Skip to contentSkip to conversation

Enable power management for ufs wlun

PATCH Applied Asutosh Das 4 participants 7 messages Apr 24, 2021 – May 11, 2021

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

v21 cover Enable power management for ufs wlun

This patch attempts to fix a deadlock in ufs while sending SSU. Recently, blk_queue_enter() added a check to not process requests if the queue is suspended. That leads to a resume of the associated device which is suspended. In ufs, that device is ufs device wlun and it's parent is ufs_hba. This resume tries to resume ufs device wlun which in turn tries to resume ufs_hba, which is already in the process of suspending, thus causing a deadlock.

This patch takes care of:
* Suspending the ufs device lun only after all other luns are suspended
* Sending SSU during ufs device wlun suspend
* Clearing uac for rpmb and ufs device wlun
* Not sending commands to the device during host suspend

v20 -> v21:
- Addressed Adrian's comments

v19 -> v20:
- Addressed Adrian's comments
- Addressed Bart's comments

v18 -> v19:
- Addressed Adrian's comments
- Fixed smatch warnings

v17 -> v18:
- Addressed Adrian's comments

v16 -> v17:
- Addressed Adrian's & Daejun's comments

v15 -> v16:
- Brought back the missing changes
* Added scsi_autopm_[get/put] to ufs_debugfs[get/put]_user_access()
* Fix ufshcd_wl_poweroff()

v14 -> v15:
- Rebased on 5.13/scsi-staging

v13 -> v14:
- Addressed Adrian's comments
* Rebased it on top of scsi-next
* Added scsi_autopm_[get/put] to ufs_debugfs[get/put]_user_access()
* Resume the device in ufshcd_remove()
* Unregister ufs_rpmb_wlun before ufs_dev_wlun
* hba->shutting_down moved to ufshcd_wl_shutdown()

v12 -> v13:
- Addressed Adrian's comments
* Paired pm_runtime_get_noresume() with pm_runtime_put()
* no rpm_autosuspend for ufs device wlun
* Moved runtime-pm init functionality to ufshcd_wl_probe()
- Addressed Bart's comments
* Expanded abbrevs in commit message

v11 -> v12:
- Addressed Adrian's comments
* Fixed ahit for Mediatek driver
* Fixed error handling in ufshcd_core_init()
* Tested this patch and the issue is still seen.

v10 -> v11:
- Fixed supplier suspending before consumer race
- Addressed Adrian's comments
* Added proper resume/suspend cb to ufshcd_auto_hibern8_update()
* Cosmetic changes to ufshcd-pci.c
* Cleaned up ufshcd_system_suspend()
* Added ufshcd_debugfs_eh_exit to ufshcd_core_init()

v9 -> v10:
- Addressed Adrian's comments
* Moved suspend/resume vops to __ufshcd_wl_[suspend/resume]()
* Added correct resume in ufs_bsg

v8 -> v9:
- Addressed Adrian's comments
* Moved link transition to __ufshcd_wl_[suspend/resume]()
* Fixed the other minor comments

v7 -> v8:
- Addressed Adrian's comments
* Removed separate autosuspend delay for ufs-device lun
* Fixed the ee handler getting scheduled during pm
* Always runtime resume in suspend_prepare()
* Added CONFIG_PM_SLEEP where needed

v6 -> v7:
* Resume the ufs device before shutting it down

v5 -> v6:
- Addressed Adrian's comments
* Added complete() cb
* Added suspend_prepare() and complete() to all drivers
* Moved suspend_prepare() and complete() to ufshcd
* .poweroff() uses ufhcd_wl_poweroff()
* Removed several forward declarations
* Moved scsi_register_driver() to ufshcd_core_init()

v4 -> v5:
- Addressed Adrian's comments
* Used the rpmb driver contributed by Adrian
* Runtime-resume the ufs device during suspend to honor spm-lvl
* Unregister the scsi_driver in ufshcd_remove()
* Currently shutdown() puts the ufs device to power-down mode so, just removed ufshcd_pci_poweroff()
* Quiesce the scsi device during shutdown instead of remove

v3 RFC -> v4:
- Addressed Bart's comments
* Except that I didn't get any checkpatch failures
- Addressed Avri's comments
- Addressed Adrian's comments
* Added a check for deepsleep power mode
* Removed a couple of forward declarations
* Didn't separate the scsi drivers because in rpmb case it just sends uac in resume and it seemed pretty neat to me.
- Added sysfs changes to resume the devices before accessing

Asutosh Das (2):
scsi: ufs: Enable power management for wlun
ufs: sysfs: Resume the proper scsi device

 drivers/scsi/ufs/cdns-pltfrm.c     |   2 +
 drivers/scsi/ufs/tc-dwc-g210-pci.c |   2 +
 drivers/scsi/ufs/ufs-debugfs.c     |   6 +-
 drivers/scsi/ufs/ufs-debugfs.h     |   2 +-
 drivers/scsi/ufs/ufs-exynos.c      |   2 +
 drivers/scsi/ufs/ufs-hisi.c        |   2 +
 drivers/scsi/ufs/ufs-mediatek.c    |  12 +-
 drivers/scsi/ufs/ufs-qcom.c        |   2 +
 drivers/scsi/ufs/ufs-sysfs.c       |  24 +-
 drivers/scsi/ufs/ufs_bsg.c         |   6 +-
 drivers/scsi/ufs/ufshcd-pci.c      |  36 +-
 drivers/scsi/ufs/ufshcd.c          | 700 ++++++++++++++++++++++++++-----------
 drivers/scsi/ufs/ufshcd.h          |  33 ++
 include/trace/events/ufs.h         |  20 ++
 14 files changed, 585 insertions(+), 264 deletions(-)
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
#1 lore raw

PATCH v21 1/2 scsi: ufs: Enable power management for wlun

During runtime-suspend of ufs host, the scsi devices are already suspended and so are the queues associated with them. But the ufs host sends SSU (START_STOP_UNIT) to wlun
during its runtime-suspend.
During the process blk_queue_enter checks if the queue is not in suspended state. If so, it waits for the queue to resume, and never comes out of it.
The commit
(d55d15a33: scsi: block: Do not accept any requests while suspended) adds the check if the queue is in suspended state in blk_queue_enter().

Call trace:
 __switch_to+0x174/0x2c4
 __schedule+0x478/0x764
 schedule+0x9c/0xe0
 blk_queue_enter+0x158/0x228
 blk_mq_alloc_request+0x40/0xa4
 blk_get_request+0x2c/0x70
 __scsi_execute+0x60/0x1c4
 ufshcd_set_dev_pwr_mode+0x124/0x1e4
 ufshcd_suspend+0x208/0x83c
 ufshcd_runtime_suspend+0x40/0x154
 ufshcd_pltfrm_runtime_suspend+0x14/0x20
 pm_generic_runtime_suspend+0x28/0x3c
 __rpm_callback+0x80/0x2a4
 rpm_suspend+0x308/0x614
 rpm_idle+0x158/0x228
 pm_runtime_work+0x84/0xac
 process_one_work+0x1f0/0x470
 worker_thread+0x26c/0x4c8
 kthread+0x13c/0x320
 ret_from_fork+0x10/0x18

Fix this by registering ufs device wlun as a scsi driver and registering it for block runtime-pm. Also make this as a supplier for all other luns. That way, this device wlun suspends after all the consumers and resumes after
hba resumes. This also registers a new scsi driver for rpmb wlun. This new driver is mostly used to clear rpmb uac.

Fixed smatch warnings:

---
 drivers/scsi/ufs/cdns-pltfrm.c     |   2 +
 drivers/scsi/ufs/tc-dwc-g210-pci.c |   2 +
 drivers/scsi/ufs/ufs-debugfs.c     |   6 +-
 drivers/scsi/ufs/ufs-debugfs.h     |   2 +-
 drivers/scsi/ufs/ufs-exynos.c      |   2 +
 drivers/scsi/ufs/ufs-hisi.c        |   2 +
 drivers/scsi/ufs/ufs-mediatek.c    |  12 +-
 drivers/scsi/ufs/ufs-qcom.c        |   2 +
 drivers/scsi/ufs/ufs_bsg.c         |   6 +-
 drivers/scsi/ufs/ufshcd-pci.c      |  36 +-
 drivers/scsi/ufs/ufshcd.c          | 700 ++++++++++++++++++++++++++-----------
 drivers/scsi/ufs/ufshcd.h          |  33 ++
 include/trace/events/ufs.h         |  20 ++
 13 files changed, 573 insertions(+), 252 deletions(-)
drivers/scsi/ufs/cdns-pltfrm.c +2 -0
index 13d9204..908ff39 100644
@@ -323,6 +323,8 @@ static const struct dev_pm_ops cdns_ufs_dev_pm_ops = {
 	.runtime_suspend = ufshcd_pltfrm_runtime_suspend,
 	.runtime_resume  = ufshcd_pltfrm_runtime_resume,
 	.runtime_idle    = ufshcd_pltfrm_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static struct platform_driver cdns_ufs_pltfrm_driver = {
drivers/scsi/ufs/tc-dwc-g210-pci.c +2 -0
index 67a6a61..ec4589a 100644
@@ -148,6 +148,8 @@ static const struct dev_pm_ops tc_dwc_g210_pci_pm_ops = {
 	.runtime_suspend = tc_dwc_g210_pci_runtime_suspend,
 	.runtime_resume  = tc_dwc_g210_pci_runtime_resume,
 	.runtime_idle    = tc_dwc_g210_pci_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static const struct pci_device_id tc_dwc_g210_pci_tbl[] = {
drivers/scsi/ufs/ufs-debugfs.c +3 -3
index ced9ef4..4e1ff20 100644
@@ -13,7 +13,7 @@ void __init ufs_debugfs_init(void)
 	ufs_debugfs_root = debugfs_create_dir("ufshcd", NULL);
 }
 
-void __exit ufs_debugfs_exit(void)
+void ufs_debugfs_exit(void)
 {
 	debugfs_remove_recursive(ufs_debugfs_root);
 }
@@ -60,14 +60,14 @@ __acquires(&hba->host_sem)
 		up(&hba->host_sem);
 		return -EBUSY;
 	}
-	pm_runtime_get_sync(hba->dev);
+	ufshcd_rpm_get_sync(hba);
 	return 0;
 }
 
 static void ufs_debugfs_put_user_access(struct ufs_hba *hba)
 __releases(&hba->host_sem)
 {
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_rpm_put_sync(hba);
 	up(&hba->host_sem);
 }
 
drivers/scsi/ufs/ufs-debugfs.h +1 -1
index 3ca29d3..97548a3 100644
@@ -9,7 +9,7 @@ struct ufs_hba;
 
 #ifdef CONFIG_DEBUG_FS
 void __init ufs_debugfs_init(void);
-void __exit ufs_debugfs_exit(void);
+void ufs_debugfs_exit(void);
 void ufs_debugfs_hba_init(struct ufs_hba *hba);
 void ufs_debugfs_hba_exit(struct ufs_hba *hba);
 void ufs_debugfs_exception_event(struct ufs_hba *hba, u16 status);
drivers/scsi/ufs/ufs-exynos.c +2 -0
index 70647ea..8b31efb 100644
@@ -1267,6 +1267,8 @@ static const struct dev_pm_ops exynos_ufs_pm_ops = {
 	.runtime_suspend = ufshcd_pltfrm_runtime_suspend,
 	.runtime_resume  = ufshcd_pltfrm_runtime_resume,
 	.runtime_idle    = ufshcd_pltfrm_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static struct platform_driver exynos_ufs_pltform = {
drivers/scsi/ufs/ufs-hisi.c +2 -0
index 0aa5813..a473349 100644
@@ -574,6 +574,8 @@ static const struct dev_pm_ops ufs_hisi_pm_ops = {
 	.runtime_suspend = ufshcd_pltfrm_runtime_suspend,
 	.runtime_resume  = ufshcd_pltfrm_runtime_resume,
 	.runtime_idle    = ufshcd_pltfrm_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static struct platform_driver ufs_hisi_pltform = {
drivers/scsi/ufs/ufs-mediatek.c +6 -6
index a981f26..9912e20 100644
@@ -810,12 +810,10 @@ static int ufs_mtk_post_link(struct ufs_hba *hba)
 	/* enable unipro clock gating feature */
 	ufs_mtk_cfg_unipro_cg(hba, true);
 
-	/* configure auto-hibern8 timer to 10ms */
-	if (ufshcd_is_auto_hibern8_supported(hba)) {
-		ufshcd_auto_hibern8_update(hba,
-			FIELD_PREP(UFSHCI_AHIBERN8_TIMER_MASK, 10) |
-			FIELD_PREP(UFSHCI_AHIBERN8_SCALE_MASK, 3));
-	}
+	/* will be configured during probe hba */
+	if (ufshcd_is_auto_hibern8_supported(hba))
+		hba->ahit = FIELD_PREP(UFSHCI_AHIBERN8_TIMER_MASK, 10) |
+			FIELD_PREP(UFSHCI_AHIBERN8_SCALE_MASK, 3);
 
 	ufs_mtk_setup_clk_gating(hba);
 
@@ -1097,6 +1095,8 @@ static const struct dev_pm_ops ufs_mtk_pm_ops = {
 	.runtime_suspend = ufshcd_pltfrm_runtime_suspend,
 	.runtime_resume  = ufshcd_pltfrm_runtime_resume,
 	.runtime_idle    = ufshcd_pltfrm_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static struct platform_driver ufs_mtk_pltform = {
drivers/scsi/ufs/ufs-qcom.c +2 -0
index 9b711d6..9fc218f 100644
@@ -1556,6 +1556,8 @@ static const struct dev_pm_ops ufs_qcom_pm_ops = {
 	.runtime_suspend = ufshcd_pltfrm_runtime_suspend,
 	.runtime_resume  = ufshcd_pltfrm_runtime_resume,
 	.runtime_idle    = ufshcd_pltfrm_runtime_idle,
+	.prepare	 = ufshcd_suspend_prepare,
+	.complete	 = ufshcd_resume_complete,
 };
 
 static struct platform_driver ufs_qcom_pltform = {
drivers/scsi/ufs/ufs_bsg.c +3 -3
index 5b2bc1a..39bf204 100644
@@ -97,7 +97,7 @@ static int ufs_bsg_request(struct bsg_job *job)
 
 	bsg_reply->reply_payload_rcv_len = 0;
 
-	pm_runtime_get_sync(hba->dev);
+	ufshcd_rpm_get_sync(hba);
 
 	msgcode = bsg_request->msgcode;
 	switch (msgcode) {
@@ -106,7 +106,7 @@ static int ufs_bsg_request(struct bsg_job *job)
 		ret = ufs_bsg_alloc_desc_buffer(hba, job, &desc_buff,
 						&desc_len, desc_op);
 		if (ret) {
-			pm_runtime_put_sync(hba->dev);
+			ufshcd_rpm_put_sync(hba);
 			goto out;
 		}
 
@@ -138,7 +138,7 @@ static int ufs_bsg_request(struct bsg_job *job)
 		break;
 	}
 
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_rpm_put_sync(hba);
 
 	if (!desc_buff)
 		goto out;
drivers/scsi/ufs/ufshcd-pci.c +5 -31
index 23ee828..e6c334b 100644
@@ -410,29 +410,6 @@ static int ufshcd_pci_resume(struct device *dev)
 	return ufshcd_system_resume(dev_get_drvdata(dev));
 }
 
-/**
- * ufshcd_pci_poweroff - suspend-to-disk poweroff function
- * @dev: pointer to PCI device handle
- *
- * Returns 0 if successful
- * Returns non-zero otherwise
- */
-static int ufshcd_pci_poweroff(struct device *dev)
-{
-	struct ufs_hba *hba = dev_get_drvdata(dev);
-	int spm_lvl = hba->spm_lvl;
-	int ret;
-
-	/*
-	 * For poweroff we need to set the UFS device to PowerDown mode.
-	 * Force spm_lvl to ensure that.
-	 */
-	hba->spm_lvl = 5;
-	ret = ufshcd_system_suspend(hba);
-	hba->spm_lvl = spm_lvl;
-	return ret;
-}
-
 #endif /* !CONFIG_PM_SLEEP */
 
 #ifdef CONFIG_PM
@@ -533,17 +510,14 @@ ufshcd_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 }
 
 static const struct dev_pm_ops ufshcd_pci_pm_ops = {
-#ifdef CONFIG_PM_SLEEP
-	.suspend	= ufshcd_pci_suspend,
-	.resume		= ufshcd_pci_resume,
-	.freeze		= ufshcd_pci_suspend,
-	.thaw		= ufshcd_pci_resume,
-	.poweroff	= ufshcd_pci_poweroff,
-	.restore	= ufshcd_pci_resume,
-#endif
 	SET_RUNTIME_PM_OPS(ufshcd_pci_runtime_suspend,
 			   ufshcd_pci_runtime_resume,
 			   ufshcd_pci_runtime_idle)
+	SET_SYSTEM_SLEEP_PM_OPS(ufshcd_pci_suspend, ufshcd_pci_resume)
+#ifdef CONFIG_PM_SLEEP
+	.prepare	= ufshcd_suspend_prepare,
+	.complete	= ufshcd_resume_complete,
+#endif
 };
 
 static const struct pci_device_id ufshcd_pci_tbl[] = {
drivers/scsi/ufs/ufshcd.c +492 -208

1,037 lines not on this page · raw message

drivers/scsi/ufs/ufshcd.h +33 -0
index 5eb66a8..b8b4fe2 100644
@@ -72,6 +72,8 @@ enum ufs_event_type {
 	UFS_EVT_LINK_STARTUP_FAIL,
 	UFS_EVT_RESUME_ERR,
 	UFS_EVT_SUSPEND_ERR,
+	UFS_EVT_WL_SUSP_ERR,
+	UFS_EVT_WL_RES_ERR,
 
 	/* abnormal events */
 	UFS_EVT_DEV_RESET,
@@ -807,6 +809,7 @@ struct ufs_hba {
 	struct list_head clk_list_head;
 
 	bool wlun_dev_clr_ua;
+	bool wlun_rpmb_clr_ua;
 
 	/* Number of requests aborts */
 	int req_abort_count;
@@ -846,6 +849,9 @@ struct ufs_hba {
 	struct delayed_work debugfs_ee_work;
 	u32 debugfs_ee_rate_limit_ms;
 #endif
+	u32 luns_avail;
+	bool complete_put;
+	bool rpmb_complete_put;
 };
 
 /* Returns true if clocks can be gated. Otherwise false */
@@ -1105,6 +1111,8 @@ int ufshcd_exec_raw_upiu_cmd(struct ufs_hba *hba,
 			     enum query_opcode desc_op);
 
 int ufshcd_wb_toggle(struct ufs_hba *hba, bool enable);
+int ufshcd_suspend_prepare(struct device *dev);
+void ufshcd_resume_complete(struct device *dev);
 
 /* Wrapper functions for safely calling variant operations */
 static inline const char *ufshcd_get_var_name(struct ufs_hba *hba)
@@ -1309,4 +1317,29 @@ static inline int ufshcd_update_ee_usr_mask(struct ufs_hba *hba,
 					&hba->ee_drv_mask, set, clr);
 }
 
+static inline int ufshcd_rpm_get_sync(struct ufs_hba *hba)
+{
+	return pm_runtime_get_sync(&hba->sdev_ufs_device->sdev_gendev);
+}
+
+static inline int ufshcd_rpm_put_sync(struct ufs_hba *hba)
+{
+	return pm_runtime_put_sync(&hba->sdev_ufs_device->sdev_gendev);
+}
+
+static inline int ufshcd_rpm_put(struct ufs_hba *hba)
+{
+	return pm_runtime_put(&hba->sdev_ufs_device->sdev_gendev);
+}
+
+static inline int ufshcd_rpmb_rpm_get_sync(struct ufs_hba *hba)
+{
+	return pm_runtime_get_sync(&hba->sdev_rpmb->sdev_gendev);
+}
+
+static inline int ufshcd_rpmb_rpm_put(struct ufs_hba *hba)
+{
+	return pm_runtime_put(&hba->sdev_rpmb->sdev_gendev);
+}
+
 #endif /* End of Header */
include/trace/events/ufs.h +20 -0
index 1cb6f1a..599739e 100644
@@ -246,6 +246,26 @@ DEFINE_EVENT(ufshcd_template, ufshcd_init,
 		      int dev_state, int link_state),
 	     TP_ARGS(dev_name, err, usecs, dev_state, link_state));
 
+DEFINE_EVENT(ufshcd_template, ufshcd_wl_suspend,
+	     TP_PROTO(const char *dev_name, int err, s64 usecs,
+		      int dev_state, int link_state),
+	     TP_ARGS(dev_name, err, usecs, dev_state, link_state));
+
+DEFINE_EVENT(ufshcd_template, ufshcd_wl_resume,
+	     TP_PROTO(const char *dev_name, int err, s64 usecs,
+		      int dev_state, int link_state),
+	     TP_ARGS(dev_name, err, usecs, dev_state, link_state));
+
+DEFINE_EVENT(ufshcd_template, ufshcd_wl_runtime_suspend,
+	     TP_PROTO(const char *dev_name, int err, s64 usecs,
+		      int dev_state, int link_state),
+	     TP_ARGS(dev_name, err, usecs, dev_state, link_state));
+
+DEFINE_EVENT(ufshcd_template, ufshcd_wl_runtime_resume,
+	     TP_PROTO(const char *dev_name, int err, s64 usecs,
+		      int dev_state, int link_state),
+	     TP_ARGS(dev_name, err, usecs, dev_state, link_state));
+
 TRACE_EVENT(ufshcd_command,
 	TP_PROTO(const char *dev_name, enum ufs_trace_str_t str_t,
 		 unsigned int tag, u32 doorbell, int transfer_len, u32 intr,
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
Reviewed-by on v21 1/2
#2 lore raw

Re: [PATCH v21 1/2] scsi: ufs: Enable power management for wlun

Quoted 40 lines · Asutosh Das

During runtime-suspend of ufs host, the scsi devices are already suspended and so are the queues associated with them. But the ufs host sends SSU (START_STOP_UNIT) to wlun
during its runtime-suspend.
During the process blk_queue_enter checks if the queue is not in suspended state. If so, it waits for the queue to resume, and never comes out of it.
The commit
(d55d15a33: scsi: block: Do not accept any requests while suspended) adds the check if the queue is in suspended state in blk_queue_enter().

Call trace:
 __switch_to+0x174/0x2c4
 __schedule+0x478/0x764
 schedule+0x9c/0xe0
 blk_queue_enter+0x158/0x228
 blk_mq_alloc_request+0x40/0xa4
 blk_get_request+0x2c/0x70
 __scsi_execute+0x60/0x1c4
 ufshcd_set_dev_pwr_mode+0x124/0x1e4
 ufshcd_suspend+0x208/0x83c
 ufshcd_runtime_suspend+0x40/0x154
 ufshcd_pltfrm_runtime_suspend+0x14/0x20
 pm_generic_runtime_suspend+0x28/0x3c
 __rpm_callback+0x80/0x2a4
 rpm_suspend+0x308/0x614
 rpm_idle+0x158/0x228
 pm_runtime_work+0x84/0xac
 process_one_work+0x1f0/0x470
 worker_thread+0x26c/0x4c8
 kthread+0x13c/0x320
 ret_from_fork+0x10/0x18

Fix this by registering ufs device wlun as a scsi driver and registering it for block runtime-pm. Also make this as a supplier for all other luns. That way, this device wlun suspends after all the consumers and resumes after
hba resumes. This also registers a new scsi driver for rpmb wlun. This new driver is mostly used to clear rpmb uac.

Quoted 7 lines

Fixed smatch warnings:

Note we don't use regulators, clocks, clock scaling or gating. Although those things do look OK.
Notwithstanding that:

#3 lore raw

PATCH v21 2/2 ufs: sysfs: Resume the proper scsi device

In reply to Asutosh Das

Resumes the actual scsi device the unit descriptor of which is being accessed instead of the hba alone.

---
 drivers/scsi/ufs/ufs-sysfs.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
drivers/scsi/ufs/ufs-sysfs.c +12 -12
index d7c3cff..4d9d4d8 100644
@@ -245,9 +245,9 @@ static ssize_t wb_on_store(struct device *dev, struct device_attribute *attr,
 		goto out;
 	}
 
-	pm_runtime_get_sync(hba->dev);
+	ufshcd_rpm_get_sync(hba);
 	res = ufshcd_wb_toggle(hba, wb_enable);
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_rpm_put_sync(hba);
 out:
 	up(&hba->host_sem);
 	return res < 0 ? res : count;
@@ -297,10 +297,10 @@ static ssize_t ufs_sysfs_read_desc_param(struct ufs_hba *hba,
 		goto out;
 	}
 
-	pm_runtime_get_sync(hba->dev);
+	ufshcd_rpm_get_sync(hba);
 	ret = ufshcd_read_desc_param(hba, desc_id, desc_index,
 				param_offset, desc_buf, param_size);
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_rpm_put_sync(hba);
 	if (ret) {
 		ret = -EINVAL;
 		goto out;
@@ -678,7 +678,7 @@ static ssize_t _name##_show(struct device *dev,				\
 		up(&hba->host_sem);					\
 		return -ENOMEM;						\
 	}								\
-	pm_runtime_get_sync(hba->dev);					\
+	ufshcd_rpm_get_sync(hba);					\
 	ret = ufshcd_query_descriptor_retry(hba,			\
 		UPIU_QUERY_OPCODE_READ_DESC, QUERY_DESC_IDN_DEVICE,	\
 		0, 0, desc_buf, &desc_len);				\
@@ -695,7 +695,7 @@ static ssize_t _name##_show(struct device *dev,				\
 		goto out;						\
 	ret = sysfs_emit(buf, "%s\n", desc_buf);			\
 out:									\
-	pm_runtime_put_sync(hba->dev);					\
+	ufshcd_rpm_put_sync(hba);					\
 	kfree(desc_buf);						\
 	up(&hba->host_sem);						\
 	return ret;							\
@@ -744,10 +744,10 @@ static ssize_t _name##_show(struct device *dev,				\
 	}								\
 	if (ufshcd_is_wb_flags(QUERY_FLAG_IDN##_uname))			\
 		index = ufshcd_wb_get_query_index(hba);			\
-	pm_runtime_get_sync(hba->dev);					\
+	ufshcd_rpm_get_sync(hba);					\
 	ret = ufshcd_query_flag(hba, UPIU_QUERY_OPCODE_READ_FLAG,	\
 		QUERY_FLAG_IDN##_uname, index, &flag);			\
-	pm_runtime_put_sync(hba->dev);					\
+	ufshcd_rpm_put_sync(hba);					\
 	if (ret) {							\
 		ret = -EINVAL;						\
 		goto out;						\
@@ -813,10 +813,10 @@ static ssize_t _name##_show(struct device *dev,				\
 	}								\
 	if (ufshcd_is_wb_attrs(QUERY_ATTR_IDN##_uname))			\
 		index = ufshcd_wb_get_query_index(hba);			\
-	pm_runtime_get_sync(hba->dev);					\
+	ufshcd_rpm_get_sync(hba);					\
 	ret = ufshcd_query_attr(hba, UPIU_QUERY_OPCODE_READ_ATTR,	\
 		QUERY_ATTR_IDN##_uname, index, 0, &value);		\
-	pm_runtime_put_sync(hba->dev);					\
+	ufshcd_rpm_put_sync(hba);					\
 	if (ret) {							\
 		ret = -EINVAL;						\
 		goto out;						\
@@ -964,10 +964,10 @@ static ssize_t dyn_cap_needed_attribute_show(struct device *dev,
 		goto out;
 	}
 
-	pm_runtime_get_sync(hba->dev);
+	ufshcd_rpm_get_sync(hba);
 	ret = ufshcd_query_attr(hba, UPIU_QUERY_OPCODE_READ_ATTR,
 		QUERY_ATTR_IDN_DYN_CAP_NEEDED, lun, 0, &value);
-	pm_runtime_put_sync(hba->dev);
+	ufshcd_rpm_put_sync(hba);
 	if (ret) {
 		ret = -EINVAL;
 		goto out;
Signature · 1 line
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.
#4 lore raw
In reply to Asutosh Das

Asutosh,

Quoted 1 line

This patch attempts to fix a deadlock in ufs while sending SSU.

Applied to 5.14/scsi-staging, thanks!

Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering
#5 lore raw
Quoted 1 line

Asutosh,

In several places there is now a reference: "Also check the description of __ufshcd_wl_suspend()" But no such description is to found.
Might want to add it in a supplementary patch?

Thanks,
Avri

Quoted 7 lines
Quoted 2 lines

This patch attempts to fix a deadlock in ufs while sending SSU.

Applied to 5.14/scsi-staging, thanks!

--
Martin K. Petersen Oracle Linux Engineering

#6 lore raw
In reply to Asutosh Das
Quoted 9 lines · Asutosh Das

This patch attempts to fix a deadlock in ufs while sending SSU. Recently, blk_queue_enter() added a check to not process requests if the queue is suspended. That leads to a resume of the associated device which is suspended. In ufs, that device is ufs device wlun and it's parent is ufs_hba. This resume tries to resume ufs device wlun which in turn tries to resume ufs_hba, which is already in the process of suspending, thus causing a deadlock.

[...]

Applied to 5.14/scsi-queue, thanks!

[1/2] scsi: ufs: Enable power management for wlun

      https://git.kernel.org/mkp/scsi/c/b294ff3e3449
[2/2] ufs: sysfs: Resume the proper scsi device
      https://git.kernel.org/mkp/scsi/c/aef80fd1da32
Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering