Skip to contentSkip to conversation

Fix up and simplify error recovery mechanism

PATCH Superseded Can Guo 2 participants 13 messages Jul 23, 2020 – Jul 28, 2020

Subsystems (2)

v6 of 11 Superseded Latest is v11 →

lore
Share & tools
lore

Can Guo

author
#0 lore raw

v6 cover Fix up and simplify error recovery mechanism

The changes have been tested with error injections of multiple error types (and all kinds of mixture of them) during runtime, e.g. hibern8 enter/ exit error, power mode change error and fatal/non-fatal error from IRQ context. During the test, error injections happen randomly across all contexts, e.g. clk scaling, clk gate/ungate, runtime suspend/resume and IRQ.

There are a few more fixes to resolve other minor problems based on the main change, such as LINERESET handling and racing btw error handler and system suspend/resume/shutdown, but they will be pushed after this series is taken, due to there are already too many lines in these changes.

Change since v5:
- Dropped change "scsi: ufs: Fix imbalanced scsi_block_reqs_cnt caused by ufshcd_hold()" as it is not quite related with this series
- Refined func ufshcd_err_handling_prepare in change "scsi: ufs: Recover hba runtime PM error in error handler"

Change since v4:
- Split the original change "ufs: ufs-qcom: Fix a few BUGs in func ufs_qcom_dump_dbg_regs()" to 2 small changes

Change since v3:
- Split the original change "scsi: ufs: Fix up and simplify error recovery mechanism" into 5 changes

Change since v2:
- Incorporate Bart's comment to change "scsi: ufs: Add checks before setting clk-gating states"
- Revised the commit msg of change "scsi: ufs: Fix up and simplify error recovery mechanism"

Change since v1:
- Fixed a compilation error in case that CONFIG_PM is N

Can Guo (8):
scsi: ufs: Add checks before setting clk-gating states
ufs: ufs-qcom: Fix race conditions caused by func
ufs_qcom_testbus_config
scsi: ufs-qcom: Fix schedule while atomic error in
ufs_qcom_dump_dbg_regs
scsi: ufs: Add some debug infos to ufshcd_print_host_state
scsi: ufs: Fix concurrency of error handler and other error recovery
paths
scsi: ufs: Recover hba runtime PM error in error handler
scsi: ufs: Move dumps in IRQ handler to error handler
scsi: ufs: Fix a racing problem btw error handler and runtime PM ops

 drivers/scsi/ufs/ufs-qcom.c  |  20 +-
 drivers/scsi/ufs/ufs-sysfs.c |   1 +
 drivers/scsi/ufs/ufshcd.c    | 494 +++++++++++++++++++++++++++----------------
 drivers/scsi/ufs/ufshcd.h    |  15 ++
 4 files changed, 333 insertions(+), 197 deletions(-)
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
author
#1 lore raw

PATCH v6 1/8 scsi: ufs: Add checks before setting clk-gating states

Clock gating features can be turned on/off selectively which means its state information is only important if it is enabled. This change makes sure that we only look at state of clk-gating if it is enabled.

---
 drivers/scsi/ufs/ufshcd.c | 17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)
drivers/scsi/ufs/ufshcd.c +14 -3
index cdff7e5..99bd3e4 100644
@@ -1839,6 +1839,8 @@ static void ufshcd_init_clk_gating(struct ufs_hba *hba)
 	if (!ufshcd_is_clkgating_allowed(hba))
 		return;
 
+	hba->clk_gating.state = CLKS_ON;
+
 	hba->clk_gating.delay_ms = 150;
 	INIT_DELAYED_WORK(&hba->clk_gating.gate_work, ufshcd_gate_work);
 	INIT_WORK(&hba->clk_gating.ungate_work, ufshcd_ungate_work);
@@ -2541,7 +2543,8 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
 		err = SCSI_MLQUEUE_HOST_BUSY;
 		goto out;
 	}
-	WARN_ON(hba->clk_gating.state != CLKS_ON);
+	WARN_ON(ufshcd_is_clkgating_allowed(hba) &&
+		(hba->clk_gating.state != CLKS_ON));
 
 	lrbp = &hba->lrb[tag];
 
@@ -8315,8 +8318,11 @@ static int ufshcd_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 		/* If link is active, device ref_clk can't be switched off */
 		__ufshcd_setup_clocks(hba, false, true);
 
-	hba->clk_gating.state = CLKS_OFF;
-	trace_ufshcd_clk_gating(dev_name(hba->dev), hba->clk_gating.state);
+	if (ufshcd_is_clkgating_allowed(hba)) {
+		hba->clk_gating.state = CLKS_OFF;
+		trace_ufshcd_clk_gating(dev_name(hba->dev),
+					hba->clk_gating.state);
+	}
 
 	/* Put the host controller in low power mode if possible */
 	ufshcd_hba_vreg_set_lpm(hba);
@@ -8456,6 +8462,11 @@ static int ufshcd_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	if (hba->clk_scaling.is_allowed)
 		ufshcd_suspend_clkscaling(hba);
 	ufshcd_setup_clocks(hba, false);
+	if (ufshcd_is_clkgating_allowed(hba)) {
+		hba->clk_gating.state = CLKS_OFF;
+		trace_ufshcd_clk_gating(dev_name(hba->dev),
+					hba->clk_gating.state);
+	}
 out:
 	hba->pm_op_in_progress = 0;
 	if (ret)
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
author
#2 lore raw

PATCH v6 2/8 ufs: ufs-qcom: Fix race conditions caused by func ufs_qcom_testbus_config

In reply to Can Guo

If ufs_qcom_dump_dbg_regs() calls ufs_qcom_testbus_config() from ufshcd_suspend/resume and/or clk gate/ungate context, pm_runtime_get_sync() and ufshcd_hold() will cause racing problems. Fix this by removing the unnecessary calls of pm_runtime_get_sync() and ufshcd_hold().

---
 drivers/scsi/ufs/ufs-qcom.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/scsi/ufs/ufs-qcom.c +0 -5
index 2e6ddb5..7da27ee 100644
@@ -1604,9 +1604,6 @@ int ufs_qcom_testbus_config(struct ufs_qcom_host *host)
 	 */
 	}
 	mask <<= offset;
-
-	pm_runtime_get_sync(host->hba->dev);
-	ufshcd_hold(host->hba, false);
 	ufshcd_rmwl(host->hba, TEST_BUS_SEL,
 		    (u32)host->testbus.select_major << 19,
 		    REG_UFS_CFG1);
@@ -1619,8 +1616,6 @@ int ufs_qcom_testbus_config(struct ufs_qcom_host *host)
 	 * committed before returning.
 	 */
 	mb();
-	ufshcd_release(host->hba);
-	pm_runtime_put_sync(host->hba->dev);
 
 	return 0;
 }
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v6 2/8
#3 lore raw

RE: [PATCH v6 2/8] ufs: ufs-qcom: Fix race conditions caused by func ufs_qcom_testbus_config

Quoted 10 lines

If ufs_qcom_dump_dbg_regs() calls ufs_qcom_testbus_config() from ufshcd_suspend/resume and/or clk gate/ungate context, pm_runtime_get_sync()
and ufshcd_hold() will cause racing problems. Fix this by removing the unnecessary calls of pm_runtime_get_sync() and ufshcd_hold().

author
#4 lore raw

PATCH v6 3/8 scsi: ufs-qcom: Fix schedule while atomic error in ufs_qcom_dump_dbg_regs

In reply to Can Guo

Dumping testbus registers needs to sleep a bit intermittently as there are too many of them. Skip them for those contexts where sleep is not allowed.

---
 drivers/scsi/ufs/ufs-qcom.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufs-qcom.c +9 -6
index 7da27ee..7831b2b 100644
@@ -1651,13 +1651,16 @@ static void ufs_qcom_dump_dbg_regs(struct ufs_hba *hba)
 	ufshcd_dump_regs(hba, REG_UFS_SYS1CLK_1US, 16 * 4,
 			 "HCI Vendor Specific Registers ");
 
-	/* sleep a bit intermittently as we are dumping too much data */
 	ufs_qcom_print_hw_debug_reg_all(hba, NULL, ufs_qcom_dump_regs_wrapper);
-	udelay(1000);
-	ufs_qcom_testbus_read(hba);
-	udelay(1000);
-	ufs_qcom_print_unipro_testbus(hba);
-	udelay(1000);
+
+	if (in_task()) {
+		/* sleep a bit intermittently as we are dumping too much data */
+		usleep_range(1000, 1100);
+		ufs_qcom_testbus_read(hba);
+		usleep_range(1000, 1100);
+		ufs_qcom_print_unipro_testbus(hba);
+		usleep_range(1000, 1100);
+	}
 }
 
 /**
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
#5 lore raw

RE: [PATCH v6 3/8] scsi: ufs-qcom: Fix schedule while atomic error in ufs_qcom_dump_dbg_regs

Quoted 25 lines

Dumping testbus registers needs to sleep a bit intermittently as there are too many of them. Skip them for those contexts where sleep is not allowed.

---
 drivers/scsi/ufs/ufs-qcom.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufs-qcom.c +0 -6
index 7da27ee..7831b2b 100644
@@ -1651,13 +1651,16 @@ static void ufs_qcom_dump_dbg_regs(struct
 ufs_hba *hba)
        ufshcd_dump_regs(hba, REG_UFS_SYS1CLK_1US, 16 * 4,
                         "HCI Vendor Specific Registers ");
 
-       /* sleep a bit intermittently as we are dumping too much data */
        ufs_qcom_print_hw_debug_reg_all(hba, NULL,
 ufs_qcom_dump_regs_wrapper);
-       udelay(1000);
-       ufs_qcom_testbus_read(hba);
-       udelay(1000);
-       ufs_qcom_print_unipro_testbus(hba);
-       udelay(1000);

drivers/scsi/ufs/ufs-qcom.c static void ufs_qcom_dump_dbg_regs(struct · line 1666

Quoted 10 lines
drivers/scsi/ufs/ufs-qcom.c +9 -0
+
+       if (in_task()) {
+               /* sleep a bit intermittently as we are dumping too much data */
+               usleep_range(1000, 1100);
+               ufs_qcom_testbus_read(hba);
+               usleep_range(1000, 1100);
+               ufs_qcom_print_unipro_testbus(hba);
+               usleep_range(1000, 1100);
+       }
 }

How about moving the intermittent sleep out of the check if preemption is disabled? And maybe then you need to switch back to uedlay?

author
#6 lore raw

Re: [PATCH v6 3/8] scsi: ufs-qcom: Fix schedule while atomic error in ufs_qcom_dump_dbg_regs

Hi Avri,

Quoted 32 lines · Avri Altman
Quoted 32 lines

Dumping testbus registers needs to sleep a bit intermittently as there are too many of them. Skip them for those contexts where sleep is not allowed.

---
 drivers/scsi/ufs/ufs-qcom.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufs-qcom.c +3 -6
index 7da27ee..7831b2b 100644
@@ -1651,13 +1651,16 @@ static void ufs_qcom_dump_dbg_regs(struct
 ufs_hba *hba)
        ufshcd_dump_regs(hba, REG_UFS_SYS1CLK_1US, 16 * 4,
                         "HCI Vendor Specific Registers ");
 
-       /* sleep a bit intermittently as we are dumping too much data 
 */
        ufs_qcom_print_hw_debug_reg_all(hba, NULL,
 ufs_qcom_dump_regs_wrapper);
-       udelay(1000);
-       ufs_qcom_testbus_read(hba);
-       udelay(1000);
-       ufs_qcom_print_unipro_testbus(hba);
-       udelay(1000);
+
+       if (in_task()) {
+               /* sleep a bit intermittently as we are dumping too 

much data */

drivers/scsi/ufs/ufs-qcom.c +2 -0
+               usleep_range(1000, 1100);
+               ufs_qcom_testbus_read(hba);
Quoted 8 lines

drivers/scsi/ufs/ufs-qcom.c static void ufs_qcom_dump_dbg_regs(struct ufs_hba *hba) · line 1664

Quoted 5 lines
drivers/scsi/ufs/ufs-qcom.c +4 -0
+               usleep_range(1000, 1100);
+               ufs_qcom_print_unipro_testbus(hba);
+               usleep_range(1000, 1100);
+       }
 }

How about moving the intermittent sleep out of the check if preemption is disabled?
And maybe then you need to switch back to uedlay?

I will just remove all the testbus prints in next version to save us time here.

Thanks,

Can Guo.

author
#7 lore raw

PATCH v6 4/8 scsi: ufs: Add some debug infos to ufshcd_print_host_state

In reply to Can Guo

The infos of the last interrupt status and its timestamp are very helpful when debug system stability issues, e.g. IRQ starvation, so add them to ufshcd_print_host_state. Meanwhile, UFS device infos like model name and its FW version also come in handy during debug. In addition, this change makes cleanup to some prints in ufshcd_print_host_regs as similar prints are already available in ufshcd_print_host_state.

---
 drivers/scsi/ufs/ufshcd.c | 31 ++++++++++++++++++-------------
 drivers/scsi/ufs/ufshcd.h |  5 +++++
 2 files changed, 23 insertions(+), 13 deletions(-)
drivers/scsi/ufs/ufshcd.c +18 -13
index 99bd3e4..eda4dc6 100644
@@ -411,15 +411,6 @@ static void ufshcd_print_err_hist(struct ufs_hba *hba,
 static void ufshcd_print_host_regs(struct ufs_hba *hba)
 {
 	ufshcd_dump_regs(hba, 0, UFSHCI_REG_SPACE_SIZE, "host_regs: ");
-	dev_err(hba->dev, "hba->ufs_version = 0x%x, hba->capabilities = 0x%x\n",
-		hba->ufs_version, hba->capabilities);
-	dev_err(hba->dev,
-		"hba->outstanding_reqs = 0x%x, hba->outstanding_tasks = 0x%x\n",
-		(u32)hba->outstanding_reqs, (u32)hba->outstanding_tasks);
-	dev_err(hba->dev,
-		"last_hibern8_exit_tstamp at %lld us, hibern8_exit_cnt = %d\n",
-		ktime_to_us(hba->ufs_stats.last_hibern8_exit_tstamp),
-		hba->ufs_stats.hibern8_exit_cnt);
 
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.pa_err, "pa_err");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.dl_err, "dl_err");
@@ -438,8 +429,6 @@ static void ufshcd_print_host_regs(struct ufs_hba *hba)
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.host_reset, "host_reset");
 	ufshcd_print_err_hist(hba, &hba->ufs_stats.task_abort, "task_abort");
 
-	ufshcd_print_clk_freqs(hba);
-
 	ufshcd_vops_dbg_register_dump(hba);
 }
 
@@ -499,6 +488,8 @@ static void ufshcd_print_tmrs(struct ufs_hba *hba, unsigned long bitmap)
 
 static void ufshcd_print_host_state(struct ufs_hba *hba)
 {
+	struct scsi_device *sdev_ufs = hba->sdev_ufs_device;
+
 	dev_err(hba->dev, "UFS Host state=%d\n", hba->ufshcd_state);
 	dev_err(hba->dev, "outstanding reqs=0x%lx tasks=0x%lx\n",
 		hba->outstanding_reqs, hba->outstanding_tasks);
@@ -511,12 +502,24 @@ static void ufshcd_print_host_state(struct ufs_hba *hba)
 	dev_err(hba->dev, "Auto BKOPS=%d, Host self-block=%d\n",
 		hba->auto_bkops_enabled, hba->host->host_self_blocked);
 	dev_err(hba->dev, "Clk gate=%d\n", hba->clk_gating.state);
+	dev_err(hba->dev,
+		"last_hibern8_exit_tstamp at %lld us, hibern8_exit_cnt=%d\n",
+		ktime_to_us(hba->ufs_stats.last_hibern8_exit_tstamp),
+		hba->ufs_stats.hibern8_exit_cnt);
+	dev_err(hba->dev, "last intr at %lld us, last intr status=0x%x\n",
+		ktime_to_us(hba->ufs_stats.last_intr_ts),
+		hba->ufs_stats.last_intr_status);
 	dev_err(hba->dev, "error handling flags=0x%x, req. abort count=%d\n",
 		hba->eh_flags, hba->req_abort_count);
-	dev_err(hba->dev, "Host capabilities=0x%x, caps=0x%x\n",
-		hba->capabilities, hba->caps);
+	dev_err(hba->dev, "hba->ufs_version=0x%x, Host capabilities=0x%x, caps=0x%x\n",
+		hba->ufs_version, hba->capabilities, hba->caps);
 	dev_err(hba->dev, "quirks=0x%x, dev. quirks=0x%x\n", hba->quirks,
 		hba->dev_quirks);
+	if (sdev_ufs)
+		dev_err(hba->dev, "UFS dev info: %.8s %.16s rev %.4s\n",
+			sdev_ufs->vendor, sdev_ufs->model, sdev_ufs->rev);
+
+	ufshcd_print_clk_freqs(hba);
 }
 
 /**
@@ -5951,6 +5954,8 @@ static irqreturn_t ufshcd_intr(int irq, void *__hba)
 
 	spin_lock(hba->host->host_lock);
 	intr_status = ufshcd_readl(hba, REG_INTERRUPT_STATUS);
+	hba->ufs_stats.last_intr_status = intr_status;
+	hba->ufs_stats.last_intr_ts = ktime_get();
 
 	/*
 	 * There could be max of hba->nutrs reqs in flight and in worst case
drivers/scsi/ufs/ufshcd.h +5 -0
index 656c069..5b2cdaf 100644
@@ -406,6 +406,8 @@ struct ufs_err_reg_hist {
 
 /**
  * struct ufs_stats - keeps usage/err statistics
+ * @last_intr_status: record the last interrupt status.
+ * @last_intr_ts: record the last interrupt timestamp.
  * @hibern8_exit_cnt: Counter to keep track of number of exits,
  *		reset this after link-startup.
  * @last_hibern8_exit_tstamp: Set time after the hibern8 exit.
@@ -425,6 +427,9 @@ struct ufs_err_reg_hist {
  * @tsk_abort: tracks task abort events
  */
 struct ufs_stats {
+	u32 last_intr_status;
+	ktime_t last_intr_ts;
+
 	u32 hibern8_exit_cnt;
 	ktime_t last_hibern8_exit_tstamp;
 
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v6 4/8
#8 lore raw

RE: [PATCH v6 4/8] scsi: ufs: Add some debug infos to ufshcd_print_host_state

Quoted 10 lines

The infos of the last interrupt status and its timestamp are very helpful when debug system stability issues, e.g. IRQ starvation, so add them to ufshcd_print_host_state. Meanwhile, UFS device infos like model name and its FW version also come in handy during debug. In addition, this change makes cleanup to some prints in ufshcd_print_host_regs as similar prints are already available in ufshcd_print_host_state.

author
#9 lore raw

PATCH v6 5/8 scsi: ufs: Fix concurrency of error handler and other error recovery paths

In reply to Can Guo

Error recovery can be invoked from multiple paths, including hibern8 enter/exit, some vendor vops, ufshcd_eh_host_reset_handler(), resume and eh_work scheduled from IRQ context. Ultimately, these paths are trying to invoke ufshcd_reset_and_restore(), in either sync or async manner.

Having both sync and async manners at the same time has some problems

- If link recovery happens during clock scaling work, acquring scaling_lock in ufshcd_exec_dev_cmd() would cause dead lock, because scaling_lock is already held by scaling work before link recovery happens.

- If link recovery happens during ungate work, ufshcd_hold() would be called recursively. Although commit 53c12d0ef6fcb
("scsi: ufs: fix error recovery after the hibern8 exit failure") [1] fixed a deadlock due to recursive calls of ufshcd_hold() by adding a check of eh_in_progress into ufshcd_hold, this check allows eh_work to run in parallel when link recovery is running.

- Similar concurrency can also happen when error recovery is invoked from eh_host_reset_handler. Although it tries to avoid that from happening by flushing eh_work before start invoking reset and restore, after flush work returns, eh_work can still be scheduled and running in parallel.

- Concurrency can even happen between eh_works. eh_work, currently queued on system_wq, is allowed to have multiple instances running in parallel, but we don't have proper protection for that.

If any of above concurrency happens, error recovery would fail and lead ufs device and host into bad states. To fix the concurrency problem, this change queues eh_work on a single threaded workqueue and remove link recovery calls from hibern8 enter/exit path. Meanwhile, make use of eh_work in eh_host_reset_handler instead of calling ufshcd_reset_and_restore. This unifies UFS error recovery mechanism.

In addition, according to the UFSHCI JEDEC spec, hibern8 enter/exit error occurs when the link is broken. This essentially applies to any power mode change operations (since they all use PACP_PWR cmds in UniPro layer). So, in this change, if a power mode change operation (including AH8 enter/exit) fails, mark link state as UIC_LINK_BROKEN_STATE and schedule the eh_work. In this case, error handler needs to do a full reset and restore to recover the link back to active. Before the link state is recovered to active, ufshcd_uic_pwr_ctrl simply returns -ENOLINK to avoid more errors.

---
 drivers/scsi/ufs/ufs-sysfs.c |   1 +
 drivers/scsi/ufs/ufshcd.c    | 268 +++++++++++++++++++++++--------------------
 drivers/scsi/ufs/ufshcd.h    |   9 ++
 3 files changed, 151 insertions(+), 127 deletions(-)
drivers/scsi/ufs/ufs-sysfs.c +1 -0
index 2d71d23..02d379f00 100644
@@ -16,6 +16,7 @@ static const char *ufschd_uic_link_state_to_string(
 	case UIC_LINK_OFF_STATE:	return "OFF";
 	case UIC_LINK_ACTIVE_STATE:	return "ACTIVE";
 	case UIC_LINK_HIBERN8_STATE:	return "HIBERN8";
+	case UIC_LINK_BROKEN_STATE:	return "BROKEN";
 	default:			return "UNKNOWN";
 	}
 }
drivers/scsi/ufs/ufshcd.c +141 -127

491 lines not on this page · raw message

drivers/scsi/ufs/ufshcd.h +9 -0
index 5b2cdaf..83e5cd9 100644
@@ -90,6 +90,7 @@ enum uic_link_state {
 	UIC_LINK_OFF_STATE	= 0, /* Link powered down or disabled */
 	UIC_LINK_ACTIVE_STATE	= 1, /* Link is in Fast/Slow/Sleep state */
 	UIC_LINK_HIBERN8_STATE	= 2, /* Link is in Hibernate state */
+	UIC_LINK_BROKEN_STATE	= 3, /* Link is in broken state */
 };
 
 #define ufshcd_is_link_off(hba) ((hba)->uic_link_state == UIC_LINK_OFF_STATE)
@@ -97,11 +98,15 @@ enum uic_link_state {
 				    UIC_LINK_ACTIVE_STATE)
 #define ufshcd_is_link_hibern8(hba) ((hba)->uic_link_state == \
 				    UIC_LINK_HIBERN8_STATE)
+#define ufshcd_is_link_broken(hba) ((hba)->uic_link_state == \
+				   UIC_LINK_BROKEN_STATE)
 #define ufshcd_set_link_off(hba) ((hba)->uic_link_state = UIC_LINK_OFF_STATE)
 #define ufshcd_set_link_active(hba) ((hba)->uic_link_state = \
 				    UIC_LINK_ACTIVE_STATE)
 #define ufshcd_set_link_hibern8(hba) ((hba)->uic_link_state = \
 				    UIC_LINK_HIBERN8_STATE)
+#define ufshcd_set_link_broken(hba) ((hba)->uic_link_state = \
+				    UIC_LINK_BROKEN_STATE)
 
 #define ufshcd_set_ufs_dev_active(h) \
 	((h)->curr_dev_pwr_mode = UFS_ACTIVE_PWR_MODE)
@@ -613,12 +618,14 @@ struct ufs_hba_variant_params {
  * @intr_mask: Interrupt Mask Bits
  * @ee_ctrl_mask: Exception event control mask
  * @is_powered: flag to check if HBA is powered
+ * @eh_wq: Workqueue that eh_work works on
  * @eh_work: Worker to handle UFS errors that require s/w attention
  * @eeh_work: Worker to handle exception events
  * @errors: HBA errors
  * @uic_error: UFS interconnect layer error status
  * @saved_err: sticky error mask
  * @saved_uic_err: sticky UIC error mask
+ * @force_reset: flag to force eh_work perform a full reset
  * @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
@@ -707,6 +714,7 @@ struct ufs_hba {
 	bool is_powered;
 
 	/* Work Queues */
+	struct workqueue_struct *eh_wq;
 	struct work_struct eh_work;
 	struct work_struct eeh_work;
 
@@ -716,6 +724,7 @@ struct ufs_hba {
 	u32 saved_err;
 	u32 saved_uic_err;
 	struct ufs_stats ufs_stats;
+	bool force_reset;
 	bool silence_err_logs;
 
 	/* Device management request data */
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
author
#10 lore raw

PATCH v6 6/8 scsi: ufs: Recover hba runtime PM error in error handler

In reply to Can Guo

Current error handler cannot work well or recover hba runtime PM error if ufshcd_suspend/resume has failed due to UFS errors, e.g. hibern8 enter/exit error or SSU cmd error. When this happens, error handler may fail doing full reset and restore because error handler always assumes that powers, IRQs and clocks are ready after pm_runtime_get_sync returns, but actually they are not if ufshcd_reusme fails [1]. Besides, if ufschd_suspend/resume fails due to UFS error, runtime PM framework saves the error value to dev.power.runtime_error. After that, hba dev runtime suspend/resume would not be invoked anymore unless runtime_error is cleared [2].

In case of ufshcd_suspend/resume fails due to UFS errors, for scenario [1], error handler cannot assume anything of pm_runtime_get_sync, meaning error handler should explicitly turn ON powers, IRQs and clocks again. To get the hba runtime PM work as regard for scenario [2], error handler can clear the runtime_error by calling pm_runtime_set_active() if full reset and restore succeeds. And, more important, if pm_runtime_set_active() returns no error, which means runtime_error has been cleared, we also need to resume those scsi devices under hba in case any of them has failed to be resumed due to hba runtime resume failure. This is to unblock blk_queue_enter in case there are bios waiting inside it.

In addition, if ufshcd_resume errors out, ufshcd_release in ufshcd_resume would be skipped. After hba runtime PM error is recovered in error handler, we should do ufshcd_release once to get clock gating back to work.

---
 drivers/scsi/ufs/ufshcd.c | 91 +++++++++++++++++++++++++++++++++++++++++++----
 drivers/scsi/ufs/ufshcd.h |  1 +
 2 files changed, 86 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +85 -6
index c2d7a90..c480823 100644
@@ -15,6 +15,7 @@
 #include <linux/of.h>
 #include <linux/bitfield.h>
 #include <linux/blk-pm.h>
+#include <linux/blkdev.h>
 #include "ufshcd.h"
 #include "ufs_quirks.h"
 #include "unipro.h"
@@ -229,6 +230,10 @@ static irqreturn_t ufshcd_intr(int irq, void *__hba);
 static int ufshcd_change_power_mode(struct ufs_hba *hba,
 			     struct ufs_pa_layer_attr *pwr_mode);
 static void ufshcd_schedule_eh_work(struct ufs_hba *hba);
+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_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);
@@ -5553,6 +5558,78 @@ static inline void ufshcd_schedule_eh_work(struct ufs_hba *hba)
 	}
 }
 
+static void ufshcd_err_handling_prepare(struct ufs_hba *hba)
+{
+	pm_runtime_get_sync(hba->dev);
+	if (pm_runtime_status_suspended(hba->dev)) {
+		/*
+		 * Don't assume anything of pm_runtime_get_sync(), if
+		 * resume fails, irq and clocks can be OFF, and powers
+		 * can be OFF or in LPM.
+		 */
+		ufshcd_setup_hba_vreg(hba, true);
+		ufshcd_setup_clocks(hba, true);
+		ufshcd_enable_irq(hba);
+		ufshcd_setup_vreg(hba, true);
+		ufshcd_config_vreg_hpm(hba, hba->vreg_info.vccq);
+		ufshcd_config_vreg_hpm(hba, hba->vreg_info.vccq2);
+		ufshcd_vops_resume(hba, UFS_RUNTIME_PM);
+	}
+	ufshcd_hold(hba, false);
+	if (ufshcd_is_clkscaling_supported(hba)) {
+		cancel_work_sync(&hba->clk_scaling.suspend_work);
+		cancel_work_sync(&hba->clk_scaling.resume_work);
+		ufshcd_suspend_clkscaling(hba);
+	}
+}
+
+static void ufshcd_err_handling_unprepare(struct ufs_hba *hba)
+{
+	/* If clk_gating is held by pm ops, release it */
+	if (pm_runtime_active(hba->dev) && hba->clk_gating.held_by_pm) {
+		hba->clk_gating.held_by_pm = false;
+		ufshcd_release(hba);
+	}
+	ufshcd_release(hba);
+	if (ufshcd_is_clkscaling_supported(hba))
+		ufshcd_resume_clkscaling(hba);
+	pm_runtime_put(hba->dev);
+}
+
+#ifdef CONFIG_PM
+static void ufshcd_recover_pm_error(struct ufs_hba *hba)
+{
+	struct Scsi_Host *shost = hba->host;
+	struct scsi_device *sdev;
+	struct request_queue *q;
+	int ret;
+
+	/*
+	 * Set RPM status of hba device to RPM_ACTIVE,
+	 * this also clears its runtime error.
+	 */
+	ret = pm_runtime_set_active(hba->dev);
+	/*
+	 * If hba device had runtime error, we also need to resume those
+	 * scsi devices under hba in case any of them has failed to be
+	 * resumed due to hba runtime resume failure. This is to unblock
+	 * blk_queue_enter in case there are bios waiting inside it.
+	 */
+	if (!ret) {
+		list_for_each_entry(sdev, &shost->__devices, siblings) {
+			q = sdev->request_queue;
+			if (q->dev && (q->rpm_status == RPM_SUSPENDED ||
+						q->rpm_status == RPM_SUSPENDING))
+				pm_request_resume(q->dev);
+		}
+	}
+}
+#else
+static inline void ufshcd_recover_pm_error(struct ufs_hba *hba)
+{
+}
+#endif
+
 /**
  * ufshcd_err_handler - handle UFS errors that require s/w attention
  * @work: pointer to work structure
@@ -5580,9 +5657,7 @@ static void ufshcd_err_handler(struct work_struct *work)
 	}
 	ufshcd_set_eh_in_progress(hba);
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
-	pm_runtime_get_sync(hba->dev);
-	ufshcd_hold(hba, false);
-
+	ufshcd_err_handling_prepare(hba);
 	spin_lock_irqsave(hba->host->host_lock, flags);
 	hba->ufshcd_state = UFSHCD_STATE_RESET;
 
@@ -5661,10 +5736,12 @@ static void ufshcd_err_handler(struct work_struct *work)
 		hba->force_reset = false;
 		spin_unlock_irqrestore(hba->host->host_lock, flags);
 		err = ufshcd_reset_and_restore(hba);
-		spin_lock_irqsave(hba->host->host_lock, flags);
 		if (err)
 			dev_err(hba->dev, "%s: reset and restore failed with err %d\n",
 					__func__, err);
+		else
+			ufshcd_recover_pm_error(hba);
+		spin_lock_irqsave(hba->host->host_lock, flags);
 	}
 
 skip_err_handling:
@@ -5679,8 +5756,7 @@ static void ufshcd_err_handler(struct work_struct *work)
 	ufshcd_clear_eh_in_progress(hba);
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	ufshcd_scsi_unblock_requests(hba);
-	ufshcd_release(hba);
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_err_handling_unprepare(hba);
 }
 
 /**
@@ -8232,6 +8308,7 @@ static int ufshcd_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	 * just gate the clocks.
 	 */
 	ufshcd_hold(hba, false);
+	hba->clk_gating.held_by_pm = true;
 	hba->clk_gating.is_suspended = true;
 
 	if (hba->clk_scaling.is_allowed) {
@@ -8351,6 +8428,7 @@ static int ufshcd_suspend(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	hba->clk_gating.is_suspended = false;
 	hba->dev_info.b_rpm_dev_flush_capable = false;
 	ufshcd_release(hba);
+	hba->clk_gating.held_by_pm = false;
 out:
 	if (hba->dev_info.b_rpm_dev_flush_capable) {
 		schedule_delayed_work(&hba->rpm_dev_flush_recheck_work,
@@ -8457,6 +8535,7 @@ static int ufshcd_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 
 	/* Schedule clock gating in case of no access to UFS device yet */
 	ufshcd_release(hba);
+	hba->clk_gating.held_by_pm = false;
 
 	goto out;
 
drivers/scsi/ufs/ufshcd.h +1 -0
index 83e5cd9..585e58b 100644
@@ -354,6 +354,7 @@ struct ufs_clk_gating {
 	struct device_attribute delay_attr;
 	struct device_attribute enable_attr;
 	bool is_enabled;
+	bool held_by_pm;
 	int active_reqs;
 	struct workqueue_struct *clk_gating_workq;
 };
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
author
#11 lore raw

PATCH v6 7/8 scsi: ufs: Move dumps in IRQ handler to error handler

In reply to Can Guo

Sometime dumps in IRQ handler are heavy enough to cause system stability issues, move them to error handler.

---
 drivers/scsi/ufs/ufshcd.c | 31 +++++++++++++++----------------
 1 file changed, 15 insertions(+), 16 deletions(-)
drivers/scsi/ufs/ufshcd.c +15 -16
index c480823..b2bafa3 100644
@@ -5682,6 +5682,21 @@ 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)) {
+		bool pr_prdt = !!(hba->saved_err & SYSTEM_BUS_FATAL_ERROR);
+
+		dev_err(hba->dev, "%s: saved_err 0x%x saved_uic_err 0x%x\n",
+				__func__, hba->saved_err, hba->saved_uic_err);
+		spin_unlock_irqrestore(hba->host->host_lock, flags);
+		ufshcd_print_host_state(hba);
+		ufshcd_print_pwr_info(hba);
+		ufshcd_print_host_regs(hba);
+		ufshcd_print_tmrs(hba, hba->outstanding_tasks);
+		ufshcd_print_trs(hba, hba->outstanding_reqs, pr_prdt);
+		spin_lock_irqsave(hba->host->host_lock, flags);
+	}
+
 	/*
 	 * if host reset is required then skip clearing the pending
 	 * transfers forcefully because they will get cleared during
@@ -5900,22 +5915,6 @@ static irqreturn_t ufshcd_check_errors(struct ufs_hba *hba)
 
 		/* block commands from scsi mid-layer */
 		ufshcd_scsi_block_requests(hba);
-
-		/* dump controller state before resetting */
-		if (hba->saved_err & (INT_FATAL_ERRORS | UIC_ERROR)) {
-			bool pr_prdt = !!(hba->saved_err &
-					SYSTEM_BUS_FATAL_ERROR);
-
-			dev_err(hba->dev, "%s: saved_err 0x%x saved_uic_err 0x%x\n",
-					__func__, hba->saved_err,
-					hba->saved_uic_err);
-
-			ufshcd_print_host_regs(hba);
-			ufshcd_print_pwr_info(hba);
-			ufshcd_print_tmrs(hba, hba->outstanding_tasks);
-			ufshcd_print_trs(hba, hba->outstanding_reqs,
-					pr_prdt);
-		}
 		ufshcd_schedule_eh_work(hba);
 		retval |= IRQ_HANDLED;
 	}
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
author
#12 lore raw

PATCH v6 8/8 scsi: ufs: Fix a racing problem btw error handler and runtime PM ops

In reply to Can Guo

Current IRQ handler blocks scsi requests before scheduling eh_work, when error handler calls pm_runtime_get_sync, if ufshcd_suspend/resume sends a scsi cmd, most likely the SSU cmd, since scsi requests are blocked, pm_runtime_get_sync() will never return because ufshcd_suspend/reusme is blocked by the scsi cmd. Some changes and code re-arrangement can be made to resolve it.

o In queuecommand path, hba->ufshcd_state check and ufshcd_send_command should stay into the same spin lock. This is to make sure that no more commands leak into doorbell after hba->ufshcd_state is changed.
o Don't block scsi requests before scheduling eh_work, let error handler block scsi requests when it is ready to start error recovery.
o Don't let scsi layer keep requeuing the scsi cmds sent from hba runtime PM ops, let them pass or fail them. Let them pass if eh_work is scheduled due to non-fatal errors. Fail them fail if eh_work is scheduled due to fatal errors, otherwise the cmds may eventually time out since UFS is in bad state, which gets error handler blocked for too long. If we fail the scsi cmds sent from hba runtime PM ops, hba runtime PM ops fails too, but it does not hurt since error handler can recover hba runtime PM error.

---
 drivers/scsi/ufs/ufshcd.c | 84 +++++++++++++++++++++++++++--------------------
 1 file changed, 49 insertions(+), 35 deletions(-)
drivers/scsi/ufs/ufshcd.c +49 -35
index b2bafa3..9c8c43f 100644
@@ -126,7 +126,8 @@ enum {
 	UFSHCD_STATE_RESET,
 	UFSHCD_STATE_ERROR,
 	UFSHCD_STATE_OPERATIONAL,
-	UFSHCD_STATE_EH_SCHEDULED,
+	UFSHCD_STATE_EH_SCHEDULED_FATAL,
+	UFSHCD_STATE_EH_SCHEDULED_NON_FATAL,
 };
 
 /* UFSHCD error handling flags */
@@ -2515,34 +2516,6 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
 	if (!down_read_trylock(&hba->clk_scaling_lock))
 		return SCSI_MLQUEUE_HOST_BUSY;
 
-	spin_lock_irqsave(hba->host->host_lock, flags);
-	switch (hba->ufshcd_state) {
-	case UFSHCD_STATE_OPERATIONAL:
-		break;
-	case UFSHCD_STATE_EH_SCHEDULED:
-	case UFSHCD_STATE_RESET:
-		err = SCSI_MLQUEUE_HOST_BUSY;
-		goto out_unlock;
-	case UFSHCD_STATE_ERROR:
-		set_host_byte(cmd, DID_ERROR);
-		cmd->scsi_done(cmd);
-		goto out_unlock;
-	default:
-		dev_WARN_ONCE(hba->dev, 1, "%s: invalid state %d\n",
-				__func__, hba->ufshcd_state);
-		set_host_byte(cmd, DID_BAD_TARGET);
-		cmd->scsi_done(cmd);
-		goto out_unlock;
-	}
-
-	/* if error handling is in progress, don't issue commands */
-	if (ufshcd_eh_in_progress(hba)) {
-		set_host_byte(cmd, DID_ERROR);
-		cmd->scsi_done(cmd);
-		goto out_unlock;
-	}
-	spin_unlock_irqrestore(hba->host->host_lock, flags);
-
 	hba->req_abort_count = 0;
 
 	err = ufshcd_hold(hba, true);
@@ -2578,11 +2551,50 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
 	/* Make sure descriptors are ready before ringing the doorbell */
 	wmb();
 
-	/* issue command to the controller */
 	spin_lock_irqsave(hba->host->host_lock, flags);
+	switch (hba->ufshcd_state) {
+	case UFSHCD_STATE_OPERATIONAL:
+	case UFSHCD_STATE_EH_SCHEDULED_NON_FATAL:
+		break;
+	case UFSHCD_STATE_EH_SCHEDULED_FATAL:
+		/*
+		 * If we are here, eh_work is either scheduled or running.
+		 * Before eh_work sets ufshcd_state to STATE_RESET, it flushes
+		 * runtime PM ops by calling pm_runtime_get_sync(). If a scsi
+		 * cmd, e.g. the SSU cmd, is sent by PM ops, it can never be
+		 * finished if we let SCSI layer keep retrying it, which gets
+		 * eh_work stuck forever. Neither can we let it pass, because
+		 * ufs now is not in good status, so the SSU cmd may eventually
+		 * time out, blocking eh_work for too long. So just let it fail.
+		 */
+		if (hba->pm_op_in_progress) {
+			hba->force_reset = true;
+			set_host_byte(cmd, DID_BAD_TARGET);
+			goto out_compl_cmd;
+		}
+	case UFSHCD_STATE_RESET:
+		err = SCSI_MLQUEUE_HOST_BUSY;
+		goto out_compl_cmd;
+	case UFSHCD_STATE_ERROR:
+		set_host_byte(cmd, DID_ERROR);
+		goto out_compl_cmd;
+	default:
+		dev_WARN_ONCE(hba->dev, 1, "%s: invalid state %d\n",
+				__func__, hba->ufshcd_state);
+		set_host_byte(cmd, DID_BAD_TARGET);
+		goto out_compl_cmd;
+	}
 	ufshcd_send_command(hba, tag);
-out_unlock:
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
+	goto out;
+
+out_compl_cmd:
+	scsi_dma_unmap(lrbp->cmd);
+	lrbp->cmd = NULL;
+	spin_unlock_irqrestore(hba->host->host_lock, flags);
+	ufshcd_release(hba);
+	if (!err)
+		cmd->scsi_done(cmd);
 out:
 	up_read(&hba->clk_scaling_lock);
 	return err;
@@ -5553,7 +5565,11 @@ static inline void ufshcd_schedule_eh_work(struct ufs_hba *hba)
 {
 	/* handle fatal errors only when link is not in error state */
 	if (hba->ufshcd_state != UFSHCD_STATE_ERROR) {
-		hba->ufshcd_state = UFSHCD_STATE_EH_SCHEDULED;
+		if (hba->force_reset || ufshcd_is_link_broken(hba) ||
+		    ufshcd_is_saved_err_fatal(hba))
+			hba->ufshcd_state = UFSHCD_STATE_EH_SCHEDULED_FATAL;
+		else
+			hba->ufshcd_state = UFSHCD_STATE_EH_SCHEDULED_NON_FATAL;
 		queue_work(hba->eh_wq, &hba->eh_work);
 	}
 }
@@ -5659,6 +5675,7 @@ static void ufshcd_err_handler(struct work_struct *work)
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	ufshcd_err_handling_prepare(hba);
 	spin_lock_irqsave(hba->host->host_lock, flags);
+	ufshcd_scsi_block_requests(hba);
 	hba->ufshcd_state = UFSHCD_STATE_RESET;
 
 	/* Complete requests that have door-bell cleared by h/w */
@@ -5912,9 +5929,6 @@ static irqreturn_t ufshcd_check_errors(struct ufs_hba *hba)
 		 */
 		hba->saved_err |= hba->errors;
 		hba->saved_uic_err |= hba->uic_error;
-
-		/* block commands from scsi mid-layer */
-		ufshcd_scsi_block_requests(hba);
 		ufshcd_schedule_eh_work(hba);
 		retval |= IRQ_HANDLED;
 	}
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.