >   */
> >  struct ufs_stats {
> >  	u32 hibern8_exit_cnt;
> > @@ -453,6 +457,12 @@ struct ufs_stats {
> >  	/* fatal errors */
> >  	struct ufs_err_reg_hist fatal_err;
> >  	struct ufs_err_reg_hist auto_hibern8_err;
> > +
> > +	/* fatal events */
> Maybe move here fatal_err as well?

OK! these could be classified as fatal errors as well.
Will fix them in next version.

> 
> > +	struct ufs_err_reg_hist task_abort_err;
> > +	struct ufs_err_reg_hist link_startup_err;
> > +	struct ufs_err_reg_hist suspend_err;
> > +	struct ufs_err_reg_hist resume_err;
> >  };
> > 
> >  /**
> > --
> > 2.18.0
> 
> 
> Thanks,
> Avri
> 

Thanks,
Stanley

RE: [PATCH v2 4/4] scsi: ufs: Add history of fatal events

[Stanley Chu]

Hi Avri,

On Wed, 2019-07-10 at 08:04 +0000, Avri Altman wrote:
> Hi Stanley,
> 
> > +					       (u32)ret);
> >  			goto out;
> > +		}
> >  	} while (ret && retries--);
> > 
> >  	if (ret)
> Here also link startup fails...

Thanks! Will track this place as well in next version.

> >   * ufshcd_update_uic_error - check and set fatal UIC error flags.
> >   * @hba: per-adapter instance
> > @@ -6043,6 +6056,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
> >  	 */
> >  	scsi_print_command(hba->lrb[tag].cmd);
> >  	if (!hba->req_abort_count) {
> > +		ufshcd_update_reg_hist(&hba->ufs_stats.task_abort_err,
> > 0);
> Here you are collecting abort events statistics, not abort errors.
> If this is what you meant, then it's not task_abort_err, but task_abort.
> And if indeed you are tracking task aborts, maybe add lun resets as well?

Good suggestion! I would add history of lun reset and host reset as well
in next version.

> > diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
> > index c6ec5c749ceb..f9f109da7f18 100644
> > --- a/drivers/scsi/ufs/ufshcd.h
> > +++ b/drivers/scsi/ufs/ufshcd.h
> > @@ -438,6 +438,10 @@ struct ufs_err_reg_hist {
> >   * @dme_err: tracks dme errors
> >   * @fatal_err: tracks fatal errors
> >   * @auto_hibern8_err: tracks auto-hibernate errors
> > + * @tsk_abort_err: tracks task abort events
> > + * @linkup_err: tracks link-startup fail events
> > + * @suspend_err: tracks suspend fail events
> > + * @resume_err: tracks resume fail events
> >   */
> >  struct ufs_stats {
> >  	u32 hibern8_exit_cnt;
> > @@ -453,6 +457,12 @@ struct ufs_stats {
> >  	/* fatal errors */
> >  	struct ufs_err_reg_hist fatal_err;
> >  	struct ufs_err_reg_hist auto_hibern8_err;
> > +
> > +	/* fatal events */
> Maybe move here fatal_err as well?

OK! these could be classified as fatal errors as well.
Will fix them in next version.

> 
> > +	struct ufs_err_reg_hist task_abort_err;
> > +	struct ufs_err_reg_hist link_startup_err;
> > +	struct ufs_err_reg_hist suspend_err;
> > +	struct ufs_err_reg_hist resume_err;
> >  };
> > 
> >  /**
> > --
> > 2.18.0
> 
> 
> Thanks,
> Avri
> 

Thanks,
Stanley




_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel