[PATCH v2 0/2] Fix handling of host-aware ZBC disks

[PATCH v2 0/2] Fix handling of host-aware ZBC disks

[Damien Le Moal]

Martin,

Two patches for this cycle (with a cc stable) to fix handling of
host-aware ZBC disks that have partitions, that is, used as regular
disks.

The first patch fixes host-aware disk initialization and command
completion processing. It also enables the use of host-aware disks as
regular disks when CONFIG_BLK_DEV_ZONED is disabled.

The second patch fixes the CONFIG_BLK_DEV_ZONED enabled configuration
so that zone append emulation is not initialized for host-aware disks
with partitions/used as regular disks. While at it, this patch also
removes a problem with sd_zbc_init_disk() error handling in
sd_revalidate_disk() by moving this function execution inside
sd_zbc_revalidate_zones().

Borislav tested the series and confirmed that it solves his problem
(thanks Borislav !)

Changes from v1:
* Rebased on rc5
* Use "if (IS_DEFINED())" instead of #ifdef in patch 1

Damien Le Moal (2):
  scsi: Fix handling of host-aware ZBC disks
  scsi: Fix ZBC disk initialization

 drivers/scsi/sd.c     | 32 ++++++++++++++-------
 drivers/scsi/sd.h     |  8 +-----
 drivers/scsi/sd_zbc.c | 66 ++++++++++++++++++++++++++-----------------
 3 files changed, 63 insertions(+), 43 deletions(-)

-- 
2.26.2

[PATCH v2 1/2] scsi: Fix handling of host-aware ZBC disks

[Damien Le Moal]

When CONFIG_BLK_DEV_ZONED is disabled, allow using host-aware ZBC
disks as regular disks. In this case, ensure that command completion
is correctly executed by changing sd_zbc_complete() to return good_bytes
instead of 0, causing a hang during device probe (endless retries).

When CONFIG_BLK_DEV_ZONED is enabled and a host-aware disk is detected
to have partitions, it will be used as a regular disk. In this case,
make sure to not do anything in sd_zbc_revalidate_zones() as that
triggers warnings.

Reported-by: Borislav Petkov <bp@alien8.de>
Fixes: b72053072c0b ("block: allow partitions on host aware zone devices")
Cc: <stable@vger.kernel.org>
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 drivers/scsi/sd.c     | 28 ++++++++++++++++++++++------
 drivers/scsi/sd.h     |  2 +-
 drivers/scsi/sd_zbc.c |  6 +++++-
 3 files changed, 28 insertions(+), 8 deletions(-)

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 95018e650f2d..e99f57bfeff4 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -2968,22 +2968,38 @@ static void sd_read_block_characteristics(struct scsi_disk *sdkp)
 	} else {
 		sdkp->zoned = (buffer[8] >> 4) & 3;
 		if (sdkp->zoned == 1 && !disk_has_partitions(sdkp->disk)) {
-			/* Host-aware */
-			q->limits.zoned = BLK_ZONED_HA;
+			/*
+			 * Host-aware disk without partition: use the disk as
+			 * such if support for zoned block devices is enabled.
+			 * Otherwise, use it as a regular disk.
+			 */
+			if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
+				q->limits.zoned = BLK_ZONED_HA;
+			else
+				q->limits.zoned = BLK_ZONED_NONE;
 		} else {
 			/*
 			 * Treat drive-managed devices and host-aware devices
 			 * with partitions as regular block devices.
 			 */
 			q->limits.zoned = BLK_ZONED_NONE;
-			if (sdkp->zoned == 2 && sdkp->first_scan)
-				sd_printk(KERN_NOTICE, sdkp,
-					  "Drive-managed SMR disk\n");
 		}
 	}
-	if (blk_queue_is_zoned(q) && sdkp->first_scan)
+
+	if (!sdkp->first_scan)
+		goto out;
+
+	if (blk_queue_is_zoned(q)) {
 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
+	} else {
+		if (sdkp->zoned == 1)
+			sd_printk(KERN_NOTICE, sdkp,
+				  "Host-aware SMR disk used as regular disk\n");
+		else if (sdkp->zoned == 2)
+			sd_printk(KERN_NOTICE, sdkp,
+				  "Drive-managed SMR disk\n");
+	}
 
  out:
 	kfree(buffer);
diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h
index 4933e7daf17d..7251434100e6 100644
--- a/drivers/scsi/sd.h
+++ b/drivers/scsi/sd.h
@@ -259,7 +259,7 @@ static inline blk_status_t sd_zbc_setup_zone_mgmt_cmnd(struct scsi_cmnd *cmd,
 static inline unsigned int sd_zbc_complete(struct scsi_cmnd *cmd,
 			unsigned int good_bytes, struct scsi_sense_hdr *sshdr)
 {
-	return 0;
+	return good_bytes;
 }
 
 static inline blk_status_t sd_zbc_prepare_zone_append(struct scsi_cmnd *cmd,
diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
index 0e94ff056bff..a739456dea02 100644
--- a/drivers/scsi/sd_zbc.c
+++ b/drivers/scsi/sd_zbc.c
@@ -667,7 +667,11 @@ int sd_zbc_revalidate_zones(struct scsi_disk *sdkp)
 	u32 max_append;
 	int ret = 0;
 
-	if (!sd_is_zoned(sdkp))
+	/*
+	 * There is nothing to do for regular disks, including host-aware disks
+	 * that have partitions.
+	 */
+	if (!blk_queue_is_zoned(q))
 		return 0;
 
 	/*
-- 
2.26.2

[PATCH v2 2/2] scsi: Fix ZBC disk initialization

[Damien Le Moal]

Make sure to call sd_zbc_init_disk() when the sdkp->zoned field is
known, that is, once sd_read_block_characteristics() is executed in
sd_revalidate_disk(), so that host-aware disks also get initialized.
To do so, move sd_zbc_init_disk() call in sd_zbc_revalidate_zones() and
make sure to execute it for all zoned disks, including for host-aware
disks used as regular disks as these disk zoned model may be changed
back to BLK_ZONED_HA when partitions are deleted.

Reported-by: Borislav Petkov <bp@alien8.de>
Fixes: 5795eb443060 ("scsi: sd_zbc: emulate ZONE_APPEND commands")
Cc: <stable@vger.kernel.org> # v5.8+
Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Tested-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 drivers/scsi/sd.c     |  4 ---
 drivers/scsi/sd.h     |  6 -----
 drivers/scsi/sd_zbc.c | 60 +++++++++++++++++++++++++------------------
 3 files changed, 35 insertions(+), 35 deletions(-)

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index e99f57bfeff4..b3c80fa47c79 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -3420,10 +3420,6 @@ static int sd_probe(struct device *dev)
 	sdkp->first_scan = 1;
 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
 
-	error = sd_zbc_init_disk(sdkp);
-	if (error)
-		goto out_free_index;
-
 	sd_revalidate_disk(gd);
 
 	gd->flags = GENHD_FL_EXT_DEVT;
diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h
index 7251434100e6..a3aad608bc38 100644
--- a/drivers/scsi/sd.h
+++ b/drivers/scsi/sd.h
@@ -215,7 +215,6 @@ static inline int sd_is_zoned(struct scsi_disk *sdkp)
 
 #ifdef CONFIG_BLK_DEV_ZONED
 
-int sd_zbc_init_disk(struct scsi_disk *sdkp);
 void sd_zbc_release_disk(struct scsi_disk *sdkp);
 int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buffer);
 int sd_zbc_revalidate_zones(struct scsi_disk *sdkp);
@@ -231,11 +230,6 @@ blk_status_t sd_zbc_prepare_zone_append(struct scsi_cmnd *cmd, sector_t *lba,
 
 #else /* CONFIG_BLK_DEV_ZONED */
 
-static inline int sd_zbc_init_disk(struct scsi_disk *sdkp)
-{
-	return 0;
-}
-
 static inline void sd_zbc_release_disk(struct scsi_disk *sdkp) {}
 
 static inline int sd_zbc_read_zones(struct scsi_disk *sdkp,
diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
index a739456dea02..cf07b7f93579 100644
--- a/drivers/scsi/sd_zbc.c
+++ b/drivers/scsi/sd_zbc.c
@@ -651,6 +651,28 @@ static void sd_zbc_print_zones(struct scsi_disk *sdkp)
 			  sdkp->zone_blocks);
 }
 
+static int sd_zbc_init_disk(struct scsi_disk *sdkp)
+{
+	sdkp->zones_wp_offset = NULL;
+	spin_lock_init(&sdkp->zones_wp_offset_lock);
+	sdkp->rev_wp_offset = NULL;
+	mutex_init(&sdkp->rev_mutex);
+	INIT_WORK(&sdkp->zone_wp_offset_work, sd_zbc_update_wp_offset_workfn);
+	sdkp->zone_wp_update_buf = kzalloc(SD_BUF_SIZE, GFP_KERNEL);
+	if (!sdkp->zone_wp_update_buf)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void sd_zbc_release_disk(struct scsi_disk *sdkp)
+{
+	kvfree(sdkp->zones_wp_offset);
+	sdkp->zones_wp_offset = NULL;
+	kfree(sdkp->zone_wp_update_buf);
+	sdkp->zone_wp_update_buf = NULL;
+}
+
 static void sd_zbc_revalidate_zones_cb(struct gendisk *disk)
 {
 	struct scsi_disk *sdkp = scsi_disk(disk);
@@ -667,6 +689,19 @@ int sd_zbc_revalidate_zones(struct scsi_disk *sdkp)
 	u32 max_append;
 	int ret = 0;
 
+	/*
+	 * For all zoned disks, initialize zone append emulation data if not
+	 * already done. This is necessary also for host-aware disks used as
+	 * regular disks due to the presence of partitions as these partitions
+	 * may be deleted and the disk zoned model changed back from
+	 * BLK_ZONED_NONE to BLK_ZONED_HA.
+	 */
+	if (sd_is_zoned(sdkp) && !sdkp->zone_wp_update_buf) {
+		ret = sd_zbc_init_disk(sdkp);
+		if (ret)
+			return ret;
+	}
+
 	/*
 	 * There is nothing to do for regular disks, including host-aware disks
 	 * that have partitions.
@@ -768,28 +803,3 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
 
 	return ret;
 }
-
-int sd_zbc_init_disk(struct scsi_disk *sdkp)
-{
-	if (!sd_is_zoned(sdkp))
-		return 0;
-
-	sdkp->zones_wp_offset = NULL;
-	spin_lock_init(&sdkp->zones_wp_offset_lock);
-	sdkp->rev_wp_offset = NULL;
-	mutex_init(&sdkp->rev_mutex);
-	INIT_WORK(&sdkp->zone_wp_offset_work, sd_zbc_update_wp_offset_workfn);
-	sdkp->zone_wp_update_buf = kzalloc(SD_BUF_SIZE, GFP_KERNEL);
-	if (!sdkp->zone_wp_update_buf)
-		return -ENOMEM;
-
-	return 0;
-}
-
-void sd_zbc_release_disk(struct scsi_disk *sdkp)
-{
-	kvfree(sdkp->zones_wp_offset);
-	sdkp->zones_wp_offset = NULL;
-	kfree(sdkp->zone_wp_update_buf);
-	sdkp->zone_wp_update_buf = NULL;
-}
-- 
2.26.2

Re: [PATCH v2 1/2] scsi: Fix handling of host-aware ZBC disks

[Christoph Hellwig]

On Mon, Sep 14, 2020 at 09:34:47AM +0900, Damien Le Moal wrote:
> When CONFIG_BLK_DEV_ZONED is disabled, allow using host-aware ZBC
> disks as regular disks. In this case, ensure that command completion
> is correctly executed by changing sd_zbc_complete() to return good_bytes
> instead of 0, causing a hang during device probe (endless retries).
> 
> When CONFIG_BLK_DEV_ZONED is enabled and a host-aware disk is detected
> to have partitions, it will be used as a regular disk. In this case,
> make sure to not do anything in sd_zbc_revalidate_zones() as that
> triggers warnings.
> 
> Reported-by: Borislav Petkov <bp@alien8.de>
> Fixes: b72053072c0b ("block: allow partitions on host aware zone devices")
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
> Tested-by: Borislav Petkov <bp@suse.de>
> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> ---
>  drivers/scsi/sd.c     | 28 ++++++++++++++++++++++------
>  drivers/scsi/sd.h     |  2 +-
>  drivers/scsi/sd_zbc.c |  6 +++++-
>  3 files changed, 28 insertions(+), 8 deletions(-)
> 
>  	} else {
>  		sdkp->zoned = (buffer[8] >> 4) & 3;
>  		if (sdkp->zoned == 1 && !disk_has_partitions(sdkp->disk)) {
> +			/*
> +			 * Host-aware disk without partition: use the disk as
> +			 * such if support for zoned block devices is enabled.
> +			 * Otherwise, use it as a regular disk.
> +			 */
> +			if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
> +				q->limits.zoned = BLK_ZONED_HA;
> +			else
> +				q->limits.zoned = BLK_ZONED_NONE;
>  		} else {
>  			/*
>  			 * Treat drive-managed devices and host-aware devices
>  			 * with partitions as regular block devices.
>  			 */
>  			q->limits.zoned = BLK_ZONED_NONE;
>  		}

I think we need to lift much of this into a block layer helper,
as the logic is way subtile.  Something like this (written in the MUA
editor, not even compile tested).

static inline void blk_queue_set_zoned(struct gendisk *disk,
		enum blk_zoned_model model)
{
	switch (model) {
	case BLK_ZONED_HM:
		WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
		break;
	/*
	 * Host aware drivers are neither fish nor fowl.  We can either
	 * treat them like a drive managed devices, in which case they
	 * aren't different from a normal block device, or we can try
	 * to take advantage of the Zone command set, but in that case
	 * we need to treat them like a host managed device.  We try
	 * the latter if there are not partitions and zoned block device
	 * set support is enabled, else we do nothing special as far as
	 * the block layer is concerned.
	 */
	case BLK_ZONED_HA:
		if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
		    disk_has_partitions(disk)) {
			model = BLK_ZONED_NONE;
		break;
	default:
		break;

	disk->queue->limits.zoned = model;
}

Then in sd do:

	if (sdkp->device->type == TYPE_ZBC) {
		blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HM);
	} else {
		sdkp->zoned = (buffer[8] >> 4) & 3;
		if (sdkp->zoned == 1)
			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HA);
		else
			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_NONE);
	}

Re: [PATCH v2 1/2] scsi: Fix handling of host-aware ZBC disks

[Damien Le Moal]

On 2020/09/14 16:20, Christoph Hellwig wrote:
> On Mon, Sep 14, 2020 at 09:34:47AM +0900, Damien Le Moal wrote:
>> When CONFIG_BLK_DEV_ZONED is disabled, allow using host-aware ZBC
>> disks as regular disks. In this case, ensure that command completion
>> is correctly executed by changing sd_zbc_complete() to return good_bytes
>> instead of 0, causing a hang during device probe (endless retries).
>>
>> When CONFIG_BLK_DEV_ZONED is enabled and a host-aware disk is detected
>> to have partitions, it will be used as a regular disk. In this case,
>> make sure to not do anything in sd_zbc_revalidate_zones() as that
>> triggers warnings.
>>
>> Reported-by: Borislav Petkov <bp@alien8.de>
>> Fixes: b72053072c0b ("block: allow partitions on host aware zone devices")
>> Cc: <stable@vger.kernel.org>
>> Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
>> Tested-by: Borislav Petkov <bp@suse.de>
>> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
>> ---
>>  drivers/scsi/sd.c     | 28 ++++++++++++++++++++++------
>>  drivers/scsi/sd.h     |  2 +-
>>  drivers/scsi/sd_zbc.c |  6 +++++-
>>  3 files changed, 28 insertions(+), 8 deletions(-)
>>
>>  	} else {
>>  		sdkp->zoned = (buffer[8] >> 4) & 3;
>>  		if (sdkp->zoned == 1 && !disk_has_partitions(sdkp->disk)) {
>> +			/*
>> +			 * Host-aware disk without partition: use the disk as
>> +			 * such if support for zoned block devices is enabled.
>> +			 * Otherwise, use it as a regular disk.
>> +			 */
>> +			if (IS_ENABLED(CONFIG_BLK_DEV_ZONED))
>> +				q->limits.zoned = BLK_ZONED_HA;
>> +			else
>> +				q->limits.zoned = BLK_ZONED_NONE;
>>  		} else {
>>  			/*
>>  			 * Treat drive-managed devices and host-aware devices
>>  			 * with partitions as regular block devices.
>>  			 */
>>  			q->limits.zoned = BLK_ZONED_NONE;
>>  		}
> 
> I think we need to lift much of this into a block layer helper,
> as the logic is way subtile.  Something like this (written in the MUA
> editor, not even compile tested).
> 
> static inline void blk_queue_set_zoned(struct gendisk *disk,
> 		enum blk_zoned_model model)
> {
> 	switch (model) {
> 	case BLK_ZONED_HM:
> 		WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED));
> 		break;
> 	/*
> 	 * Host aware drivers are neither fish nor fowl.  We can either
> 	 * treat them like a drive managed devices, in which case they
> 	 * aren't different from a normal block device, or we can try
> 	 * to take advantage of the Zone command set, but in that case
> 	 * we need to treat them like a host managed device.  We try
> 	 * the latter if there are not partitions and zoned block device
> 	 * set support is enabled, else we do nothing special as far as
> 	 * the block layer is concerned.
> 	 */
> 	case BLK_ZONED_HA:
> 		if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) ||
> 		    disk_has_partitions(disk)) {
> 			model = BLK_ZONED_NONE;
> 		break;
> 	default:
> 		break;
> 
> 	disk->queue->limits.zoned = model;
> }
> 
> Then in sd do:
> 
> 	if (sdkp->device->type == TYPE_ZBC) {
> 		blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HM);
> 	} else {
> 		sdkp->zoned = (buffer[8] >> 4) & 3;
> 		if (sdkp->zoned == 1)
> 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HA);
> 		else
> 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_NONE);
> 	}
> 

Yes, that's nice. Will send something along these lines. But since this is a bug
fix for the current cycle & stable, we probably should keep the patch as is and
add the improvement on top for 5.10, no ?

-- 
Damien Le Moal
Western Digital Research

Re: [PATCH v2 1/2] scsi: Fix handling of host-aware ZBC disks

[hch]

On Mon, Sep 14, 2020 at 09:03:49AM +0000, Damien Le Moal wrote:
> Yes, that's nice. Will send something along these lines. But since this is a bug
> fix for the current cycle & stable, we probably should keep the patch as is and
> add the improvement on top for 5.10, no ?

I'd much rather also add the trivial helper for 5.9 - otherwise we'll
need to pull current mainline into the for-5.10 tree and create
all kinds of mess.  And just adding the helper and using it for sd
only will have no side effects elsewhere.

Re: [PATCH v2 1/2] scsi: Fix handling of host-aware ZBC disks

[Damien Le Moal]

On 2020/09/14 23:20, hch@infradead.org wrote:
> On Mon, Sep 14, 2020 at 09:03:49AM +0000, Damien Le Moal wrote:
>> Yes, that's nice. Will send something along these lines. But since this is a bug
>> fix for the current cycle & stable, we probably should keep the patch as is and
>> add the improvement on top for 5.10, no ?
> 
> I'd much rather also add the trivial helper for 5.9 - otherwise we'll
> need to pull current mainline into the for-5.10 tree and create
> all kinds of mess.  And just adding the helper and using it for sd
> only will have no side effects elsewhere.
> 

OK. Sending a v3 with this added.

-- 
Damien Le Moal
Western Digital Research