[RFC PATCH 0/2] enable sector size > PAGE_SIZE for scsi

[RFC PATCH 0/2] enable sector size > PAGE_SIZE for scsi

[sw.prabhu6]

From: Swarna Prabhu <sw.prabhu6@gmail.com>

Hi All,

Now that block layer can support block size > PAGE_SIZE, enable the
same for scsi devices. There was one issue with write_same16 and
write_same10 command, which is fixed as a part of the series.

Testing:
  1. Test suite: xfs and generic from fstest + QEMU emulated block
    device(scsi and nvme)
  - fstest Config for patched xfs 16k block size [xfs_reflink_16k_scsi]
    TEST_DEV=/dev/sda
    SCRATCH_DEV_POOL="/dev/sdb"
    MKFS_OPTIONS='-f -m reflink=1,rmapbt=1, -i sparse=1, -b size=16384, 
    -s size=16384'
  - Generic test results 
    Baseline: 6.18-rc5 kernel + nvme 16k logical block size
    Patched: 6.18-rc5 kernel + scsi 16k logical block size
    6 failures seen on generic tests and the same seen on baseline 
    No regressions 
  - XFS tests results 
    Baseline: 6.18-rc6 kernel + nvme16k logical block size
    Patched: 6.18-rc6 kernel + sci 16k logical block size
    198 failures seen on patched and baseline
    No regressions 
  - Tests similar to above was run on 32k logical block size
    No regressions seen on xfs tests. 
    Generic/778 (add sudden shutdown tests for multi block atomic writes
    )test is failing on scsi 32k block size due to atomic write taking
    too long to start, but is passing on the nvme 32k block size setup. 
    Since sysfs shows atomic write granularity not enforced for scsi device
    it seems that XFS issues atomic writes slower. Upon increasing the 
    wait timeout from 10s to 40s for the first atomic write in 
    'start_atomic_write_and_shutdown' generic/778 seems to pass for 
    32k scsi sector size.  

 2. blktests: scsi and block layer sub tests 
    Baseline: tests with scsi 4k block size 
    Patched:  tests with patched kernel + 16k scsi block size
  - scsi tests - no regressions.
  - block tests - Disabled block/010 and block/011 
    Hard to reproduce(sporadic) kernel hang seen while running 
    block layer tests ie at block/003 due to fio task being blocked
    by a mutex held by udev event. Not seen while running it
    individually. 
    Rootcause - block/001 stress the scsi debug devices by adding,
    tearing, removing the devices in parallel that triggers udevs.
    This races with fio task triggered by block/003. 
    Added udevadm settle and small delay between block/001 and 
    block/003 fixes this and using this as a temporary fix, while
    we continue to evaluate. Note: block/002 being skipped.

Swarna Prabhu (2):
  scsi: fix write_same16 and write_same10 for sector size > PAGE_SIZE
  scsi: enable sector_size > PAGE_SIZE

 drivers/scsi/scsi_debug.c |  9 ++-------
 drivers/scsi/sd.c         | 19 +++++++++++++------
 2 files changed, 15 insertions(+), 13 deletions(-)

-- 
2.51.0

[PATCH 1/2] scsi: fix write_same16 and write_same10 for sector size > PAGE_SIZE

[sw.prabhu6]

From: Swarna Prabhu <sw.prabhu6@gmail.com>

The WRITE SAME(16) and WRITE SAME(10) scsi commands uses
a page from a dedicated mempool('sd_page_pool') for its
payload. This pool was initialized to allocate single
pages, which was sufficient as long as the device sector
size did not exceed the PAGE_SIZE.

Given that block layer now supports block size upto
64K ie beyond PAGE_SIZE, adapt sd_set_special_bvec()
to accommodate that.

Signed-off-by: Swarna Prabhu <s.prabhu@samsung.com>
Co-developed-by: Pankaj Raghav <p.raghav@samsung.com>
Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
---
Note: We are allocating pages of order aligned to 
BLK_MAX_BLOCK_SIZE for the mempool page allocator 
'sd_page_pool' all the time. This is because we only
know that a bigger sector size device is attached at 
sd_probe and it might be too late to reallocate mempool
with order > 0.

 drivers/scsi/sd.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 0252d3f6bed1..c3502fcba1bb 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -895,11 +895,20 @@ static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
 {
 	struct page *page;
+	struct scsi_device *sdp = scsi_disk(rq->q->disk)->device;
+	unsigned sector_size = sdp->sector_size;
+	unsigned int nr_pages = DIV_ROUND_UP(sector_size, PAGE_SIZE);
+	int n = 0;
 
 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
 	if (!page)
 		return NULL;
-	clear_highpage(page);
+
+	do {
+		clear_highpage(page + n);
+		n++;
+	} while (n < nr_pages);
+
 	bvec_set_page(&rq->special_vec, page, data_len, 0);
 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
 	return bvec_virt(&rq->special_vec);
@@ -4368,7 +4377,7 @@ static int __init init_sd(void)
 	if (err)
 		goto err_out;
 
-	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
+	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, get_order(BLK_MAX_BLOCK_SIZE));
 	if (!sd_page_pool) {
 		printk(KERN_ERR "sd: can't init discard page pool\n");
 		err = -ENOMEM;
-- 
2.51.0

Re: [PATCH 1/2] scsi: fix write_same16 and write_same10 for sector size > PAGE_SIZE

[Bart Van Assche]

On 12/1/25 4:15 PM, sw.prabhu6@gmail.com wrote:
> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index 0252d3f6bed1..c3502fcba1bb 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -895,11 +895,20 @@ static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
>   static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
>   {
>   	struct page *page;
> +	struct scsi_device *sdp = scsi_disk(rq->q->disk)->device;
> +	unsigned sector_size = sdp->sector_size;
> +	unsigned int nr_pages = DIV_ROUND_UP(sector_size, PAGE_SIZE);
> +	int n = 0;
>   
>   	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
>   	if (!page)
>   		return NULL;
> -	clear_highpage(page);
> +
> +	do {
> +		clear_highpage(page + n);
> +		n++;
> +	} while (n < nr_pages);
> +
>   	bvec_set_page(&rq->special_vec, page, data_len, 0);
>   	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
>   	return bvec_virt(&rq->special_vec);
> @@ -4368,7 +4377,7 @@ static int __init init_sd(void)
>   	if (err)
>   		goto err_out;
>   
> -	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
> +	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, get_order(BLK_MAX_BLOCK_SIZE));
>   	if (!sd_page_pool) {
>   		printk(KERN_ERR "sd: can't init discard page pool\n");
>   		err = -ENOMEM;

Have the above changes been tested? I don't see any changes for this
code from drivers/scsi/sd.c:

	if (sector_size != 512 &&
	    sector_size != 1024 &&
	    sector_size != 2048 &&
	    sector_size != 4096) {
		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
			  sector_size);
		[ ... ]
	}

Thanks,

Bart.

Re: [PATCH 1/2] scsi: fix write_same16 and write_same10 for sector size > PAGE_SIZE

[Swarna Prabhu]

On Tue, Dec 02, 2025 at 02:28:47PM -1000, Bart Van Assche wrote:
> On 12/1/25 4:15 PM, sw.prabhu6@gmail.com wrote:
> > diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> > index 0252d3f6bed1..c3502fcba1bb 100644
> > --- a/drivers/scsi/sd.c
> > +++ b/drivers/scsi/sd.c
> > @@ -895,11 +895,20 @@ static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
> >   static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
> >   {
> >   	struct page *page;
> > +	struct scsi_device *sdp = scsi_disk(rq->q->disk)->device;
> > +	unsigned sector_size = sdp->sector_size;
> > +	unsigned int nr_pages = DIV_ROUND_UP(sector_size, PAGE_SIZE);
> > +	int n = 0;
> >   	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
> >   	if (!page)
> >   		return NULL;
> > -	clear_highpage(page);
> > +
> > +	do {
> > +		clear_highpage(page + n);
> > +		n++;
> > +	} while (n < nr_pages);
> > +
> >   	bvec_set_page(&rq->special_vec, page, data_len, 0);
> >   	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
> >   	return bvec_virt(&rq->special_vec);
> > @@ -4368,7 +4377,7 @@ static int __init init_sd(void)
> >   	if (err)
> >   		goto err_out;
> > -	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
> > +	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, get_order(BLK_MAX_BLOCK_SIZE));
> >   	if (!sd_page_pool) {
> >   		printk(KERN_ERR "sd: can't init discard page pool\n");
> >   		err = -ENOMEM;
> 
> Have the above changes been tested? 
 
  Yes, the test details are included in the cover letter

  Thanks
  Swarna

[PATCH 2/2] scsi: enable sector_size > PAGE_SIZE

[sw.prabhu6]

From: Swarna Prabhu <sw.prabhu6@gmail.com>

Now that block layer can support block size > PAGE_SIZE
and the issue with WRITE SAME(16) and WRITE SAME(10) are
fixed for sector sizes > PAGE_SIZE, enable sector sizes
> PAGE_SIZE in scsi and scsi_debug.

Signed-off-by: Swarna Prabhu <s.prabhu@samsung.com>
---
 drivers/scsi/scsi_debug.c | 9 ++-------
 drivers/scsi/sd.c         | 6 ++----
 2 files changed, 4 insertions(+), 11 deletions(-)

diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
index b2ab97be5db3..b5839e62d3bb 100644
--- a/drivers/scsi/scsi_debug.c
+++ b/drivers/scsi/scsi_debug.c
@@ -8459,13 +8459,8 @@ static int __init scsi_debug_init(void)
 	} else if (sdebug_ndelay > 0)
 		sdebug_jdelay = JDELAY_OVERRIDDEN;
 
-	switch (sdebug_sector_size) {
-	case  512:
-	case 1024:
-	case 2048:
-	case 4096:
-		break;
-	default:
+	if (sdebug_sector_size < 512 || sdebug_sector_size > BLK_MAX_BLOCK_SIZE ||
+	    !is_power_of_2(sdebug_sector_size)) {
 		pr_err("invalid sector_size %d\n", sdebug_sector_size);
 		return -EINVAL;
 	}
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index c3502fcba1bb..f2eac79d7263 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -2889,10 +2889,8 @@ sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
 			  "assuming 512.\n");
 	}
 
-	if (sector_size != 512 &&
-	    sector_size != 1024 &&
-	    sector_size != 2048 &&
-	    sector_size != 4096) {
+	if (sector_size < 512 || sector_size > BLK_MAX_BLOCK_SIZE ||
+	    !is_power_of_2(sector_size)) {
 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
 			  sector_size);
 		/*
-- 
2.51.0

Re: [PATCH 2/2] scsi: enable sector_size > PAGE_SIZE

[Bart Van Assche]

On 12/1/25 4:15 PM, sw.prabhu6@gmail.com wrote:
> diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
> index b2ab97be5db3..b5839e62d3bb 100644
> --- a/drivers/scsi/scsi_debug.c
> +++ b/drivers/scsi/scsi_debug.c
> @@ -8459,13 +8459,8 @@ static int __init scsi_debug_init(void)
>   	} else if (sdebug_ndelay > 0)
>   		sdebug_jdelay = JDELAY_OVERRIDDEN;
>   
> -	switch (sdebug_sector_size) {
> -	case  512:
> -	case 1024:
> -	case 2048:
> -	case 4096:
> -		break;
> -	default:
> +	if (sdebug_sector_size < 512 || sdebug_sector_size > BLK_MAX_BLOCK_SIZE ||
> +	    !is_power_of_2(sdebug_sector_size)) {
>   		pr_err("invalid sector_size %d\n", sdebug_sector_size);
>   		return -EINVAL;
>   	}
> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index c3502fcba1bb..f2eac79d7263 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -2889,10 +2889,8 @@ sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
>   			  "assuming 512.\n");
>   	}
>   
> -	if (sector_size != 512 &&
> -	    sector_size != 1024 &&
> -	    sector_size != 2048 &&
> -	    sector_size != 4096) {
> +	if (sector_size < 512 || sector_size > BLK_MAX_BLOCK_SIZE ||
> +	    !is_power_of_2(sector_size)) {
>   		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
>   			  sector_size);
>   		/*

Please reorganize this patch series into one patch for the scsi_debug
driver and another patch for the sd driver.

Thanks,

Bart.

Re: [PATCH 2/2] scsi: enable sector_size > PAGE_SIZE

[Swarna Prabhu]

On Tue, Dec 02, 2025 at 02:30:19PM -1000, Bart Van Assche wrote:
> On 12/1/25 4:15 PM, sw.prabhu6@gmail.com wrote:
> > diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
> > index b2ab97be5db3..b5839e62d3bb 100644
> > --- a/drivers/scsi/scsi_debug.c
> > +++ b/drivers/scsi/scsi_debug.c
> > @@ -8459,13 +8459,8 @@ static int __init scsi_debug_init(void)
> >   	} else if (sdebug_ndelay > 0)
> >   		sdebug_jdelay = JDELAY_OVERRIDDEN;
> > -	switch (sdebug_sector_size) {
> > -	case  512:
> > -	case 1024:
> > -	case 2048:
> > -	case 4096:
> > -		break;
> > -	default:
> > +	if (sdebug_sector_size < 512 || sdebug_sector_size > BLK_MAX_BLOCK_SIZE ||
> > +	    !is_power_of_2(sdebug_sector_size)) {
> >   		pr_err("invalid sector_size %d\n", sdebug_sector_size);
> >   		return -EINVAL;
> >   	}
> > diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> > index c3502fcba1bb..f2eac79d7263 100644
> > --- a/drivers/scsi/sd.c
> > +++ b/drivers/scsi/sd.c
> > @@ -2889,10 +2889,8 @@ sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
> >   			  "assuming 512.\n");
> >   	}
> > -	if (sector_size != 512 &&
> > -	    sector_size != 1024 &&
> > -	    sector_size != 2048 &&
> > -	    sector_size != 4096) {
> > +	if (sector_size < 512 || sector_size > BLK_MAX_BLOCK_SIZE ||
> > +	    !is_power_of_2(sector_size)) {
> >   		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
> >   			  sector_size);
> >   		/*
> 
> Please reorganize this patch series into one patch for the scsi_debug
> driver and another patch for the sd driver.
> 
  Sure, will do. 

  Thanks
  Swarna