ublk: enable zoned storage support
Subsystems (2)
v11 cover ublk: enable zoned storage support
From: Andreas Hindborg <[email protected]>
Hi All,
This patch set adds zoned storage support to ublk. The first two patches do some house cleaning in preparation for the last patch. The last patch adds support for report_zones and the following operations:
- REQ_OP_ZONE_OPEN
- REQ_OP_ZONE_CLOSE
- REQ_OP_ZONE_FINISH
- REQ_OP_ZONE_RESET
- REQ_OP_ZONE_APPEND
A user space component based on ubdsrv is available for testing [1] with the "loop" target.
Read/write and zone operations are tested with zenfs [2].
The zone append path is tested with fio -> zonefs -> ublk -> null_blk.
The series is based on v6.5-rc4.
Changes for v11
- Style fixes
- Use kvmalloc() instead of __vmalloc() for zone report buffer
[1] https://github.com/metaspace/ubdsrv/tree/2966e5f9637b5856d4a4273ae113e31b1c53ff98
[2] https://github.com/westerndigitalcorporation/zenfs
[3] https://git.kernel.dk/linux.git
Andreas Hindborg (3):
ublk: add helper to check if device supports user copy
ublk: move check for empty address field on command submission
ublk: enable zoned storage support
drivers/block/ublk_drv.c | 348 ++++++++++++++++++++++++++++++++-- include/uapi/linux/ublk_cmd.h | 63 +++++- 2 files changed, 383 insertions(+), 28 deletions(-)
base-commit: 5d0c230f1de8c7515b6567d9afba1f196fb4e2f4
Signature · 1 line
2.41.0
PATCH v11 1/3 ublk: add helper to check if device supports user copy
From: Andreas Hindborg <[email protected]>
This will be used by ublk zoned storage support.
--- drivers/block/ublk_drv.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/block/ublk_drv.c +6 -1
@@ -185,6 +185,11 @@ struct ublk_params_header {__u32 types; }; +static inline bool ublk_dev_is_user_copy (const struct ublk_device *ub) +{ +return ub->dev_info.flags &UBLK_F_USER_COPY ; +} +static inline void __ublk_complete_rq (struct request *req);static void ublk_complete_rq (struct kref *ref);
@@ -2038,7 +2043,7 @@ static int ublk_ctrl_add_dev(struct io_uring_cmd *cmd)UBLK_F_URING_CMD_COMP_IN_TASK ;/* GET_DATA isn't needed any more with USER_COPY */ -if (ub->dev_info.flags &UBLK_F_USER_COPY ) +if (ublk_dev_is_user_copy (ub)) ub->dev_info.flags &= ~UBLK_F_NEED_GET_DATA ;/* We are not ready to support zero copy */
Signature · 1 line
2.41.0
PATCH v11 2/3 ublk: move check for empty address field on command submission
In reply to Andreas Hindborg (Samsung)From: Andreas Hindborg <[email protected]>
In preparation for zoned storage support, move the check for empty addr field into the command handler case statement. Note that the check makes no sense for UBLK_IO_NEED_GET_DATA because the addr field must always be set for this command.
--- drivers/block/ublk_drv.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-)
drivers/block/ublk_drv.c +9 -5
@@ -1419,11 +1419,6 @@ static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd, ^ (_IOC_NR(cmd_op) == UBLK_IO_NEED_GET_DATA)) goto out; -if (ublk_support_user_copy (ubq) && ub_cmd->addr) { - ret = -EINVAL ; - goto out; - } - ret =ublk_check_cmd_op (cmd_op);if (ret) goto out;
@@ -1450,6 +1445,10 @@ static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd,*/ if (!ub_cmd->addr && !ublk_need_get_data (ubq)) goto out; + }else if (ub_cmd->addr) { +/* User copy requires addr to be unset */ + ret = -EINVAL ; + goto out; }ublk_fill_io_cmd (io, cmd, ub_cmd->addr);
@@ -1469,7 +1468,12 @@ static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd,if (!ub_cmd->addr && (!ublk_need_get_data (ubq) ||req_op (req) ==REQ_OP_READ )) goto out; + }else if (ub_cmd->addr) { +/* User copy requires addr to be unset */ + ret = -EINVAL ; + goto out; } +ublk_fill_io_cmd (io, cmd, ub_cmd->addr);ublk_commit_completion (ub, ub_cmd);break ;
Signature · 1 line
2.41.0
Re: [PATCH v11 2/3] ublk: move check for empty address field on command submission
Quoted 9 lines · Andreas Hindborg (Samsung)
From: Andreas Hindborg <[email protected]>
In preparation for zoned storage support, move the check for empty
addrfield into the command handler case statement. Note that the check makes no sense forUBLK_IO_NEED_GET_DATAbecause theaddrfield must always be set for this command.Signed-off-by Andreas Hindborg <[email protected]>
Thanks,
PATCH v11 3/3 ublk: enable zoned storage support
In reply to Andreas Hindborg (Samsung)From: Andreas Hindborg <[email protected]>
Add zoned storage support to ublk: report_zones and operations:
- REQ_OP_ZONE_OPEN
- REQ_OP_ZONE_CLOSE
- REQ_OP_ZONE_FINISH
- REQ_OP_ZONE_RESET
- REQ_OP_ZONE_APPEND
The zone append feature uses the addr field of struct ublksrv_io_cmd to communicate ALBA back to the kernel. Therefore ublk must be used with the user copy feature (UBLK_F_USER_COPY) for zoned storage support to be available. Without this feature, ublk will not allow zoned storage support.
--- drivers/block/ublk_drv.c | 331 ++++++++++++++++++++++++++++++++-- include/uapi/linux/ublk_cmd.h | 63 ++++++- 2 files changed, 370 insertions(+), 24 deletions(-)
drivers/block/ublk_drv.c +316 -15
@@ -56,16 +56,21 @@ | UBLK_F_USER_RECOVERY_REISSUE \ | UBLK_F_UNPRIVILEGED_DEV \ | UBLK_F_CMD_IOCTL_ENCODE \ - | UBLK_F_USER_COPY) + | UBLK_F_USER_COPY \ + | UBLK_F_ZONED)/* All UBLK_PARAM_TYPE_* should be included here */ -#define UBLK_PARAM_TYPE_ALL (UBLK_PARAM_TYPE_BASIC | \ - UBLK_PARAM_TYPE_DISCARD | UBLK_PARAM_TYPE_DEVT) +#define UBLK_PARAM_TYPE_ALL \ + (UBLK_PARAM_TYPE_BASIC | UBLK_PARAM_TYPE_DISCARD | \ + UBLK_PARAM_TYPE_DEVT | UBLK_PARAM_TYPE_ZONED)struct ublk_rq_data {struct llist_node node;struct kref ref; +__u64 sector; +__u32 operation; +__u32 nr_zones; };struct ublk_uring_cmd_pdu {
@@ -185,11 +190,263 @@ struct ublk_params_header {__u32 types; }; +static inline unsigned int ublk_req_build_flags (struct request *req); +static inline struct ublksrv_io_desc *ublk_get_iod (struct ublk_queue *ubq, +int tag); +static inline bool ublk_dev_is_user_copy (const struct ublk_device *ub) {return ub->dev_info.flags &UBLK_F_USER_COPY ; } +static inline bool ublk_dev_is_zoned (const struct ublk_device *ub) +{ +return ub->dev_info.flags &UBLK_F_ZONED ; +} + +static inline bool ublk_queue_is_zoned (struct ublk_queue *ubq) +{ +return ubq->flags &UBLK_F_ZONED ; +} + +#ifdef CONFIG_BLK_DEV_ZONED + +static int ublk_get_nr_zones (const struct ublk_device *ub) +{ +const struct ublk_param_basic *p = &ub->params.basic; + +/* Zone size is a power of 2 */ +return p->dev_sectors >>ilog2 (p->chunk_sectors); +} + +static int ublk_revalidate_disk_zones (struct ublk_device *ub) +{ +return blk_revalidate_disk_zones (ub->ub_disk,NULL ); +} + +static int ublk_dev_param_zoned_validate (const struct ublk_device *ub) +{ +const struct ublk_param_zoned *p = &ub->params.zoned; +int nr_zones; + +if (!ublk_dev_is_zoned (ub)) +return -EINVAL ; + +if (!p->max_zone_append_sectors) +return -EINVAL ; + + nr_zones =ublk_get_nr_zones (ub); + +if (p->max_active_zones > nr_zones) +return -EINVAL ; + +if (p->max_open_zones > nr_zones) +return -EINVAL ; + +return 0 ; +} + +static int ublk_dev_param_zoned_apply (struct ublk_device *ub) +{ +const struct ublk_param_zoned *p = &ub->params.zoned; + +disk_set_zoned (ub->ub_disk,BLK_ZONED_HM ); +blk_queue_required_elevator_features (ub->ub_disk->queue, +ELEVATOR_F_ZBD_SEQ_WRITE ); +disk_set_max_active_zones (ub->ub_disk, p->max_active_zones); +disk_set_max_open_zones (ub->ub_disk, p->max_open_zones); +blk_queue_max_zone_append_sectors (ub->ub_disk->queue, p->max_zone_append_sectors); + + ub->ub_disk->nr_zones =ublk_get_nr_zones (ub); + +return 0 ; +} + +/* Based on virtblk_alloc_report_buffer */ +static void *ublk_alloc_report_buffer (struct ublk_device *ublk, +unsigned int nr_zones,size_t *buflen) +{ +struct request_queue *q = ublk->ub_disk->queue; +size_t bufsize; +void *buf; + + nr_zones = min_t(unsigned int, nr_zones, + ublk->ub_disk->nr_zones); + + bufsize = nr_zones *sizeof (struct blk_zone ); + bufsize = +min_t (size_t, bufsize,queue_max_hw_sectors (q) <<SECTOR_SHIFT ); + +while (bufsize >=sizeof (struct blk_zone )) { + buf =kvmalloc (bufsize,GFP_KERNEL | __GFP_NORETRY); +if (buf) { + *buflen = bufsize; +return buf; + } + bufsize >>=1 ; + } + + *buflen =0 ; +return NULL ; +} + +static int ublk_report_zones (struct gendisk *disk,sector_t sector, +unsigned int nr_zones,report_zones_cb cb,void *data) +{ +struct ublk_device *ub = disk->private_data; +unsigned int zone_size_sectors = disk->queue->limits.chunk_sectors; +unsigned int first_zone = sector >>ilog2 (zone_size_sectors); +unsigned int done_zones =0 ; +unsigned int max_zones_per_request; +int ret; +struct blk_zone *buffer; +size_t buffer_length; + + nr_zones = min_t(unsigned int, ub->ub_disk->nr_zones - first_zone, + nr_zones); + + buffer =ublk_alloc_report_buffer (ub, nr_zones, &buffer_length); +if (!buffer) +return -ENOMEM ; + + max_zones_per_request = buffer_length /sizeof (struct blk_zone ); + +while (done_zones < nr_zones) { +unsigned int remaining_zones = nr_zones - done_zones; +unsigned int zones_in_request = + min_t(unsigned int, remaining_zones, max_zones_per_request); +struct request *req; +struct ublk_rq_data *pdu; +blk_status_t status; + +memset (buffer,0 , buffer_length); + + req =blk_mq_alloc_request (disk->queue,REQ_OP_DRV_IN ,0 ); +if (IS_ERR (req)) { + ret =PTR_ERR (req); + goto out; + } + + pdu =blk_mq_rq_to_pdu (req); + pdu->operation =UBLK_IO_OP_REPORT_ZONES ; + pdu->sector = sector; + pdu->nr_zones = zones_in_request; + + ret =blk_rq_map_kern (disk->queue, req, buffer, buffer_length, +GFP_KERNEL ); +if (ret) { +blk_mq_free_request (req); + goto out; + } + + status =blk_execute_rq (req,0 ); + ret =blk_status_to_errno (status); +blk_mq_free_request (req); +if (ret) + goto out; + +for (unsigned int i =0 ; i < zones_in_request; i++) { +struct blk_zone *zone = buffer + i; + +/* A zero length zone means no more zones in this response */ +if (!zone->len) +break ; + + ret =cb (zone, i, data); +if (ret) + goto out; + + done_zones++; + sector += zone_size_sectors; + + } + } + + ret = done_zones; + +out: +kvfree (buffer); +return ret; +} + +static blk_status_t ublk_setup_iod_zoned (struct ublk_queue *ubq, +struct request *req) +{ +struct ublksrv_io_desc *iod =ublk_get_iod (ubq, req->tag); +struct ublk_io *io = &ubq->ios[req->tag]; +struct ublk_rq_data *pdu =blk_mq_rq_to_pdu (req); +u32 ublk_op; + +switch (req_op (req)) { +case REQ_OP_ZONE_OPEN : + ublk_op =UBLK_IO_OP_ZONE_OPEN ; +break ; +case REQ_OP_ZONE_CLOSE : + ublk_op =UBLK_IO_OP_ZONE_CLOSE ; +break ; +case REQ_OP_ZONE_FINISH : + ublk_op =UBLK_IO_OP_ZONE_FINISH ; +break ; +case REQ_OP_ZONE_RESET : + ublk_op =UBLK_IO_OP_ZONE_RESET ; +break ; +case REQ_OP_ZONE_APPEND : + ublk_op =UBLK_IO_OP_ZONE_APPEND ; +break ; +case REQ_OP_DRV_IN : + ublk_op = pdu->operation; +switch (ublk_op) { +case UBLK_IO_OP_REPORT_ZONES : + iod->op_flags = ublk_op |ublk_req_build_flags (req); + iod->nr_zones = pdu->nr_zones; + iod->start_sector = pdu->sector; +return BLK_STS_OK ; +default : +return BLK_STS_IOERR ; + } +case REQ_OP_ZONE_RESET_ALL : +case REQ_OP_DRV_OUT : +/* We do not support reset_all and drv_out */ +return BLK_STS_NOTSUPP ; +default : +return BLK_STS_IOERR ; + } + + iod->op_flags = ublk_op |ublk_req_build_flags (req); + iod->nr_sectors =blk_rq_sectors (req); + iod->start_sector =blk_rq_pos (req); + iod->addr = io->addr; + +return BLK_STS_OK ; +} + +#else + +#define ublk_report_zones (NULL) + +static int ublk_dev_param_zoned_validate (const struct ublk_device *ub) +{ +return -EOPNOTSUPP ; +} + +static int ublk_dev_param_zoned_apply (struct ublk_device *ub) +{ +return -EOPNOTSUPP ; +} + +static int ublk_revalidate_disk_zones (struct ublk_device *ub) +{ +return 0 ; +} + +static blk_status_t ublk_setup_iod_zoned (struct ublk_queue *ubq, +struct request *req) +{ +return -EOPNOTSUPP ; +} + +#endif +static inline void __ublk_complete_rq (struct request *req);static void ublk_complete_rq (struct kref *ref);
@@ -286,6 +543,9 @@ static int ublk_validate_params(const struct ublk_device *ub)if (p->max_sectors > (ub->dev_info.max_io_buf_bytes >>9 ))return -EINVAL ; + +if (ublk_dev_is_zoned (ub) && !p->chunk_sectors) +return -EINVAL ; }else return -EINVAL ;
@@ -304,6 +564,11 @@ static int ublk_validate_params(const struct ublk_device *ub)if (ub->params.types &UBLK_PARAM_TYPE_DEVT )return -EINVAL ; +if (ub->params.types &UBLK_PARAM_TYPE_ZONED ) +return ublk_dev_param_zoned_validate (ub); +else if (ublk_dev_is_zoned (ub)) +return -EINVAL ; +return 0 ; }
@@ -317,6 +582,9 @@ static int ublk_apply_params(struct ublk_device *ub)if (ub->params.types &UBLK_PARAM_TYPE_DISCARD )ublk_dev_param_discard_apply (ub); +if (ub->params.types &UBLK_PARAM_TYPE_ZONED ) +return ublk_dev_param_zoned_apply (ub); +return 0 ; }
@@ -487,6 +755,7 @@ static const struct block_device_operations ub_fops = { .owner = THIS_MODULE, .open = ublk_open, .free_disk = ublk_free_disk, + .report_zones = ublk_report_zones, };#define UBLK_MAX_PIN_PAGES 32
@@ -601,7 +870,8 @@ static inline bool ublk_need_map_req(const struct request *req)static inline bool ublk_need_unmap_req (const struct request *req) { -return ublk_rq_has_data (req) &&req_op (req) ==REQ_OP_READ ; +return ublk_rq_has_data (req) && + (req_op (req) ==REQ_OP_READ ||req_op (req) ==REQ_OP_DRV_IN ); } static int ublk_map_io(const struct ublk_queue *ubq, const struct request *req,
@@ -685,8 +955,13 @@ static blk_status_t ublk_setup_iod(struct ublk_queue *ubq, struct request *req) {struct ublksrv_io_desc *iod =ublk_get_iod (ubq, req->tag);struct ublk_io *io = &ubq->ios[req->tag]; +enum req_op op =req_op (req);u32 ublk_op; +if (!ublk_queue_is_zoned (ubq) && + (op_is_zone_mgmt (op) || op ==REQ_OP_ZONE_APPEND )) +return -EIO ; +switch (req_op (req)) {case REQ_OP_READ : ublk_op =UBLK_IO_OP_READ ;
@@ -704,6 +979,8 @@ static blk_status_t ublk_setup_iod(struct ublk_queue *ubq, struct request *req) ublk_op =UBLK_IO_OP_WRITE_ZEROES ;break ;default : +if (ublk_queue_is_zoned (ubq)) +return ublk_setup_iod_zoned (ubq, req);return BLK_STS_IOERR ; }
@@ -756,7 +1033,8 @@ static inline void __ublk_complete_rq(struct request *req)* * Both the two needn't unmap. */ -if (req_op (req) !=REQ_OP_READ &&req_op (req) !=REQ_OP_WRITE ) +if (req_op (req) !=REQ_OP_READ &&req_op (req) !=REQ_OP_WRITE && +req_op (req) !=REQ_OP_DRV_IN ) goto exit;/* for READ request, writing data in iod->addr to rq buffers */
@@ -1120,6 +1398,9 @@ static void ublk_commit_completion(struct ublk_device *ub,/* find the io request and complete */ req =blk_mq_tag_to_rq (ub->tag_set.tags[qid], tag); +if (req_op (req) ==REQ_OP_ZONE_APPEND ) + req->__sector = ub_cmd->zone_append_lba; +if (req &&likely (!blk_should_fake_timeout (req->q)))ublk_put_req_ref (ubq, req); }
@@ -1468,8 +1749,11 @@ static int __ublk_ch_uring_cmd(struct io_uring_cmd *cmd,if (!ub_cmd->addr && (!ublk_need_get_data (ubq) ||req_op (req) ==REQ_OP_READ )) goto out; - }else if (ub_cmd->addr) { -/* User copy requires addr to be unset */ + }else if (req_op (req) !=REQ_OP_ZONE_APPEND && ub_cmd->addr) { +/* +* User copy requires addr to be unset when command is +* not zone append +*/ ret = -EINVAL ; goto out; }
@@ -1546,11 +1830,14 @@ static inline bool ublk_check_ubuf_dir(const struct request *req, int ubuf_dir) {/* copy ubuf to request pages */ -if (req_op (req) ==REQ_OP_READ && ubuf_dir ==ITER_SOURCE ) +if ((req_op (req) ==REQ_OP_READ ||req_op (req) ==REQ_OP_DRV_IN ) && + ubuf_dir ==ITER_SOURCE )return true;/* copy request pages to ubuf */ -if (req_op (req) ==REQ_OP_WRITE && ubuf_dir ==ITER_DEST ) +if ((req_op (req) ==REQ_OP_WRITE || +req_op (req) ==REQ_OP_ZONE_APPEND ) && + ubuf_dir ==ITER_DEST )return true;return false;
@@ -1890,17 +2177,24 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd)get_device (&ub->cdev_dev); ub->dev_info.state =UBLK_S_DEV_LIVE ; + +if (ublk_dev_is_zoned (ub)) { + ret =ublk_revalidate_disk_zones (ub); +if (ret) + goto out_put_cdev; + } + ret =add_disk (disk); +if (ret) + goto out_put_cdev; + +set_bit (UB_STATE_USED , &ub->state); + +out_put_cdev:if (ret) { -/* -* Has to drop the reference since ->free_disk won't be -* called in case of add_disk failure. -*/ ub->dev_info.state =UBLK_S_DEV_DEAD ;ublk_put_device (ub); - goto out_put_disk; } -set_bit (UB_STATE_USED , &ub->state); out_put_disk:if (ret)put_disk (disk);
@@ -2050,6 +2344,13 @@ static int ublk_ctrl_add_dev(struct io_uring_cmd *cmd)if (ublk_dev_is_user_copy (ub)) ub->dev_info.flags &= ~UBLK_F_NEED_GET_DATA ; +/* Zoned storage support requires user copy feature */ +if (ublk_dev_is_zoned (ub) && + (!IS_ENABLED (CONFIG_BLK_DEV_ZONED ) || !ublk_dev_is_user_copy (ub))) { + ret = -EINVAL ; + goto out_free_dev_number; + } +/* We are not ready to support zero copy */ ub->dev_info.flags &= ~UBLK_F_SUPPORT_ZERO_COPY ;
include/uapi/linux/ublk_cmd.h +54 -9
104 lines not on this page · raw message
Signature · 1 line
2.41.0
Quoted 10 lines · Andreas Hindborg (Samsung)
From: Andreas Hindborg <[email protected]>
Add zoned storage support to ublk:
report_zonesand operations:
-REQ_OP_ZONE_OPEN
-REQ_OP_ZONE_CLOSE
-REQ_OP_ZONE_FINISH
-REQ_OP_ZONE_RESET
-REQ_OP_ZONE_APPEND
Quoted 6 lines
The zone append feature uses the
addrfield ofstruct ublksrv_io_cmdto communicate ALBA back to the kernel. Therefore ublk must be used with the user copy feature (UBLK_F_USER_COPY) for zoned storage support to be available. Without this feature, ublk will not allow zoned storage support.Signed-off-by Andreas Hindborg <[email protected]>
Thanks,
Quoted 11 lines · Andreas Hindborg (Samsung)
This patch set adds zoned storage support to
ublk. The first two patches do some house cleaning in preparation for the last patch. The last patch adds support forreport_zonesand the following operations:-
REQ_OP_ZONE_OPEN
-REQ_OP_ZONE_CLOSE
-REQ_OP_ZONE_FINISH
-REQ_OP_ZONE_RESET
-REQ_OP_ZONE_APPEND[...]
Applied, thanks!
[1/3] ublk: add helper to check if device supports user copy
commit: 9d4ed6d46272bb60036a6539aae74eafcbbb3d2d
[2/3] ublk: move check for empty address field on command submission
commit: 1a6e88b9593b63ccdfe1d84e3f99dd91e4f8d490
[3/3] ublk: enable zoned storage support
commit: 29802d7ca33bc0a75c9da2a143eeed4f9e99fca4Best regards,
Signature · 1 line
Jens Axboe