[RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Kiwoong Kim]

Some SoC specific might need command history for
various reasons, such as stacking command contexts
in system memory to check for debugging in the future
or scaling some DVFS knobs to boost IO throughput.

What you would do with the information could be
variant per SoC vendor.

Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
---
 drivers/scsi/ufs/ufshcd.c | 4 ++++
 drivers/scsi/ufs/ufshcd.h | 8 ++++++++
 2 files changed, 12 insertions(+)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 52abe82..0eae3ce 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
 	/* issue command to the controller */
 	spin_lock_irqsave(hba->host->host_lock, flags);
 	ufshcd_vops_setup_xfer_req(hba, tag, true);
+	if (cmd)
+		ufshcd_vops_cmd_log(hba, cmd, 1);
 	ufshcd_send_command(hba, tag);
 out_unlock:
 	spin_unlock_irqrestore(hba->host->host_lock, flags);
@@ -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
 			/* Mark completed command as NULL in LRB */
 			lrbp->cmd = NULL;
 			lrbp->compl_time_stamp = ktime_get();
+			ufshcd_vops_cmd_log(hba, cmd, 2);
+
 			/* Do not touch lrbp after scsi done */
 			cmd->scsi_done(cmd);
 			__ufshcd_release(hba);
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index c774012..80c4f0d 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -307,6 +307,7 @@ struct ufs_hba_variant_ops {
 	void	(*config_scaling_param)(struct ufs_hba *hba,
 					struct devfreq_dev_profile *profile,
 					void *data);
+	void	(*cmd_log)(struct ufs_hba *hba, struct scsi_cmnd *cmd, int enter);
 };
 
 /* clock gating state  */
@@ -1137,6 +1138,13 @@ static inline void ufshcd_vops_config_scaling_param(struct ufs_hba *hba,
 		hba->vops->config_scaling_param(hba, profile, data);
 }
 
+static inline void ufshcd_vops_cmd_log(struct ufs_hba *hba,
+					 struct scsi_cmnd *cmd, int enter)
+{
+	if (hba->vops && hba->vops->cmd_log)
+		hba->vops->cmd_log(hba, cmd, enter);
+}
+
 extern struct ufs_pm_lvl_states ufs_pm_lvl_states[];
 
 /*
-- 
2.7.4

[RFC PATCH v1 2/2] exynos-ufs: support command history

[Kiwoong Kim]

This patch is to record contexts of incoming commands
in a circular queue. ufshcd.c has already some function
tracer calls to get command history but ftrace would be
gone when system dies before you get the information,
such as panic cases.

Setting EN_PRINT_UFS_LOG to one enables to print
information in the circular queue whenever ufshcd.c
invokes a dbg_register_dump callback.

Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
---
 drivers/scsi/ufs/Makefile         |   2 +-
 drivers/scsi/ufs/ufs-exynos-dbg.c | 208 ++++++++++++++++++++++++++++++++++++++
 drivers/scsi/ufs/ufs-exynos-if.h  |  19 ++++
 drivers/scsi/ufs/ufs-exynos.c     |  43 ++++++++
 drivers/scsi/ufs/ufs-exynos.h     |  13 +++
 5 files changed, 284 insertions(+), 1 deletion(-)
 create mode 100644 drivers/scsi/ufs/ufs-exynos-dbg.c
 create mode 100644 drivers/scsi/ufs/ufs-exynos-if.h

diff --git a/drivers/scsi/ufs/Makefile b/drivers/scsi/ufs/Makefile
index f0c5b95..d9e4da7 100644
--- a/drivers/scsi/ufs/Makefile
+++ b/drivers/scsi/ufs/Makefile
@@ -4,7 +4,7 @@ obj-$(CONFIG_SCSI_UFS_DWC_TC_PCI) += tc-dwc-g210-pci.o ufshcd-dwc.o tc-dwc-g210.
 obj-$(CONFIG_SCSI_UFS_DWC_TC_PLATFORM) += tc-dwc-g210-pltfrm.o ufshcd-dwc.o tc-dwc-g210.o
 obj-$(CONFIG_SCSI_UFS_CDNS_PLATFORM) += cdns-pltfrm.o
 obj-$(CONFIG_SCSI_UFS_QCOM) += ufs-qcom.o
-obj-$(CONFIG_SCSI_UFS_EXYNOS) += ufs-exynos.o
+obj-$(CONFIG_SCSI_UFS_EXYNOS) += ufs-exynos.o ufs-exynos-dbg.o
 obj-$(CONFIG_SCSI_UFSHCD) += ufshcd-core.o
 ufshcd-core-y				+= ufshcd.o ufs-sysfs.o
 ufshcd-core-$(CONFIG_SCSI_UFS_BSG)	+= ufs_bsg.o
diff --git a/drivers/scsi/ufs/ufs-exynos-dbg.c b/drivers/scsi/ufs/ufs-exynos-dbg.c
new file mode 100644
index 0000000..2265af7
--- /dev/null
+++ b/drivers/scsi/ufs/ufs-exynos-dbg.c
@@ -0,0 +1,208 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * UFS Exynos debugging functions
+ *
+ * Copyright (C) 2020 Samsung Electronics Co., Ltd.
+ * Author: Kiwoong Kim <kwmad.kim@samsung.com>
+ *
+ */
+#include <linux/platform_device.h>
+#include <linux/module.h>
+#include "ufshcd.h"
+#include "ufs-exynos.h"
+#include "ufs-exynos-if.h"
+
+#define EN_PRINT_UFS_LOG 1
+
+/* Structure for ufs cmd logging */
+#define MAX_CMD_LOGS    32
+
+struct cmd_data {
+	unsigned int tag;
+	unsigned int sct;
+	unsigned long lba;
+	u64 start_time;
+	u64 end_time;
+	u64 outstanding_reqs;
+	int retries;
+	unsigned char op;
+};
+
+struct ufs_cmd_info {
+	u32 total;
+	u32 last;
+	struct cmd_data data[MAX_CMD_LOGS];
+	struct cmd_data *pdata[32];	/* Currently, 32 slots */
+};
+
+#define DBG_NUM_OF_HOSTS	1
+struct ufs_s_dbg_mgr {
+	struct ufs_exynos_handle *handle;
+	int active;
+	u64 first_time;
+	u64 time;
+
+	/* cmd log */
+	struct ufs_cmd_info cmd_info;
+	struct cmd_data cmd_log;		/* temp buffer to put */
+	spinlock_t cmd_lock;
+};
+static struct ufs_s_dbg_mgr ufs_s_dbg[DBG_NUM_OF_HOSTS];
+static int ufs_s_dbg_mgr_idx;
+
+static void ufs_s_print_cmd_log(struct ufs_s_dbg_mgr *mgr, struct device *dev)
+{
+	struct ufs_cmd_info *cmd_info = &mgr->cmd_info;
+	struct cmd_data *data = cmd_info->data;
+	u32 i;
+	u32 last;
+	u32 max = MAX_CMD_LOGS;
+	unsigned long flags;
+	u32 total;
+
+	spin_lock_irqsave(&mgr->cmd_lock, flags);
+	total = cmd_info->total;
+	if (cmd_info->total < max)
+		max = cmd_info->total;
+	last = (cmd_info->last + MAX_CMD_LOGS - 1) % MAX_CMD_LOGS;
+	spin_unlock_irqrestore(&mgr->cmd_lock, flags);
+
+	dev_err(dev, ":---------------------------------------------------\n");
+	dev_err(dev, ":\t\tSCSI CMD(%u)\n", total - 1);
+	dev_err(dev, ":---------------------------------------------------\n");
+	dev_err(dev, ":OP, TAG, LBA, SCT, RETRIES, STIME, ETIME, REQS\n\n");
+
+	for (i = 0 ; i < max ; i++, data++) {
+		dev_err(dev, ": 0x%02x, %02d, 0x%08lx, 0x%04x, %d, %llu, %llu, 0x%llx %s",
+				data->op,
+				data->tag,
+				data->lba,
+				data->sct,
+				data->retries,
+				data->start_time,
+				data->end_time,
+				data->outstanding_reqs,
+				((last == i) ? "<--" : " "));
+		if (last == i)
+			dev_err(dev, "\n");
+	}
+}
+
+static void ufs_s_put_cmd_log(struct ufs_s_dbg_mgr *mgr,
+			      struct cmd_data *cmd_data)
+{
+	struct ufs_cmd_info *cmd_info = &mgr->cmd_info;
+	unsigned long flags;
+	struct cmd_data *pdata;
+
+	spin_lock_irqsave(&mgr->cmd_lock, flags);
+	pdata = &cmd_info->data[cmd_info->last];
+	++cmd_info->total;
+	++cmd_info->last;
+	cmd_info->last = cmd_info->last % MAX_CMD_LOGS;
+	spin_unlock_irqrestore(&mgr->cmd_lock, flags);
+
+	pdata->op = cmd_data->op;
+	pdata->tag = cmd_data->tag;
+	pdata->lba = cmd_data->lba;
+	pdata->sct = cmd_data->sct;
+	pdata->retries = cmd_data->retries;
+	pdata->start_time = cmd_data->start_time;
+	pdata->end_time = 0;
+	pdata->outstanding_reqs = cmd_data->outstanding_reqs;
+	cmd_info->pdata[cmd_data->tag] = pdata;
+}
+
+/*
+ * EXTERNAL FUNCTIONS
+ *
+ * There are two classes that are to initialize data structures for debug
+ * and to define actual behavior.
+ */
+void exynos_ufs_dump_info(struct ufs_exynos_handle *handle, struct device *dev)
+{
+	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
+
+	if (mgr->active == 0)
+		goto out;
+
+	mgr->time = cpu_clock(raw_smp_processor_id());
+
+#if EN_PRINT_UFS_LOG
+	ufs_s_print_cmd_log(mgr, dev);
+#endif
+
+	if (mgr->first_time == 0ULL)
+		mgr->first_time = mgr->time;
+out:
+	return;
+}
+
+void exynos_ufs_cmd_log_start(struct ufs_exynos_handle *handle,
+			      struct ufs_hba *hba, struct scsi_cmnd *cmd)
+{
+	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
+	int cpu = raw_smp_processor_id();
+	struct cmd_data *cmd_log = &mgr->cmd_log;	/* temp buffer to put */
+	unsigned long lba = (cmd->cmnd[2] << 24) |
+					(cmd->cmnd[3] << 16) |
+					(cmd->cmnd[4] << 8) |
+					(cmd->cmnd[5] << 0);
+	unsigned int sct = (cmd->cmnd[7] << 8) |
+					(cmd->cmnd[8] << 0);
+
+	if (mgr->active == 0)
+		return;
+
+	cmd_log->start_time = cpu_clock(cpu);
+	cmd_log->op = cmd->cmnd[0];
+	cmd_log->tag = cmd->request->tag;
+
+	/* This function runtime is protected by spinlock from outside */
+	cmd_log->outstanding_reqs = hba->outstanding_reqs;
+
+	/* unmap */
+	if (cmd->cmnd[0] != UNMAP)
+		cmd_log->lba = lba;
+
+	cmd_log->sct = sct;
+	cmd_log->retries = cmd->allowed;
+
+	ufs_s_put_cmd_log(mgr, cmd_log);
+}
+
+void exynos_ufs_cmd_log_end(struct ufs_exynos_handle *handle,
+			    struct ufs_hba *hba, struct scsi_cmnd *cmd)
+{
+	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
+	struct ufs_cmd_info *cmd_info = &mgr->cmd_info;
+	int cpu = raw_smp_processor_id();
+	int tag = cmd->request->tag;
+
+	if (mgr->active == 0)
+		return;
+
+	cmd_info->pdata[tag]->end_time = cpu_clock(cpu);
+}
+
+int exynos_ufs_init_dbg(struct ufs_exynos_handle *handle)
+{
+	struct ufs_s_dbg_mgr *mgr;
+
+	if (ufs_s_dbg_mgr_idx >= DBG_NUM_OF_HOSTS)
+		return -1;
+
+	mgr = &ufs_s_dbg[ufs_s_dbg_mgr_idx++];
+	handle->private = (void *)mgr;
+	mgr->handle = handle;
+	mgr->active = 1;
+
+	/* cmd log */
+	spin_lock_init(&mgr->cmd_lock);
+
+	return 0;
+}
+MODULE_AUTHOR("Kiwoong Kim <kwmad.kim@samsung.com>");
+MODULE_DESCRIPTION("Exynos UFS debug information");
+MODULE_LICENSE("GPL");
+MODULE_VERSION("0.1");
diff --git a/drivers/scsi/ufs/ufs-exynos-if.h b/drivers/scsi/ufs/ufs-exynos-if.h
new file mode 100644
index 0000000..33f7187
--- /dev/null
+++ b/drivers/scsi/ufs/ufs-exynos-if.h
@@ -0,0 +1,19 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * UFS Exynos debugging functions
+ *
+ * Copyright (C) 2020 Samsung Electronics Co., Ltd.
+ * Author: Kiwoong Kim <kwmad.kim@samsung.com>
+ *
+ */
+#ifndef _UFS_EXYNOS_IF_H_
+#define _UFS_EXYNOS_IF_H_
+
+#define UFS_VS_MMIO_VER 1
+
+/* more members would be added in the future */
+struct ufs_exynos_handle {
+	void *private;
+};
+
+#endif /* _UFS_EXYNOS_IF_H_ */
diff --git a/drivers/scsi/ufs/ufs-exynos.c b/drivers/scsi/ufs/ufs-exynos.c
index 440f2af..50437db 100644
--- a/drivers/scsi/ufs/ufs-exynos.c
+++ b/drivers/scsi/ufs/ufs-exynos.c
@@ -1008,6 +1008,12 @@ static int exynos_ufs_init(struct ufs_hba *hba)
 		goto out;
 	exynos_ufs_specify_phy_time_attr(ufs);
 	exynos_ufs_config_smu(ufs);
+
+	/* init dbg */
+	ret = exynos_ufs_init_dbg(ufs->handle);
+	if (ret)
+		return ret;
+	spin_lock_init(&ufs->dbg_lock);
 	return 0;
 
 phy_off:
@@ -1210,6 +1216,41 @@ static int exynos_ufs_resume(struct ufs_hba *hba, enum ufs_pm_op pm_op)
 	return 0;
 }
 
+static void exynos_ufs_dbg_register_dump(struct ufs_hba *hba)
+{
+	struct exynos_ufs *ufs = ufshcd_get_variant(hba);
+	unsigned long flags;
+
+	spin_lock_irqsave(&ufs->dbg_lock, flags);
+	if (ufs->under_dump == 0)
+		ufs->under_dump = 1;
+	else {
+		spin_unlock_irqrestore(&ufs->dbg_lock, flags);
+		goto out;
+	}
+	spin_unlock_irqrestore(&ufs->dbg_lock, flags);
+
+	exynos_ufs_dump_info(ufs->handle, hba->dev);
+
+	spin_lock_irqsave(&ufs->dbg_lock, flags);
+	ufs->under_dump = 0;
+	spin_unlock_irqrestore(&ufs->dbg_lock, flags);
+out:
+	return;
+}
+
+static void exynos_ufs_cmd_log(struct ufs_hba *hba,
+					 struct scsi_cmnd *cmd, int enter)
+{
+	struct exynos_ufs *ufs = ufshcd_get_variant(hba);
+	struct ufs_exynos_handle *handle = ufs->handle;
+
+	if (enter == 1)
+		exynos_ufs_cmd_log_start(handle, hba, cmd);
+	else if (enter == 2)
+		exynos_ufs_cmd_log_end(handle, hba, cmd);
+}
+
 static struct ufs_hba_variant_ops ufs_hba_exynos_ops = {
 	.name				= "exynos_ufs",
 	.init				= exynos_ufs_init,
@@ -1221,6 +1262,8 @@ static struct ufs_hba_variant_ops ufs_hba_exynos_ops = {
 	.hibern8_notify			= exynos_ufs_hibern8_notify,
 	.suspend			= exynos_ufs_suspend,
 	.resume				= exynos_ufs_resume,
+	.dbg_register_dump		= exynos_ufs_dbg_register_dump,
+	.cmd_log			= exynos_ufs_cmd_log,
 };
 
 static int exynos_ufs_probe(struct platform_device *pdev)
diff --git a/drivers/scsi/ufs/ufs-exynos.h b/drivers/scsi/ufs/ufs-exynos.h
index 76d6e39..14347fb 100644
--- a/drivers/scsi/ufs/ufs-exynos.h
+++ b/drivers/scsi/ufs/ufs-exynos.h
@@ -8,6 +8,7 @@
 
 #ifndef _UFS_EXYNOS_H_
 #define _UFS_EXYNOS_H_
+#include "ufs-exynos-if.h"
 
 /*
  * UNIPRO registers
@@ -212,6 +213,10 @@ struct exynos_ufs {
 #define EXYNOS_UFS_OPT_BROKEN_AUTO_CLK_CTRL	BIT(2)
 #define EXYNOS_UFS_OPT_BROKEN_RX_SEL_IDX	BIT(3)
 #define EXYNOS_UFS_OPT_USE_SW_HIBERN8_TIMER	BIT(4)
+
+	struct ufs_exynos_handle *handle;
+	spinlock_t dbg_lock;
+	int under_dump;
 };
 
 #define for_each_ufs_rx_lane(ufs, i) \
@@ -284,4 +289,12 @@ struct exynos_ufs_uic_attr exynos7_uic_attr = {
 	.rx_hs_g3_prep_sync_len_cap	= PREP_LEN(0xf),
 	.pa_dbg_option_suite		= 0x30103,
 };
+
+/* public function declarations */
+void exynos_ufs_dump_info(struct ufs_exynos_handle *handle, struct device *dev);
+void exynos_ufs_cmd_log_start(struct ufs_exynos_handle *handle,
+			      struct ufs_hba *hba, struct scsi_cmnd *cmd);
+void exynos_ufs_cmd_log_end(struct ufs_exynos_handle *handle,
+			    struct ufs_hba *hba, struct scsi_cmnd *cmd);
+int exynos_ufs_init_dbg(struct ufs_exynos_handle *handle);
 #endif /* _UFS_EXYNOS_H_ */
-- 
2.7.4

Re: [RFC PATCH v1 2/2] exynos-ufs: support command history

[Bart Van Assche]

> +#define EN_PRINT_UFS_LOG 1

Since this macro controls debug code, please make this configurable at
runtime, e.g. as a kernel module parameter or by using the dynamic debug
mechanism.

> +/* Structure for ufs cmd logging */
> +#define MAX_CMD_LOGS    32

Please clarify in the comment above this constant how this constant has
been chosen. Is this constant e.g. related to the queue depth? Is this
constant per UFS device or is it a maximum for all UFS devices?

> +struct cmd_data {
> +	unsigned int tag;
> +	unsigned int sct;
> +	unsigned long lba;
> +	u64 start_time;
> +	u64 end_time;
> +	u64 outstanding_reqs;
> +	int retries;
> +	unsigned char op;
> +};

Please use data types that explicitly specify the width for members like
e.g. the lba (u64 instead of unsigned long). Please also use u8 instead
of unsigned char for 'op' since 'op' is not used to store any kind of
ASCII character.

> +struct ufs_cmd_info {
> +	u32 total;
> +	u32 last;
> +	struct cmd_data data[MAX_CMD_LOGS];
> +	struct cmd_data *pdata[32];	/* Currently, 32 slots */
> +};

What are 'slots'? Why 32?

> +#define DBG_NUM_OF_HOSTS	1
> +struct ufs_s_dbg_mgr {
> +	struct ufs_exynos_handle *handle;
> +	int active;
> +	u64 first_time;
> +	u64 time;
> +
> +	/* cmd log */
> +	struct ufs_cmd_info cmd_info;
> +	struct cmd_data cmd_log;		/* temp buffer to put */
> +	spinlock_t cmd_lock;
> +};

Please add a comment above this structure that explains the role of this
data structure.

> +static struct ufs_s_dbg_mgr ufs_s_dbg[DBG_NUM_OF_HOSTS];

A static array? That's suspicious. Should that array perhaps be a member
of another UFS data structure, e.g. the UFS host or device data structure?

> +static int ufs_s_dbg_mgr_idx;

What does this variable represent?

> +	for (i = 0 ; i < max ; i++, data++) {
> +		dev_err(dev, ": 0x%02x, %02d, 0x%08lx, 0x%04x, %d, %llu, %llu, 0x%llx %s",
> +				data->op,
> +				data->tag,
> +				data->lba,
> +				data->sct,
> +				data->retries,
> +				data->start_time,
> +				data->end_time,
> +				data->outstanding_reqs,
> +				((last == i) ? "<--" : " "));

Please follow the same coding style as elsewhere in the Linux kernel and
specify multiple arguments per line (up to the current column limit).
Please also align the arguments either with the opening parentheses or
indent these by one tab as requested in the Linux kernel coding style
document.

> +/*
> + * EXTERNAL FUNCTIONS
> + *
> + * There are two classes that are to initialize data structures for debug
> + * and to define actual behavior.
> + */
> +void exynos_ufs_dump_info(struct ufs_exynos_handle *handle, struct device *dev)
> +{
> +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> +
> +	if (mgr->active == 0)
> +		goto out;
> +
> +	mgr->time = cpu_clock(raw_smp_processor_id());
> +
> +#if EN_PRINT_UFS_LOG
> +	ufs_s_print_cmd_log(mgr, dev);
> +#endif
> +
> +	if (mgr->first_time == 0ULL)
> +		mgr->first_time = mgr->time;
> +out:
> +	return;
> +}

Using cpu_clock() without storing the CPU on which it has been sampled
seems wrong to me. Is higher accuracy than a single jiffy required? If
not, how about using 'jiffies' instead? From clock.h:

/*
 * As outlined in clock.c, provides a fast, high resolution, nanosecond
 * time source that is monotonic per cpu argument and has bounded drift
 * between cpus.
 *
 * ######################### BIG FAT WARNING ##########################
 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
 * # go backwards !!                                                  #
 * ####################################################################
 */

> +void exynos_ufs_cmd_log_start(struct ufs_exynos_handle *handle,
> +			      struct ufs_hba *hba, struct scsi_cmnd *cmd)
> +{
> +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> +	int cpu = raw_smp_processor_id();
> +	struct cmd_data *cmd_log = &mgr->cmd_log;	/* temp buffer to put */
> +	unsigned long lba = (cmd->cmnd[2] << 24) |
> +					(cmd->cmnd[3] << 16) |
> +					(cmd->cmnd[4] << 8) |
> +					(cmd->cmnd[5] << 0);
> +	unsigned int sct = (cmd->cmnd[7] << 8) |
> +					(cmd->cmnd[8] << 0);

Aargh! SCSI command parsing ... Has it been considered to use
blk_rq_pos(cmd->req) and blk_rq_bytes(cmd->req) instead?

> +int exynos_ufs_init_dbg(struct ufs_exynos_handle *handle)
> +{
> +	struct ufs_s_dbg_mgr *mgr;
> +
> +	if (ufs_s_dbg_mgr_idx >= DBG_NUM_OF_HOSTS)
> +		return -1;
> +
> +	mgr = &ufs_s_dbg[ufs_s_dbg_mgr_idx++];
> +	handle->private = (void *)mgr;
> +	mgr->handle = handle;
> +	mgr->active = 1;

Can the '(void *)' cast above be left out?

> +#define UFS_VS_MMIO_VER 1

What does this constant represent? Please add a comment.

Thanks,

Bart.

RE: [RFC PATCH v1 2/2] exynos-ufs: support command history

[Kiwoong Kim]

> > +#define EN_PRINT_UFS_LOG 1
> 
> Since this macro controls debug code, please make this configurable at
> runtime, e.g. as a kernel module parameter or by using the dynamic debug
> mechanism.
> 
> > +/* Structure for ufs cmd logging */
> > +#define MAX_CMD_LOGS    32

Got it.

> 
> Please clarify in the comment above this constant how this constant has
> been chosen. Is this constant e.g. related to the queue depth? Is this
> constant per UFS device or is it a maximum for all UFS devices?
> 
> > +struct cmd_data {
> > +	unsigned int tag;
> > +	unsigned int sct;
> > +	unsigned long lba;
> > +	u64 start_time;
> > +	u64 end_time;
> > +	u64 outstanding_reqs;
> > +	int retries;
> > +	unsigned char op;
> > +};
> 
> Please use data types that explicitly specify the width for members like
> e.g. the lba (u64 instead of unsigned long). Please also use u8 instead
> of unsigned char for '
op' since 'op' is not used to store any kind of
> ASCII character. 

Got it.

> 
> > +struct ufs_cmd_info {
> > +	u32 total;
> > +	u32 last;
> > +	struct cmd_data data[MAX_CMD_LOGS];
> > +	struct cmd_data *pdata[32];	/* Currently, 32 slots */
> > +};
> 
> What are 'slots'? Why 32? 

I meant tag number and assumed that CAP.NUTRS be 32.
With 32, you can see all the command context with full outstanding situations.
Of course, there is a possibility that the value increases in the next UFS versions.
So, to run automatically, yes, I should have made this get CAP.NUTRS.
However, to do that, I have to add another call after enabling host.

> 
> > +#define DBG_NUM_OF_HOSTS	1
> > +struct ufs_s_dbg_mgr {
> > +	struct ufs_exynos_handle *handle;
> > +	int active;
> > +	u64 first_time;
> > +	u64 time;
> > +
> > +	/* cmd log */
> > +	struct ufs_cmd_info cmd_info;
> > +	struct cmd_data cmd_log;		/* temp buffer to put */
> > +	spinlock_t cmd_lock;
> > +};
> 
> Please add a comment above this structure that explains the role of this
> data structure. 

Got it.

> 
> > +static struct ufs_s_dbg_mgr ufs_s_dbg[DBG_NUM_OF_HOSTS];
> 
> A static array? That's suspicious. Should that array perhaps be a member
> of another UFS data structure, e.g. the UFS host or device data structure? 

Got it.

> 
> > +static int ufs_s_dbg_mgr_idx;
> 
> What does this variable represent? 

I considered it again and will remove.

> 
> > +	for (i = 0 ; i < max ; i++, data++) {
> > +		dev_err(dev, ": 0x%02x, %02d, 0x%08lx,
> 0x%04x, %d, %llu, %llu, 0x%llx %s",
> > +				data->op,
> > +				data->tag,
> > +				data->lba,
> > +				data->sct,
> > +				data->retries,
> > +				data->start_time,
> > +				data->end_time,
> > +				data->outstanding_reqs,
> > +				((last == i) ? "<--" : " "));
> 
> Please follow the same coding style as elsewhere in the Linux kernel and
> specify multiple arguments per line (up to the current column limit).
> Please also align the arguments either with the opening parentheses or
> indent these by one tab as requested in the Linux kernel coding style
> document. 

Got it.

> 
> > +/*
> > + * EXTERNAL FUNCTIONS
> > + *
> > + * There are two classes that are to initialize data structures for
> debug
> > + * and to define actual behavior.
> > + */
> > +void exynos_ufs_dump_info(struct ufs_exynos_handle *handle, struct
> device *dev)
> > +{
> > +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> > +
> > +	if (mgr->active == 0)
> > +		goto out;
> > +
> > +	mgr->time = cpu_clock(raw_smp_processor_id());
> > +
> > +#if EN_PRINT_UFS_LOG
> > +	ufs_s_print_cmd_log(mgr, dev);
> > +#endif
> > +
> > +	if (mgr->first_time == 0ULL)
> > +		mgr->first_time = mgr->time;
> > +out:
> > +	return;
> > +}
> 
> Using cpu_clock() without storing the CPU on which it has been sampled
> seems wrong to me. Is higher accuracy than a single jiffy required? If
> not, how about using 'jiffies' instead? From clock.h:
> 

I think jiffies isn't proper for the original purpose.

> /*
>  * As outlined in clock.c, provides a fast, high resolution, nanosecond
>  * time source that is monotonic per cpu argument and has bounded drift
>  * between cpus.
>  *
>  * ######################### BIG FAT WARNING ##########################
>  * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
>  * # go backwards !!                                                  #
>  * ####################################################################
>  */
> 
> > +void exynos_ufs_cmd_log_start(struct ufs_exynos_handle *handle,
> > +			      struct ufs_hba *hba, struct scsi_cmnd *cmd)
> > +{
> > +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> > +	int cpu = raw_smp_processor_id();
> > +	struct cmd_data *cmd_log = &mgr->cmd_log;	/* temp buffer to put
> */
> > +	unsigned long lba = (cmd->cmnd[2] << 24) |
> > +					(cmd->cmnd[3] << 16) |
> > +					(cmd->cmnd[4] << 8) |
> > +					(cmd->cmnd[5] << 0);
> > +	unsigned int sct = (cmd->cmnd[7] << 8) |
> > +					(cmd->cmnd[8] << 0);
> 
> Aargh! SCSI command parsing ... Has it been considered to use
> blk_rq_pos(cmd->req) and blk_rq_bytes(cmd->req) instead?

Sure and I have recognized those things for years.
UFS depends on SCSI and regarding to actual usages.
Thus, I don't feel that sort of parsing unless there is another macro to parse SCSI commands.
As I know, nothing exits.
And blk_rq_pos doesn't represent actual lba. That represents in-partition addressing.
With this, I have to refer to the start address of partitions and I don’t think
that isn't suitable in UFS driver.

> 
> > +int exynos_ufs_init_dbg(struct ufs_exynos_handle *handle)
> > +{
> > +	struct ufs_s_dbg_mgr *mgr;
> > +
> > +	if (ufs_s_dbg_mgr_idx >= DBG_NUM_OF_HOSTS)
> > +		return -1;
> > +
> > +	mgr = &ufs_s_dbg[ufs_s_dbg_mgr_idx++];
> > +	handle->private = (void *)mgr;
> > +	mgr->handle = handle;
> > +	mgr->active = 1;
> 
> Can the '(void *)' cast above be left out? 

I didn't want to share the details of this structure with ufs-exynos.c.
I don’t feel that it's better.

> 
> > +#define UFS_VS_MMIO_VER 1
> 
> What does this constant represent? Please add a comment. 

I considered it again and will remove.

> 
> Thanks,
> 
> Bart. 

Thank you for your comments even if this is a RFS patch.


Thanks.
Kiwoong Kim

RE: [RFC PATCH v1 2/2] exynos-ufs: support command history

[Kiwoong Kim]

> > +#define EN_PRINT_UFS_LOG 1
> 
> Since this macro controls debug code, please make this configurable at
> runtime, e.g. as a kernel module parameter or by using the dynamic debug
> mechanism.

Got it.

> 
> > +/* Structure for ufs cmd logging */
> > +#define MAX_CMD_LOGS    32
> 
> Please clarify in the comment above this constant how this constant has
> been chosen. Is this constant e.g. related to the queue depth? Is this
> constant per UFS device or is it a maximum for all UFS devices?
> 
> > +struct cmd_data {
> > +	unsigned int tag;
> > +	unsigned int sct;
> > +	unsigned long lba;
> > +	u64 start_time;
> > +	u64 end_time;
> > +	u64 outstanding_reqs;
> > +	int retries;
> > +	unsigned char op;
> > +};
> 
> Please use data types that explicitly specify the width for members like
> e.g. the lba (u64 instead of unsigned long). Please also use u8 instead of
> unsigned char for 'op' since 'op' is not used to store any kind of ASCII
> character.

Got it.

> 
> > +struct ufs_cmd_info {
> > +	u32 total;
> > +	u32 last;
> > +	struct cmd_data data[MAX_CMD_LOGS];
> > +	struct cmd_data *pdata[32];	/* Currently, 32 slots */
> > +};
> 
> What are 'slots'? Why 32?

I meant tag number and assumed that CAP.NUTRS be 32.
With 32, you can see all the command context with full outstanding situations.
Of course, there is a possibility that the value increases in the next UFS versions.
So, to run automatically, yes, I should have made this get CAP.NUTRS.
However, to do that, I have to add another call after enabling host.

> 
> > +#define DBG_NUM_OF_HOSTS	1
> > +struct ufs_s_dbg_mgr {
> > +	struct ufs_exynos_handle *handle;
> > +	int active;
> > +	u64 first_time;
> > +	u64 time;
> > +
> > +	/* cmd log */
> > +	struct ufs_cmd_info cmd_info;
> > +	struct cmd_data cmd_log;		/* temp buffer to put */
> > +	spinlock_t cmd_lock;
> > +};
> 
> Please add a comment above this structure that explains the role of this
> data structure.

Got it.

> 
> > +static struct ufs_s_dbg_mgr ufs_s_dbg[DBG_NUM_OF_HOSTS];
> 
> A static array? That's suspicious. Should that array perhaps be a member
> of another UFS data structure, e.g. the UFS host or device data structure?
> 
> > +static int ufs_s_dbg_mgr_idx;

I considered it again and will remove.

> 
> What does this variable represent?
> 
> > +	for (i = 0 ; i < max ; i++, data++) {
> > +		dev_err(dev, ": 0x%02x, %02d, 0x%08lx,
> 0x%04x, %d, %llu, %llu, 0x%llx %s",
> > +				data->op,
> > +				data->tag,
> > +				data->lba,
> > +				data->sct,
> > +				data->retries,
> > +				data->start_time,
> > +				data->end_time,
> > +				data->outstanding_reqs,
> > +				((last == i) ? "<--" : " "));
> 
> Please follow the same coding style as elsewhere in the Linux kernel and
> specify multiple arguments per line (up to the current column limit).
> Please also align the arguments either with the opening parentheses or
> indent these by one tab as requested in the Linux kernel coding style
> document.
> 

Got it.

> > +/*
> > + * EXTERNAL FUNCTIONS
> > + *
> > + * There are two classes that are to initialize data structures for
> > +debug
> > + * and to define actual behavior.
> > + */
> > +void exynos_ufs_dump_info(struct ufs_exynos_handle *handle, struct
> > +device *dev) {
> > +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> > +
> > +	if (mgr->active == 0)
> > +		goto out;
> > +
> > +	mgr->time = cpu_clock(raw_smp_processor_id());
> > +
> > +#if EN_PRINT_UFS_LOG
> > +	ufs_s_print_cmd_log(mgr, dev);
> > +#endif
> > +
> > +	if (mgr->first_time == 0ULL)
> > +		mgr->first_time = mgr->time;
> > +out:
> > +	return;
> > +}
> 
> Using cpu_clock() without storing the CPU on which it has been sampled
> seems wrong to me. Is higher accuracy than a single jiffy required? If not,
> how about using 'jiffies' instead? From clock.h:

I think jiffies isn't proper for the original purpose.

> 
> /*
>  * As outlined in clock.c, provides a fast, high resolution, nanosecond
>  * time source that is monotonic per cpu argument and has bounded drift
>  * between cpus.
>  *
>  * ######################### BIG FAT WARNING ##########################
>  * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
>  * # go backwards !!                                                  #
>  * ####################################################################
>  */
> 
> > +void exynos_ufs_cmd_log_start(struct ufs_exynos_handle *handle,
> > +			      struct ufs_hba *hba, struct scsi_cmnd *cmd) {
> > +	struct ufs_s_dbg_mgr *mgr = (struct ufs_s_dbg_mgr *)handle->private;
> > +	int cpu = raw_smp_processor_id();
> > +	struct cmd_data *cmd_log = &mgr->cmd_log;	/* temp buffer to put
> */
> > +	unsigned long lba = (cmd->cmnd[2] << 24) |
> > +					(cmd->cmnd[3] << 16) |
> > +					(cmd->cmnd[4] << 8) |
> > +					(cmd->cmnd[5] << 0);
> > +	unsigned int sct = (cmd->cmnd[7] << 8) |
> > +					(cmd->cmnd[8] << 0);
> 
> Aargh! SCSI command parsing ... Has it been considered to use
> blk_rq_pos(cmd->req) and blk_rq_bytes(cmd->req) instead?

Sure and I have recognized those things for years.
UFS depends on SCSI and regarding to actual usages.
Thus, I don't feel that sort of parsing unless there is another macro to parse SCSI commands.
As I know, nothing exits.
And blk_rq_pos doesn't represent actual lba. That represents in-partition addressing.
With this, I have to refer to the start address of partitions and I don’t think that isn't suitable in UFS driver.

> 
> > +int exynos_ufs_init_dbg(struct ufs_exynos_handle *handle) {
> > +	struct ufs_s_dbg_mgr *mgr;
> > +
> > +	if (ufs_s_dbg_mgr_idx >= DBG_NUM_OF_HOSTS)
> > +		return -1;
> > +
> > +	mgr = &ufs_s_dbg[ufs_s_dbg_mgr_idx++];
> > +	handle->private = (void *)mgr;
> > +	mgr->handle = handle;
> > +	mgr->active = 1;
> 
> Can the '(void *)' cast above be left out?

I didn't want to share the details of this structure with ufs-exynos.c.
I don’t feel that it's better.

> 
> > +#define UFS_VS_MMIO_VER 1
> 
> What does this constant represent? Please add a comment.

I considered it again and will remove.

> 
> Thanks,
> 
> Bart.

Thank you for your comments even if this is a RFS patch.


Thanks.
Kiwoong Kim

Re: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Bart Van Assche]

On 2020-06-19 23:53, Kiwoong Kim wrote:
> Some SoC specific might need command history for
> various reasons, such as stacking command contexts
> in system memory to check for debugging in the future
> or scaling some DVFS knobs to boost IO throughput.
> 
> What you would do with the information could be
> variant per SoC vendor.

Isn't this something that should be done in an I/O scheduler instead of
in a SCSI LLD? I don't like the idea behind this patch series at all.

Bart.

RE: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Kiwoong Kim]

> On 2020-06-19 23:53, Kiwoong Kim wrote:
> > Some SoC specific might need command history for various reasons, such
> > as stacking command contexts in system memory to check for debugging
> > in the future or scaling some DVFS knobs to boost IO throughput.
> >
> > What you would do with the information could be variant per SoC
> > vendor.
> 
> Isn't this something that should be done in an I/O scheduler instead of in
> a SCSI LLD? I don't like the idea behind this patch series at all.
> 
> Bart.

If you could get the information, Many would exploit it for their respective purposes.
But, it's important for the information to contain accurate timestamps when the driver
hooks it, if you're trying to figure out something wrong.

As for scaling DVFS knobs to boost UFS throughput, locating in the ufs driver is more
Beneficial because SoC vendors have their own power domains and thus lead to make
different way of what to scale up to boost. If it's populated in block layer, there is
no way to introduce boosting per SoC.

Re: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Bart Van Assche]

On 2020-06-22 19:28, Kiwoong Kim wrote:
> If you could get the information, Many would exploit it for their respective purposes.
> But, it's important for the information to contain accurate timestamps when the driver
> hooks it, if you're trying to figure out something wrong.
> 
> As for scaling DVFS knobs to boost UFS throughput, locating in the ufs driver is more
> Beneficial because SoC vendors have their own power domains and thus lead to make
> different way of what to scale up to boost. If it's populated in block layer, there is
> no way to introduce boosting per SoC.

Thanks for the clarification. Unfortunately we do not yet have an API
for sharing information between I/O schedulers and block drivers ...

Bart.

Re: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Stanley Chu]

Hi Kiwoong,

Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月20日 週六 下午3:00寫道：
>
> Some SoC specific might need command history for
> various reasons, such as stacking command contexts
> in system memory to check for debugging in the future
> or scaling some DVFS knobs to boost IO throughput.
>
> What you would do with the information could be
> variant per SoC vendor.
>
> Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> ---
>  drivers/scsi/ufs/ufshcd.c | 4 ++++
>  drivers/scsi/ufs/ufshcd.h | 8 ++++++++
>  2 files changed, 12 insertions(+)
>
> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index 52abe82..0eae3ce 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd)
>         /* issue command to the controller */
>         spin_lock_irqsave(hba->host->host_lock, flags);
>         ufshcd_vops_setup_xfer_req(hba, tag, true);
> +       if (cmd)
> +               ufshcd_vops_cmd_log(hba, cmd, 1);
>         ufshcd_send_command(hba, tag);
>  out_unlock:
>         spin_unlock_irqrestore(hba->host->host_lock, flags);
> @@ -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct ufs_hba *hba,
>                         /* Mark completed command as NULL in LRB */
>                         lrbp->cmd = NULL;
>                         lrbp->compl_time_stamp = ktime_get();
> +                       ufshcd_vops_cmd_log(hba, cmd, 2);
> +
>                         /* Do not touch lrbp after scsi done */
>                         cmd->scsi_done(cmd);
>                         __ufshcd_release(hba);

If your cmd_log vop callbacks are only existed in
"ufshcd_queuecommand" and "ufshcd_transfer_req_compl", perhaps you
could re-use "ufshcd_vops_setup_xfer_req()" and an added
"ufshcd_vops_compl_req()" instead of a brand new
"ufshcd_vops_cmd_log()" ?

Thanks,
Stanley Chu

RE: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Kiwoong Kim]

> Hi Kiwoong,
> 
> Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月20日 週六 下午3:00寫道：
> >
> > Some SoC specific might need command history for various reasons, such
> > as stacking command contexts in system memory to check for debugging
> > in the future or scaling some DVFS knobs to boost IO throughput.
> >
> > What you would do with the information could be variant per SoC
> > vendor.
> >
> > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > ---
> >  drivers/scsi/ufs/ufshcd.c | 4 ++++
> >  drivers/scsi/ufs/ufshcd.h | 8 ++++++++
> >  2 files changed, 12 insertions(+)
> >
> > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > index 52abe82..0eae3ce 100644
> > --- a/drivers/scsi/ufs/ufshcd.c
> > +++ b/drivers/scsi/ufs/ufshcd.c
> > @@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct Scsi_Host
> *host, struct scsi_cmnd *cmd)
> >         /* issue command to the controller */
> >         spin_lock_irqsave(hba->host->host_lock, flags);
> >         ufshcd_vops_setup_xfer_req(hba, tag, true);
> > +       if (cmd)
> > +               ufshcd_vops_cmd_log(hba, cmd, 1);
> >         ufshcd_send_command(hba, tag);
> >  out_unlock:
> >         spin_unlock_irqrestore(hba->host->host_lock, flags); @@
> > -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct
> ufs_hba *hba,
> >                         /* Mark completed command as NULL in LRB */
> >                         lrbp->cmd = NULL;
> >                         lrbp->compl_time_stamp = ktime_get();
> > +                       ufshcd_vops_cmd_log(hba, cmd, 2);
> > +
> >                         /* Do not touch lrbp after scsi done */
> >                         cmd->scsi_done(cmd);
> >                         __ufshcd_release(hba);
> 
> If your cmd_log vop callbacks are only existed in "ufshcd_queuecommand"
> and "ufshcd_transfer_req_compl", perhaps you could re-use
> "ufshcd_vops_setup_xfer_req()" and an added "ufshcd_vops_compl_req()"
> instead of a brand new "ufshcd_vops_cmd_log()" ?
> 
> Thanks,
> Stanley Chu

Currently, ufshcd_vops_setup_xfer_req doesn't get scsi_cmnd variable.
Actually, when introduced this callback first, I was willing to make it do that
but someone gave me another idea. Then do you agree to change argument set of the function?

And I can't find ufshcd_vops_compl_req in 5.9/scsi-queue. Could you let me know where to find?

Thank you for your opinions.

Re: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Stanley Chu]

Hi Kiwoong,

Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月26日 週五 下午7:42寫道：
>
> > Hi Kiwoong,
> >
> > Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月20日 週六 下午3:00寫道：
> > >
> > > Some SoC specific might need command history for various reasons, such
> > > as stacking command contexts in system memory to check for debugging
> > > in the future or scaling some DVFS knobs to boost IO throughput.
> > >
> > > What you would do with the information could be variant per SoC
> > > vendor.
> > >
> > > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > > ---
> > >  drivers/scsi/ufs/ufshcd.c | 4 ++++
> > >  drivers/scsi/ufs/ufshcd.h | 8 ++++++++
> > >  2 files changed, 12 insertions(+)
> > >
> > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > > index 52abe82..0eae3ce 100644
> > > --- a/drivers/scsi/ufs/ufshcd.c
> > > +++ b/drivers/scsi/ufs/ufshcd.c
> > > @@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct Scsi_Host
> > *host, struct scsi_cmnd *cmd)
> > >         /* issue command to the controller */
> > >         spin_lock_irqsave(hba->host->host_lock, flags);
> > >         ufshcd_vops_setup_xfer_req(hba, tag, true);
> > > +       if (cmd)
> > > +               ufshcd_vops_cmd_log(hba, cmd, 1);
> > >         ufshcd_send_command(hba, tag);
> > >  out_unlock:
> > >         spin_unlock_irqrestore(hba->host->host_lock, flags); @@
> > > -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct
> > ufs_hba *hba,
> > >                         /* Mark completed command as NULL in LRB */
> > >                         lrbp->cmd = NULL;
> > >                         lrbp->compl_time_stamp = ktime_get();
> > > +                       ufshcd_vops_cmd_log(hba, cmd, 2);
> > > +
> > >                         /* Do not touch lrbp after scsi done */
> > >                         cmd->scsi_done(cmd);
> > >                         __ufshcd_release(hba);
> >
> > If your cmd_log vop callbacks are only existed in "ufshcd_queuecommand"
> > and "ufshcd_transfer_req_compl", perhaps you could re-use
> > "ufshcd_vops_setup_xfer_req()" and an added "ufshcd_vops_compl_req()"
> > instead of a brand new "ufshcd_vops_cmd_log()" ?
> >
> > Thanks,
> > Stanley Chu
>
> Currently, ufshcd_vops_setup_xfer_req doesn't get scsi_cmnd variable.

You could get scsi_cmnd by hba->lrb[tag].cmd if is_scsi_cmd is true in
your ufshcd_vops_setup_xfer_req vendor implementation.

> Actually, when introduced this callback first, I was willing to make it do that
> but someone gave me another idea. Then do you agree to change argument set of the function?
>
> And I can't find ufshcd_vops_compl_req in 5.9/scsi-queue. Could you let me know where to find?
>

Sorry to not describing clearly. What I mean is that you could "add"
ufshcd_vops_compl_xfer_req vop callback in your patch.

Thanks,
Stanley Chu

RE: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Kiwoong Kim]

> Hi Kiwoong,
> 
> Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月26日 週五 下午7:42寫道：
> >
> > > Hi Kiwoong,
> > >
> > > Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月20日 週六 下午3:00寫道：
> > > >
> > > > Some SoC specific might need command history for various reasons,
> such
> > > > as stacking command contexts in system memory to check for debugging
> > > > in the future or scaling some DVFS knobs to boost IO throughput.
> > > >
> > > > What you would do with the information could be variant per SoC
> > > > vendor.
> > > >
> > > > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > > > ---
> > > >  drivers/scsi/ufs/ufshcd.c | 4 ++++
> > > >  drivers/scsi/ufs/ufshcd.h | 8 ++++++++
> > > >  2 files changed, 12 insertions(+)
> > > >
> > > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > > > index 52abe82..0eae3ce 100644
> > > > --- a/drivers/scsi/ufs/ufshcd.c
> > > > +++ b/drivers/scsi/ufs/ufshcd.c
> > > > @@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct
> Scsi_Host
> > > *host, struct scsi_cmnd *cmd)
> > > >         /* issue command to the controller */
> > > >         spin_lock_irqsave(hba->host->host_lock, flags);
> > > >         ufshcd_vops_setup_xfer_req(hba, tag, true);
> > > > +       if (cmd)
> > > > +               ufshcd_vops_cmd_log(hba, cmd, 1);
> > > >         ufshcd_send_command(hba, tag);
> > > >  out_unlock:
> > > >         spin_unlock_irqrestore(hba->host->host_lock, flags); @@
> > > > -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct
> > > ufs_hba *hba,
> > > >                         /* Mark completed command as NULL in LRB */
> > > >                         lrbp->cmd = NULL;
> > > >                         lrbp->compl_time_stamp = ktime_get();
> > > > +                       ufshcd_vops_cmd_log(hba, cmd, 2);
> > > > +
> > > >                         /* Do not touch lrbp after scsi done */
> > > >                         cmd->scsi_done(cmd);
> > > >                         __ufshcd_release(hba);
> > >
> > > If your cmd_log vop callbacks are only existed in
> "ufshcd_queuecommand"
> > > and "ufshcd_transfer_req_compl", perhaps you could re-use
> > > "ufshcd_vops_setup_xfer_req()" and an added "ufshcd_vops_compl_req()"
> > > instead of a brand new "ufshcd_vops_cmd_log()" ?
> > >
> > > Thanks,
> > > Stanley Chu
> >
> > Currently, ufshcd_vops_setup_xfer_req doesn't get scsi_cmnd variable.
> 
> You could get scsi_cmnd by hba->lrb[tag].cmd if is_scsi_cmd is true in
> your ufshcd_vops_setup_xfer_req vendor implementation.
> 
> > Actually, when introduced this callback first, I was willing to make it
> do that
> > but someone gave me another idea. Then do you agree to change argument
> set of the function?
> >
> > And I can't find ufshcd_vops_compl_req in 5.9/scsi-queue. Could you let
> me know where to find?
> >
> 
> Sorry to not describing clearly. What I mean is that you could "add"
> ufshcd_vops_compl_xfer_req vop callback in your patch.
> 
> Thanks,
> Stanley Chu

Got it and I'll refer to what you said.
Thanks.
Kiwoong Kim

RE: [RFC PATCH v1 1/2] ufs: introduce callbacks to get command information

[Kiwoong Kim]

> Hi Kiwoong,
> 
> Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月26日 週五 下午7:42寫道：
> >
> > > Hi Kiwoong,
> > >
> > > Kiwoong Kim <kwmad.kim@samsung.com> 於 2020年6月20日 週六 下午3:00寫道：
> > > >
> > > > Some SoC specific might need command history for various reasons,
> such
> > > > as stacking command contexts in system memory to check for debugging
> > > > in the future or scaling some DVFS knobs to boost IO throughput.
> > > >
> > > > What you would do with the information could be variant per SoC
> > > > vendor.
> > > >
> > > > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > > > ---
> > > >  drivers/scsi/ufs/ufshcd.c | 4 ++++
> > > >  drivers/scsi/ufs/ufshcd.h | 8 ++++++++
> > > >  2 files changed, 12 insertions(+)
> > > >
> > > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > > > index 52abe82..0eae3ce 100644
> > > > --- a/drivers/scsi/ufs/ufshcd.c
> > > > +++ b/drivers/scsi/ufs/ufshcd.c
> > > > @@ -2545,6 +2545,8 @@ static int ufshcd_queuecommand(struct
> Scsi_Host
> > > *host, struct scsi_cmnd *cmd)
> > > >         /* issue command to the controller */
> > > >         spin_lock_irqsave(hba->host->host_lock, flags);
> > > >         ufshcd_vops_setup_xfer_req(hba, tag, true);
> > > > +       if (cmd)
> > > > +               ufshcd_vops_cmd_log(hba, cmd, 1);
> > > >         ufshcd_send_command(hba, tag);
> > > >  out_unlock:
> > > >         spin_unlock_irqrestore(hba->host->host_lock, flags); @@
> > > > -4890,6 +4892,8 @@ static void __ufshcd_transfer_req_compl(struct
> > > ufs_hba *hba,
> > > >                         /* Mark completed command as NULL in LRB */
> > > >                         lrbp->cmd = NULL;
> > > >                         lrbp->compl_time_stamp = ktime_get();
> > > > +                       ufshcd_vops_cmd_log(hba, cmd, 2);
> > > > +
> > > >                         /* Do not touch lrbp after scsi done */
> > > >                         cmd->scsi_done(cmd);
> > > >                         __ufshcd_release(hba);
> > >
> > > If your cmd_log vop callbacks are only existed in
> "ufshcd_queuecommand"
> > > and "ufshcd_transfer_req_compl", perhaps you could re-use
> > > "ufshcd_vops_setup_xfer_req()" and an added "ufshcd_vops_compl_req()"
> > > instead of a brand new "ufshcd_vops_cmd_log()" ?
> > >
> > > Thanks,
> > > Stanley Chu
> >
> > Currently, ufshcd_vops_setup_xfer_req doesn't get scsi_cmnd variable.
> 
> You could get scsi_cmnd by hba->lrb[tag].cmd if is_scsi_cmd is true in
> your ufshcd_vops_setup_xfer_req vendor implementation.
> 
> > Actually, when introduced this callback first, I was willing to make it
> do that
> > but someone gave me another idea. Then do you agree to change argument
> set of the function?
> >
> > And I can't find ufshcd_vops_compl_req in 5.9/scsi-queue. Could you let
> me know where to find?
> >
> 
> Sorry to not describing clearly. What I mean is that you could "add"
> ufshcd_vops_compl_xfer_req vop callback in your patch.
> 
> Thanks,
> Stanley Chu 

Got it and I'll refer to what you said.

Thanks.
Kiwoong Kim