enable sector size > PAGE_SIZE for scsi
v3 cover enable sector size > PAGE_SIZE for scsi
From: Swarna Prabhu <[email protected]>
Hi All,
This is v3 series sent based on the review comments received on v2 [1]. This patchset enables sector sizes > PAGE_SIZE for
sd driver and scsi_debug driver since block layer can support block size > PAGE_SIZE. There was one issue with write_same16 and write_same10 command, which is fixed as a part of the series.
Changes since v2:
- create a helper function to initialize a large page mempool at 'sd_probe' when the first device with sector size > 4k is detected for sd driver with ensuring atomicity.
- create a helper function for safe destruction of the large page mempool and use that in 'sd_probe' if the device fails at probe after the mempool is successfully created.
- Utilize the helper function for safe destruction of the large page mempool in 'sd_remove' when the last device with sector size > PAGE_SIZE is detached from the system.
- Replace while with for loop to 'clear_page' in 'sd_set_special_bvec' function in sd driver.
- Modified the git commit title and message to remove the fix tag for for scsi sd driver patch.
- Rebased the changes on latest origin/master branch.
Thanks to Damien for review on the v2 series.
Testing:
-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: 6.19-rc8 kernel + nvme 16k logical block size
Patched: 6.19-rc8 kernel + scsi 16k logical block size No regressions introduced by the patch.
- XFS tests results
Baseline: 6.19-rc8 kernel + nvme 16k logical block size
Patched: 6.19-rc8 kernel + scsi 16k logical block size No regressions introduced by the patch
- Blktests results
scsi and block layer tests with 16k logical block size.
Baseline: vanilla kernel + scsi 4k
No regressions seen by the patch.
Link to v2: https://lore.kernel.org/all/[email protected]/ [1]
Swarna Prabhu (2):
scsi: sd: enable sector size > PAGE_SIZE in scsi sd driver
scsi: scsi_debug: enable sdebug_sector_size > PAGE_SIZE
drivers/scsi/scsi_debug.c | 8 +--- drivers/scsi/sd.c | 80 +++++++++++++++++++++++++++++++++------ 2 files changed, 69 insertions(+), 19 deletions(-)
Signature · 1 line
2.39.5
PATCH v3 1/2 scsi: sd: enable sector size > PAGE_SIZE in scsi sd driver
From: Swarna Prabhu <[email protected]>
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, initialize large page pool in 'sd_probe()' if a higher sector device is attached ensuring atomicity. Adapt 'sd_set_special_bvec()' to use large page pool when a higher sector size device is attached. Hence enable sector sizes > PAGE_SIZE in scsi sd driver.
--- drivers/scsi/sd.c | 80 ++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 68 insertions(+), 12 deletions(-)
drivers/scsi/sd.c +68 -12
@@ -107,8 +107,11 @@ static void sd_config_write_same(struct scsi_disk *sdkp,static void sd_revalidate_disk (struct gendisk *);static DEFINE_IDA (sd_index_ida); +static DEFINE_MUTEX (sd_mutex_lock);static mempool_t *sd_page_pool; +static mempool_t *sd_large_page_pool; +static atomic_t sd_large_page_pool_users =ATOMIC_INIT (0 );static struct lock_class_key sd_bio_compl_lkclass;static const char *sd_cache_types[] = {
@@ -116,6 +119,33 @@ static const char *sd_cache_types[] = {"write back, no read (daft)" }; +static int sd_large_pool_create_lock (void ) +{ +mutex_lock (&sd_mutex_lock); +if (!sd_large_page_pool) { + sd_large_page_pool =mempool_create_page_pool ( +SD_MEMPOOL_SIZE ,get_order (BLK_MAX_BLOCK_SIZE )); +if (!sd_large_page_pool) { +printk (KERN_ERR "sd: can't create large page mempool\n" ); +mutex_unlock (&sd_mutex_lock); +return -ENOMEM ; + } + } +atomic_inc (&sd_large_page_pool_users); +mutex_unlock (&sd_mutex_lock); +return 0 ; +} + +static void sd_large_pool_destroy_lock (void ) +{ +mutex_lock (&sd_mutex_lock); +if (atomic_dec_and_test (&sd_large_page_pool_users)) { +mempool_destroy (sd_large_page_pool); + sd_large_page_pool =NULL ; + } +mutex_unlock (&sd_mutex_lock); +} +static void sd_disable_discard (struct scsi_disk *sdkp) { sdkp->provisioning_mode =SD_LBP_DISABLE ;
@@ -928,14 +958,24 @@ static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,return protect; } -static void *sd_set_special_bvec (struct request *rq,unsigned int data_len) +static void *sd_set_special_bvec (struct scsi_cmnd *cmd,unsigned int data_len) {struct page *page; +struct request *rq =scsi_cmd_to_rq (cmd); +struct scsi_device *sdp = cmd->device; +unsigned sector_size = sdp->sector_size; +unsigned int nr_pages =DIV_ROUND_UP (sector_size,PAGE_SIZE ); +int n; - page =mempool_alloc (sd_page_pool,GFP_ATOMIC ); +if (sector_size >PAGE_SIZE ) + page =mempool_alloc (sd_large_page_pool,GFP_ATOMIC ); +else + page =mempool_alloc (sd_page_pool,GFP_ATOMIC );if (!page)return NULL ; -clear_highpage (page); + +for (n =0 ; n < nr_pages; n++) +clear_highpage (page + n);bvec_set_page (&rq->special_vec, page, data_len,0 ); rq->rq_flags |=RQF_SPECIAL_PAYLOAD ;return bvec_virt (&rq->special_vec);
@@ -951,7 +991,7 @@ static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)unsigned int data_len =24 ;char *buf; - buf =sd_set_special_bvec (rq, data_len); + buf =sd_set_special_bvec (cmd, data_len);if (!buf)return BLK_STS_RESOURCE ;
@@ -1040,7 +1080,7 @@ static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,u32 nr_blocks =sectors_to_logical (sdp,blk_rq_sectors (rq));u32 data_len = sdp->sector_size; -if (!sd_set_special_bvec (rq, data_len)) +if (!sd_set_special_bvec (cmd, data_len))return BLK_STS_RESOURCE ; cmd->cmd_len =16 ;
@@ -1067,7 +1107,7 @@ static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,u32 nr_blocks =sectors_to_logical (sdp,blk_rq_sectors (rq));u32 data_len = sdp->sector_size; -if (!sd_set_special_bvec (rq, data_len)) +if (!sd_set_special_bvec (cmd, data_len))return BLK_STS_RESOURCE ; cmd->cmd_len =10 ;
@@ -1513,9 +1553,15 @@ static blk_status_t sd_init_command(struct scsi_cmnd *cmd)static void sd_uninit_command (struct scsi_cmnd *SCpnt) {struct request *rq =scsi_cmd_to_rq (SCpnt); +struct scsi_device *sdp = SCpnt->device; +unsigned sector_size = sdp->sector_size; -if (rq->rq_flags &RQF_SPECIAL_PAYLOAD ) -mempool_free (rq->special_vec.bv_page, sd_page_pool); +if (rq->rq_flags &RQF_SPECIAL_PAYLOAD ) { +if (sector_size >PAGE_SIZE ) +mempool_free (rq->special_vec.bv_page, sd_large_page_pool); +else +mempool_free (rq->special_vec.bv_page, sd_page_pool); + } }static bool sd_need_revalidate (struct gendisk *disk,struct scsi_disk *sdkp)
@@ -2912,10 +2958,7 @@ sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim, "Sector size 0 reported, assuming 512.\n"); } -if (sector_size !=512 && - sector_size !=1024 && - sector_size !=2048 && - sector_size !=4096 ) { +if (blk_validate_block_size (sector_size)) {sd_printk (KERN_NOTICE , sdkp,"Unsupported sector size %d.\n" , sector_size);/*
@@ -4043,6 +4086,12 @@ static int sd_probe(struct scsi_device *sdp) sdkp->max_medium_access_timeouts =SD_MAX_MEDIUM_TIMEOUTS ;sd_revalidate_disk (gd); +if (sdp->sector_size >PAGE_SIZE ) { +if (sd_large_pool_create_lock ()) { + error = -ENOMEM ; + goto out_free_index; + } + }if (sdp->removable) { gd->flags |=GENHD_FL_REMOVABLE ;
@@ -4060,6 +4109,8 @@ static int sd_probe(struct scsi_device *sdp)if (error) {device_unregister (&sdkp->disk_dev);put_disk (gd); +if (sdp->sector_size >PAGE_SIZE ) +sd_large_pool_destroy_lock (); goto out; }
@@ -4212,6 +4263,9 @@ static void sd_remove(struct scsi_device *sdp)sd_shutdown (sdp);put_disk (sdkp->disk); + +if (sdp->sector_size >PAGE_SIZE ) +sd_large_pool_destroy_lock (); }static inline bool sd_do_start_stop (struct scsi_device *sdev,bool runtime)
@@ -4435,6 +4489,8 @@ static void __exit exit_sd(void)scsi_unregister_driver (&sd_template);mempool_destroy (sd_page_pool); +if (sd_large_page_pool) +mempool_destroy (sd_large_page_pool);class_unregister (&sd_disk_class);
Signature · 1 line
2.39.5
Re: [PATCH v3 1/2] scsi: sd: enable sector size > PAGE_SIZE in scsi sd driver
Quoted 15 lines · On 2/14/26 10:18, [email protected] wrote:
From: Swarna Prabhu <[email protected]>
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 thePAGE_SIZE.Given that block layer now supports block size upto 64K ie beyond
PAGE_SIZE, initialize large page pool in 'sd_probe()' if a higher sector device is attached ensuring atomicity. Adapt 'sd_set_special_bvec()' to use large page pool when a higher sector size device is attached. Hence enable sector sizes >PAGE_SIZEin scsi sd driver.
Quoted 3 lines
Signed-off-by Swarna Prabhu <[email protected]>Co-developed-by Pankaj Raghav <[email protected]>Signed-off-by Pankaj Raghav <[email protected]>
One nit below. With that fixed, feel free to add:
Quoted 21 lines
--- drivers/scsi/sd.c | 80 ++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 68 insertions(+), 12 deletions(-)drivers/scsi/sd.c +3
-0@@ -107,8 +107,11 @@ static void sd_config_write_same(struct scsi_disk *sdkp,static void sd_revalidate_disk (struct gendisk *);static DEFINE_IDA (sd_index_ida); +static DEFINE_MUTEX (sd_mutex_lock);static mempool_t *sd_page_pool; +static mempool_t *sd_large_page_pool; +static atomic_t sd_large_page_pool_users =ATOMIC_INIT (0 );static struct lock_class_key sd_bio_compl_lkclass;static const char *sd_cache_types[] = {
drivers/scsi/sd.c static const char *sd_cache_types[] = { · line 122
Quoted 4 lines
drivers/scsi/sd.c +1
-0@@ -116,6 +119,33 @@ static const char *sd_cache_types[] = { "write back, no read (daft)" }; +static int sd_large_pool_create_lock (void )
I do not see the point of the "_lock" suffix since you do not have a "no lock" variant of this function. So let's drop that suffix.
drivers/scsi/sd.c static const char *sd_cache_types[] = { · line 139
Quoted 1 line
drivers/scsi/sd.c +1
-0+static void sd_large_pool_destroy_lock (void )
Same here.
Signature · 2 lines
Damien Le Moal Western Digital Research
PATCH v3 2/2 scsi: scsi_debug: enable sdebug_sector_size > PAGE_SIZE
In reply to [email protected]From: Swarna Prabhu <[email protected]>
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 sdebug_sector_size
Quoted 1 line
PAGE_SIZEinscsi_debug.
--- drivers/scsi/scsi_debug.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-)
drivers/scsi/scsi_debug.c +1 -7
@@ -8495,13 +8495,7 @@ 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 (blk_validate_block_size (sdebug_sector_size)) {pr_err ("invalid sector_size %d\n" , sdebug_sector_size);return -EINVAL ; }
Signature · 1 line
2.39.5
Re: [PATCH v3 2/2] scsi: scsi_debug: enable sdebug_sector_size > PAGE_SIZE
Quoted 8 lines · On 2/14/26 10:18, [email protected] wrote:
From: Swarna Prabhu <[email protected]>
Now that block layer can support block size >
PAGE_SIZEand the issue withWRITE_SAME(16)andWRITE_SAME(10)are fixed for sector sizes >PAGE_SIZE, enablesdebug_sector_sizeQuoted 1 line
PAGE_SIZEinscsi_debug.Signed-off-by Swarna Prabhu <[email protected]>
Signature · 2 lines
Damien Le Moal Western Digital Research
Hannes Reinecke
Reviewed-byRe: [PATCH v3 2/2] scsi: scsi_debug: enable sdebug_sector_size > PAGE_SIZE
Quoted 13 lines · Damien Le Moal
Quoted 8 lines · On 2/14/26 10:18, [email protected] wrote:
From: Swarna Prabhu <[email protected]>
Now that block layer can support block size >
PAGE_SIZEand the issue withWRITE_SAME(16)andWRITE_SAME(10)are fixed for sector sizes >PAGE_SIZE, enablesdebug_sector_sizeQuoted 1 line
PAGE_SIZEinscsi_debug.Signed-off-by Swarna Prabhu <[email protected]>Reviewed-by Damien Le Moal <[email protected]>
Cheers,
Hannes
Signature · 4 lines
Dr. Hannes Reinecke Kernel Storage Architect [email protected] +49 911 74053 688 SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich