Skip to contentSkip to conversation

propose some UFS fixes

PATCH Superseded Jaegeuk Kim 3 participants 17 messages Oct 20, 2020 – Oct 26, 2020

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

Change log from v1:
- remove clkgating_enable check in __ufshcd_release
- use __uhfshcd_release instead of active_req.

#1 lore raw

PATCH v2 1/5 scsi: ufs: atomic update for clkgating_enable

From: Jaegeuk Kim <[email protected]>

When giving a stress test which enables/disables clkgating, we hit device timeout sometimes. This patch avoids subtle racy condition to address it.

Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +6 -6
index b8f573a02713..7344353a9167 100644
@@ -1807,19 +1807,19 @@ static ssize_t ufshcd_clkgate_enable_store(struct device *dev,
 		return -EINVAL;
 
 	value = !!value;
+
+	spin_lock_irqsave(hba->host->host_lock, flags);
 	if (value == hba->clk_gating.is_enabled)
 		goto out;
 
-	if (value) {
-		ufshcd_release(hba);
-	} else {
-		spin_lock_irqsave(hba->host->host_lock, flags);
+	if (value)
+		__ufshcd_release(hba);
+	else
 		hba->clk_gating.active_reqs++;
-		spin_unlock_irqrestore(hba->host->host_lock, flags);
-	}
 
 	hba->clk_gating.is_enabled = value;
 out:
+	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	return count;
 }
 
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
Reviewed-by on v2 1/5
#2 lore raw

Re: [PATCH v2 1/5] scsi: ufs: atomic update for clkgating_enable

Quoted 9 lines · Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

When giving a stress test which enables/disables clkgating, we hit device timeout sometimes. This patch avoids subtle racy condition to address it.

Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>

Next time can you have a cover letter in case of multiple patches?

Thanks,

Can Guo.

Quoted 34 lines
---
 drivers/scsi/ufs/ufshcd.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +6 -6
index b8f573a02713..7344353a9167 100644
@@ -1807,19 +1807,19 @@ static ssize_t
 ufshcd_clkgate_enable_store(struct device *dev,
 		return -EINVAL;
 
 	value = !!value;
+
+	spin_lock_irqsave(hba->host->host_lock, flags);
 	if (value == hba->clk_gating.is_enabled)
 		goto out;
 
-	if (value) {
-		ufshcd_release(hba);
-	} else {
-		spin_lock_irqsave(hba->host->host_lock, flags);
+	if (value)
+		__ufshcd_release(hba);
+	else
 		hba->clk_gating.active_reqs++;
-		spin_unlock_irqrestore(hba->host->host_lock, flags);
-	}
 
 	hba->clk_gating.is_enabled = value;
 out:
+	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	return count;
 }
#3 lore raw

Re: [PATCH v2 1/5] scsi: ufs: atomic update for clkgating_enable

Quoted 8 lines · Can Guo
Quoted 7 lines · Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

When giving a stress test which enables/disables clkgating, we hit device
timeout sometimes. This patch avoids subtle racy condition to address it.

Quoted 8 lines
Quoted 4 lines
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>

Next time can you have a cover letter in case of multiple patches?

Ah, it seems I had to cc you here as well.

https://lore.kernel.org/linux-scsi/[email protected]/T/#mb4e43f3bd03a6b7bc136bea21ac805041c1417a2

Quoted 39 lines

Thanks,

Can Guo.

Quoted 34 lines
---
 drivers/scsi/ufs/ufshcd.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +6 -6
index b8f573a02713..7344353a9167 100644
@@ -1807,19 +1807,19 @@ static ssize_t
 ufshcd_clkgate_enable_store(struct device *dev,
 		return -EINVAL;
 
 	value = !!value;
+
+	spin_lock_irqsave(hba->host->host_lock, flags);
 	if (value == hba->clk_gating.is_enabled)
 		goto out;
 
-	if (value) {
-		ufshcd_release(hba);
-	} else {
-		spin_lock_irqsave(hba->host->host_lock, flags);
+	if (value)
+		__ufshcd_release(hba);
+	else
 		hba->clk_gating.active_reqs++;
-		spin_unlock_irqrestore(hba->host->host_lock, flags);
-	}
 
 	hba->clk_gating.is_enabled = value;
 out:
+	spin_unlock_irqrestore(hba->host->host_lock, flags);
 	return count;
 }
#4 lore raw

PATCH v2 2/5 scsi: ufs: clear UAC for FFU and RPMB LUNs

In reply to Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

In order to conduct FFU or RPMB operations, UFS needs to clear UAC. This patch clears it explicitly, so that we could get no failure given early execution.

Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 70 +++++++++++++++++++++++++++++++++++----
 drivers/scsi/ufs/ufshcd.h |  1 +
 2 files changed, 65 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +64 -6
index 7344353a9167..feb10ebf7a35 100644
@@ -7057,7 +7057,6 @@ static inline void ufshcd_blk_pm_runtime_init(struct scsi_device *sdev)
 static int ufshcd_scsi_add_wlus(struct ufs_hba *hba)
 {
 	int ret = 0;
-	struct scsi_device *sdev_rpmb;
 	struct scsi_device *sdev_boot;
 
 	hba->sdev_ufs_device = __scsi_add_device(hba->host, 0, 0,
@@ -7070,14 +7069,14 @@ static int ufshcd_scsi_add_wlus(struct ufs_hba *hba)
 	ufshcd_blk_pm_runtime_init(hba->sdev_ufs_device);
 	scsi_device_put(hba->sdev_ufs_device);
 
-	sdev_rpmb = __scsi_add_device(hba->host, 0, 0,
+	hba->sdev_rpmb = __scsi_add_device(hba->host, 0, 0,
 		ufshcd_upiu_wlun_to_scsi_wlun(UFS_UPIU_RPMB_WLUN), NULL);
-	if (IS_ERR(sdev_rpmb)) {
-		ret = PTR_ERR(sdev_rpmb);
+	if (IS_ERR(hba->sdev_rpmb)) {
+		ret = PTR_ERR(hba->sdev_rpmb);
 		goto remove_sdev_ufs_device;
 	}
-	ufshcd_blk_pm_runtime_init(sdev_rpmb);
-	scsi_device_put(sdev_rpmb);
+	ufshcd_blk_pm_runtime_init(hba->sdev_rpmb);
+	scsi_device_put(hba->sdev_rpmb);
 
 	sdev_boot = __scsi_add_device(hba->host, 0, 0,
 		ufshcd_upiu_wlun_to_scsi_wlun(UFS_UPIU_BOOT_WLUN), NULL);
@@ -7601,6 +7600,63 @@ static int ufshcd_add_lus(struct ufs_hba *hba)
 	return ret;
 }
 
+static int
+ufshcd_send_request_sense(struct ufs_hba *hba, struct scsi_device *sdp);
+
+static int ufshcd_clear_ua_wlun(struct ufs_hba *hba, u8 wlun)
+{
+	struct scsi_device *sdp;
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(hba->host->host_lock, flags);
+	if (wlun  == UFS_UPIU_UFS_DEVICE_WLUN)
+		sdp = hba->sdev_ufs_device;
+	else if (wlun  == UFS_UPIU_RPMB_WLUN)
+		sdp = hba->sdev_rpmb;
+	else
+		BUG_ON(1);
+	if (sdp) {
+		ret = scsi_device_get(sdp);
+		if (!ret && !scsi_device_online(sdp)) {
+			ret = -ENODEV;
+			scsi_device_put(sdp);
+		}
+	} else {
+		ret = -ENODEV;
+	}
+	spin_unlock_irqrestore(hba->host->host_lock, flags);
+	if (ret)
+		goto out_err;
+
+	ret = ufshcd_send_request_sense(hba, sdp);
+	scsi_device_put(sdp);
+out_err:
+	if (ret)
+		dev_err(hba->dev, "%s: UAC clear LU=%x ret = %d\n",
+				__func__, wlun, ret);
+	return ret;
+}
+
+static int ufshcd_clear_ua_wluns(struct ufs_hba *hba)
+{
+	int ret = 0;
+
+	if (!hba->wlun_dev_clr_ua)
+		goto out;
+
+	ret = ufshcd_clear_ua_wlun(hba, UFS_UPIU_UFS_DEVICE_WLUN);
+	if (!ret)
+		ret = ufshcd_clear_ua_wlun(hba, UFS_UPIU_RPMB_WLUN);
+	if (!ret)
+		hba->wlun_dev_clr_ua = false;
+out:
+	if (ret)
+		dev_err(hba->dev, "%s: Failed to clear UAC WLUNS ret = %d\n",
+				__func__, ret);
+	return ret;
+}
+
 /**
  * ufshcd_probe_hba - probe hba to detect device and initialize
  * @hba: per-adapter instance
@@ -7720,6 +7776,8 @@ static void ufshcd_async_scan(void *data, async_cookie_t cookie)
 		pm_runtime_put_sync(hba->dev);
 		ufshcd_exit_clk_scaling(hba);
 		ufshcd_hba_exit(hba);
+	} else {
+		ufshcd_clear_ua_wluns(hba);
 	}
 }
 
drivers/scsi/ufs/ufshcd.h +1 -0
index 47eb1430274c..718881d038f5 100644
@@ -681,6 +681,7 @@ struct ufs_hba {
 	 * "UFS device" W-LU.
 	 */
 	struct scsi_device *sdev_ufs_device;
+	struct scsi_device *sdev_rpmb;
 
 	enum ufs_dev_pwr_mode curr_dev_pwr_mode;
 	enum uic_link_state uic_link_state;
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
#5 lore raw

PATCH v2 3/5 scsi: ufs: use WQ_HIGHPRI for gating work

In reply to Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

Must have WQ_MEM_RECLAIM
``WQ_MEM_RECLAIM``
All wq which might be used in the memory reclaim paths **MUST** have this flag set. The wq is guaranteed to have at least one execution context regardless of memory pressure.

Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index feb10ebf7a35..0858c0b55eac 100644
@@ -1867,7 +1867,7 @@ static void ufshcd_init_clk_gating(struct ufs_hba *hba)
 	snprintf(wq_name, ARRAY_SIZE(wq_name), "ufs_clk_gating_%d",
 		 hba->host->host_no);
 	hba->clk_gating.clk_gating_workq = alloc_ordered_workqueue(wq_name,
-							   WQ_MEM_RECLAIM);
+					WQ_MEM_RECLAIM | WQ_HIGHPRI);
 
 	hba->clk_gating.is_enabled = true;
 
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
#6 lore raw

Re: [PATCH v2 3/5] scsi: ufs: use WQ_HIGHPRI for gating work

Quoted 8 lines · Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

Must have WQ_MEM_RECLAIM
``WQ_MEM_RECLAIM``
All wq which might be used in the memory reclaim paths **MUST** have this flag set. The wq is guaranteed to have at least one execution context regardless of memory pressure.

You misunderstood my point. I meant you need to give more info about why we are adding WQ_HIGHPRI flag but not why WQ_MEM_RECLAIM must be there.

Thanks,

Can Guo.

Quoted 21 lines
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index feb10ebf7a35..0858c0b55eac 100644
@@ -1867,7 +1867,7 @@ static void ufshcd_init_clk_gating(struct ufs_hba 
 *hba)
 	snprintf(wq_name, ARRAY_SIZE(wq_name), "ufs_clk_gating_%d",
 		 hba->host->host_no);
 	hba->clk_gating.clk_gating_workq = alloc_ordered_workqueue(wq_name,
-							   WQ_MEM_RECLAIM);
+					WQ_MEM_RECLAIM | WQ_HIGHPRI);
 
 	hba->clk_gating.is_enabled = true;
#7 lore raw

Re: [PATCH v2 3/5] scsi: ufs: use WQ_HIGHPRI for gating work

Quoted 10 lines · Can Guo
Quoted 8 lines · Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

Must have WQ_MEM_RECLAIM
``WQ_MEM_RECLAIM``
All wq which might be used in the memory reclaim paths **MUST** have this flag set. The wq is guaranteed to have at least one execution context regardless of memory pressure.

Quoted 2 lines

You misunderstood my point. I meant you need to give more info about why we are adding WQ_HIGHPRI flag but not why WQ_MEM_RECLAIM must be there.

Oh, I thought that WQ_HIGHPRI is telling everything tho.

Quoted 26 lines

Thanks,

Can Guo.

Quoted 21 lines
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index feb10ebf7a35..0858c0b55eac 100644
@@ -1867,7 +1867,7 @@ static void ufshcd_init_clk_gating(struct ufs_hba
 *hba)
 	snprintf(wq_name, ARRAY_SIZE(wq_name), "ufs_clk_gating_%d",
 		 hba->host->host_no);
 	hba->clk_gating.clk_gating_workq = alloc_ordered_workqueue(wq_name,
-							   WQ_MEM_RECLAIM);
+					WQ_MEM_RECLAIM | WQ_HIGHPRI);
 
 	hba->clk_gating.is_enabled = true;
#8 lore raw

PATCH v2 4/5 scsi: add more contexts in the ufs tracepoints

In reply to Jaegeuk Kim

From: Jaegeuk Kim <[email protected]>

This adds user-friendly tracepoints with group id.

---
 drivers/scsi/ufs/ufshcd.c  |  6 ++++--
 include/trace/events/ufs.h | 21 +++++++++++++++++----
 2 files changed, 21 insertions(+), 6 deletions(-)
drivers/scsi/ufs/ufshcd.c +4 -2
index 0858c0b55eac..b5ca0effe636 100644
@@ -348,7 +348,7 @@ static void ufshcd_add_command_trace(struct ufs_hba *hba,
 		unsigned int tag, const char *str)
 {
 	sector_t lba = -1;
-	u8 opcode = 0;
+	u8 opcode = 0, group_id = 0;
 	u32 intr, doorbell;
 	struct ufshcd_lrb *lrbp = &hba->lrb[tag];
 	struct scsi_cmnd *cmd = lrbp->cmd;
@@ -374,13 +374,15 @@ static void ufshcd_add_command_trace(struct ufs_hba *hba,
 				lba = cmd->request->bio->bi_iter.bi_sector;
 			transfer_len = be32_to_cpu(
 				lrbp->ucd_req_ptr->sc.exp_data_transfer_len);
+			if (opcode == WRITE_10)
+				group_id = lrbp->cmd->cmnd[6];
 		}
 	}
 
 	intr = ufshcd_readl(hba, REG_INTERRUPT_STATUS);
 	doorbell = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
 	trace_ufshcd_command(dev_name(hba->dev), str, tag,
-				doorbell, transfer_len, intr, lba, opcode);
+			doorbell, transfer_len, intr, lba, opcode, group_id);
 }
 
 static void ufshcd_print_clk_freqs(struct ufs_hba *hba)
include/trace/events/ufs.h +17 -4
index 84841b3a7ffd..50654f352639 100644
@@ -11,6 +11,15 @@
 
 #include <linux/tracepoint.h>
 
+#define str_opcode(opcode)						\
+	__print_symbolic(opcode,					\
+		{ WRITE_16,		"WRITE_16" },			\
+		{ WRITE_10,		"WRITE_10" },			\
+		{ READ_16,		"READ_16" },			\
+		{ READ_10,		"READ_10" },			\
+		{ SYNCHRONIZE_CACHE,	"SYNC" },			\
+		{ UNMAP,		"UNMAP" })
+
 #define UFS_LINK_STATES			\
 	EM(UIC_LINK_OFF_STATE)		\
 	EM(UIC_LINK_ACTIVE_STATE)	\
@@ -215,9 +224,10 @@ DEFINE_EVENT(ufshcd_template, ufshcd_init,
 TRACE_EVENT(ufshcd_command,
 	TP_PROTO(const char *dev_name, const char *str, unsigned int tag,
 			u32 doorbell, int transfer_len, u32 intr, u64 lba,
-			u8 opcode),
+			u8 opcode, u8 group_id),
 
-	TP_ARGS(dev_name, str, tag, doorbell, transfer_len, intr, lba, opcode),
+	TP_ARGS(dev_name, str, tag, doorbell, transfer_len,
+				intr, lba, opcode, group_id),
 
 	TP_STRUCT__entry(
 		__string(dev_name, dev_name)
@@ -228,6 +238,7 @@ TRACE_EVENT(ufshcd_command,
 		__field(u32, intr)
 		__field(u64, lba)
 		__field(u8, opcode)
+		__field(u8, group_id)
 	),
 
 	TP_fast_assign(
@@ -239,13 +250,15 @@ TRACE_EVENT(ufshcd_command,
 		__entry->intr = intr;
 		__entry->lba = lba;
 		__entry->opcode = opcode;
+		__entry->group_id = group_id;
 	),
 
 	TP_printk(
-		"%s: %s: tag: %u, DB: 0x%x, size: %d, IS: %u, LBA: %llu, opcode: 0x%x",
+		"%s: %s: tag: %u, DB: 0x%x, size: %d, IS: %u, LBA: %llu, opcode: 0x%x (%s), group_id: 0x%x",
 		__get_str(str), __get_str(dev_name), __entry->tag,
 		__entry->doorbell, __entry->transfer_len,
-		__entry->intr, __entry->lba, (u32)__entry->opcode
+		__entry->intr, __entry->lba, (u32)__entry->opcode,
+		str_opcode(__entry->opcode), (u32)__entry->group_id
 	)
 );
 
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
#9 lore raw

PATCH v2 5/5 scsi: ufs: fix clkgating on/off correctly

In reply to Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work will check it again.

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
 
Signature · 1 line
2.29.0.rc1.297.gfa9743e501-goog
#10 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 4 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However, they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag assigned from block layer is still in use, but tags are released asynchronously (through block softirq), meaning it does not reflect the real occupation of UFS host. That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still return true.

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(), but having the check of it in gate_work() can still prevent gating from happening. The current change works for you maybe because the tags are release before hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter or somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use() returns true. What do you think?

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#11 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 19 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However, they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still return true.

Quoted 10 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from happening.
The current change works for you maybe because the tags are release before hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

I don't think this breaks clkgating, but fix the wrong condition check which prevented gate_work at all. As you mentioned, even if this schedules gate_work by racy conditions, gate_work will handle it as a last resort.

Quoted 38 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#12 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 31 lines · On 2020-10-21 12:52, [email protected] wrote:
Quoted 19 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However, they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still return true.

Quoted 10 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from happening.
The current change works for you maybe because the tags are release before hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

Quoted 4 lines

I don't think this breaks clkgating, but fix the wrong condition check which prevented gate_work at all. As you mentioned, even if this schedules gate_work by racy conditions, gate_work will handle it as a last resort.

If clocks cannot be gated after the last task is cleared from UFS host, then clk gating is broken, no? Assume UFS has completed the last task in its queue, as this change says, ufshcd_any_tag_in_use() is preventing ufshcd_release() from invoking gate_work(). Similarly, ufshcd_any_tag_in_use() can prevent gate_work() from doing its real work - disabling the clocks. Do you agree?

        if (hba->clk_gating.active_reqs
                || hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL
                || ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks
                || hba->active_uic_cmd || hba->uic_async_done)
                goto rel_lock;

Thanks,

Can Guo.

Quoted 38 lines
Quoted 38 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#13 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 56 lines · Can Guo
Quoted 38 lines · On 2020-10-21 12:52, [email protected] wrote:
Quoted 23 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However,
they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag
assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of
UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still
return true.

Quoted 13 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from
happening.
The current change works for you maybe because the tags are release before
hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter
or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

Quoted 6 lines

I don't think this breaks clkgating, but fix the wrong condition check which
prevented gate_work at all. As you mentioned, even if this schedules gate_work
by racy conditions, gate_work will handle it as a last resort.

If clocks cannot be gated after the last task is cleared from UFS host, then clk gating
is broken, no? Assume UFS has completed the last task in its queue, as this change says,
ufshcd_any_tag_in_use() is preventing ufshcd_release() from invoking gate_work().
Similarly, ufshcd_any_tag_in_use() can prevent gate_work() from doing its real work -
disabling the clocks. Do you agree?

Quoted 5 lines
        if (hba->clk_gating.active_reqs
                || hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL
                || ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks
                || hba->active_uic_cmd || hba->uic_async_done)
                goto rel_lock;

I see the point, but this happens only when clkgate_delay_ms is too short to give enough time for releasing tag. If it's correctly set, I think there'd be no problem, unless softirq was delayed by other RT threads which is just a corner case tho.

Quoted 43 lines

Thanks,

Can Guo.

Quoted 38 lines
Quoted 38 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#14 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 64 lines · Jaegeuk Kim
Quoted 56 lines · Can Guo
Quoted 38 lines · On 2020-10-21 12:52, [email protected] wrote:
Quoted 23 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However,
they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag
assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of
UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still
return true.

Quoted 13 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from
happening.
The current change works for you maybe because the tags are release before
hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter
or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

Quoted 6 lines

I don't think this breaks clkgating, but fix the wrong condition check which
prevented gate_work at all. As you mentioned, even if this schedules gate_work
by racy conditions, gate_work will handle it as a last resort.

If clocks cannot be gated after the last task is cleared from UFS host, then clk gating
is broken, no? Assume UFS has completed the last task in its queue, as this change says,
ufshcd_any_tag_in_use() is preventing ufshcd_release() from invoking gate_work().
Similarly, ufshcd_any_tag_in_use() can prevent gate_work() from doing its real work -
disabling the clocks. Do you agree?

Quoted 6 lines
        if (hba->clk_gating.active_reqs
                || hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL
                || ufshcd_any_tag_in_use(hba) || 
hba->outstanding_tasks
                || hba->active_uic_cmd || hba->uic_async_done)
                goto rel_lock;
Quoted 5 lines

I see the point, but this happens only when clkgate_delay_ms is too short to give enough time for releasing tag. If it's correctly set, I think there'd be no problem, unless softirq was delayed by other RT threads which is just a corner case tho.

Yes, we are fixing corner cases, aren't we? I thought you would like to address it since you are fixing clk gating.

Regards,

Can Guo.

Quoted 43 lines
Quoted 43 lines

Thanks,

Can Guo.

Quoted 38 lines
Quoted 38 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#15 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

Quoted 77 lines · Can Guo
Quoted 67 lines · Jaegeuk Kim
Quoted 59 lines · Can Guo
Quoted 38 lines · On 2020-10-21 12:52, [email protected] wrote:
Quoted 23 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However,
they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag
assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of
UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still
return true.

Quoted 13 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from
happening.
The current change works for you maybe because the tags are release before
hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter
or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

Quoted 6 lines

I don't think this breaks clkgating, but fix the wrong condition check which
prevented gate_work at all. As you mentioned, even if this schedules gate_work
by racy conditions, gate_work will handle it as a last resort.

If clocks cannot be gated after the last task is cleared from UFS host, then
clk gating
is broken, no? Assume UFS has completed the last task in its queue, as this
change says,
ufshcd_any_tag_in_use() is preventing ufshcd_release() from invoking gate_work().
Similarly, ufshcd_any_tag_in_use() can prevent gate_work() from doing its
real work -
disabling the clocks. Do you agree?

Quoted 6 lines
        if (hba->clk_gating.active_reqs
                || hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL
                || ufshcd_any_tag_in_use(hba) ||
hba->outstanding_tasks
                || hba->active_uic_cmd || hba->uic_async_done)
                goto rel_lock;
Quoted 8 lines

I see the point, but this happens only when clkgate_delay_ms is too short
to give enough time for releasing tag. If it's correctly set, I think there'd
be no problem, unless softirq was delayed by other RT threads which is just
a corner case tho.

Quoted 2 lines

Yes, we are fixing corner cases, aren't we? I thought you would like to address it since you are fixing clk gating.

I think that can be fixed by a separate patch which controls delay_ms when user tries to change it from default 150 ms?

Quoted 48 lines

Regards,

Can Guo.

Quoted 43 lines
Quoted 43 lines

Thanks,

Can Guo.

Quoted 38 lines
Quoted 38 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 29 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;
#16 lore raw

Re: [PATCH v2 5/5] scsi: ufs: fix clkgating on/off correctly

In reply to Jaegeuk Kim
Quoted 106 lines · Jaegeuk Kim
Quoted 56 lines · Can Guo
Quoted 38 lines · On 2020-10-21 12:52, [email protected] wrote:
Quoted 23 lines · Can Guo
Quoted 5 lines · Jaegeuk Kim

The below call stack prevents clk_gating at every IO completion. We can remove the condition, ufshcd_any_tag_in_use(), since clkgating_work
will check it again.

I think checking ufshcd_any_tag_in_use() in either ufshcd_release() or gate_work() can break UFS clk gating's functionality.

ufshcd_any_tag_in_use() was introduced to replace hba->lrb_in_use. However,
they are not exactly same - ufshcd_any_tag_in_use() returns true if any tag
assigned from block layer is still in use, but tags are released asynchronously
(through block softirq), meaning it does not reflect the real occupation of
UFS host.
That is after UFS host finishes all tasks, ufshcd_any_tag_in_use() can still
return true.

Quoted 13 lines

This change only removes the check of ufshcd_any_tag_in_use() in ufshcd_release(),
but having the check of it in gate_work() can still prevent gating from
happening.
The current change works for you maybe because the tags are release before
hba->clk_gating.delay_ms expires, but if hba->clk_gating.delay_ms is shorter
or
somehow block softirq is retarded, gate_work() may have chance to see ufshcd_any_tag_in_use()
returns true. What do you think?

Quoted 6 lines

I don't think this breaks clkgating, but fix the wrong condition check which
prevented gate_work at all. As you mentioned, even if this schedules gate_work
by racy conditions, gate_work will handle it as a last resort.

If clocks cannot be gated after the last task is cleared from UFS host, then clk gating
is broken, no? Assume UFS has completed the last task in its queue, as this change says,
ufshcd_any_tag_in_use() is preventing ufshcd_release() from invoking gate_work().
Similarly, ufshcd_any_tag_in_use() can prevent gate_work() from doing its real work -
disabling the clocks. Do you agree?

Quoted 6 lines
        if (hba->clk_gating.active_reqs
                || hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL
                || ufshcd_any_tag_in_use(hba) || 
hba->outstanding_tasks
                || hba->active_uic_cmd || hba->uic_async_done)
                goto rel_lock;

I see the point, but this happens only when clkgate_delay_ms is too short to give enough time for releasing tag. If it's correctly set, I think there'd be no problem, unless softirq was delayed by other RT threads which is just a corner case tho.

Quoted 37 lines

Thanks,

Can Guo.

Quoted 32 lines
Quoted 32 lines

Thanks,

Can Guo.

In __ufshcd_transfer_req_compl Ihba->lrb_in_use is cleared immediately when UFS driver finishes all tasks

Quoted 23 lines

ufshcd_complete_requests(struct ufs_hba *hba)
ufshcd_transfer_req_compl()

    __ufshcd_transfer_req_compl()
      __ufshcd_release(hba)
        if (ufshcd_any_tag_in_use() == 1)
           return;
  ufshcd_tmc_handler(hba);
    blk_mq_tagset_busy_iter();
Cc Alim Akhtar <[email protected]>
Cc Avri Altman <[email protected]>
Cc Can Guo <[email protected]>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +0 -0
index b5ca0effe636..cecbd4ace8b4 100644
@@ -1746,7 +1746,7 @@ static void __ufshcd_release(struct ufs_hba *hba)
 
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

drivers/scsi/ufs/ufshcd.c static void __ufshcd_release(struct ufs_hba *hba) · line 1751

Quoted 6 lines
drivers/scsi/ufs/ufshcd.c +1 -1
 	if (hba->clk_gating.active_reqs || hba->clk_gating.is_suspended ||
 	    hba->ufshcd_state != UFSHCD_STATE_OPERATIONAL ||
-	    ufshcd_any_tag_in_use(hba) || hba->outstanding_tasks ||
+	    hba->outstanding_tasks ||
 	    hba->active_uic_cmd || hba->uic_async_done)
 		return;

How about checking outstanding_reqs as well, say in ufshcd_any_tag_in_use() ?

-asd