[PATCH v4 0/3] Changes for ufshcd.c

[PATCH v4 0/3] Changes for ufshcd.c

[Bean Huo]

From: Bean Huo <beanhuo@micron.com>

v3--v4:
    1. Change patch 1/3 title
    2. Fix one typo in patch 3/3
    3. Fix coding type issue in  patch 2/3 based on Bart's comment

v2--v3:
    1. Add patch 3/3
    2. Add remove unused parameter param_offset in patch 1/3 based on Daejun's comment
    3. Further optimized patch 2/3 based on Bart's comment

v1--v2:
    1. change in ufshcd_lu_init() in  patch 2/2:
        1) Remove duplicate parameter initialization
        2) Move lun parameter initialization before kmalloc()


Bean Huo (3):
  scsi: ufs: core: Revert "WB is only available on LUN #0 to #7"
  scsi: ufs: core: Cleanup ufshcd_slave_alloc()
  scsi: ufs: core: Use is_visible to control UFS unit descriptor sysfs
    nodes

 drivers/ufs/core/ufs-sysfs.c   |  21 ++++-
 drivers/ufs/core/ufshcd-priv.h |   6 +-
 drivers/ufs/core/ufshcd.c      | 152 ++++++++++++---------------------
 3 files changed, 74 insertions(+), 105 deletions(-)

-- 
2.34.1

[PATCH v4 1/3] scsi: ufs: core: Revert "WB is only available on LUN #0 to #7"

[Bean Huo]

From: Bean Huo <beanhuo@micron.com>

One newer 'commit d3d9c4570285 ("scsi: ufs: Fix memory corruption by
ufshcd_read_desc_param()")' has properly fixed stack overflow issue,
'commit a2fca52ee640 ("scsi: ufs: WB is only available on LUN #0 to #7")'
is no longer required, we can revert it now.

Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Bean Huo <beanhuo@micron.com>
---
 drivers/ufs/core/ufs-sysfs.c   | 3 +--
 drivers/ufs/core/ufshcd-priv.h | 6 +-----
 drivers/ufs/core/ufshcd.c      | 2 +-
 3 files changed, 3 insertions(+), 8 deletions(-)

diff --git a/drivers/ufs/core/ufs-sysfs.c b/drivers/ufs/core/ufs-sysfs.c
index 53aea56d1de1..eb6b278c4e79 100644
--- a/drivers/ufs/core/ufs-sysfs.c
+++ b/drivers/ufs/core/ufs-sysfs.c
@@ -1234,8 +1234,7 @@ static ssize_t _pname##_show(struct device *dev,			\
 	struct scsi_device *sdev = to_scsi_device(dev);			\
 	struct ufs_hba *hba = shost_priv(sdev->host);			\
 	u8 lun = ufshcd_scsi_to_upiu_lun(sdev->lun);			\
-	if (!ufs_is_valid_unit_desc_lun(&hba->dev_info, lun,		\
-				_duname##_DESC_PARAM##_puname))		\
+	if (!ufs_is_valid_unit_desc_lun(&hba->dev_info, lun))		\
 		return -EINVAL;						\
 	return ufs_sysfs_read_desc_param(hba, QUERY_DESC_IDN_##_duname,	\
 		lun, _duname##_DESC_PARAM##_puname, buf, _size);	\
diff --git a/drivers/ufs/core/ufshcd-priv.h b/drivers/ufs/core/ufshcd-priv.h
index f68ca33f6ac7..a9e8e1f5afe7 100644
--- a/drivers/ufs/core/ufshcd-priv.h
+++ b/drivers/ufs/core/ufshcd-priv.h
@@ -293,16 +293,12 @@ static inline int ufshcd_rpm_put(struct ufs_hba *hba)
  * @lun: LU number to check
  * @return: true if the lun has a matching unit descriptor, false otherwise
  */
-static inline bool ufs_is_valid_unit_desc_lun(struct ufs_dev_info *dev_info,
-		u8 lun, u8 param_offset)
+static inline bool ufs_is_valid_unit_desc_lun(struct ufs_dev_info *dev_info, u8 lun)
 {
 	if (!dev_info || !dev_info->max_lu_supported) {
 		pr_err("Max General LU supported by UFS isn't initialized\n");
 		return false;
 	}
-	/* WB is available only for the logical unit from 0 to 7 */
-	if (param_offset == UNIT_DESC_PARAM_WB_BUF_ALLOC_UNITS)
-		return lun < UFS_UPIU_MAX_WB_LUN_ID;
 	return lun == UFS_UPIU_RPMB_WLUN || (lun < dev_info->max_lu_supported);
 }
 
diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index c8f0fe740005..8cdba9a54870 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -3606,7 +3606,7 @@ static inline int ufshcd_read_unit_desc_param(struct ufs_hba *hba,
 	 * Unit descriptors are only available for general purpose LUs (LUN id
 	 * from 0 to 7) and RPMB Well known LU.
 	 */
-	if (!ufs_is_valid_unit_desc_lun(&hba->dev_info, lun, param_offset))
+	if (!ufs_is_valid_unit_desc_lun(&hba->dev_info, lun))
 		return -EOPNOTSUPP;
 
 	return ufshcd_read_desc_param(hba, QUERY_DESC_IDN_UNIT, lun,
-- 
2.34.1

RE: [PATCH v4 1/3] scsi: ufs: core: Revert "WB is only available on LUN #0 to #7"

[Arthur Simchaev]

> -----Original Message-----
> From: Bean Huo <beanhuo@iokpp.de>
> Sent: Sunday, October 23, 2022 12:37 AM
> To: alim.akhtar@samsung.com; Avri Altman <Avri.Altman@wdc.com>;
> asutoshd@codeaurora.org; jejb@linux.ibm.com;
> martin.petersen@oracle.com; stanley.chu@mediatek.com;
> beanhuo@micron.com; bvanassche@acm.org; tomas.winkler@intel.com;
> cang@codeaurora.org; daejun7.park@samsung.com
> Cc: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org; Jaegeuk Kim
> <jaegeuk@kernel.org>
> Subject: [PATCH v4 1/3] scsi: ufs: core: Revert "WB is only available on LUN #0
> to #7"
> 
> CAUTION: This email originated from outside of Western Digital. Do not click
> on links or open attachments unless you recognize the sender and know that
> the content is safe.
> 
> 
> From: Bean Huo <beanhuo@micron.com>
> 
> One newer 'commit d3d9c4570285 ("scsi: ufs: Fix memory corruption by
> ufshcd_read_desc_param()")' has properly fixed stack overflow issue,
> 'commit a2fca52ee640 ("scsi: ufs: WB is only available on LUN #0 to #7")'
> is no longer required, we can revert it now.
> 
> Cc: Jaegeuk Kim <jaegeuk@kernel.org>
> Signed-off-by: Bean Huo <beanhuo@micron.com>

Reviewed-by: Arthur Simchaev <arthur.simchaev@wdc.com>

Re: [PATCH v4 1/3] scsi: ufs: core: Revert "WB is only available on LUN #0 to #7"

[Bart Van Assche]

On 10/22/22 14:36, Bean Huo wrote:
> One newer 'commit d3d9c4570285 ("scsi: ufs: Fix memory corruption by

One -> A?

> ufshcd_read_desc_param()")' has properly fixed stack overflow issue,
> 'commit a2fca52ee640 ("scsi: ufs: WB is only available on LUN #0 to #7")'
> is no longer required, we can revert it now.

Anyway:

Reviewed-by: Bart Van Assche <bvanassche@acm.org>

[PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bean Huo]

From: Bean Huo <beanhuo@micron.com>

Combine ufshcd_get_lu_power_on_wp_status() and ufshcd_set_queue_depth()
into one single ufshcd_lu_init(), so that we only need to read the LUN
descriptor once.

Signed-off-by: Bean Huo <beanhuo@micron.com>
---
 drivers/ufs/core/ufshcd.c | 150 ++++++++++++++------------------------
 1 file changed, 53 insertions(+), 97 deletions(-)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 8cdba9a54870..8599780022fb 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -4859,100 +4859,6 @@ static int ufshcd_verify_dev_init(struct ufs_hba *hba)
 	return err;
 }
 
-/**
- * ufshcd_set_queue_depth - set lun queue depth
- * @sdev: pointer to SCSI device
- *
- * Read bLUQueueDepth value and activate scsi tagged command
- * queueing. For WLUN, queue depth is set to 1. For best-effort
- * cases (bLUQueueDepth = 0) the queue depth is set to a maximum
- * value that host can queue.
- */
-static void ufshcd_set_queue_depth(struct scsi_device *sdev)
-{
-	int ret = 0;
-	u8 lun_qdepth;
-	struct ufs_hba *hba;
-
-	hba = shost_priv(sdev->host);
-
-	lun_qdepth = hba->nutrs;
-	ret = ufshcd_read_unit_desc_param(hba,
-					  ufshcd_scsi_to_upiu_lun(sdev->lun),
-					  UNIT_DESC_PARAM_LU_Q_DEPTH,
-					  &lun_qdepth,
-					  sizeof(lun_qdepth));
-
-	/* Some WLUN doesn't support unit descriptor */
-	if (ret == -EOPNOTSUPP)
-		lun_qdepth = 1;
-	else if (!lun_qdepth)
-		/* eventually, we can figure out the real queue depth */
-		lun_qdepth = hba->nutrs;
-	else
-		lun_qdepth = min_t(int, lun_qdepth, hba->nutrs);
-
-	dev_dbg(hba->dev, "%s: activate tcq with queue depth %d\n",
-			__func__, lun_qdepth);
-	scsi_change_queue_depth(sdev, lun_qdepth);
-}
-
-/*
- * ufshcd_get_lu_wp - returns the "b_lu_write_protect" from UNIT DESCRIPTOR
- * @hba: per-adapter instance
- * @lun: UFS device lun id
- * @b_lu_write_protect: pointer to buffer to hold the LU's write protect info
- *
- * Returns 0 in case of success and b_lu_write_protect status would be returned
- * @b_lu_write_protect parameter.
- * Returns -ENOTSUPP if reading b_lu_write_protect is not supported.
- * Returns -EINVAL in case of invalid parameters passed to this function.
- */
-static int ufshcd_get_lu_wp(struct ufs_hba *hba,
-			    u8 lun,
-			    u8 *b_lu_write_protect)
-{
-	int ret;
-
-	if (!b_lu_write_protect)
-		ret = -EINVAL;
-	/*
-	 * According to UFS device spec, RPMB LU can't be write
-	 * protected so skip reading bLUWriteProtect parameter for
-	 * it. For other W-LUs, UNIT DESCRIPTOR is not available.
-	 */
-	else if (lun >= hba->dev_info.max_lu_supported)
-		ret = -ENOTSUPP;
-	else
-		ret = ufshcd_read_unit_desc_param(hba,
-					  lun,
-					  UNIT_DESC_PARAM_LU_WR_PROTECT,
-					  b_lu_write_protect,
-					  sizeof(*b_lu_write_protect));
-	return ret;
-}
-
-/**
- * ufshcd_get_lu_power_on_wp_status - get LU's power on write protect
- * status
- * @hba: per-adapter instance
- * @sdev: pointer to SCSI device
- *
- */
-static inline void ufshcd_get_lu_power_on_wp_status(struct ufs_hba *hba,
-						    const struct scsi_device *sdev)
-{
-	if (hba->dev_info.f_power_on_wp_en &&
-	    !hba->dev_info.is_lu_power_on_wp) {
-		u8 b_lu_write_protect;
-
-		if (!ufshcd_get_lu_wp(hba, ufshcd_scsi_to_upiu_lun(sdev->lun),
-				      &b_lu_write_protect) &&
-		    (b_lu_write_protect == UFS_LU_POWER_ON_WP))
-			hba->dev_info.is_lu_power_on_wp = true;
-	}
-}
-
 /**
  * ufshcd_setup_links - associate link b/w device wlun and other luns
  * @sdev: pointer to SCSI device
@@ -4990,6 +4896,58 @@ static void ufshcd_setup_links(struct ufs_hba *hba, struct scsi_device *sdev)
 	}
 }
 
+/**
+ * ufshcd_lu_init - Initialize the relevant parameters of the LU
+ * @hba: per-adapter instance
+ * @sdev: pointer to SCSI device
+ */
+static void ufshcd_lu_init(struct ufs_hba *hba, struct scsi_device *sdev)
+{
+	int len = hba->desc_size[QUERY_DESC_IDN_UNIT];
+	u8 lun = ufshcd_scsi_to_upiu_lun(sdev->lun);
+	u8 lun_qdepth = hba->nutrs;
+	u8 *desc_buf;
+	int ret;
+
+	desc_buf = kzalloc(len, GFP_KERNEL);
+	if (!desc_buf)
+		goto set_qdepth;
+
+	ret = ufshcd_read_unit_desc_param(hba, lun, 0, desc_buf, len);
+	if (ret < 0) {
+		if (ret == -EOPNOTSUPP)
+			/* If LU doesn't support unit descriptor, its queue depth is set to 1 */
+			lun_qdepth = 1;
+		kfree(desc_buf);
+		goto set_qdepth;
+	}
+
+	if (desc_buf[UNIT_DESC_PARAM_LU_Q_DEPTH]) {
+		/*
+		 * In per-LU queueing architecture, bLUQueueDepth will not be 0, then we will
+		 * use the smaller between UFSHCI CAP.NUTRS and UFS LU bLUQueueDepth
+		 */
+		lun_qdepth = min_t(int, desc_buf[UNIT_DESC_PARAM_LU_Q_DEPTH], hba->nutrs);
+	}
+	/*
+	 * According to UFS device specification, the write protection mode is only supported by
+	 * normal LU, not supported by WLUN.
+	 */
+	if (hba->dev_info.f_power_on_wp_en && lun < hba->dev_info.max_lu_supported &&
+	    !hba->dev_info.is_lu_power_on_wp &&
+	    desc_buf[UNIT_DESC_PARAM_LU_WR_PROTECT] == UFS_LU_POWER_ON_WP)
+		hba->dev_info.is_lu_power_on_wp = true;
+
+	kfree(desc_buf);
+set_qdepth:
+	/*
+	 * For WLUNs that don't support unit descriptor, queue depth is set to 1. For LUs whose
+	 * bLUQueueDepth == 0, the queue depth is set to a maximum value that host can queue.
+	 */
+	dev_dbg(hba->dev, "Set LU %x queue depth %d\n", lun, lun_qdepth);
+	scsi_change_queue_depth(sdev, lun_qdepth);
+}
+
 /**
  * ufshcd_slave_alloc - handle initial SCSI device configurations
  * @sdev: pointer to SCSI device
@@ -5017,9 +4975,7 @@ static int ufshcd_slave_alloc(struct scsi_device *sdev)
 	/* WRITE_SAME command is not supported */
 	sdev->no_write_same = 1;
 
-	ufshcd_set_queue_depth(sdev);
-
-	ufshcd_get_lu_power_on_wp_status(hba, sdev);
+	ufshcd_lu_init(hba, sdev);
 
 	ufshcd_setup_links(hba, sdev);
 
-- 
2.34.1

RE: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Arthur Simchaev]

> -----Original Message-----
> From: Bean Huo <beanhuo@iokpp.de>
> Sent: Sunday, October 23, 2022 12:37 AM
> To: alim.akhtar@samsung.com; Avri Altman <Avri.Altman@wdc.com>;
> asutoshd@codeaurora.org; jejb@linux.ibm.com;
> martin.petersen@oracle.com; stanley.chu@mediatek.com;
> beanhuo@micron.com; bvanassche@acm.org; tomas.winkler@intel.com;
> cang@codeaurora.org; daejun7.park@samsung.com
> Cc: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org
> Subject: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()
> 
> CAUTION: This email originated from outside of Western Digital. Do not click
> on links or open attachments unless you recognize the sender and know that
> the content is safe.
> 
> 
> From: Bean Huo <beanhuo@micron.com>
> 
> Combine ufshcd_get_lu_power_on_wp_status() and
> ufshcd_set_queue_depth()
> into one single ufshcd_lu_init(), so that we only need to read the LUN
> descriptor once.
> 
> Signed-off-by: Bean Huo <beanhuo@micron.com>

Reviewed-by: Arthur Simchaev <arthur.simchaev@wdc.com>

Re: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bart Van Assche]

On 10/22/22 14:36, Bean Huo wrote:
> +	u8 lun_qdepth = hba->nutrs;

How about changing this into hba->nutrs - 1 since one tag is reserved 
('reserved_slot')?

Thanks,

Bart.

Re: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bean Huo]

On Mon, 2022-10-24 at 16:25 -0700, Bart Van Assche wrote:
> On 10/22/22 14:36, Bean Huo wrote:
> > +       u8 lun_qdepth = hba->nutrs;
> 
> How about changing this into hba->nutrs - 1 since one tag is reserved
> ('reserved_slot')?
> 
> Thanks,


Bart,

If having a look at 

scsi_change_queue_depth() {

  depth = min_t(int, depth, scsi_device_max_queue_depth(sdev));
	...
}

so the finnal depth per Lu is still 31.  I did test and debug, this is
true.



+set_qdepth:
+	/*
+	 * For WLUNs that don't support unit descriptor, queue depth
is set to 1. For LUs whose
+	 * bLUQueueDepth == 0, the queue depth is set to a maximum
value that host can queue.
+	 */
+	dev_dbg(hba->dev, "Set LU %x queue depth %d\n", lun,
lun_qdepth);
+	scsi_change_queue_depth(sdev, lun_qdepth); 



so lun_qdepth is 32, but scsi_change_queue_depth() will return 31,
since we set sdev->host->can_queue is 31.

If you need me to change the patch, let me know, it is no problem for
me to change to: 


u8 lun_qdepth = hba->nutrs - UFSHCD_NUM_RESERVED.

or

u8 lun_qdepth = UFSHCD_CMD_PER_LUN.


Kind regards,
Bean

Re: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bart Van Assche]

On 10/25/22 04:44, Bean Huo wrote:
> On Mon, 2022-10-24 at 16:25 -0700, Bart Van Assche wrote:
> If having a look at
> 
> scsi_change_queue_depth() {
> 
>    depth = min_t(int, depth, scsi_device_max_queue_depth(sdev));
> 	...
> }
> 
> so the final depth per Lu is still 31.  I did test and debug, this is
> true.

Hi Bean,

Thanks for the clarification. I will add my Reviewed-by to this patch.

Bart.

Re: [PATCH v4 2/3] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bart Van Assche]

On 10/22/22 14:36, Bean Huo wrote:
> Combine ufshcd_get_lu_power_on_wp_status() and ufshcd_set_queue_depth()
> into one single ufshcd_lu_init(), so that we only need to read the LUN
> descriptor once.

Reviewed-by: Bart Van Assche <bvanassche@acm.org>

[PATCH v4 3/3] scsi: ufs: core: Use is_visible to control UFS unit descriptor sysfs nodes

[Bean Huo]

From: Bean Huo <beanhuo@micron.com>

UFS Boot and Device W-LUs do not have unit descriptors, and RPMB does not
support WB, we can use is_visible() to control which nodes are visible
and which are not.

Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Bean Huo <beanhuo@micron.com>
---
 drivers/ufs/core/ufs-sysfs.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/ufs/core/ufs-sysfs.c b/drivers/ufs/core/ufs-sysfs.c
index eb6b278c4e79..883f0e44b54e 100644
--- a/drivers/ufs/core/ufs-sysfs.c
+++ b/drivers/ufs/core/ufs-sysfs.c
@@ -1285,9 +1285,27 @@ static struct attribute *ufs_sysfs_unit_descriptor[] = {
 	NULL,
 };
 
+static umode_t ufs_unit_descriptor_is_visible(struct kobject *kobj, struct attribute *attr, int n)
+{
+	struct device *dev = container_of(kobj, struct device, kobj);
+	struct scsi_device *sdev = to_scsi_device(dev);
+	u8 lun = ufshcd_scsi_to_upiu_lun(sdev->lun);
+	umode_t mode = attr->mode;
+
+	if (lun == UFS_UPIU_BOOT_WLUN || lun == UFS_UPIU_UFS_DEVICE_WLUN)
+		/* Boot and device WLUN have no unit descriptors */
+		mode = 0;
+	if (lun == UFS_UPIU_RPMB_WLUN && attr == &dev_attr_wb_buf_alloc_units.attr)
+		mode = 0;
+
+	return mode;
+}
+
+
 const struct attribute_group ufs_sysfs_unit_descriptor_group = {
 	.name = "unit_descriptor",
 	.attrs = ufs_sysfs_unit_descriptor,
+	.is_visible = ufs_unit_descriptor_is_visible,
 };
 
 static ssize_t dyn_cap_needed_attribute_show(struct device *dev,
-- 
2.34.1

RE: [PATCH v4 3/3] scsi: ufs: core: Use is_visible to control UFS unit descriptor sysfs nodes

[Arthur Simchaev]

> -----Original Message-----
> From: Bean Huo <beanhuo@iokpp.de>
> Sent: Sunday, October 23, 2022 12:37 AM
> To: alim.akhtar@samsung.com; Avri Altman <Avri.Altman@wdc.com>;
> asutoshd@codeaurora.org; jejb@linux.ibm.com;
> martin.petersen@oracle.com; stanley.chu@mediatek.com;
> beanhuo@micron.com; bvanassche@acm.org; tomas.winkler@intel.com;
> cang@codeaurora.org; daejun7.park@samsung.com
> Cc: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org
> Subject: [PATCH v4 3/3] scsi: ufs: core: Use is_visible to control UFS unit
> descriptor sysfs nodes
> 
> CAUTION: This email originated from outside of Western Digital. Do not click
> on links or open attachments unless you recognize the sender and know that
> the content is safe.
> 
> 
> From: Bean Huo <beanhuo@micron.com>
> 
> UFS Boot and Device W-LUs do not have unit descriptors, and RPMB does not
> support WB, we can use is_visible() to control which nodes are visible
> and which are not.
> 
> Reviewed-by: Bart Van Assche <bvanassche@acm.org>
> Signed-off-by: Bean Huo <beanhuo@micron.com>

Reviewed-by: Arthur Simchaev <arthur.simchaev@wdc.com>