Skip to contentSkip to conversation

scsi: ufs: some cleanups and make the delay for host enabling customizable

PATCH Superseded Stanley Chu 3 participants 14 messages Mar 16, 2020 – Mar 18, 2020

Subsystems (2)

v6 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#0 lore raw

v6 cover scsi: ufs: some cleanups and make the delay for host enabling customizable

v5 -> v6
- Drop patch #2 "scsi: ufs: remove init_prefetch_data in struct ufs_hba" in v5 because Can Guo has similar cleanup earlier in patch "scsi: ufs: Do not rely on prefetched data"

v4 -> v5
- Fix patch #7: Fix typo "initializatoin" in title

v3 -> v4
- In patch #8, fix incorrect condition of customized delay for host enabling

v2 -> v3
- Remove /arch/arm64/configs/defconfig chnage because it is for local test only

v1 -> v2
- Add patch #1 "scsi: ufs: fix uninitialized tx_lanes in ufshcd_disable_tx_lcc"
- Remove struct ufs_init_prefetch in patch #2 "scsi: ufs: remove init_prefetch_data in struct ufs_hba"
- Introduce common delay function in patch #4
- Replace all delay places by common delay function in ufs-mediatek in patch #5
- Use common delay function instead for host enabling delay in patch #6
- Add patch #7 "scsi: ufs: make HCE polling more compact to improve initializatoin latency"
- In patch #8, customize the delay in ufs_mtk_hce_enable_notify callback instead of ufs_mtk_init (Avri)

Stanley Chu (7):
scsi: ufs: fix uninitialized tx_lanes in ufshcd_disable_tx_lcc()
scsi: ufs: use an enum for host capabilities
scsi: ufs: introduce common delay function
scsi: ufs-mediatek: replace all delay places by common delay function
scsi: ufs: allow customized delay for host enabling
scsi: ufs: make HCE polling more compact to improve initialization
latency
scsi: ufs-mediatek: customize the delay for host enabling

 drivers/scsi/ufs/ufs-mediatek.c | 64 +++++++++++++++++++------------
 drivers/scsi/ufs/ufs-mediatek.h |  1 +
 drivers/scsi/ufs/ufshcd.c       | 32 ++++++++++------
 drivers/scsi/ufs/ufshcd.h       | 67 +++++++++++++++++++--------------
 4 files changed, 100 insertions(+), 64 deletions(-)
Signature · 1 line
2.18.0
#1 lore raw

PATCH v6 1/7 scsi: ufs: fix uninitialized tx_lanes in ufshcd_disable_tx_lcc()

In ufshcd_disable_tx_lcc(), if ufshcd_dme_get() or ufshcd_dme_peer_get() get fail, uninitialized variable "tx_lanes" may be used as unexpected lane ID for DME configuration.

Fix this issue by initializing "tx_lanes".

---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +1 -1
index 5698f1164a5e..314e808b0d4e 100644
@@ -4315,7 +4315,7 @@ EXPORT_SYMBOL_GPL(ufshcd_hba_enable);
 
 static int ufshcd_disable_tx_lcc(struct ufs_hba *hba, bool peer)
 {
-	int tx_lanes, i, err = 0;
+	int tx_lanes = 0, i, err = 0;
 
 	if (!peer)
 		ufshcd_dme_get(hba, UIC_ARG_MIB(PA_CONNECTEDTXDATALANES),
Signature · 1 line
2.18.0
#2 lore raw

PATCH v6 2/7 scsi: ufs: use an enum for host capabilities

In reply to Stanley Chu

Use an enum to specify the host capabilities instead of #defines inside the structure definition.

---
 drivers/scsi/ufs/ufshcd.h | 65 ++++++++++++++++++++++-----------------
 1 file changed, 37 insertions(+), 28 deletions(-)
drivers/scsi/ufs/ufshcd.h +37 -28
index 5c10777154fc..52425371082a 100644
@@ -510,6 +510,43 @@ enum ufshcd_quirks {
 	UFSHCD_QUIRK_BROKEN_UFS_HCI_VERSION		= 1 << 5,
 };
 
+enum ufshcd_caps {
+	/* Allow dynamic clk gating */
+	UFSHCD_CAP_CLK_GATING				= 1 << 0,
+
+	/* Allow hiberb8 with clk gating */
+	UFSHCD_CAP_HIBERN8_WITH_CLK_GATING		= 1 << 1,
+
+	/* Allow dynamic clk scaling */
+	UFSHCD_CAP_CLK_SCALING				= 1 << 2,
+
+	/* Allow auto bkops to enabled during runtime suspend */
+	UFSHCD_CAP_AUTO_BKOPS_SUSPEND			= 1 << 3,
+
+	/*
+	 * This capability allows host controller driver to use the UFS HCI's
+	 * interrupt aggregation capability.
+	 * CAUTION: Enabling this might reduce overall UFS throughput.
+	 */
+	UFSHCD_CAP_INTR_AGGR				= 1 << 4,
+
+	/*
+	 * This capability allows the device auto-bkops to be always enabled
+	 * except during suspend (both runtime and suspend).
+	 * Enabling this capability means that device will always be allowed
+	 * to do background operation when it's active but it might degrade
+	 * the performance of ongoing read/write operations.
+	 */
+	UFSHCD_CAP_KEEP_AUTO_BKOPS_ENABLED_EXCEPT_SUSPEND = 1 << 5,
+
+	/*
+	 * This capability allows host controller driver to automatically
+	 * enable runtime power management by itself instead of waiting
+	 * for userspace to control the power management.
+	 */
+	UFSHCD_CAP_RPM_AUTOSUSPEND			= 1 << 6,
+};
+
 /**
  * struct ufs_hba - per adapter private structure
  * @mmio_base: UFSHCI base register address
@@ -664,34 +701,6 @@ struct ufs_hba {
 	struct ufs_clk_gating clk_gating;
 	/* Control to enable/disable host capabilities */
 	u32 caps;
-	/* Allow dynamic clk gating */
-#define UFSHCD_CAP_CLK_GATING	(1 << 0)
-	/* Allow hiberb8 with clk gating */
-#define UFSHCD_CAP_HIBERN8_WITH_CLK_GATING (1 << 1)
-	/* Allow dynamic clk scaling */
-#define UFSHCD_CAP_CLK_SCALING	(1 << 2)
-	/* Allow auto bkops to enabled during runtime suspend */
-#define UFSHCD_CAP_AUTO_BKOPS_SUSPEND (1 << 3)
-	/*
-	 * This capability allows host controller driver to use the UFS HCI's
-	 * interrupt aggregation capability.
-	 * CAUTION: Enabling this might reduce overall UFS throughput.
-	 */
-#define UFSHCD_CAP_INTR_AGGR (1 << 4)
-	/*
-	 * This capability allows the device auto-bkops to be always enabled
-	 * except during suspend (both runtime and suspend).
-	 * Enabling this capability means that device will always be allowed
-	 * to do background operation when it's active but it might degrade
-	 * the performance of ongoing read/write operations.
-	 */
-#define UFSHCD_CAP_KEEP_AUTO_BKOPS_ENABLED_EXCEPT_SUSPEND (1 << 5)
-	/*
-	 * This capability allows host controller driver to automatically
-	 * enable runtime power management by itself instead of waiting
-	 * for userspace to control the power management.
-	 */
-#define UFSHCD_CAP_RPM_AUTOSUSPEND (1 << 6)
 
 	struct devfreq *devfreq;
 	struct ufs_clk_scaling clk_scaling;
Signature · 1 line
2.18.0
#3 lore raw

PATCH v6 3/7 scsi: ufs: introduce common delay function

In reply to Stanley Chu

Introduce common delay function to collect all delay requirements to simplify driver and take choices of udelay and usleep_range into consideration.

---
 drivers/scsi/ufs/ufshcd.c | 27 ++++++++++++++++++---------
 drivers/scsi/ufs/ufshcd.h |  1 +
 2 files changed, 19 insertions(+), 9 deletions(-)
drivers/scsi/ufs/ufshcd.c +18 -9
index 314e808b0d4e..9fea346f7d22 100644
@@ -597,6 +597,18 @@ static void ufshcd_print_pwr_info(struct ufs_hba *hba)
 		 hba->pwr_info.hs_rate);
 }
 
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);
+
 /*
  * ufshcd_wait_for_register - wait for register value to change
  * @hba - per-adapter interface
@@ -620,10 +632,7 @@ int ufshcd_wait_for_register(struct ufs_hba *hba, u32 reg, u32 mask,
 	val = val & mask;
 
 	while ((ufshcd_readl(hba, reg) & mask) != val) {
-		if (can_sleep)
-			usleep_range(interval_us, interval_us + 50);
-		else
-			udelay(interval_us);
+		ufshcd_wait_us(interval_us, 50, can_sleep);
 		if (time_after(jiffies, timeout)) {
 			if ((ufshcd_readl(hba, reg) & mask) != val)
 				err = -ETIMEDOUT;
@@ -3565,7 +3574,7 @@ static inline void ufshcd_add_delay_before_dme_cmd(struct ufs_hba *hba)
 	}
 
 	/* allow sleep for extra 50us if needed */
-	usleep_range(min_sleep_time_us, min_sleep_time_us + 50);
+	ufshcd_wait_us(min_sleep_time_us, 50, true);
 }
 
 /**
@@ -4289,7 +4298,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 	 * instruction might be read back.
 	 * This delay can be changed based on the controller.
 	 */
-	usleep_range(1000, 1100);
+	ufshcd_wait_us(1000, 100, true);
 
 	/* wait for the host controller to complete initialization */
 	retry = 10;
@@ -4301,7 +4310,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 				"Controller enable failed\n");
 			return -EIO;
 		}
-		usleep_range(5000, 5100);
+		ufshcd_wait_us(5000, 100, true);
 	}
 
 	/* enable UIC related interrupts */
@@ -6224,7 +6233,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 			reg = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
 			if (reg & (1 << tag)) {
 				/* sleep for max. 200us to stabilize */
-				usleep_range(100, 200);
+				ufshcd_wait_us(100, 100, true);
 				continue;
 			}
 			/* command completed already */
@@ -7786,7 +7795,7 @@ static void ufshcd_vreg_set_lpm(struct ufs_hba *hba)
 	 */
 	if (!ufshcd_is_link_active(hba) &&
 	    hba->dev_quirks & UFS_DEVICE_QUIRK_DELAY_BEFORE_LPM)
-		usleep_range(2000, 2100);
+		ufshcd_wait_us(2000, 100, true);
 
 	/*
 	 * If UFS device is either in UFS_Sleep turn off VCC rail to save some
drivers/scsi/ufs/ufshcd.h +1 -0
index 52425371082a..842f0223f5e5 100644
@@ -792,6 +792,7 @@ int ufshcd_init(struct ufs_hba * , void __iomem * , unsigned int);
 int ufshcd_make_hba_operational(struct ufs_hba *hba);
 void ufshcd_remove(struct ufs_hba *);
 int ufshcd_uic_hibern8_exit(struct ufs_hba *hba);
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep);
 int ufshcd_wait_for_register(struct ufs_hba *hba, u32 reg, u32 mask,
 				u32 val, unsigned long interval_us,
 				unsigned long timeout_ms, bool can_sleep);
Signature · 5 lines
2.18.0
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Reviewed-by on v6 3/7
#4 lore raw

Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

Quoted 6 lines · Stanley Chu

Introduce common delay function to collect all delay requirements to simplify driver and take choices of udelay and usleep_range into consideration.

Quoted 103 lines
---
 drivers/scsi/ufs/ufshcd.c | 27 ++++++++++++++++++---------
 drivers/scsi/ufs/ufshcd.h |  1 +
 2 files changed, 19 insertions(+), 9 deletions(-)
drivers/scsi/ufs/ufshcd.c +12 -0
index 314e808b0d4e..9fea346f7d22 100644
@@ -597,6 +597,18 @@ static void ufshcd_print_pwr_info(struct ufs_hba 
 *hba)
 		 hba->pwr_info.hs_rate);
 }
 
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool 
 can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);
+
 /*

* ufshcd_wait_for_register - wait for register value to change
* @hba - per-adapter interface

drivers/scsi/ufs/ufshcd.c +1 -4
@@ -620,10 +632,7 @@ int ufshcd_wait_for_register(struct ufs_hba *hba,
 u32 reg, u32 mask,
 	val = val & mask;
 
 	while ((ufshcd_readl(hba, reg) & mask) != val) {
-		if (can_sleep)
-			usleep_range(interval_us, interval_us + 50);
-		else
-			udelay(interval_us);
+		ufshcd_wait_us(interval_us, 50, can_sleep);
 		if (time_after(jiffies, timeout)) {
 			if ((ufshcd_readl(hba, reg) & mask) != val)
drivers/scsi/ufs/ufshcd.c +1 -1
 				err = -ETIMEDOUT;
@@ -3565,7 +3574,7 @@ static inline void
 ufshcd_add_delay_before_dme_cmd(struct ufs_hba *hba)
 	}
 
 	/* allow sleep for extra 50us if needed */
-	usleep_range(min_sleep_time_us, min_sleep_time_us + 50);
+	ufshcd_wait_us(min_sleep_time_us, 50, true);
 }
 

/**

drivers/scsi/ufs/ufshcd.c +4 -4
@@ -4289,7 +4298,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 	 * instruction might be read back.
 	 * This delay can be changed based on the controller.
 	 */
-	usleep_range(1000, 1100);
+	ufshcd_wait_us(1000, 100, true);
 
 	/* wait for the host controller to complete initialization */
 	retry = 10;
@@ -4301,7 +4310,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 				"Controller enable failed\n");
 			return -EIO;
 		}
-		usleep_range(5000, 5100);
+		ufshcd_wait_us(5000, 100, true);
 	}
 
 	/* enable UIC related interrupts */
@@ -6224,7 +6233,7 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
 			reg = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
 			if (reg & (1 << tag)) {
 				/* sleep for max. 200us to stabilize */
-				usleep_range(100, 200);
+				ufshcd_wait_us(100, 100, true);
 				continue;
 			}
 			/* command completed already */
@@ -7786,7 +7795,7 @@ static void ufshcd_vreg_set_lpm(struct ufs_hba 
 *hba)
 	 */
 	if (!ufshcd_is_link_active(hba) &&
 	    hba->dev_quirks & UFS_DEVICE_QUIRK_DELAY_BEFORE_LPM)
-		usleep_range(2000, 2100);
+		ufshcd_wait_us(2000, 100, true);
 
 	/*

* If UFS device is either in UFS_Sleep turn off VCC rail to save
some

drivers/scsi/ufs/ufshcd.h +1 -0
index 52425371082a..842f0223f5e5 100644
@@ -792,6 +792,7 @@ int ufshcd_init(struct ufs_hba * , void __iomem *
 , unsigned int);
 int ufshcd_make_hba_operational(struct ufs_hba *hba);
 void ufshcd_remove(struct ufs_hba *);
 int ufshcd_uic_hibern8_exit(struct ufs_hba *hba);
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool 
 can_sleep);
 int ufshcd_wait_for_register(struct ufs_hba *hba, u32 reg, u32 mask,

drivers/scsi/ufs/ufshcd.h int ufshcd_init(struct ufs_hba * , void __iomem * , unsigned int); · line 798

Quoted 2 lines
drivers/scsi/ufs/ufshcd.h +0 -0
 				u32 val, unsigned long interval_us,
 				unsigned long timeout_ms, bool can_sleep);

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#5 lore raw

Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

In reply to Stanley Chu

drivers/scsi/ufs/ufshcd.c static void ufshcd_print_pwr_info(struct ufs_hba *hba) · line 610

Quoted 11 lines · Stanley Chu
drivers/scsi/ufs/ufshcd.c +11 -0
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);

I don't like this function because I think it makes the UFS code harder to read instead of easier. The 'can_sleep' argument is only set by one caller which I think is a strong argument to remove that argument again and to move the code that depends on that argument from the above function into the caller. Additionally, it is not possible to comprehend what a ufshcd_wait_us() call does without looking up the function definition to see what the meaning of the third argument is.

Please drop this patch.

Thanks,

Bart.

#6 lore raw

Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

Hi Bart,

Quoted 21 lines · Bart Van Assche

drivers/scsi/ufs/ufshcd.c static void ufshcd_print_pwr_info(struct ufs_hba *hba) · line 610

Quoted 11 lines · Stanley Chu
drivers/scsi/ufs/ufshcd.c +11 -0
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);

I don't like this function because I think it makes the UFS code harder to read instead of easier. The 'can_sleep' argument is only set by one caller which I think is a strong argument to remove that argument again and to move the code that depends on that argument from the above function into the caller. Additionally, it is not possible to comprehend what a ufshcd_wait_us() call does without looking up the function definition to see what the meaning of the third argument is.

Quoted 1 line

Please drop this patch.

Thanks for your review and comments.

If the problem is the third argument 'can_sleep' which makes the code not be easily comprehensible, how about just removing 'can_sleep' from this function and keeping left parts because this function provides good flexibility to users to choose udelay or usleep_range according to the 'us' argument?

Thanks,
Stanley Chu

Quoted 5 lines

Thanks,

Bart.

#7 lore raw

Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

Quoted 24 lines · Stanley Chu
Quoted 21 lines · Bart Van Assche

drivers/scsi/ufs/ufshcd.c static void ufshcd_print_pwr_info(struct ufs_hba *hba) · line 610

Quoted 11 lines · Stanley Chu
drivers/scsi/ufs/ufshcd.c +11 -0
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);

I don't like this function because I think it makes the UFS code harder to read instead of easier. The 'can_sleep' argument is only set by one caller which I think is a strong argument to remove that argument again and to move the code that depends on that argument from the above function into the caller. Additionally, it is not possible to comprehend what a ufshcd_wait_us() call does without looking up the function definition to see what the meaning of the third argument is.

Quoted 1 line

Please drop this patch.

Quoted 7 lines

Thanks for your review and comments.

If the problem is the third argument 'can_sleep' which makes the code not be easily comprehensible, how about just removing 'can_sleep' from this function and keeping left parts because this function provides good flexibility to users to choose udelay or usleep_range according to the 'us' argument?

Hi Stanley,

I think that we need to get rid of 'can_sleep' across the entire UFS driver. As far as I can see the only context from which 'can_sleep' is set to true is ufshcd_host_reset_and_restore() and 'can_sleep' is set to true because ufshcd_hba_stop() is called with a spinlock held. Do you agree that it is wrong to call udelay() while holding a spinlock() and that doing so has a bad impact on the energy consumption of the UFS driver? Has it already been considered to use another mechanism to serialize REG_CONTROLLER_ENABLE changes, e.g. a mutex?

Thanks,

Bart.

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#8 lore raw

Re: [SPAM]Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

Hi Bart,

Quoted 35 lines · Bart Van Assche
Quoted 24 lines · Stanley Chu
Quoted 21 lines · Bart Van Assche

drivers/scsi/ufs/ufshcd.c static void ufshcd_print_pwr_info(struct ufs_hba *hba) · line 610

Quoted 11 lines · Stanley Chu
drivers/scsi/ufs/ufshcd.c +11 -0
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);

I don't like this function because I think it makes the UFS code harder to read instead of easier. The 'can_sleep' argument is only set by one caller which I think is a strong argument to remove that argument again and to move the code that depends on that argument from the above function into the caller. Additionally, it is not possible to comprehend what a ufshcd_wait_us() call does without looking up the function definition to see what the meaning of the third argument is.

Quoted 1 line

Please drop this patch.

Quoted 7 lines

Thanks for your review and comments.

If the problem is the third argument 'can_sleep' which makes the code not be easily comprehensible, how about just removing 'can_sleep' from this function and keeping left parts because this function provides good flexibility to users to choose udelay or usleep_range according to the 'us' argument?

Hi Stanley,

Quoted 7 lines

I think that we need to get rid of 'can_sleep' across the entire UFS driver. As far as I can see the only context from which 'can_sleep' is set to true is ufshcd_host_reset_and_restore() and 'can_sleep' is set to true because ufshcd_hba_stop() is called with a spinlock held. Do you agree that it is wrong to call udelay() while holding a spinlock() and that doing so has a bad impact on the energy consumption of the UFS driver?

Thanks for your positive suggestion.

Indeed using udelay() with spinlock held may have performance or power consumption concerns. However the concern in ufshcd_hba_stop() could be ignored in most cases since the execution period of changing bit 0 in REG_CONTROLLER_ENABLE from 1 to 0 shall be very fast. In my local environment, it could have only several 'ns' latency thus udelay() was never executed during the stress test. The delay here may be required for rare cases that host is under an abnormal state.

Quoted 2 lines

Has it already been considered to use another mechanism to serialize REG_CONTROLLER_ENABLE changes, e.g. a mutex?

I think mutex is not suitable for REG_CONTROLLER_ENABLE case because stopping host (by what ufshcd_hba_stop does) will reset doorbell bits in the same time by host, and those doorbell bits are looked up by UFS interrupt routine for request completion flow as well.

I agree that "can_sleep" can be improved and may have other optimized way but this is beyond this patch set. I would like to remove the "can_sleep" related modification from this patch set first.

Thanks,
Stanley Chu
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#9 lore raw

Re: [PATCH v6 3/7] scsi: ufs: introduce common delay function

In reply to Bart Van Assche

(Sorry to resend this mail with "[SPAM]" prefix removed in title)

Hi Bart,

Quoted 35 lines · Bart Van Assche
Quoted 24 lines · Stanley Chu
Quoted 21 lines · Bart Van Assche

drivers/scsi/ufs/ufshcd.c static void ufshcd_print_pwr_info(struct ufs_hba *hba) · line 610

Quoted 11 lines · Stanley Chu
drivers/scsi/ufs/ufshcd.c +11 -0
+void ufshcd_wait_us(unsigned long us, unsigned long tolerance, bool can_sleep)
+{
+	if (!us)
+		return;
+
+	if (us < 10 || !can_sleep)
+		udelay(us);
+	else
+		usleep_range(us, us + tolerance);
+}
+EXPORT_SYMBOL_GPL(ufshcd_wait_us);

I don't like this function because I think it makes the UFS code harder to read instead of easier. The 'can_sleep' argument is only set by one caller which I think is a strong argument to remove that argument again and to move the code that depends on that argument from the above function into the caller. Additionally, it is not possible to comprehend what a ufshcd_wait_us() call does without looking up the function definition to see what the meaning of the third argument is.

Quoted 1 line

Please drop this patch.

Quoted 7 lines

Thanks for your review and comments.

If the problem is the third argument 'can_sleep' which makes the code not be easily comprehensible, how about just removing 'can_sleep' from this function and keeping left parts because this function provides good flexibility to users to choose udelay or usleep_range according to the 'us' argument?

Hi Stanley,

Quoted 7 lines

I think that we need to get rid of 'can_sleep' across the entire UFS driver. As far as I can see the only context from which 'can_sleep' is set to true is ufshcd_host_reset_and_restore() and 'can_sleep' is set to true because ufshcd_hba_stop() is called with a spinlock held. Do you agree that it is wrong to call udelay() while holding a spinlock() and that doing so has a bad impact on the energy consumption of the UFS driver?

Thanks for your positive suggestion.

Indeed using udelay() with spinlock held may have performance or power consumption concerns. However the concern in ufshcd_hba_stop() could be ignored in most cases since the execution period of changing bit 0 in REG_CONTROLLER_ENABLE from 1 to 0 shall be very fast. In my local environment, it could have only several 'ns' latency thus udelay() was never executed during the stress test. The delay here may be required for rare cases that host is under an abnormal state.

Quoted 2 lines

Has it already been considered to use another mechanism to serialize REG_CONTROLLER_ENABLE changes, e.g. a mutex?

I think mutex is not suitable for REG_CONTROLLER_ENABLE case because stopping host (by what ufshcd_hba_stop does) will reset doorbell bits in the same time by host, and those doorbell bits are looked up by UFS interrupt routine for request completion flow as well.

I agree that "can_sleep" can be improved and may have other optimized way but this is beyond this patch set. I would like to remove the "can_sleep" related modification from this patch set first.

Thanks,
Stanley Chu

#10 lore raw

PATCH v6 4/7 scsi: ufs-mediatek: replace all delay places by common delay function

In reply to Stanley Chu

A common delay function is introduced in UFS core driver, thus ufs-mediatek can use it to replace all delay codes.

---
 drivers/scsi/ufs/ufs-mediatek.c | 21 +++++----------------
 1 file changed, 5 insertions(+), 16 deletions(-)
drivers/scsi/ufs/ufs-mediatek.c +5 -16
index 3b0e575d7460..0ff6781654fd 100644
@@ -100,17 +100,6 @@ static int ufs_mtk_bind_mphy(struct ufs_hba *hba)
 	return err;
 }
 
-static void ufs_mtk_udelay(unsigned long us)
-{
-	if (!us)
-		return;
-
-	if (us < 10)
-		udelay(us);
-	else
-		usleep_range(us, us + 10);
-}
-
 static int ufs_mtk_setup_ref_clk(struct ufs_hba *hba, bool on)
 {
 	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
@@ -123,7 +112,7 @@ static int ufs_mtk_setup_ref_clk(struct ufs_hba *hba, bool on)
 
 	if (on) {
 		ufs_mtk_ref_clk_notify(on, res);
-		ufs_mtk_udelay(host->ref_clk_ungating_wait_us);
+		ufshcd_wait_us(host->ref_clk_ungating_wait_us, 10, true);
 		ufshcd_writel(hba, REFCLK_REQUEST, REG_UFS_REFCLK_CTRL);
 	} else {
 		ufshcd_writel(hba, REFCLK_RELEASE, REG_UFS_REFCLK_CTRL);
@@ -138,7 +127,7 @@ static int ufs_mtk_setup_ref_clk(struct ufs_hba *hba, bool on)
 		if (((value & REFCLK_ACK) >> 1) == (value & REFCLK_REQUEST))
 			goto out;
 
-		usleep_range(100, 200);
+		ufshcd_wait_us(100, 100, true);
 	} while (time_before(jiffies, timeout));
 
 	dev_err(hba->dev, "missing ack of refclk req, reg: 0x%x\n", value);
@@ -150,7 +139,7 @@ static int ufs_mtk_setup_ref_clk(struct ufs_hba *hba, bool on)
 out:
 	host->ref_clk_enabled = on;
 	if (!on) {
-		ufs_mtk_udelay(host->ref_clk_gating_wait_us);
+		ufshcd_wait_us(host->ref_clk_gating_wait_us, 10, true);
 		ufs_mtk_ref_clk_notify(on, res);
 	}
 
@@ -430,12 +419,12 @@ static void ufs_mtk_device_reset(struct ufs_hba *hba)
 	 *
 	 * To be on safe side, keep the reset low for at least 10us.
 	 */
-	usleep_range(10, 15);
+	ufshcd_wait_us(10, 5, true);
 
 	ufs_mtk_device_reset_ctrl(1, res);
 
 	/* Some devices may need time to respond to rst_n */
-	usleep_range(10000, 15000);
+	ufshcd_wait_us(10000, 5000, true);
 
 	dev_info(hba->dev, "device reset done\n");
 }
Signature · 1 line
2.18.0
#11 lore raw

PATCH v6 5/7 scsi: ufs: allow customized delay for host enabling

In reply to Stanley Chu

Currently a 1 ms delay is applied before polling CONTROLLER_ENABLE bit. This delay may not be required or can be changed in different controllers. Make the delay as a changeable value in struct ufs_hba to allow it customized by vendors.

---
 drivers/scsi/ufs/ufshcd.c | 3 ++-
 drivers/scsi/ufs/ufshcd.h | 1 +
 2 files changed, 3 insertions(+), 1 deletion(-)
drivers/scsi/ufs/ufshcd.c +2 -1
index 9fea346f7d22..78b6ac6fcc4e 100644
@@ -4298,7 +4298,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 	 * instruction might be read back.
 	 * This delay can be changed based on the controller.
 	 */
-	ufshcd_wait_us(1000, 100, true);
+	ufshcd_wait_us(hba->hba_enable_delay_us, 100, true);
 
 	/* wait for the host controller to complete initialization */
 	retry = 10;
@@ -8421,6 +8421,7 @@ int ufshcd_init(struct ufs_hba *hba, void __iomem *mmio_base, unsigned int irq)
 
 	hba->mmio_base = mmio_base;
 	hba->irq = irq;
+	hba->hba_enable_delay_us = 1000;
 
 	err = ufshcd_hba_init(hba);
 	if (err)
drivers/scsi/ufs/ufshcd.h +1 -0
index 842f0223f5e5..b7111925d899 100644
@@ -663,6 +663,7 @@ struct ufs_hba {
 	u32 eh_flags;
 	u32 intr_mask;
 	u16 ee_ctrl_mask;
+	u16 hba_enable_delay_us;
 	bool is_powered;
 	struct ufs_init_prefetch init_prefetch_data;
 
Signature · 1 line
2.18.0
#12 lore raw

PATCH v6 6/7 scsi: ufs: make HCE polling more compact to improve initialization latency

In reply to Stanley Chu

Reduce the waiting period between each HCE (Host Controller Enable) polling from 5 ms to 1 ms. In the same time, increase the maximum polling times to make "total polling time" unchanged approximately.

This change could make HCE initializatoin faster to improve latency of ufshcd initialization, error recovery, and resume behaviors.

---
 drivers/scsi/ufs/ufshcd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/scsi/ufs/ufshcd.c +2 -2
index 78b6ac6fcc4e..c06a15df2f40 100644
@@ -4301,7 +4301,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 	ufshcd_wait_us(hba->hba_enable_delay_us, 100, true);
 
 	/* wait for the host controller to complete initialization */
-	retry = 10;
+	retry = 50;
 	while (ufshcd_is_hba_active(hba)) {
 		if (retry) {
 			retry--;
@@ -4310,7 +4310,7 @@ int ufshcd_hba_enable(struct ufs_hba *hba)
 				"Controller enable failed\n");
 			return -EIO;
 		}
-		ufshcd_wait_us(5000, 100, true);
+		ufshcd_wait_us(1000, 100, true);
 	}
 
 	/* enable UIC related interrupts */
Signature · 1 line
2.18.0
#13 lore raw

PATCH v6 7/7 scsi: ufs-mediatek: customize the delay for host enabling

In reply to Stanley Chu

MediaTek platform and UFS controller can dynamically customize the delay for host enabling according to different scenarios.

For example, if UniPro enters lower-power mode, such delay can be minimized, otherwise longer delay shall be expected.

---
 drivers/scsi/ufs/ufs-mediatek.c | 43 ++++++++++++++++++++++++++-------
 drivers/scsi/ufs/ufs-mediatek.h |  1 +
 2 files changed, 35 insertions(+), 9 deletions(-)
drivers/scsi/ufs/ufs-mediatek.c +34 -9
index 0ff6781654fd..c0fd7d2e4d0d 100644
@@ -30,11 +30,6 @@
 #define ufs_mtk_device_reset_ctrl(high, res) \
 	ufs_mtk_smc(UFS_MTK_SIP_DEVICE_RESET, high, res)
 
-#define ufs_mtk_unipro_powerdown(hba, powerdown) \
-	ufshcd_dme_set(hba, \
-		       UIC_ARG_MIB_SEL(VS_UNIPROPOWERDOWNCONTROL, 0), \
-		       powerdown)
-
 static void ufs_mtk_cfg_unipro_cg(struct ufs_hba *hba, bool enable)
 {
 	u32 tmp;
@@ -71,6 +66,21 @@ static void ufs_mtk_cfg_unipro_cg(struct ufs_hba *hba, bool enable)
 	}
 }
 
+static int ufs_mtk_hce_enable_notify(struct ufs_hba *hba,
+				     enum ufs_notify_change_status status)
+{
+	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
+
+	if (status == PRE_CHANGE) {
+		if (host->unipro_lpm)
+			hba->hba_enable_delay_us = 0;
+		else
+			hba->hba_enable_delay_us = 600;
+	}
+
+	return 0;
+}
+
 static int ufs_mtk_bind_mphy(struct ufs_hba *hba)
 {
 	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
@@ -324,12 +334,26 @@ static int ufs_mtk_pwr_change_notify(struct ufs_hba *hba,
 	return ret;
 }
 
+static int ufs_mtk_unipro_set_pm(struct ufs_hba *hba, u32 lpm)
+{
+	int ret;
+	struct ufs_mtk_host *host = ufshcd_get_variant(hba);
+
+	ret = ufshcd_dme_set(hba,
+			     UIC_ARG_MIB_SEL(VS_UNIPROPOWERDOWNCONTROL, 0),
+			     lpm);
+	if (!ret)
+		host->unipro_lpm = lpm;
+
+	return ret;
+}
+
 static int ufs_mtk_pre_link(struct ufs_hba *hba)
 {
 	int ret;
 	u32 tmp;
 
-	ufs_mtk_unipro_powerdown(hba, 0);
+	ufs_mtk_unipro_set_pm(hba, 0);
 
 	/*
 	 * Setting PA_Local_TX_LCC_Enable to 0 before link startup
@@ -437,7 +461,7 @@ static int ufs_mtk_link_set_hpm(struct ufs_hba *hba)
 	if (err)
 		return err;
 
-	err = ufs_mtk_unipro_powerdown(hba, 0);
+	err = ufs_mtk_unipro_set_pm(hba, 0);
 	if (err)
 		return err;
 
@@ -458,10 +482,10 @@ static int ufs_mtk_link_set_lpm(struct ufs_hba *hba)
 {
 	int err;
 
-	err = ufs_mtk_unipro_powerdown(hba, 1);
+	err = ufs_mtk_unipro_set_pm(hba, 1);
 	if (err) {
 		/* Resume UniPro state for following error recovery */
-		ufs_mtk_unipro_powerdown(hba, 0);
+		ufs_mtk_unipro_set_pm(hba, 0);
 		return err;
 	}
 
@@ -552,6 +576,7 @@ static struct ufs_hba_variant_ops ufs_hba_mtk_vops = {
 	.name                = "mediatek.ufshci",
 	.init                = ufs_mtk_init,
 	.setup_clocks        = ufs_mtk_setup_clocks,
+	.hce_enable_notify   = ufs_mtk_hce_enable_notify,
 	.link_startup_notify = ufs_mtk_link_startup_notify,
 	.pwr_change_notify   = ufs_mtk_pwr_change_notify,
 	.apply_dev_quirks    = ufs_mtk_apply_dev_quirks,
drivers/scsi/ufs/ufs-mediatek.h +1 -0
index 4c787b99fe41..5bbd3e9cbae2 100644
@@ -91,6 +91,7 @@ enum {
 struct ufs_mtk_host {
 	struct ufs_hba *hba;
 	struct phy *mphy;
+	bool unipro_lpm;
 	bool ref_clk_enabled;
 	u16 ref_clk_ungating_wait_us;
 	u16 ref_clk_gating_wait_us;
Signature · 1 line
2.18.0