* [dm-devel] [PATCH 0/2] block: blk_interposer v3 @ 2021-01-28 17:12 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: pavel.tide, sergei.shtepa Hi all, I`m ready to suggest the blk_interposer again. blk_interposer allows to intercept bio requests, remap bio to another devices or add new bios. This version has support from device mapper. For the dm-linear device creation command, the `noexcl` parameter has been added, which allows to open block devices without FMODE_EXCL mode. It allows to create dm-linear device on a block device with an already mounted file system. The new ioctl DM_DEV_REMAP allows to enable and disable bio interception. Thus, it is possible to add the dm-device to the block layer stack without reconfiguring and rebooting. Sergei Shtepa (2): block: blk_interposer [dm] blk_interposer for dm-linear block/bio.c | 2 + block/blk-core.c | 29 +++ block/blk-mq.c | 13 ++ block/genhd.c | 82 ++++++++ drivers/md/dm-core.h | 46 +++- drivers/md/dm-ioctl.c | 39 ++++ drivers/md/dm-linear.c | 17 +- drivers/md/dm-table.c | 12 +- drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- drivers/md/dm.h | 2 +- include/linux/blk-mq.h | 1 + include/linux/blk_types.h | 6 +- include/linux/device-mapper.h | 7 + include/linux/genhd.h | 19 ++ include/uapi/linux/dm-ioctl.h | 15 +- 15 files changed, 643 insertions(+), 30 deletions(-) -- 2.20.1 -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH 0/2] block: blk_interposer v3 @ 2021-01-28 17:12 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: sergei.shtepa, pavel.tide Hi all, I`m ready to suggest the blk_interposer again. blk_interposer allows to intercept bio requests, remap bio to another devices or add new bios. This version has support from device mapper. For the dm-linear device creation command, the `noexcl` parameter has been added, which allows to open block devices without FMODE_EXCL mode. It allows to create dm-linear device on a block device with an already mounted file system. The new ioctl DM_DEV_REMAP allows to enable and disable bio interception. Thus, it is possible to add the dm-device to the block layer stack without reconfiguring and rebooting. Sergei Shtepa (2): block: blk_interposer [dm] blk_interposer for dm-linear block/bio.c | 2 + block/blk-core.c | 29 +++ block/blk-mq.c | 13 ++ block/genhd.c | 82 ++++++++ drivers/md/dm-core.h | 46 +++- drivers/md/dm-ioctl.c | 39 ++++ drivers/md/dm-linear.c | 17 +- drivers/md/dm-table.c | 12 +- drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- drivers/md/dm.h | 2 +- include/linux/blk-mq.h | 1 + include/linux/blk_types.h | 6 +- include/linux/device-mapper.h | 7 + include/linux/genhd.h | 19 ++ include/uapi/linux/dm-ioctl.h | 15 +- 15 files changed, 643 insertions(+), 30 deletions(-) -- 2.20.1 ^ permalink raw reply [flat|nested] 24+ messages in thread
* [dm-devel] [PATCH 1/2] block: blk_interposer 2021-01-28 17:12 ` Sergei Shtepa @ 2021-01-28 17:12 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: pavel.tide, sergei.shtepa The block layer interposer allows to intercept bio requests. This allows to connect device mapper and other kernel modules to the block device stack on the fly. changes: * new BIO_INTERPOSED bio flag. * new function __submit_bio_interposed() implements the interposers logic. * new function blk_mq_is_queue_frozen() allow to assert that the queue is frozen. * functions blk_interposer_attach() and blk_interposer_detach() allow to attach and detach interposers. Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> --- block/bio.c | 2 + block/blk-core.c | 29 ++++++++++++++ block/blk-mq.c | 13 +++++++ block/genhd.c | 82 +++++++++++++++++++++++++++++++++++++++ include/linux/blk-mq.h | 1 + include/linux/blk_types.h | 6 ++- include/linux/genhd.h | 19 +++++++++ 7 files changed, 150 insertions(+), 2 deletions(-) diff --git a/block/bio.c b/block/bio.c index 1f2cc1fbe283..f6f135eb84b5 100644 --- a/block/bio.c +++ b/block/bio.c @@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src) bio_set_flag(bio, BIO_CLONED); if (bio_flagged(bio_src, BIO_THROTTLED)) bio_set_flag(bio, BIO_THROTTLED); + if (bio_flagged(bio_src, BIO_INTERPOSED)) + bio_set_flag(bio, BIO_INTERPOSED); bio->bi_opf = bio_src->bi_opf; bio->bi_ioprio = bio_src->bi_ioprio; bio->bi_write_hint = bio_src->bi_write_hint; diff --git a/block/blk-core.c b/block/blk-core.c index 7663a9b94b80..07ec82d8fe57 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1032,6 +1032,32 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio) return ret; } +static blk_qc_t __submit_bio_interposed(struct bio *bio) +{ + struct bio_list bio_list[2] = { }; + blk_qc_t ret = BLK_QC_T_NONE; + + current->bio_list = bio_list; + if (likely(bio_queue_enter(bio) == 0)) { + struct gendisk *disk = bio->bi_disk; + + bio_set_flag(bio, BIO_INTERPOSED); + if (likely(blk_has_interposer(disk))) + disk->interposer->ip_submit_bio(bio); + else /* interposer was removed */ + bio_list_add(¤t->bio_list[0], bio); + + blk_queue_exit(disk->queue); + } + current->bio_list = NULL; + + /* Resubmit remaining bios */ + while ((bio = bio_list_pop(&bio_list[0]))) + ret = submit_bio_noacct(bio); + + return ret; +} + /** * submit_bio_noacct - re-submit a bio to the block device layer for I/O * @bio: The bio describing the location in memory and on the device. @@ -1057,6 +1083,9 @@ blk_qc_t submit_bio_noacct(struct bio *bio) return BLK_QC_T_NONE; } + if (blk_has_interposer(bio->bi_disk) && + !bio_flagged(bio, BIO_INTERPOSED)) + return __submit_bio_interposed(bio); if (!bio->bi_disk->fops->submit_bio) return __submit_bio_noacct_mq(bio); return __submit_bio_noacct(bio); diff --git a/block/blk-mq.c b/block/blk-mq.c index f285a9123a8b..924ec26fae5f 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -161,6 +161,19 @@ int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, } EXPORT_SYMBOL_GPL(blk_mq_freeze_queue_wait_timeout); + +bool blk_mq_is_queue_frozen(struct request_queue *q) +{ + bool ret; + + mutex_lock(&q->mq_freeze_lock); + ret = percpu_ref_is_dying(&q->q_usage_counter) && percpu_ref_is_zero(&q->q_usage_counter); + mutex_unlock(&q->mq_freeze_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(blk_mq_is_queue_frozen); + /* * Guarantee no request is in use, so we can change any data structure of * the queue afterward. diff --git a/block/genhd.c b/block/genhd.c index 419548e92d82..d3459582f56c 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -30,6 +30,7 @@ static struct kobject *block_depr; DECLARE_RWSEM(bdev_lookup_sem); +DEFINE_MUTEX(bdev_interposer_mutex); /* for extended dynamic devt allocation, currently only one major is used */ #define NR_EXT_DEVT (1 << MINORBITS) @@ -2148,3 +2149,84 @@ static void disk_release_events(struct gendisk *disk) WARN_ON_ONCE(disk->ev && disk->ev->block != 1); kfree(disk->ev); } + +/** + * blk_interposer_attach - Attach interposer to disk + * @disk: target disk + * @interposer: block device interposer + * @ip_submit_bio: hook for submit_bio() + * + * Returns: + * -EINVAL if @interposer is NULL. + * -EPERM if queue is not frozen. + * -EBUSY if the block device already has @interposer. + * -EALREADY if the block device already has @interposer with same callback. + * + * Disk must be frozen by blk_mq_freeze_queue(). + */ +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio) +{ + int ret = 0; + + if (!interposer) + return -EINVAL; + + if (!blk_mq_is_queue_frozen(disk->queue)) + return -EPERM; + + mutex_lock(&bdev_interposer_mutex); + if (blk_has_interposer(disk)) { + if (disk->interposer->ip_submit_bio == ip_submit_bio) + ret = -EALREADY; + else + ret = -EBUSY; + goto out; + } + + interposer->ip_submit_bio = ip_submit_bio; + interposer->disk = disk; + + disk->interposer = interposer; +out: + mutex_unlock(&bdev_interposer_mutex); + + return ret; +} +EXPORT_SYMBOL_GPL(blk_interposer_attach); + +/** + * blk_interposer_detach - Detach interposer from disk + * @interposer: block device interposer + * @ip_submit_bio: hook for submit_bio() + * + * Disk must be frozen by blk_mq_freeze_queue(). + */ +void blk_interposer_detach(struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio) +{ + struct gendisk *disk; + + if (WARN_ON(!interposer)) + return; + + mutex_lock(&bdev_interposer_mutex); + + /* Check if the interposer is still active. */ + disk = interposer->disk; + if (WARN_ON(!disk)) + goto out; + + if (WARN_ON(!blk_mq_is_queue_frozen(disk->queue))) + goto out; + + /* Check if it is really our interposer. */ + if (WARN_ON(disk->interposer->ip_submit_bio != ip_submit_bio)) + goto out; + + disk->interposer = NULL; + interposer->disk = NULL; +out: + mutex_unlock(&bdev_interposer_mutex); +} +EXPORT_SYMBOL_GPL(blk_interposer_detach); diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h index d705b174d346..9d1e8c4e922e 100644 --- a/include/linux/blk-mq.h +++ b/include/linux/blk-mq.h @@ -525,6 +525,7 @@ void blk_freeze_queue_start(struct request_queue *q); void blk_mq_freeze_queue_wait(struct request_queue *q); int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, unsigned long timeout); +bool blk_mq_is_queue_frozen(struct request_queue *q); int blk_mq_map_queues(struct blk_mq_queue_map *qmap); void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index 866f74261b3b..6c1351d7b73f 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -227,7 +227,7 @@ struct bio { * top bits REQ_OP. Use * accessors. */ - unsigned short bi_flags; /* status, etc and bvec pool number */ + unsigned int bi_flags; /* status, etc and bvec pool number */ unsigned short bi_ioprio; unsigned short bi_write_hint; blk_status_t bi_status; @@ -304,6 +304,8 @@ enum { * of this bio. */ BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ BIO_TRACKED, /* set if bio goes through the rq_qos path */ + BIO_INTERPOSED, /* bio has been interposed and can be moved to + * a different disk */ BIO_FLAG_LAST }; @@ -322,7 +324,7 @@ enum { * freed. */ #define BVEC_POOL_BITS (3) -#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) +#define BVEC_POOL_OFFSET (32 - BVEC_POOL_BITS) #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) # error "BVEC_POOL_BITS is too small" diff --git a/include/linux/genhd.h b/include/linux/genhd.h index 809aaa32d53c..8094af3a3db9 100644 --- a/include/linux/genhd.h +++ b/include/linux/genhd.h @@ -134,6 +134,14 @@ struct blk_integrity { unsigned char tag_size; }; +struct blk_interposer; +typedef void (*ip_submit_bio_t) (struct bio *bio); + +struct blk_interposer { + ip_submit_bio_t ip_submit_bio; + struct gendisk *disk; +}; + struct gendisk { /* major, first_minor and minors are input parameters only, * don't use directly. Use disk_devt() and disk_max_parts(). @@ -158,6 +166,7 @@ struct gendisk { const struct block_device_operations *fops; struct request_queue *queue; + struct blk_interposer *interposer; void *private_data; int flags; @@ -346,4 +355,14 @@ static inline void printk_all_partitions(void) } #endif /* CONFIG_BLOCK */ +/* + * block layer interposer + */ +#define blk_has_interposer(d) ((d)->interposer != NULL) + +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio); +void blk_interposer_detach(struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio); + #endif /* _LINUX_GENHD_H */ -- 2.20.1 -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH 1/2] block: blk_interposer @ 2021-01-28 17:12 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: sergei.shtepa, pavel.tide The block layer interposer allows to intercept bio requests. This allows to connect device mapper and other kernel modules to the block device stack on the fly. changes: * new BIO_INTERPOSED bio flag. * new function __submit_bio_interposed() implements the interposers logic. * new function blk_mq_is_queue_frozen() allow to assert that the queue is frozen. * functions blk_interposer_attach() and blk_interposer_detach() allow to attach and detach interposers. Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> --- block/bio.c | 2 + block/blk-core.c | 29 ++++++++++++++ block/blk-mq.c | 13 +++++++ block/genhd.c | 82 +++++++++++++++++++++++++++++++++++++++ include/linux/blk-mq.h | 1 + include/linux/blk_types.h | 6 ++- include/linux/genhd.h | 19 +++++++++ 7 files changed, 150 insertions(+), 2 deletions(-) diff --git a/block/bio.c b/block/bio.c index 1f2cc1fbe283..f6f135eb84b5 100644 --- a/block/bio.c +++ b/block/bio.c @@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src) bio_set_flag(bio, BIO_CLONED); if (bio_flagged(bio_src, BIO_THROTTLED)) bio_set_flag(bio, BIO_THROTTLED); + if (bio_flagged(bio_src, BIO_INTERPOSED)) + bio_set_flag(bio, BIO_INTERPOSED); bio->bi_opf = bio_src->bi_opf; bio->bi_ioprio = bio_src->bi_ioprio; bio->bi_write_hint = bio_src->bi_write_hint; diff --git a/block/blk-core.c b/block/blk-core.c index 7663a9b94b80..07ec82d8fe57 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1032,6 +1032,32 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio) return ret; } +static blk_qc_t __submit_bio_interposed(struct bio *bio) +{ + struct bio_list bio_list[2] = { }; + blk_qc_t ret = BLK_QC_T_NONE; + + current->bio_list = bio_list; + if (likely(bio_queue_enter(bio) == 0)) { + struct gendisk *disk = bio->bi_disk; + + bio_set_flag(bio, BIO_INTERPOSED); + if (likely(blk_has_interposer(disk))) + disk->interposer->ip_submit_bio(bio); + else /* interposer was removed */ + bio_list_add(¤t->bio_list[0], bio); + + blk_queue_exit(disk->queue); + } + current->bio_list = NULL; + + /* Resubmit remaining bios */ + while ((bio = bio_list_pop(&bio_list[0]))) + ret = submit_bio_noacct(bio); + + return ret; +} + /** * submit_bio_noacct - re-submit a bio to the block device layer for I/O * @bio: The bio describing the location in memory and on the device. @@ -1057,6 +1083,9 @@ blk_qc_t submit_bio_noacct(struct bio *bio) return BLK_QC_T_NONE; } + if (blk_has_interposer(bio->bi_disk) && + !bio_flagged(bio, BIO_INTERPOSED)) + return __submit_bio_interposed(bio); if (!bio->bi_disk->fops->submit_bio) return __submit_bio_noacct_mq(bio); return __submit_bio_noacct(bio); diff --git a/block/blk-mq.c b/block/blk-mq.c index f285a9123a8b..924ec26fae5f 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -161,6 +161,19 @@ int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, } EXPORT_SYMBOL_GPL(blk_mq_freeze_queue_wait_timeout); + +bool blk_mq_is_queue_frozen(struct request_queue *q) +{ + bool ret; + + mutex_lock(&q->mq_freeze_lock); + ret = percpu_ref_is_dying(&q->q_usage_counter) && percpu_ref_is_zero(&q->q_usage_counter); + mutex_unlock(&q->mq_freeze_lock); + + return ret; +} +EXPORT_SYMBOL_GPL(blk_mq_is_queue_frozen); + /* * Guarantee no request is in use, so we can change any data structure of * the queue afterward. diff --git a/block/genhd.c b/block/genhd.c index 419548e92d82..d3459582f56c 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -30,6 +30,7 @@ static struct kobject *block_depr; DECLARE_RWSEM(bdev_lookup_sem); +DEFINE_MUTEX(bdev_interposer_mutex); /* for extended dynamic devt allocation, currently only one major is used */ #define NR_EXT_DEVT (1 << MINORBITS) @@ -2148,3 +2149,84 @@ static void disk_release_events(struct gendisk *disk) WARN_ON_ONCE(disk->ev && disk->ev->block != 1); kfree(disk->ev); } + +/** + * blk_interposer_attach - Attach interposer to disk + * @disk: target disk + * @interposer: block device interposer + * @ip_submit_bio: hook for submit_bio() + * + * Returns: + * -EINVAL if @interposer is NULL. + * -EPERM if queue is not frozen. + * -EBUSY if the block device already has @interposer. + * -EALREADY if the block device already has @interposer with same callback. + * + * Disk must be frozen by blk_mq_freeze_queue(). + */ +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio) +{ + int ret = 0; + + if (!interposer) + return -EINVAL; + + if (!blk_mq_is_queue_frozen(disk->queue)) + return -EPERM; + + mutex_lock(&bdev_interposer_mutex); + if (blk_has_interposer(disk)) { + if (disk->interposer->ip_submit_bio == ip_submit_bio) + ret = -EALREADY; + else + ret = -EBUSY; + goto out; + } + + interposer->ip_submit_bio = ip_submit_bio; + interposer->disk = disk; + + disk->interposer = interposer; +out: + mutex_unlock(&bdev_interposer_mutex); + + return ret; +} +EXPORT_SYMBOL_GPL(blk_interposer_attach); + +/** + * blk_interposer_detach - Detach interposer from disk + * @interposer: block device interposer + * @ip_submit_bio: hook for submit_bio() + * + * Disk must be frozen by blk_mq_freeze_queue(). + */ +void blk_interposer_detach(struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio) +{ + struct gendisk *disk; + + if (WARN_ON(!interposer)) + return; + + mutex_lock(&bdev_interposer_mutex); + + /* Check if the interposer is still active. */ + disk = interposer->disk; + if (WARN_ON(!disk)) + goto out; + + if (WARN_ON(!blk_mq_is_queue_frozen(disk->queue))) + goto out; + + /* Check if it is really our interposer. */ + if (WARN_ON(disk->interposer->ip_submit_bio != ip_submit_bio)) + goto out; + + disk->interposer = NULL; + interposer->disk = NULL; +out: + mutex_unlock(&bdev_interposer_mutex); +} +EXPORT_SYMBOL_GPL(blk_interposer_detach); diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h index d705b174d346..9d1e8c4e922e 100644 --- a/include/linux/blk-mq.h +++ b/include/linux/blk-mq.h @@ -525,6 +525,7 @@ void blk_freeze_queue_start(struct request_queue *q); void blk_mq_freeze_queue_wait(struct request_queue *q); int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, unsigned long timeout); +bool blk_mq_is_queue_frozen(struct request_queue *q); int blk_mq_map_queues(struct blk_mq_queue_map *qmap); void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index 866f74261b3b..6c1351d7b73f 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -227,7 +227,7 @@ struct bio { * top bits REQ_OP. Use * accessors. */ - unsigned short bi_flags; /* status, etc and bvec pool number */ + unsigned int bi_flags; /* status, etc and bvec pool number */ unsigned short bi_ioprio; unsigned short bi_write_hint; blk_status_t bi_status; @@ -304,6 +304,8 @@ enum { * of this bio. */ BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ BIO_TRACKED, /* set if bio goes through the rq_qos path */ + BIO_INTERPOSED, /* bio has been interposed and can be moved to + * a different disk */ BIO_FLAG_LAST }; @@ -322,7 +324,7 @@ enum { * freed. */ #define BVEC_POOL_BITS (3) -#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) +#define BVEC_POOL_OFFSET (32 - BVEC_POOL_BITS) #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) # error "BVEC_POOL_BITS is too small" diff --git a/include/linux/genhd.h b/include/linux/genhd.h index 809aaa32d53c..8094af3a3db9 100644 --- a/include/linux/genhd.h +++ b/include/linux/genhd.h @@ -134,6 +134,14 @@ struct blk_integrity { unsigned char tag_size; }; +struct blk_interposer; +typedef void (*ip_submit_bio_t) (struct bio *bio); + +struct blk_interposer { + ip_submit_bio_t ip_submit_bio; + struct gendisk *disk; +}; + struct gendisk { /* major, first_minor and minors are input parameters only, * don't use directly. Use disk_devt() and disk_max_parts(). @@ -158,6 +166,7 @@ struct gendisk { const struct block_device_operations *fops; struct request_queue *queue; + struct blk_interposer *interposer; void *private_data; int flags; @@ -346,4 +355,14 @@ static inline void printk_all_partitions(void) } #endif /* CONFIG_BLOCK */ +/* + * block layer interposer + */ +#define blk_has_interposer(d) ((d)->interposer != NULL) + +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio); +void blk_interposer_detach(struct blk_interposer *interposer, + const ip_submit_bio_t ip_submit_bio); + #endif /* _LINUX_GENHD_H */ -- 2.20.1 ^ permalink raw reply related [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 1/2] block: blk_interposer 2021-01-28 17:12 ` Sergei Shtepa @ 2021-01-29 1:21 ` Damien Le Moal -1 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 1:21 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > The block layer interposer allows to intercept bio requests. > This allows to connect device mapper and other kernel modules > to the block device stack on the fly. > > changes: > * new BIO_INTERPOSED bio flag. > * new function __submit_bio_interposed() implements the interposers > logic. > * new function blk_mq_is_queue_frozen() allow to assert that > the queue is frozen. > * functions blk_interposer_attach() and blk_interposer_detach() > allow to attach and detach interposers. The changelog should not be part of the commit message. If you need a changelog for a single patch, add it between the commit message end "---" and the patch stats. git will ignore that part. > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > --- > block/bio.c | 2 + > block/blk-core.c | 29 ++++++++++++++ > block/blk-mq.c | 13 +++++++ > block/genhd.c | 82 +++++++++++++++++++++++++++++++++++++++ > include/linux/blk-mq.h | 1 + > include/linux/blk_types.h | 6 ++- > include/linux/genhd.h | 19 +++++++++ > 7 files changed, 150 insertions(+), 2 deletions(-) > > diff --git a/block/bio.c b/block/bio.c > index 1f2cc1fbe283..f6f135eb84b5 100644 > --- a/block/bio.c > +++ b/block/bio.c > @@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src) > bio_set_flag(bio, BIO_CLONED); > if (bio_flagged(bio_src, BIO_THROTTLED)) > bio_set_flag(bio, BIO_THROTTLED); > + if (bio_flagged(bio_src, BIO_INTERPOSED)) > + bio_set_flag(bio, BIO_INTERPOSED); > bio->bi_opf = bio_src->bi_opf; > bio->bi_ioprio = bio_src->bi_ioprio; > bio->bi_write_hint = bio_src->bi_write_hint; > diff --git a/block/blk-core.c b/block/blk-core.c > index 7663a9b94b80..07ec82d8fe57 100644 > --- a/block/blk-core.c > +++ b/block/blk-core.c > @@ -1032,6 +1032,32 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio) > return ret; > } > > +static blk_qc_t __submit_bio_interposed(struct bio *bio) > +{ > + struct bio_list bio_list[2] = { }; > + blk_qc_t ret = BLK_QC_T_NONE; > + > + current->bio_list = bio_list; > + if (likely(bio_queue_enter(bio) == 0)) { > + struct gendisk *disk = bio->bi_disk; > + > + bio_set_flag(bio, BIO_INTERPOSED); > + if (likely(blk_has_interposer(disk))) > + disk->interposer->ip_submit_bio(bio); Why do you check again blk_has_interposer() here ? That is checked already in submit_bio_noacct() and the interposer attach/detach cannot run without the queue frozen. So I do not think you need to check again. If you do, then you definitely have a race condition here. > + else /* interposer was removed */ > + bio_list_add(¤t->bio_list[0], bio); > + > + blk_queue_exit(disk->queue); > + } > + current->bio_list = NULL; > + > + /* Resubmit remaining bios */ > + while ((bio = bio_list_pop(&bio_list[0]))) > + ret = submit_bio_noacct(bio); > + > + return ret; > +} > + > /** > * submit_bio_noacct - re-submit a bio to the block device layer for I/O > * @bio: The bio describing the location in memory and on the device. > @@ -1057,6 +1083,9 @@ blk_qc_t submit_bio_noacct(struct bio *bio) > return BLK_QC_T_NONE; > } > > + if (blk_has_interposer(bio->bi_disk) && > + !bio_flagged(bio, BIO_INTERPOSED)) > + return __submit_bio_interposed(bio); OK. I *think* that this is to handle stacked devices, right ? Otherwise, this condition does not make much sense. Why not just: if (blk_has_interposer(bio->bi_disk)) return __submit_bio_interposed(bio); So at least adding some comments here may be good. > if (!bio->bi_disk->fops->submit_bio) > return __submit_bio_noacct_mq(bio); > return __submit_bio_noacct(bio); > diff --git a/block/blk-mq.c b/block/blk-mq.c > index f285a9123a8b..924ec26fae5f 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -161,6 +161,19 @@ int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > } > EXPORT_SYMBOL_GPL(blk_mq_freeze_queue_wait_timeout); > > + > +bool blk_mq_is_queue_frozen(struct request_queue *q) > +{ > + bool ret; > + > + mutex_lock(&q->mq_freeze_lock); > + ret = percpu_ref_is_dying(&q->q_usage_counter) && percpu_ref_is_zero(&q->q_usage_counter); > + mutex_unlock(&q->mq_freeze_lock); > + > + return ret; > +} > +EXPORT_SYMBOL_GPL(blk_mq_is_queue_frozen); Maybe move this change to its own patch preceding this one ? > + > /* > * Guarantee no request is in use, so we can change any data structure of > * the queue afterward. > diff --git a/block/genhd.c b/block/genhd.c > index 419548e92d82..d3459582f56c 100644 > --- a/block/genhd.c > +++ b/block/genhd.c > @@ -30,6 +30,7 @@ > static struct kobject *block_depr; > > DECLARE_RWSEM(bdev_lookup_sem); > +DEFINE_MUTEX(bdev_interposer_mutex); > > /* for extended dynamic devt allocation, currently only one major is used */ > #define NR_EXT_DEVT (1 << MINORBITS) > @@ -2148,3 +2149,84 @@ static void disk_release_events(struct gendisk *disk) > WARN_ON_ONCE(disk->ev && disk->ev->block != 1); > kfree(disk->ev); > } > + > +/** > + * blk_interposer_attach - Attach interposer to disk > + * @disk: target disk > + * @interposer: block device interposer > + * @ip_submit_bio: hook for submit_bio() > + * > + * Returns: > + * -EINVAL if @interposer is NULL. > + * -EPERM if queue is not frozen. > + * -EBUSY if the block device already has @interposer. > + * -EALREADY if the block device already has @interposer with same callback. > + * > + * Disk must be frozen by blk_mq_freeze_queue(). > + */ > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio) > +{ > + int ret = 0; > + > + if (!interposer) > + return -EINVAL; Is this really necessary ? If some user of this function has interposer == NULL, that caller needs debugging... > + > + if (!blk_mq_is_queue_frozen(disk->queue)) > + return -EPERM; Why not do the queue freeze here ? > + > + mutex_lock(&bdev_interposer_mutex); > + if (blk_has_interposer(disk)) { > + if (disk->interposer->ip_submit_bio == ip_submit_bio) > + ret = -EALREADY; > + else > + ret = -EBUSY; > + goto out; > + } > + > + interposer->ip_submit_bio = ip_submit_bio; > + interposer->disk = disk; > + > + disk->interposer = interposer; > +out: > + mutex_unlock(&bdev_interposer_mutex); > + > + return ret; > +} > +EXPORT_SYMBOL_GPL(blk_interposer_attach); > + > +/** > + * blk_interposer_detach - Detach interposer from disk > + * @interposer: block device interposer > + * @ip_submit_bio: hook for submit_bio() > + * > + * Disk must be frozen by blk_mq_freeze_queue(). > + */ > +void blk_interposer_detach(struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio) > +{ The interface is weird. Why not passing the gendisk ? > + struct gendisk *disk; > + > + if (WARN_ON(!interposer)) > + return; Same comment as above. This should not be necessary. > + > + mutex_lock(&bdev_interposer_mutex); > + > + /* Check if the interposer is still active. */ > + disk = interposer->disk; > + if (WARN_ON(!disk)) > + goto out; > + > + if (WARN_ON(!blk_mq_is_queue_frozen(disk->queue))) > + goto out; > + > + /* Check if it is really our interposer. */ > + if (WARN_ON(disk->interposer->ip_submit_bio != ip_submit_bio)) > + goto out; > + > + disk->interposer = NULL; > + interposer->disk = NULL; > +out: > + mutex_unlock(&bdev_interposer_mutex); > +} > +EXPORT_SYMBOL_GPL(blk_interposer_detach); > diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h > index d705b174d346..9d1e8c4e922e 100644 > --- a/include/linux/blk-mq.h > +++ b/include/linux/blk-mq.h > @@ -525,6 +525,7 @@ void blk_freeze_queue_start(struct request_queue *q); > void blk_mq_freeze_queue_wait(struct request_queue *q); > int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > unsigned long timeout); > +bool blk_mq_is_queue_frozen(struct request_queue *q); > > int blk_mq_map_queues(struct blk_mq_queue_map *qmap); > void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); > diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h > index 866f74261b3b..6c1351d7b73f 100644 > --- a/include/linux/blk_types.h > +++ b/include/linux/blk_types.h > @@ -227,7 +227,7 @@ struct bio { > * top bits REQ_OP. Use > * accessors. > */ > - unsigned short bi_flags; /* status, etc and bvec pool number */ > + unsigned int bi_flags; /* status, etc and bvec pool number */ > unsigned short bi_ioprio; > unsigned short bi_write_hint; > blk_status_t bi_status; > @@ -304,6 +304,8 @@ enum { > * of this bio. */ > BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ > BIO_TRACKED, /* set if bio goes through the rq_qos path */ > + BIO_INTERPOSED, /* bio has been interposed and can be moved to > + * a different disk */ > BIO_FLAG_LAST > }; > > @@ -322,7 +324,7 @@ enum { > * freed. > */ > #define BVEC_POOL_BITS (3) > -#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) > +#define BVEC_POOL_OFFSET (32 - BVEC_POOL_BITS) > #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) > #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) > # error "BVEC_POOL_BITS is too small" > diff --git a/include/linux/genhd.h b/include/linux/genhd.h > index 809aaa32d53c..8094af3a3db9 100644 > --- a/include/linux/genhd.h > +++ b/include/linux/genhd.h > @@ -134,6 +134,14 @@ struct blk_integrity { > unsigned char tag_size; > }; > > +struct blk_interposer; This is not needed. > +typedef void (*ip_submit_bio_t) (struct bio *bio); This hides the definition of the submit callback for no good reasons that I can see. Since the callback has a simple interface, I would prefer this to be dropped. > + > +struct blk_interposer { > + ip_submit_bio_t ip_submit_bio; > + struct gendisk *disk; If you fix the interface of the detach function, you should not need this field. > +}; > + > struct gendisk { > /* major, first_minor and minors are input parameters only, > * don't use directly. Use disk_devt() and disk_max_parts(). > @@ -158,6 +166,7 @@ struct gendisk { > > const struct block_device_operations *fops; > struct request_queue *queue; > + struct blk_interposer *interposer; > void *private_data; > > int flags; > @@ -346,4 +355,14 @@ static inline void printk_all_partitions(void) > } > #endif /* CONFIG_BLOCK */ > > +/* > + * block layer interposer > + */ > +#define blk_has_interposer(d) ((d)->interposer != NULL) Please make this an inline function. > + > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio); > +void blk_interposer_detach(struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio); > + > #endif /* _LINUX_GENHD_H */ > -- Damien Le Moal Western Digital Research -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 1/2] block: blk_interposer @ 2021-01-29 1:21 ` Damien Le Moal 0 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 1:21 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > The block layer interposer allows to intercept bio requests. > This allows to connect device mapper and other kernel modules > to the block device stack on the fly. > > changes: > * new BIO_INTERPOSED bio flag. > * new function __submit_bio_interposed() implements the interposers > logic. > * new function blk_mq_is_queue_frozen() allow to assert that > the queue is frozen. > * functions blk_interposer_attach() and blk_interposer_detach() > allow to attach and detach interposers. The changelog should not be part of the commit message. If you need a changelog for a single patch, add it between the commit message end "---" and the patch stats. git will ignore that part. > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > --- > block/bio.c | 2 + > block/blk-core.c | 29 ++++++++++++++ > block/blk-mq.c | 13 +++++++ > block/genhd.c | 82 +++++++++++++++++++++++++++++++++++++++ > include/linux/blk-mq.h | 1 + > include/linux/blk_types.h | 6 ++- > include/linux/genhd.h | 19 +++++++++ > 7 files changed, 150 insertions(+), 2 deletions(-) > > diff --git a/block/bio.c b/block/bio.c > index 1f2cc1fbe283..f6f135eb84b5 100644 > --- a/block/bio.c > +++ b/block/bio.c > @@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src) > bio_set_flag(bio, BIO_CLONED); > if (bio_flagged(bio_src, BIO_THROTTLED)) > bio_set_flag(bio, BIO_THROTTLED); > + if (bio_flagged(bio_src, BIO_INTERPOSED)) > + bio_set_flag(bio, BIO_INTERPOSED); > bio->bi_opf = bio_src->bi_opf; > bio->bi_ioprio = bio_src->bi_ioprio; > bio->bi_write_hint = bio_src->bi_write_hint; > diff --git a/block/blk-core.c b/block/blk-core.c > index 7663a9b94b80..07ec82d8fe57 100644 > --- a/block/blk-core.c > +++ b/block/blk-core.c > @@ -1032,6 +1032,32 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio) > return ret; > } > > +static blk_qc_t __submit_bio_interposed(struct bio *bio) > +{ > + struct bio_list bio_list[2] = { }; > + blk_qc_t ret = BLK_QC_T_NONE; > + > + current->bio_list = bio_list; > + if (likely(bio_queue_enter(bio) == 0)) { > + struct gendisk *disk = bio->bi_disk; > + > + bio_set_flag(bio, BIO_INTERPOSED); > + if (likely(blk_has_interposer(disk))) > + disk->interposer->ip_submit_bio(bio); Why do you check again blk_has_interposer() here ? That is checked already in submit_bio_noacct() and the interposer attach/detach cannot run without the queue frozen. So I do not think you need to check again. If you do, then you definitely have a race condition here. > + else /* interposer was removed */ > + bio_list_add(¤t->bio_list[0], bio); > + > + blk_queue_exit(disk->queue); > + } > + current->bio_list = NULL; > + > + /* Resubmit remaining bios */ > + while ((bio = bio_list_pop(&bio_list[0]))) > + ret = submit_bio_noacct(bio); > + > + return ret; > +} > + > /** > * submit_bio_noacct - re-submit a bio to the block device layer for I/O > * @bio: The bio describing the location in memory and on the device. > @@ -1057,6 +1083,9 @@ blk_qc_t submit_bio_noacct(struct bio *bio) > return BLK_QC_T_NONE; > } > > + if (blk_has_interposer(bio->bi_disk) && > + !bio_flagged(bio, BIO_INTERPOSED)) > + return __submit_bio_interposed(bio); OK. I *think* that this is to handle stacked devices, right ? Otherwise, this condition does not make much sense. Why not just: if (blk_has_interposer(bio->bi_disk)) return __submit_bio_interposed(bio); So at least adding some comments here may be good. > if (!bio->bi_disk->fops->submit_bio) > return __submit_bio_noacct_mq(bio); > return __submit_bio_noacct(bio); > diff --git a/block/blk-mq.c b/block/blk-mq.c > index f285a9123a8b..924ec26fae5f 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -161,6 +161,19 @@ int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > } > EXPORT_SYMBOL_GPL(blk_mq_freeze_queue_wait_timeout); > > + > +bool blk_mq_is_queue_frozen(struct request_queue *q) > +{ > + bool ret; > + > + mutex_lock(&q->mq_freeze_lock); > + ret = percpu_ref_is_dying(&q->q_usage_counter) && percpu_ref_is_zero(&q->q_usage_counter); > + mutex_unlock(&q->mq_freeze_lock); > + > + return ret; > +} > +EXPORT_SYMBOL_GPL(blk_mq_is_queue_frozen); Maybe move this change to its own patch preceding this one ? > + > /* > * Guarantee no request is in use, so we can change any data structure of > * the queue afterward. > diff --git a/block/genhd.c b/block/genhd.c > index 419548e92d82..d3459582f56c 100644 > --- a/block/genhd.c > +++ b/block/genhd.c > @@ -30,6 +30,7 @@ > static struct kobject *block_depr; > > DECLARE_RWSEM(bdev_lookup_sem); > +DEFINE_MUTEX(bdev_interposer_mutex); > > /* for extended dynamic devt allocation, currently only one major is used */ > #define NR_EXT_DEVT (1 << MINORBITS) > @@ -2148,3 +2149,84 @@ static void disk_release_events(struct gendisk *disk) > WARN_ON_ONCE(disk->ev && disk->ev->block != 1); > kfree(disk->ev); > } > + > +/** > + * blk_interposer_attach - Attach interposer to disk > + * @disk: target disk > + * @interposer: block device interposer > + * @ip_submit_bio: hook for submit_bio() > + * > + * Returns: > + * -EINVAL if @interposer is NULL. > + * -EPERM if queue is not frozen. > + * -EBUSY if the block device already has @interposer. > + * -EALREADY if the block device already has @interposer with same callback. > + * > + * Disk must be frozen by blk_mq_freeze_queue(). > + */ > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio) > +{ > + int ret = 0; > + > + if (!interposer) > + return -EINVAL; Is this really necessary ? If some user of this function has interposer == NULL, that caller needs debugging... > + > + if (!blk_mq_is_queue_frozen(disk->queue)) > + return -EPERM; Why not do the queue freeze here ? > + > + mutex_lock(&bdev_interposer_mutex); > + if (blk_has_interposer(disk)) { > + if (disk->interposer->ip_submit_bio == ip_submit_bio) > + ret = -EALREADY; > + else > + ret = -EBUSY; > + goto out; > + } > + > + interposer->ip_submit_bio = ip_submit_bio; > + interposer->disk = disk; > + > + disk->interposer = interposer; > +out: > + mutex_unlock(&bdev_interposer_mutex); > + > + return ret; > +} > +EXPORT_SYMBOL_GPL(blk_interposer_attach); > + > +/** > + * blk_interposer_detach - Detach interposer from disk > + * @interposer: block device interposer > + * @ip_submit_bio: hook for submit_bio() > + * > + * Disk must be frozen by blk_mq_freeze_queue(). > + */ > +void blk_interposer_detach(struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio) > +{ The interface is weird. Why not passing the gendisk ? > + struct gendisk *disk; > + > + if (WARN_ON(!interposer)) > + return; Same comment as above. This should not be necessary. > + > + mutex_lock(&bdev_interposer_mutex); > + > + /* Check if the interposer is still active. */ > + disk = interposer->disk; > + if (WARN_ON(!disk)) > + goto out; > + > + if (WARN_ON(!blk_mq_is_queue_frozen(disk->queue))) > + goto out; > + > + /* Check if it is really our interposer. */ > + if (WARN_ON(disk->interposer->ip_submit_bio != ip_submit_bio)) > + goto out; > + > + disk->interposer = NULL; > + interposer->disk = NULL; > +out: > + mutex_unlock(&bdev_interposer_mutex); > +} > +EXPORT_SYMBOL_GPL(blk_interposer_detach); > diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h > index d705b174d346..9d1e8c4e922e 100644 > --- a/include/linux/blk-mq.h > +++ b/include/linux/blk-mq.h > @@ -525,6 +525,7 @@ void blk_freeze_queue_start(struct request_queue *q); > void blk_mq_freeze_queue_wait(struct request_queue *q); > int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > unsigned long timeout); > +bool blk_mq_is_queue_frozen(struct request_queue *q); > > int blk_mq_map_queues(struct blk_mq_queue_map *qmap); > void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); > diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h > index 866f74261b3b..6c1351d7b73f 100644 > --- a/include/linux/blk_types.h > +++ b/include/linux/blk_types.h > @@ -227,7 +227,7 @@ struct bio { > * top bits REQ_OP. Use > * accessors. > */ > - unsigned short bi_flags; /* status, etc and bvec pool number */ > + unsigned int bi_flags; /* status, etc and bvec pool number */ > unsigned short bi_ioprio; > unsigned short bi_write_hint; > blk_status_t bi_status; > @@ -304,6 +304,8 @@ enum { > * of this bio. */ > BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ > BIO_TRACKED, /* set if bio goes through the rq_qos path */ > + BIO_INTERPOSED, /* bio has been interposed and can be moved to > + * a different disk */ > BIO_FLAG_LAST > }; > > @@ -322,7 +324,7 @@ enum { > * freed. > */ > #define BVEC_POOL_BITS (3) > -#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) > +#define BVEC_POOL_OFFSET (32 - BVEC_POOL_BITS) > #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) > #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) > # error "BVEC_POOL_BITS is too small" > diff --git a/include/linux/genhd.h b/include/linux/genhd.h > index 809aaa32d53c..8094af3a3db9 100644 > --- a/include/linux/genhd.h > +++ b/include/linux/genhd.h > @@ -134,6 +134,14 @@ struct blk_integrity { > unsigned char tag_size; > }; > > +struct blk_interposer; This is not needed. > +typedef void (*ip_submit_bio_t) (struct bio *bio); This hides the definition of the submit callback for no good reasons that I can see. Since the callback has a simple interface, I would prefer this to be dropped. > + > +struct blk_interposer { > + ip_submit_bio_t ip_submit_bio; > + struct gendisk *disk; If you fix the interface of the detach function, you should not need this field. > +}; > + > struct gendisk { > /* major, first_minor and minors are input parameters only, > * don't use directly. Use disk_devt() and disk_max_parts(). > @@ -158,6 +166,7 @@ struct gendisk { > > const struct block_device_operations *fops; > struct request_queue *queue; > + struct blk_interposer *interposer; > void *private_data; > > int flags; > @@ -346,4 +355,14 @@ static inline void printk_all_partitions(void) > } > #endif /* CONFIG_BLOCK */ > > +/* > + * block layer interposer > + */ > +#define blk_has_interposer(d) ((d)->interposer != NULL) Please make this an inline function. > + > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio); > +void blk_interposer_detach(struct blk_interposer *interposer, > + const ip_submit_bio_t ip_submit_bio); > + > #endif /* _LINUX_GENHD_H */ > -- Damien Le Moal Western Digital Research ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 1/2] block: blk_interposer 2021-01-29 1:21 ` Damien Le Moal (?) @ 2021-01-29 15:58 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-29 15:58 UTC (permalink / raw) To: Damien Le Moal Cc: snitzer@redhat.com, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com The 01/29/2021 04:21, Damien Le Moal wrote: > On 2021/01/29 2:23, Sergei Shtepa wrote: > > The block layer interposer allows to intercept bio requests. > > This allows to connect device mapper and other kernel modules > > to the block device stack on the fly. > > > > changes: > > * new BIO_INTERPOSED bio flag. > > * new function __submit_bio_interposed() implements the interposers > > logic. > > * new function blk_mq_is_queue_frozen() allow to assert that > > the queue is frozen. > > * functions blk_interposer_attach() and blk_interposer_detach() > > allow to attach and detach interposers. > > The changelog should not be part of the commit message. If you need a changelog > for a single patch, add it between the commit message end "---" and the patch > stats. git will ignore that part. > > > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > > --- > > block/bio.c | 2 + > > block/blk-core.c | 29 ++++++++++++++ > > block/blk-mq.c | 13 +++++++ > > block/genhd.c | 82 +++++++++++++++++++++++++++++++++++++++ > > include/linux/blk-mq.h | 1 + > > include/linux/blk_types.h | 6 ++- > > include/linux/genhd.h | 19 +++++++++ > > 7 files changed, 150 insertions(+), 2 deletions(-) > > > > diff --git a/block/bio.c b/block/bio.c > > index 1f2cc1fbe283..f6f135eb84b5 100644 > > --- a/block/bio.c > > +++ b/block/bio.c > > @@ -684,6 +684,8 @@ void __bio_clone_fast(struct bio *bio, struct bio *bio_src) > > bio_set_flag(bio, BIO_CLONED); > > if (bio_flagged(bio_src, BIO_THROTTLED)) > > bio_set_flag(bio, BIO_THROTTLED); > > + if (bio_flagged(bio_src, BIO_INTERPOSED)) > > + bio_set_flag(bio, BIO_INTERPOSED); > > bio->bi_opf = bio_src->bi_opf; > > bio->bi_ioprio = bio_src->bi_ioprio; > > bio->bi_write_hint = bio_src->bi_write_hint; > > diff --git a/block/blk-core.c b/block/blk-core.c > > index 7663a9b94b80..07ec82d8fe57 100644 > > --- a/block/blk-core.c > > +++ b/block/blk-core.c > > @@ -1032,6 +1032,32 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio) > > return ret; > > } > > > > +static blk_qc_t __submit_bio_interposed(struct bio *bio) > > +{ > > + struct bio_list bio_list[2] = { }; > > + blk_qc_t ret = BLK_QC_T_NONE; > > + > > + current->bio_list = bio_list; > > + if (likely(bio_queue_enter(bio) == 0)) { > > + struct gendisk *disk = bio->bi_disk; > > + > > + bio_set_flag(bio, BIO_INTERPOSED); > > + if (likely(blk_has_interposer(disk))) > > + disk->interposer->ip_submit_bio(bio); > > Why do you check again blk_has_interposer() here ? That is checked already in > submit_bio_noacct() and the interposer attach/detach cannot run without the > queue frozen. So I do not think you need to check again. If you do, then you > definitely have a race condition here. There is no problem here. The first call to blk_has_interposer() allows not to call the __submit_bio_interposed() function if the interposer is not specified. Changing the disk->interpreter field is blocked by the bio_queue_enter() and blk_queue_exit() functions. Therefore, a second check of blk_has_interposer() is needed after calling bio_queue_enter(). But I think setting the BIO_INTERPOSED flag should be moved below, just before calling ip_submit_bio(). > > > + else /* interposer was removed */ > > + bio_list_add(¤t->bio_list[0], bio); > > + > > + blk_queue_exit(disk->queue); > > + } > > + current->bio_list = NULL; > > + > > + /* Resubmit remaining bios */ > > + while ((bio = bio_list_pop(&bio_list[0]))) > > + ret = submit_bio_noacct(bio); > > + > > + return ret; > > +} > > + > > /** > > * submit_bio_noacct - re-submit a bio to the block device layer for I/O > > * @bio: The bio describing the location in memory and on the device. > > @@ -1057,6 +1083,9 @@ blk_qc_t submit_bio_noacct(struct bio *bio) > > return BLK_QC_T_NONE; > > } > > > > + if (blk_has_interposer(bio->bi_disk) && > > + !bio_flagged(bio, BIO_INTERPOSED)) > > + return __submit_bio_interposed(bio); > > OK. I *think* that this is to handle stacked devices, right ? Otherwise, this > condition does not make much sense. Why not just: > > if (blk_has_interposer(bio->bi_disk)) > return __submit_bio_interposed(bio); > > So at least adding some comments here may be good. Checking the BIO_INTERPOSED flag is necessary here. The device mapper can re-send an intercepted request, or generate new ones that should not be intercepted. I'll think about a comment for this line. > > > if (!bio->bi_disk->fops->submit_bio) > > return __submit_bio_noacct_mq(bio); > > return __submit_bio_noacct(bio); > > diff --git a/block/blk-mq.c b/block/blk-mq.c > > index f285a9123a8b..924ec26fae5f 100644 > > --- a/block/blk-mq.c > > +++ b/block/blk-mq.c > > @@ -161,6 +161,19 @@ int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > > } > > EXPORT_SYMBOL_GPL(blk_mq_freeze_queue_wait_timeout); > > > > + > > +bool blk_mq_is_queue_frozen(struct request_queue *q) > > +{ > > + bool ret; > > + > > + mutex_lock(&q->mq_freeze_lock); > > + ret = percpu_ref_is_dying(&q->q_usage_counter) && percpu_ref_is_zero(&q->q_usage_counter); > > + mutex_unlock(&q->mq_freeze_lock); > > + > > + return ret; > > +} > > +EXPORT_SYMBOL_GPL(blk_mq_is_queue_frozen); > > Maybe move this change to its own patch preceding this one ? Yes, if it makes the review easier. > > > + > > /* > > * Guarantee no request is in use, so we can change any data structure of > > * the queue afterward. > > diff --git a/block/genhd.c b/block/genhd.c > > index 419548e92d82..d3459582f56c 100644 > > --- a/block/genhd.c > > +++ b/block/genhd.c > > @@ -30,6 +30,7 @@ > > static struct kobject *block_depr; > > > > DECLARE_RWSEM(bdev_lookup_sem); > > +DEFINE_MUTEX(bdev_interposer_mutex); > > > > /* for extended dynamic devt allocation, currently only one major is used */ > > #define NR_EXT_DEVT (1 << MINORBITS) > > @@ -2148,3 +2149,84 @@ static void disk_release_events(struct gendisk *disk) > > WARN_ON_ONCE(disk->ev && disk->ev->block != 1); > > kfree(disk->ev); > > } > > + > > +/** > > + * blk_interposer_attach - Attach interposer to disk > > + * @disk: target disk > > + * @interposer: block device interposer > > + * @ip_submit_bio: hook for submit_bio() > > + * > > + * Returns: > > + * -EINVAL if @interposer is NULL. > > + * -EPERM if queue is not frozen. > > + * -EBUSY if the block device already has @interposer. > > + * -EALREADY if the block device already has @interposer with same callback. > > + * > > + * Disk must be frozen by blk_mq_freeze_queue(). > > + */ > > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > > + const ip_submit_bio_t ip_submit_bio) > > +{ > > + int ret = 0; > > + > > + if (!interposer) > > + return -EINVAL; > > Is this really necessary ? If some user of this function has interposer == NULL, > that caller needs debugging... Right. I`ll add WARN_ON() at this place for debugging, but avoid the kernel drop. > > > + > > + if (!blk_mq_is_queue_frozen(disk->queue)) > > + return -EPERM; > > Why not do the queue freeze here ? In the very first version, the freeze was here. But the state when the queue is frozen is also very convenient for executing other code. In order not to create new locks, the freeze was placed in dm_interposer_attach_dev(). > > > + > > + mutex_lock(&bdev_interposer_mutex); > > + if (blk_has_interposer(disk)) { > > + if (disk->interposer->ip_submit_bio == ip_submit_bio) > > + ret = -EALREADY; > > + else > > + ret = -EBUSY; > > + goto out; > > + } > > + > > + interposer->ip_submit_bio = ip_submit_bio; > > + interposer->disk = disk; > > + > > + disk->interposer = interposer; > > +out: > > + mutex_unlock(&bdev_interposer_mutex); > > + > > + return ret; > > +} > > +EXPORT_SYMBOL_GPL(blk_interposer_attach); > > + > > +/** > > + * blk_interposer_detach - Detach interposer from disk > > + * @interposer: block device interposer > > + * @ip_submit_bio: hook for submit_bio() > > + * > > + * Disk must be frozen by blk_mq_freeze_queue(). > > + */ > > +void blk_interposer_detach(struct blk_interposer *interposer, > > + const ip_submit_bio_t ip_submit_bio) > > +{ > > The interface is weird. Why not passing the gendisk ? Passing interposer will be more safety. The kernel module (device mapper) owns the "interposer" object and passes it to the function. In addition, need to reset the interposer->disk. This is a sign that interposer has been detached. > > > + struct gendisk *disk; > > + > > + if (WARN_ON(!interposer)) > > + return; > > Same comment as above. This should not be necessary. I think it`s necessary in case detach will be called twice. It allow to avoid the kernel drop. > > > + > > + mutex_lock(&bdev_interposer_mutex); > > + > > + /* Check if the interposer is still active. */ > > + disk = interposer->disk; > > + if (WARN_ON(!disk)) > > + goto out; > > + > > + if (WARN_ON(!blk_mq_is_queue_frozen(disk->queue))) > > + goto out; > > + > > + /* Check if it is really our interposer. */ > > + if (WARN_ON(disk->interposer->ip_submit_bio != ip_submit_bio)) > > + goto out; > > + > > + disk->interposer = NULL; > > + interposer->disk = NULL; > > +out: > > + mutex_unlock(&bdev_interposer_mutex); > > +} > > +EXPORT_SYMBOL_GPL(blk_interposer_detach); > > diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h > > index d705b174d346..9d1e8c4e922e 100644 > > --- a/include/linux/blk-mq.h > > +++ b/include/linux/blk-mq.h > > @@ -525,6 +525,7 @@ void blk_freeze_queue_start(struct request_queue *q); > > void blk_mq_freeze_queue_wait(struct request_queue *q); > > int blk_mq_freeze_queue_wait_timeout(struct request_queue *q, > > unsigned long timeout); > > +bool blk_mq_is_queue_frozen(struct request_queue *q); > > > > int blk_mq_map_queues(struct blk_mq_queue_map *qmap); > > void blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, int nr_hw_queues); > > diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h > > index 866f74261b3b..6c1351d7b73f 100644 > > --- a/include/linux/blk_types.h > > +++ b/include/linux/blk_types.h > > @@ -227,7 +227,7 @@ struct bio { > > * top bits REQ_OP. Use > > * accessors. > > */ > > - unsigned short bi_flags; /* status, etc and bvec pool number */ > > + unsigned int bi_flags; /* status, etc and bvec pool number */ > > unsigned short bi_ioprio; > > unsigned short bi_write_hint; > > blk_status_t bi_status; > > @@ -304,6 +304,8 @@ enum { > > * of this bio. */ > > BIO_CGROUP_ACCT, /* has been accounted to a cgroup */ > > BIO_TRACKED, /* set if bio goes through the rq_qos path */ > > + BIO_INTERPOSED, /* bio has been interposed and can be moved to > > + * a different disk */ > > BIO_FLAG_LAST > > }; > > > > @@ -322,7 +324,7 @@ enum { > > * freed. > > */ > > #define BVEC_POOL_BITS (3) > > -#define BVEC_POOL_OFFSET (16 - BVEC_POOL_BITS) > > +#define BVEC_POOL_OFFSET (32 - BVEC_POOL_BITS) > > #define BVEC_POOL_IDX(bio) ((bio)->bi_flags >> BVEC_POOL_OFFSET) > > #if (1<< BVEC_POOL_BITS) < (BVEC_POOL_NR+1) > > # error "BVEC_POOL_BITS is too small" > > diff --git a/include/linux/genhd.h b/include/linux/genhd.h > > index 809aaa32d53c..8094af3a3db9 100644 > > --- a/include/linux/genhd.h > > +++ b/include/linux/genhd.h > > @@ -134,6 +134,14 @@ struct blk_integrity { > > unsigned char tag_size; > > }; > > > > +struct blk_interposer; > > This is not needed. Rigth. > > > +typedef void (*ip_submit_bio_t) (struct bio *bio); > > This hides the definition of the submit callback for no good reasons that I can > see. Since the callback has a simple interface, I would prefer this to be dropped. May be. But the ip_submit_bio_t type is used to describe the parameter of the blk_interposer_attach() function. I would prefer that the types of these objects are guaranteed to match. > > > + > > +struct blk_interposer { > > + ip_submit_bio_t ip_submit_bio; > > + struct gendisk *disk; > > If you fix the interface of the detach function, you should not need this field. No, I probably won't delete this field. This field is a good indicator of whether the detach has occurred or not. It allows not to detach someone else's "interposer" in case of an erroneous re-call of detach. > > > +}; > > + > > struct gendisk { > > /* major, first_minor and minors are input parameters only, > > * don't use directly. Use disk_devt() and disk_max_parts(). > > @@ -158,6 +166,7 @@ struct gendisk { > > > > const struct block_device_operations *fops; > > struct request_queue *queue; > > + struct blk_interposer *interposer; > > void *private_data; > > > > int flags; > > @@ -346,4 +355,14 @@ static inline void printk_all_partitions(void) > > } > > #endif /* CONFIG_BLOCK */ > > > > +/* > > + * block layer interposer > > + */ > > +#define blk_has_interposer(d) ((d)->interposer != NULL) > > Please make this an inline function. I prefer inline functions, but I would prefer to stay it as it defined by Hannes. In addition, such a define is useful for determining whether the kernel supports blk_interposer or not. > > > + > > +int blk_interposer_attach(struct gendisk *disk, struct blk_interposer *interposer, > > + const ip_submit_bio_t ip_submit_bio); > > +void blk_interposer_detach(struct blk_interposer *interposer, > > + const ip_submit_bio_t ip_submit_bio); > > + > > #endif /* _LINUX_GENHD_H */ > > > > > -- > Damien Le Moal > Western Digital Research > -- Sergei Shtepa Veeam Software developer. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear 2021-01-28 17:12 ` Sergei Shtepa @ 2021-01-28 17:12 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: pavel.tide, sergei.shtepa Implement a block interposer for device-mapper to attach to an existing block layer stack. Using the interposer, we can connect the dm-linear to a device with a mounted file system. changes: * the new dm_interposer structure contains blk_interposer to intercept bio from the interposed disk and interval tree of block devices on this disk. * the new interval tree for device mapper. * the dm_submit_bio_interposer_fn() function implements the bio interception logic. * the functions dm_interposer_attach_dev() & dm_interposer_detach_dev() allow to attach and detach devices to dm_interposer. * the new parameter 'noexcl' allows to create dm-linear to device with an already mounted file system. * the non_exclusive parameter in dm_target structure - it`s a sign that target device should be opened without FMODE_EXCL mode. * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to a regular block device. Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> --- drivers/md/dm-core.h | 46 +++- drivers/md/dm-ioctl.c | 39 ++++ drivers/md/dm-linear.c | 17 +- drivers/md/dm-table.c | 12 +- drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- drivers/md/dm.h | 2 +- include/linux/device-mapper.h | 7 + include/uapi/linux/dm-ioctl.h | 15 +- 8 files changed, 493 insertions(+), 28 deletions(-) diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h index 086d293c2b03..0f870b1d4be4 100644 --- a/drivers/md/dm-core.h +++ b/drivers/md/dm-core.h @@ -13,7 +13,7 @@ #include <linux/ktime.h> #include <linux/genhd.h> #include <linux/blk-mq.h> - +#include <linux/rbtree.h> #include <trace/events/block.h> #include "dm.h" @@ -109,6 +109,9 @@ struct mapped_device { bool init_tio_pdu:1; struct srcu_struct io_barrier; + + /* interposer device for remap */ + struct dm_interposed_dev *ip_dev; }; void disable_discard(struct mapped_device *md); @@ -164,6 +167,47 @@ struct dm_table { struct dm_md_mempools *mempools; }; +/* + * Interval tree for device mapper + */ +struct dm_rb_range { + struct rb_node node; + sector_t start; /* start sector of rb node */ + sector_t last; /* end sector of rb node */ + sector_t _subtree_last; /* highest sector in subtree of rb node */ +}; + +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); + +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); + +/* + * For connecting blk_interposer and dm-targets devices. + */ +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); + +struct dm_interposed_dev { + struct gendisk *disk; + struct dm_rb_range node; + void *context; + dm_interpose_bio_t dm_interpose_bio; + + atomic64_t ip_cnt; /*for debug purpose*/ +}; + +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, + sector_t ofs, sector_t len, + void *context, + dm_interpose_bio_t dm_interpose_bio_t); +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); + +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); +int dm_remap_uninstall(struct mapped_device *md); + static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) { return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index 5e306bba4375..2944d442c256 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para return r; } +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) +{ + char *donor_device_name = (char *)params; + + return dm_remap_install(md, donor_device_name); +} +static int dev_remap_finish(struct mapped_device *md) +{ + return dm_remap_uninstall(md); +} + +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) +{ + int ret = 0; + struct mapped_device *md; + void *bin_data; + struct dm_remap_param *remap_param; + + md = find_device(param); + if (!md) + return -ENXIO; + + bin_data = (void *)(param) + param->data_start; + remap_param = bin_data; + + if (remap_param->cmd == REMAP_START_CMD) + ret = dev_remap_start(md, remap_param->params); + else if (remap_param->cmd == REMAP_FINISH_CMD) + ret = dev_remap_finish(md); + else { + DMWARN("Invalid remap command, %d", remap_param->cmd); + ret = -EINVAL; + } + + dm_put(md); + return ret; +} + /* * The ioctl parameter block consists of two parts, a dm_ioctl struct * followed by a data buffer. This flag is set if the second part, @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, {DM_GET_TARGET_VERSION, 0, get_target_version}, + {DM_DEV_REMAP_CMD, 0, dev_remap}, }; if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c index 00774b5d7668..ffb8b5ca4d10 100644 --- a/drivers/md/dm-linear.c +++ b/drivers/md/dm-linear.c @@ -28,12 +28,13 @@ struct linear_c { */ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) { + fmode_t mode; struct linear_c *lc; unsigned long long tmp; char dummy; int ret; - if (argc != 2) { + if ((argc < 2) || (argc > 3)) { ti->error = "Invalid argument count"; return -EINVAL; } @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) } lc->start = tmp; - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); + ti->non_exclusive = false; + if (argc > 2) { + if (strcmp("noexcl", argv[2]) == 0) + ti->non_exclusive = true; + else if (strcmp("excl", argv[2]) == 0) + ti->non_exclusive = false; + else { + ti->error = "Invalid exclusive option"; + return -EINVAL; + } + } + + ret = dm_get_device(ti, argv[0], mode, &lc->dev); if (ret) { ti->error = "Device lookup failed"; goto bad; diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 4acf2342f7ad..f15bc2171f25 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, * device and not to touch the existing bdev field in case * it is accessed concurrently. */ -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, struct mapped_device *md) { int r; @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, old_dev = dd->dm_dev; - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, dd->dm_dev->mode | new_mode, &new_dev); if (r) return r; @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, if (!dd) return -ENOMEM; - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); + if (r) { kfree(dd); return r; } @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, list_add(&dd->list, &t->devices); goto out; - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { - r = upgrade_mode(dd, mode, t->md); + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); if (r) return r; } diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 7bac564f3faa..3b871d98b7b6 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -28,6 +28,7 @@ #include <linux/refcount.h> #include <linux/part_stat.h> #include <linux/blk-crypto.h> +#include <linux/interval_tree_generic.h> #define DM_MSG_PREFIX "core" @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; atomic_t dm_global_event_nr = ATOMIC_INIT(0); DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ + void dm_issue_global_event(void) { atomic_inc(&dm_global_event_nr); @@ -162,6 +165,26 @@ struct table_device { struct dm_dev dm_dev; }; +/* + * Device mapper`s interposer. + */ +struct dm_interposer { + struct blk_interposer blk_ip; + struct mapped_device *md; + + struct kref kref; + struct rw_semaphore ip_devs_lock; + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ +}; + +/* + * Interval tree for device mapper + */ +#define START(node) ((node)->start) +#define LAST(node) ((node)->last) +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, + START, LAST,, dm_rb); + /* * Bio-based DM's mempools' reserved IOs set by the user. */ @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) rcu_read_unlock(); } +static void dm_submit_bio_interposer_fn(struct bio *bio) +{ + struct dm_interposer *ip; + unsigned int noio_flag = 0; + sector_t start; + sector_t last; + struct dm_rb_range *node; + + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); + start = bio->bi_iter.bi_sector; + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); + + noio_flag = memalloc_noio_save(); + down_read(&ip->ip_devs_lock); + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); + while (node) { + struct dm_interposed_dev *ip_dev = + container_of(node, struct dm_interposed_dev, node); + + atomic64_inc(&ip_dev->ip_cnt); + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); + + node = dm_rb_iter_next(node, start, last); + } + up_read(&ip->ip_devs_lock); + memalloc_noio_restore(noio_flag); +} + +static void free_interposer(struct kref *kref) +{ + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); + + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); + + kfree(ip); +} + +static struct dm_interposer *new_interposer(struct gendisk *disk) +{ + int ret = 0; + struct dm_interposer *ip; + + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); + if (!ip) + return ERR_PTR(-ENOMEM); + + kref_init(&ip->kref); + init_rwsem(&ip->ip_devs_lock); + ip->ip_devs_root = RB_ROOT_CACHED; + + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); + if (ret) { + DMERR("Failed to attack blk_interposer"); + kref_put(&ip->kref, free_interposer); + return ERR_PTR(ret); + } + + return ip; +} + +static struct dm_interposer *get_interposer(struct gendisk *disk) +{ + struct dm_interposer *ip; + + if (!blk_has_interposer(disk)) + return NULL; + + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { + DMERR("Disks interposer slot already occupied."); + return ERR_PTR(-EBUSY); + } + + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); + + kref_get(&ip->kref); + return ip; +} + +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, + void *context, dm_interpose_bio_t dm_interpose_bio) +{ + sector_t start = ofs; + sector_t last = ofs + len - 1; + struct dm_interposed_dev *ip_dev = NULL; + + /* Allocate new ip_dev */ + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); + if (!ip_dev) + return NULL; + + ip_dev->disk = disk; + ip_dev->node.start = start; + ip_dev->node.last = last; + + ip_dev->context = context; + ip_dev->dm_interpose_bio = dm_interpose_bio; + + atomic64_set(&ip_dev->ip_cnt, 0); + + return ip_dev; +} + +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) +{ + kfree(ip_dev); +} + +static inline void dm_disk_freeze(struct gendisk *disk) +{ + blk_mq_freeze_queue(disk->queue); + blk_mq_quiesce_queue(disk->queue); +} + +static inline void dm_disk_unfreeze(struct gendisk *disk) +{ + blk_mq_unquiesce_queue(disk->queue); + blk_mq_unfreeze_queue(disk->queue); +} + +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) +{ + int ret = 0; + struct dm_interposer *ip = NULL; + unsigned int noio_flag = 0; + + if (!ip_dev) + return -EINVAL; + + dm_disk_freeze(ip_dev->disk); + mutex_lock(&interposer_mutex); + noio_flag = memalloc_noio_save(); + + ip = get_interposer(ip_dev->disk); + if (ip == NULL) + ip = new_interposer(ip_dev->disk); + if (IS_ERR(ip)) { + ret = PTR_ERR(ip); + goto out; + } + + /* Attach dm_interposed_dev to dm_interposer */ + down_write(&ip->ip_devs_lock); + do { + struct dm_rb_range *node; + + /* checking that ip_dev already exists for this region */ + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); + if (node) { + DMERR("Disk part form [%llu] to [%llu] already have interposer", + node->start, node->last); + + ret = -EBUSY; + break; + } + + /* insert ip_dev to ip tree */ + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); + /* increment ip reference counter */ + kref_get(&ip->kref); + } while (false); + up_write(&ip->ip_devs_lock); + + kref_put(&ip->kref, free_interposer); + +out: + memalloc_noio_restore(noio_flag); + mutex_unlock(&interposer_mutex); + dm_disk_unfreeze(ip_dev->disk); + + return ret; +} + +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) +{ + int ret = 0; + struct dm_interposer *ip = NULL; + unsigned int noio_flag = 0; + + if (!ip_dev) + return -EINVAL; + + dm_disk_freeze(ip_dev->disk); + mutex_lock(&interposer_mutex); + noio_flag = memalloc_noio_save(); + + ip = get_interposer(ip_dev->disk); + if (IS_ERR(ip)) { + ret = PTR_ERR(ip); + DMERR("Interposer not found"); + goto out; + } + if (unlikely(ip == NULL)) { + ret = -ENXIO; + DMERR("Interposer not found"); + goto out; + } + + down_write(&ip->ip_devs_lock); + do { + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); + /* the reference counter here cannot be zero */ + kref_put(&ip->kref, free_interposer); + + } while (false); + up_write(&ip->ip_devs_lock); + + /* detach and free interposer if it`s not needed */ + kref_put(&ip->kref, free_interposer); +out: + memalloc_noio_restore(noio_flag); + mutex_unlock(&interposer_mutex); + dm_disk_unfreeze(ip_dev->disk); + + return ret; +} + +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) +{ + struct mapped_device *md = context; + + /* Set acceptor device. */ + bio->bi_disk = md->disk; + + /* Remap disks offset */ + bio->bi_iter.bi_sector -= node->start; + + /* + * bio should be resubmitted. + * We can just add bio to bio_list of the current process. + * current->bio_list must be initialized when this function is called. + * If call submit_bio_noacct(), the bio will be checked twice. + */ + BUG_ON(!current->bio_list); + bio_list_add(¤t->bio_list[0], bio); +} + +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) +{ + int ret = 0; + struct block_device *donor_bdev; + fmode_t mode = FMODE_READ | FMODE_WRITE; + + DMDEBUG("Dm remap install for mapped device %s and donor device %s", + md->name, donor_device_name); + + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); + if (IS_ERR(donor_bdev)) { + DMERR("Cannot open device [%s]", donor_device_name); + return PTR_ERR(donor_bdev); + } + + do { + sector_t ofs = get_start_sect(donor_bdev); + sector_t len = bdev_nr_sectors(donor_bdev); + + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); + if (!md->ip_dev) { + ret = -ENOMEM; + break; + } + + DMDEBUG("New interposed device 0x%p", md->ip_dev); + ret = dm_interposer_attach_dev(md->ip_dev); + if (ret) { + dm_interposer_free_dev(md->ip_dev); + + md->ip_dev = NULL; + DMERR("Failed to attach dm interposer"); + break; + } + + DMDEBUG("Attached successfully."); + } while (false); + + blkdev_put(donor_bdev, mode); + + return ret; +} + +int dm_remap_uninstall(struct mapped_device *md) +{ + int ret = 0; + + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); + + if (!md->ip_dev) { + DMERR("Cannot detach dm interposer"); + return -EINVAL; + } + + ret = dm_interposer_detach_dev(md->ip_dev); + if (ret) { + DMERR("Failed to detach dm interposer"); + return ret; + } + + DMDEBUG("Detached successfully. %llu bios was interposed", + atomic64_read(&md->ip_dev->ip_cnt)); + dm_interposer_free_dev(md->ip_dev); + md->ip_dev = NULL; + + return 0; +} + static char *_dm_claim_ptr = "I belong to device-mapper"; /* * Open a table device so we can use it as a map destination. */ static int open_table_device(struct table_device *td, dev_t dev, - struct mapped_device *md) + struct mapped_device *md, bool non_exclusive) { struct block_device *bdev; - - int r; + int ret; BUG_ON(td->dm_dev.bdev); - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); - if (IS_ERR(bdev)) - return PTR_ERR(bdev); + if (non_exclusive) + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); + else + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); - r = bd_link_disk_holder(bdev, dm_disk(md)); - if (r) { - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); - return r; + if (IS_ERR(bdev)) { + ret = PTR_ERR(bdev); + if (ret != -EBUSY) + return ret; + } + + if (!non_exclusive) { + ret = bd_link_disk_holder(bdev, dm_disk(md)); + if (ret) { + blkdev_put(bdev, td->dm_dev.mode); + return ret; + } } td->dm_dev.bdev = bdev; @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md if (!td->dm_dev.bdev) return; - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); + if (td->dm_dev.mode & FMODE_EXCL) + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); + + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); + put_dax(td->dm_dev.dax_dev); td->dm_dev.bdev = NULL; td->dm_dev.dax_dev = NULL; } static struct table_device *find_table_device(struct list_head *l, dev_t dev, - fmode_t mode) + fmode_t mode, bool non_exclusive) { struct table_device *td; list_for_each_entry(td, l, list) - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) + if (td->dm_dev.bdev->bd_dev == dev && + td->dm_dev.mode == mode && + td->dm_dev.non_exclusive == non_exclusive) return td; return NULL; } -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, struct dm_dev **result) { int r; struct table_device *td; mutex_lock(&md->table_devices_lock); - td = find_table_device(&md->table_devices, dev, mode); + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); if (!td) { td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); if (!td) { @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, td->dm_dev.mode = mode; td->dm_dev.bdev = NULL; - if ((r = open_table_device(td, dev, md))) { + r = open_table_device(td, dev, md, non_exclusive); + if (r) { mutex_unlock(&md->table_devices_lock); kfree(td); return r; @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) might_sleep(); + if (md->ip_dev) { + if (dm_interposer_detach_dev(md->ip_dev)) + DMERR("Failed to detach dm interposer"); + + dm_interposer_free_dev(md->ip_dev); + md->ip_dev = NULL; + } + spin_lock(&_minor_lock); idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); set_bit(DMF_FREEING, &md->flags); diff --git a/drivers/md/dm.h b/drivers/md/dm.h index fffe1e289c53..7bf20fb2de74 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); int dm_cancel_deferred_remove(struct mapped_device *md); int dm_request_based(struct mapped_device *md); -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, struct dm_dev **result); void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h index 61a66fb8ebb3..70002363bfc0 100644 --- a/include/linux/device-mapper.h +++ b/include/linux/device-mapper.h @@ -150,6 +150,7 @@ struct dm_dev { struct block_device *bdev; struct dax_device *dax_dev; fmode_t mode; + bool non_exclusive; char name[16]; }; @@ -325,6 +326,12 @@ struct dm_target { * whether or not its underlying devices have support. */ bool discards_supported:1; + + /* + * Set if this target needs to open device without FMODE_EXCL + * mode. + */ + bool non_exclusive:1; }; void *dm_per_bio_data(struct bio *bio, size_t data_size); diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h index 4933b6b67b85..08d7dbff80f4 100644 --- a/include/uapi/linux/dm-ioctl.h +++ b/include/uapi/linux/dm-ioctl.h @@ -214,6 +214,15 @@ struct dm_target_msg { char message[0]; }; +enum { + REMAP_START_CMD = 1, + REMAP_FINISH_CMD, +}; + +struct dm_remap_param { + uint8_t cmd; + uint8_t params[0]; +}; /* * If you change this make sure you make the corresponding change * to dm-ioctl.c:lookup_ioctl() @@ -244,6 +253,7 @@ enum { DM_DEV_SET_GEOMETRY_CMD, DM_DEV_ARM_POLL_CMD, DM_GET_TARGET_VERSION_CMD, + DM_DEV_REMAP_CMD }; #define DM_IOCTL 0xfd @@ -259,6 +269,7 @@ enum { #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) @@ -272,9 +283,9 @@ enum { #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) #define DM_VERSION_MAJOR 4 -#define DM_VERSION_MINOR 43 +#define DM_VERSION_MINOR 44 #define DM_VERSION_PATCHLEVEL 0 -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" /* Status bits */ #define DM_READONLY_FLAG (1 << 0) /* In/Out */ -- 2.20.1 -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH 2/2] [dm] blk_interposer for dm-linear @ 2021-01-28 17:12 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-28 17:12 UTC (permalink / raw) To: hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: sergei.shtepa, pavel.tide Implement a block interposer for device-mapper to attach to an existing block layer stack. Using the interposer, we can connect the dm-linear to a device with a mounted file system. changes: * the new dm_interposer structure contains blk_interposer to intercept bio from the interposed disk and interval tree of block devices on this disk. * the new interval tree for device mapper. * the dm_submit_bio_interposer_fn() function implements the bio interception logic. * the functions dm_interposer_attach_dev() & dm_interposer_detach_dev() allow to attach and detach devices to dm_interposer. * the new parameter 'noexcl' allows to create dm-linear to device with an already mounted file system. * the non_exclusive parameter in dm_target structure - it`s a sign that target device should be opened without FMODE_EXCL mode. * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to a regular block device. Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> --- drivers/md/dm-core.h | 46 +++- drivers/md/dm-ioctl.c | 39 ++++ drivers/md/dm-linear.c | 17 +- drivers/md/dm-table.c | 12 +- drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- drivers/md/dm.h | 2 +- include/linux/device-mapper.h | 7 + include/uapi/linux/dm-ioctl.h | 15 +- 8 files changed, 493 insertions(+), 28 deletions(-) diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h index 086d293c2b03..0f870b1d4be4 100644 --- a/drivers/md/dm-core.h +++ b/drivers/md/dm-core.h @@ -13,7 +13,7 @@ #include <linux/ktime.h> #include <linux/genhd.h> #include <linux/blk-mq.h> - +#include <linux/rbtree.h> #include <trace/events/block.h> #include "dm.h" @@ -109,6 +109,9 @@ struct mapped_device { bool init_tio_pdu:1; struct srcu_struct io_barrier; + + /* interposer device for remap */ + struct dm_interposed_dev *ip_dev; }; void disable_discard(struct mapped_device *md); @@ -164,6 +167,47 @@ struct dm_table { struct dm_md_mempools *mempools; }; +/* + * Interval tree for device mapper + */ +struct dm_rb_range { + struct rb_node node; + sector_t start; /* start sector of rb node */ + sector_t last; /* end sector of rb node */ + sector_t _subtree_last; /* highest sector in subtree of rb node */ +}; + +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); + +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); + +/* + * For connecting blk_interposer and dm-targets devices. + */ +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); + +struct dm_interposed_dev { + struct gendisk *disk; + struct dm_rb_range node; + void *context; + dm_interpose_bio_t dm_interpose_bio; + + atomic64_t ip_cnt; /*for debug purpose*/ +}; + +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, + sector_t ofs, sector_t len, + void *context, + dm_interpose_bio_t dm_interpose_bio_t); +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); + +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); +int dm_remap_uninstall(struct mapped_device *md); + static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) { return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index 5e306bba4375..2944d442c256 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para return r; } +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) +{ + char *donor_device_name = (char *)params; + + return dm_remap_install(md, donor_device_name); +} +static int dev_remap_finish(struct mapped_device *md) +{ + return dm_remap_uninstall(md); +} + +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) +{ + int ret = 0; + struct mapped_device *md; + void *bin_data; + struct dm_remap_param *remap_param; + + md = find_device(param); + if (!md) + return -ENXIO; + + bin_data = (void *)(param) + param->data_start; + remap_param = bin_data; + + if (remap_param->cmd == REMAP_START_CMD) + ret = dev_remap_start(md, remap_param->params); + else if (remap_param->cmd == REMAP_FINISH_CMD) + ret = dev_remap_finish(md); + else { + DMWARN("Invalid remap command, %d", remap_param->cmd); + ret = -EINVAL; + } + + dm_put(md); + return ret; +} + /* * The ioctl parameter block consists of two parts, a dm_ioctl struct * followed by a data buffer. This flag is set if the second part, @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, {DM_GET_TARGET_VERSION, 0, get_target_version}, + {DM_DEV_REMAP_CMD, 0, dev_remap}, }; if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c index 00774b5d7668..ffb8b5ca4d10 100644 --- a/drivers/md/dm-linear.c +++ b/drivers/md/dm-linear.c @@ -28,12 +28,13 @@ struct linear_c { */ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) { + fmode_t mode; struct linear_c *lc; unsigned long long tmp; char dummy; int ret; - if (argc != 2) { + if ((argc < 2) || (argc > 3)) { ti->error = "Invalid argument count"; return -EINVAL; } @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) } lc->start = tmp; - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); + ti->non_exclusive = false; + if (argc > 2) { + if (strcmp("noexcl", argv[2]) == 0) + ti->non_exclusive = true; + else if (strcmp("excl", argv[2]) == 0) + ti->non_exclusive = false; + else { + ti->error = "Invalid exclusive option"; + return -EINVAL; + } + } + + ret = dm_get_device(ti, argv[0], mode, &lc->dev); if (ret) { ti->error = "Device lookup failed"; goto bad; diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 4acf2342f7ad..f15bc2171f25 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, * device and not to touch the existing bdev field in case * it is accessed concurrently. */ -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, struct mapped_device *md) { int r; @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, old_dev = dd->dm_dev; - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, dd->dm_dev->mode | new_mode, &new_dev); if (r) return r; @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, if (!dd) return -ENOMEM; - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); + if (r) { kfree(dd); return r; } @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, list_add(&dd->list, &t->devices); goto out; - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { - r = upgrade_mode(dd, mode, t->md); + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); if (r) return r; } diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 7bac564f3faa..3b871d98b7b6 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -28,6 +28,7 @@ #include <linux/refcount.h> #include <linux/part_stat.h> #include <linux/blk-crypto.h> +#include <linux/interval_tree_generic.h> #define DM_MSG_PREFIX "core" @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; atomic_t dm_global_event_nr = ATOMIC_INIT(0); DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ + void dm_issue_global_event(void) { atomic_inc(&dm_global_event_nr); @@ -162,6 +165,26 @@ struct table_device { struct dm_dev dm_dev; }; +/* + * Device mapper`s interposer. + */ +struct dm_interposer { + struct blk_interposer blk_ip; + struct mapped_device *md; + + struct kref kref; + struct rw_semaphore ip_devs_lock; + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ +}; + +/* + * Interval tree for device mapper + */ +#define START(node) ((node)->start) +#define LAST(node) ((node)->last) +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, + START, LAST,, dm_rb); + /* * Bio-based DM's mempools' reserved IOs set by the user. */ @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) rcu_read_unlock(); } +static void dm_submit_bio_interposer_fn(struct bio *bio) +{ + struct dm_interposer *ip; + unsigned int noio_flag = 0; + sector_t start; + sector_t last; + struct dm_rb_range *node; + + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); + start = bio->bi_iter.bi_sector; + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); + + noio_flag = memalloc_noio_save(); + down_read(&ip->ip_devs_lock); + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); + while (node) { + struct dm_interposed_dev *ip_dev = + container_of(node, struct dm_interposed_dev, node); + + atomic64_inc(&ip_dev->ip_cnt); + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); + + node = dm_rb_iter_next(node, start, last); + } + up_read(&ip->ip_devs_lock); + memalloc_noio_restore(noio_flag); +} + +static void free_interposer(struct kref *kref) +{ + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); + + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); + + kfree(ip); +} + +static struct dm_interposer *new_interposer(struct gendisk *disk) +{ + int ret = 0; + struct dm_interposer *ip; + + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); + if (!ip) + return ERR_PTR(-ENOMEM); + + kref_init(&ip->kref); + init_rwsem(&ip->ip_devs_lock); + ip->ip_devs_root = RB_ROOT_CACHED; + + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); + if (ret) { + DMERR("Failed to attack blk_interposer"); + kref_put(&ip->kref, free_interposer); + return ERR_PTR(ret); + } + + return ip; +} + +static struct dm_interposer *get_interposer(struct gendisk *disk) +{ + struct dm_interposer *ip; + + if (!blk_has_interposer(disk)) + return NULL; + + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { + DMERR("Disks interposer slot already occupied."); + return ERR_PTR(-EBUSY); + } + + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); + + kref_get(&ip->kref); + return ip; +} + +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, + void *context, dm_interpose_bio_t dm_interpose_bio) +{ + sector_t start = ofs; + sector_t last = ofs + len - 1; + struct dm_interposed_dev *ip_dev = NULL; + + /* Allocate new ip_dev */ + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); + if (!ip_dev) + return NULL; + + ip_dev->disk = disk; + ip_dev->node.start = start; + ip_dev->node.last = last; + + ip_dev->context = context; + ip_dev->dm_interpose_bio = dm_interpose_bio; + + atomic64_set(&ip_dev->ip_cnt, 0); + + return ip_dev; +} + +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) +{ + kfree(ip_dev); +} + +static inline void dm_disk_freeze(struct gendisk *disk) +{ + blk_mq_freeze_queue(disk->queue); + blk_mq_quiesce_queue(disk->queue); +} + +static inline void dm_disk_unfreeze(struct gendisk *disk) +{ + blk_mq_unquiesce_queue(disk->queue); + blk_mq_unfreeze_queue(disk->queue); +} + +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) +{ + int ret = 0; + struct dm_interposer *ip = NULL; + unsigned int noio_flag = 0; + + if (!ip_dev) + return -EINVAL; + + dm_disk_freeze(ip_dev->disk); + mutex_lock(&interposer_mutex); + noio_flag = memalloc_noio_save(); + + ip = get_interposer(ip_dev->disk); + if (ip == NULL) + ip = new_interposer(ip_dev->disk); + if (IS_ERR(ip)) { + ret = PTR_ERR(ip); + goto out; + } + + /* Attach dm_interposed_dev to dm_interposer */ + down_write(&ip->ip_devs_lock); + do { + struct dm_rb_range *node; + + /* checking that ip_dev already exists for this region */ + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); + if (node) { + DMERR("Disk part form [%llu] to [%llu] already have interposer", + node->start, node->last); + + ret = -EBUSY; + break; + } + + /* insert ip_dev to ip tree */ + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); + /* increment ip reference counter */ + kref_get(&ip->kref); + } while (false); + up_write(&ip->ip_devs_lock); + + kref_put(&ip->kref, free_interposer); + +out: + memalloc_noio_restore(noio_flag); + mutex_unlock(&interposer_mutex); + dm_disk_unfreeze(ip_dev->disk); + + return ret; +} + +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) +{ + int ret = 0; + struct dm_interposer *ip = NULL; + unsigned int noio_flag = 0; + + if (!ip_dev) + return -EINVAL; + + dm_disk_freeze(ip_dev->disk); + mutex_lock(&interposer_mutex); + noio_flag = memalloc_noio_save(); + + ip = get_interposer(ip_dev->disk); + if (IS_ERR(ip)) { + ret = PTR_ERR(ip); + DMERR("Interposer not found"); + goto out; + } + if (unlikely(ip == NULL)) { + ret = -ENXIO; + DMERR("Interposer not found"); + goto out; + } + + down_write(&ip->ip_devs_lock); + do { + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); + /* the reference counter here cannot be zero */ + kref_put(&ip->kref, free_interposer); + + } while (false); + up_write(&ip->ip_devs_lock); + + /* detach and free interposer if it`s not needed */ + kref_put(&ip->kref, free_interposer); +out: + memalloc_noio_restore(noio_flag); + mutex_unlock(&interposer_mutex); + dm_disk_unfreeze(ip_dev->disk); + + return ret; +} + +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) +{ + struct mapped_device *md = context; + + /* Set acceptor device. */ + bio->bi_disk = md->disk; + + /* Remap disks offset */ + bio->bi_iter.bi_sector -= node->start; + + /* + * bio should be resubmitted. + * We can just add bio to bio_list of the current process. + * current->bio_list must be initialized when this function is called. + * If call submit_bio_noacct(), the bio will be checked twice. + */ + BUG_ON(!current->bio_list); + bio_list_add(¤t->bio_list[0], bio); +} + +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) +{ + int ret = 0; + struct block_device *donor_bdev; + fmode_t mode = FMODE_READ | FMODE_WRITE; + + DMDEBUG("Dm remap install for mapped device %s and donor device %s", + md->name, donor_device_name); + + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); + if (IS_ERR(donor_bdev)) { + DMERR("Cannot open device [%s]", donor_device_name); + return PTR_ERR(donor_bdev); + } + + do { + sector_t ofs = get_start_sect(donor_bdev); + sector_t len = bdev_nr_sectors(donor_bdev); + + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); + if (!md->ip_dev) { + ret = -ENOMEM; + break; + } + + DMDEBUG("New interposed device 0x%p", md->ip_dev); + ret = dm_interposer_attach_dev(md->ip_dev); + if (ret) { + dm_interposer_free_dev(md->ip_dev); + + md->ip_dev = NULL; + DMERR("Failed to attach dm interposer"); + break; + } + + DMDEBUG("Attached successfully."); + } while (false); + + blkdev_put(donor_bdev, mode); + + return ret; +} + +int dm_remap_uninstall(struct mapped_device *md) +{ + int ret = 0; + + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); + + if (!md->ip_dev) { + DMERR("Cannot detach dm interposer"); + return -EINVAL; + } + + ret = dm_interposer_detach_dev(md->ip_dev); + if (ret) { + DMERR("Failed to detach dm interposer"); + return ret; + } + + DMDEBUG("Detached successfully. %llu bios was interposed", + atomic64_read(&md->ip_dev->ip_cnt)); + dm_interposer_free_dev(md->ip_dev); + md->ip_dev = NULL; + + return 0; +} + static char *_dm_claim_ptr = "I belong to device-mapper"; /* * Open a table device so we can use it as a map destination. */ static int open_table_device(struct table_device *td, dev_t dev, - struct mapped_device *md) + struct mapped_device *md, bool non_exclusive) { struct block_device *bdev; - - int r; + int ret; BUG_ON(td->dm_dev.bdev); - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); - if (IS_ERR(bdev)) - return PTR_ERR(bdev); + if (non_exclusive) + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); + else + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); - r = bd_link_disk_holder(bdev, dm_disk(md)); - if (r) { - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); - return r; + if (IS_ERR(bdev)) { + ret = PTR_ERR(bdev); + if (ret != -EBUSY) + return ret; + } + + if (!non_exclusive) { + ret = bd_link_disk_holder(bdev, dm_disk(md)); + if (ret) { + blkdev_put(bdev, td->dm_dev.mode); + return ret; + } } td->dm_dev.bdev = bdev; @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md if (!td->dm_dev.bdev) return; - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); + if (td->dm_dev.mode & FMODE_EXCL) + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); + + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); + put_dax(td->dm_dev.dax_dev); td->dm_dev.bdev = NULL; td->dm_dev.dax_dev = NULL; } static struct table_device *find_table_device(struct list_head *l, dev_t dev, - fmode_t mode) + fmode_t mode, bool non_exclusive) { struct table_device *td; list_for_each_entry(td, l, list) - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) + if (td->dm_dev.bdev->bd_dev == dev && + td->dm_dev.mode == mode && + td->dm_dev.non_exclusive == non_exclusive) return td; return NULL; } -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, struct dm_dev **result) { int r; struct table_device *td; mutex_lock(&md->table_devices_lock); - td = find_table_device(&md->table_devices, dev, mode); + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); if (!td) { td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); if (!td) { @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, td->dm_dev.mode = mode; td->dm_dev.bdev = NULL; - if ((r = open_table_device(td, dev, md))) { + r = open_table_device(td, dev, md, non_exclusive); + if (r) { mutex_unlock(&md->table_devices_lock); kfree(td); return r; @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) might_sleep(); + if (md->ip_dev) { + if (dm_interposer_detach_dev(md->ip_dev)) + DMERR("Failed to detach dm interposer"); + + dm_interposer_free_dev(md->ip_dev); + md->ip_dev = NULL; + } + spin_lock(&_minor_lock); idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); set_bit(DMF_FREEING, &md->flags); diff --git a/drivers/md/dm.h b/drivers/md/dm.h index fffe1e289c53..7bf20fb2de74 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); int dm_cancel_deferred_remove(struct mapped_device *md); int dm_request_based(struct mapped_device *md); -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, struct dm_dev **result); void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h index 61a66fb8ebb3..70002363bfc0 100644 --- a/include/linux/device-mapper.h +++ b/include/linux/device-mapper.h @@ -150,6 +150,7 @@ struct dm_dev { struct block_device *bdev; struct dax_device *dax_dev; fmode_t mode; + bool non_exclusive; char name[16]; }; @@ -325,6 +326,12 @@ struct dm_target { * whether or not its underlying devices have support. */ bool discards_supported:1; + + /* + * Set if this target needs to open device without FMODE_EXCL + * mode. + */ + bool non_exclusive:1; }; void *dm_per_bio_data(struct bio *bio, size_t data_size); diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h index 4933b6b67b85..08d7dbff80f4 100644 --- a/include/uapi/linux/dm-ioctl.h +++ b/include/uapi/linux/dm-ioctl.h @@ -214,6 +214,15 @@ struct dm_target_msg { char message[0]; }; +enum { + REMAP_START_CMD = 1, + REMAP_FINISH_CMD, +}; + +struct dm_remap_param { + uint8_t cmd; + uint8_t params[0]; +}; /* * If you change this make sure you make the corresponding change * to dm-ioctl.c:lookup_ioctl() @@ -244,6 +253,7 @@ enum { DM_DEV_SET_GEOMETRY_CMD, DM_DEV_ARM_POLL_CMD, DM_GET_TARGET_VERSION_CMD, + DM_DEV_REMAP_CMD }; #define DM_IOCTL 0xfd @@ -259,6 +269,7 @@ enum { #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) @@ -272,9 +283,9 @@ enum { #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) #define DM_VERSION_MAJOR 4 -#define DM_VERSION_MINOR 43 +#define DM_VERSION_MINOR 44 #define DM_VERSION_PATCHLEVEL 0 -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" /* Status bits */ #define DM_READONLY_FLAG (1 << 0) /* In/Out */ -- 2.20.1 ^ permalink raw reply related [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear 2021-01-28 17:12 ` Sergei Shtepa @ 2021-01-29 1:46 ` Damien Le Moal -1 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 1:46 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > Implement a block interposer for device-mapper to attach > to an existing block layer stack. Using the interposer, > we can connect the dm-linear to a device with a mounted > file system. > > changes: > * the new dm_interposer structure contains blk_interposer > to intercept bio from the interposed disk and interval tree > of block devices on this disk. > * the new interval tree for device mapper. > * the dm_submit_bio_interposer_fn() function implements > the bio interception logic. > * the functions dm_interposer_attach_dev() & > dm_interposer_detach_dev() allow to attach and detach devices > to dm_interposer. > * the new parameter 'noexcl' allows to create dm-linear to device > with an already mounted file system. > * the non_exclusive parameter in dm_target structure - it`s a sign > that target device should be opened without FMODE_EXCL mode. > * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to > a regular block device. Same comment about changelog as in the previous patch. > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > --- > drivers/md/dm-core.h | 46 +++- > drivers/md/dm-ioctl.c | 39 ++++ > drivers/md/dm-linear.c | 17 +- > drivers/md/dm-table.c | 12 +- > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > drivers/md/dm.h | 2 +- > include/linux/device-mapper.h | 7 + > include/uapi/linux/dm-ioctl.h | 15 +- > 8 files changed, 493 insertions(+), 28 deletions(-) > > diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h > index 086d293c2b03..0f870b1d4be4 100644 > --- a/drivers/md/dm-core.h > +++ b/drivers/md/dm-core.h > @@ -13,7 +13,7 @@ > #include <linux/ktime.h> > #include <linux/genhd.h> > #include <linux/blk-mq.h> > - whiteline change. > +#include <linux/rbtree.h> > #include <trace/events/block.h> > > #include "dm.h" > @@ -109,6 +109,9 @@ struct mapped_device { > bool init_tio_pdu:1; > > struct srcu_struct io_barrier; > + > + /* interposer device for remap */ > + struct dm_interposed_dev *ip_dev; > }; > > void disable_discard(struct mapped_device *md); > @@ -164,6 +167,47 @@ struct dm_table { > struct dm_md_mempools *mempools; > }; > > +/* > + * Interval tree for device mapper > + */ > +struct dm_rb_range { > + struct rb_node node; > + sector_t start; /* start sector of rb node */ > + sector_t last; /* end sector of rb node */ > + sector_t _subtree_last; /* highest sector in subtree of rb node */ > +}; > + > +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); > +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); > + > +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); > +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); > + > +/* > + * For connecting blk_interposer and dm-targets devices. Is this comment about the callback or the structure ? I think the latter, so it is in the worng place. Please also add a comment for the callback definition explaining what it should be doing. > + */ > +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); > + > +struct dm_interposed_dev { > + struct gendisk *disk; > + struct dm_rb_range node; > + void *context; > + dm_interpose_bio_t dm_interpose_bio; > + > + atomic64_t ip_cnt; /*for debug purpose*/ > +}; > + > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, > + sector_t ofs, sector_t len, > + void *context, > + dm_interpose_bio_t dm_interpose_bio_t); > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); > + > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); > +int dm_remap_uninstall(struct mapped_device *md); > + > static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) > { > return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; > diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c > index 5e306bba4375..2944d442c256 100644 > --- a/drivers/md/dm-ioctl.c > +++ b/drivers/md/dm-ioctl.c > @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para > return r; > } > > +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) > +{ > + char *donor_device_name = (char *)params; > + > + return dm_remap_install(md, donor_device_name); > +} > +static int dev_remap_finish(struct mapped_device *md) > +{ > + return dm_remap_uninstall(md); > +} > + > +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) > +{ > + int ret = 0; > + struct mapped_device *md; > + void *bin_data; > + struct dm_remap_param *remap_param; > + > + md = find_device(param); > + if (!md) > + return -ENXIO; > + > + bin_data = (void *)(param) + param->data_start; > + remap_param = bin_data; > + > + if (remap_param->cmd == REMAP_START_CMD) > + ret = dev_remap_start(md, remap_param->params); > + else if (remap_param->cmd == REMAP_FINISH_CMD) > + ret = dev_remap_finish(md); > + else { > + DMWARN("Invalid remap command, %d", remap_param->cmd); > + ret = -EINVAL; > + } > + > + dm_put(md); > + return ret; > +} > + > /* > * The ioctl parameter block consists of two parts, a dm_ioctl struct > * followed by a data buffer. This flag is set if the second part, > @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) > {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, > {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, > {DM_GET_TARGET_VERSION, 0, get_target_version}, > + {DM_DEV_REMAP_CMD, 0, dev_remap}, > }; > > if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) > diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c > index 00774b5d7668..ffb8b5ca4d10 100644 > --- a/drivers/md/dm-linear.c > +++ b/drivers/md/dm-linear.c > @@ -28,12 +28,13 @@ struct linear_c { > */ > static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > { > + fmode_t mode; > struct linear_c *lc; > unsigned long long tmp; > char dummy; > int ret; > > - if (argc != 2) { > + if ((argc < 2) || (argc > 3)) { > ti->error = "Invalid argument count"; > return -EINVAL; > } > @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > } > lc->start = tmp; > > - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); > + ti->non_exclusive = false; > + if (argc > 2) { > + if (strcmp("noexcl", argv[2]) == 0) > + ti->non_exclusive = true; > + else if (strcmp("excl", argv[2]) == 0) > + ti->non_exclusive = false; It already is false. > + else { > + ti->error = "Invalid exclusive option"; > + return -EINVAL; > + } > + } > + > + ret = dm_get_device(ti, argv[0], mode, &lc->dev); Where is mode initialized ? Why remove dm_table_get_mode(ti->table) ? > if (ret) { > ti->error = "Device lookup failed"; > goto bad; I would prefer to see this change to dm-linear in its own patch, following this one, with a clear explanation in the commit message how this change relates to interposer since the explanation for this "exclusive" change is nowhere to be seen. Also please check if there is a file describing dm-linear options under Documentation/ (I can't remember if there is one). If there is one, it will need to be updated too. > diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c > index 4acf2342f7ad..f15bc2171f25 100644 > --- a/drivers/md/dm-table.c > +++ b/drivers/md/dm-table.c > @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, > * device and not to touch the existing bdev field in case > * it is accessed concurrently. > */ > -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, > struct mapped_device *md) > { > int r; > @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > old_dev = dd->dm_dev; > > - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, > + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, > dd->dm_dev->mode | new_mode, &new_dev); > if (r) > return r; > @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > if (!dd) > return -ENOMEM; > > - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { > + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); > + if (r) { > kfree(dd); > return r; > } > @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > list_add(&dd->list, &t->devices); > goto out; > > - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { > - r = upgrade_mode(dd, mode, t->md); > + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && > + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { > + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); > if (r) > return r; > } > diff --git a/drivers/md/dm.c b/drivers/md/dm.c > index 7bac564f3faa..3b871d98b7b6 100644 > --- a/drivers/md/dm.c > +++ b/drivers/md/dm.c > @@ -28,6 +28,7 @@ > #include <linux/refcount.h> > #include <linux/part_stat.h> > #include <linux/blk-crypto.h> > +#include <linux/interval_tree_generic.h> > > #define DM_MSG_PREFIX "core" > > @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; > atomic_t dm_global_event_nr = ATOMIC_INIT(0); > DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); > > +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ Why not dm_interposer_mutex as the name ? And the comment is not very useful: a mutex is always for synchronizing :) > + > void dm_issue_global_event(void) > { > atomic_inc(&dm_global_event_nr); > @@ -162,6 +165,26 @@ struct table_device { > struct dm_dev dm_dev; > }; > > +/* > + * Device mapper`s interposer. > + */ > +struct dm_interposer { > + struct blk_interposer blk_ip; > + struct mapped_device *md; > + > + struct kref kref; > + struct rw_semaphore ip_devs_lock; > + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ > +}; > + > +/* > + * Interval tree for device mapper > + */ > +#define START(node) ((node)->start) > +#define LAST(node) ((node)->last) > +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, > + START, LAST,, dm_rb); > + > /* > * Bio-based DM's mempools' reserved IOs set by the user. > */ > @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) > rcu_read_unlock(); > } > > +static void dm_submit_bio_interposer_fn(struct bio *bio) > +{ > + struct dm_interposer *ip; > + unsigned int noio_flag = 0; > + sector_t start; > + sector_t last; > + struct dm_rb_range *node; > + > + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); > + start = bio->bi_iter.bi_sector; > + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); > + > + noio_flag = memalloc_noio_save(); > + down_read(&ip->ip_devs_lock); > + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); > + while (node) { > + struct dm_interposed_dev *ip_dev = > + container_of(node, struct dm_interposed_dev, node); > + > + atomic64_inc(&ip_dev->ip_cnt); > + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); > + > + node = dm_rb_iter_next(node, start, last); > + } > + up_read(&ip->ip_devs_lock); > + memalloc_noio_restore(noio_flag); > +} > + > +static void free_interposer(struct kref *kref) > +{ > + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); > + > + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); No queue freeze ? > + > + kfree(ip); > +} > + > +static struct dm_interposer *new_interposer(struct gendisk *disk) > +{ > + int ret = 0; > + struct dm_interposer *ip; > + > + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); > + if (!ip) > + return ERR_PTR(-ENOMEM); > + > + kref_init(&ip->kref); > + init_rwsem(&ip->ip_devs_lock); > + ip->ip_devs_root = RB_ROOT_CACHED; > + > + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); No queue freeze ? > + if (ret) { > + DMERR("Failed to attack blk_interposer"); > + kref_put(&ip->kref, free_interposer); > + return ERR_PTR(ret); > + } > + > + return ip; > +} > + > +static struct dm_interposer *get_interposer(struct gendisk *disk) > +{ > + struct dm_interposer *ip; > + > + if (!blk_has_interposer(disk)) > + return NULL; > + > + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { > + DMERR("Disks interposer slot already occupied."); > + return ERR_PTR(-EBUSY); This is weird... If there is an interposer, why not get a ref on that one. That is what the function name suggests at least. > + } > + > + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); > + > + kref_get(&ip->kref); > + return ip; > +} > + > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, > + void *context, dm_interpose_bio_t dm_interpose_bio) > +{ > + sector_t start = ofs; > + sector_t last = ofs + len - 1; > + struct dm_interposed_dev *ip_dev = NULL; > + > + /* Allocate new ip_dev */ > + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); > + if (!ip_dev) > + return NULL; > + > + ip_dev->disk = disk; > + ip_dev->node.start = start; > + ip_dev->node.last = last; > + > + ip_dev->context = context; > + ip_dev->dm_interpose_bio = dm_interpose_bio; > + > + atomic64_set(&ip_dev->ip_cnt, 0); > + > + return ip_dev; > +} > + > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) > +{ > + kfree(ip_dev); > +} Make this inline may be ? > + > +static inline void dm_disk_freeze(struct gendisk *disk) > +{ > + blk_mq_freeze_queue(disk->queue); > + blk_mq_quiesce_queue(disk->queue); I think you can replace this with blk_mq_freeze_queue_wait(). > +} > + > +static inline void dm_disk_unfreeze(struct gendisk *disk) > +{ > + blk_mq_unquiesce_queue(disk->queue); > + blk_mq_unfreeze_queue(disk->queue); > +} > + > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) > +{ > + int ret = 0; > + struct dm_interposer *ip = NULL; > + unsigned int noio_flag = 0; > + > + if (!ip_dev) > + return -EINVAL; > + > + dm_disk_freeze(ip_dev->disk); > + mutex_lock(&interposer_mutex); > + noio_flag = memalloc_noio_save(); > + > + ip = get_interposer(ip_dev->disk); > + if (ip == NULL) > + ip = new_interposer(ip_dev->disk); > + if (IS_ERR(ip)) { > + ret = PTR_ERR(ip); > + goto out; > + } > + > + /* Attach dm_interposed_dev to dm_interposer */ > + down_write(&ip->ip_devs_lock); > + do { > + struct dm_rb_range *node; > + > + /* checking that ip_dev already exists for this region */ > + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); > + if (node) { > + DMERR("Disk part form [%llu] to [%llu] already have interposer", > + node->start, node->last); > + > + ret = -EBUSY; > + break; > + } > + > + /* insert ip_dev to ip tree */ > + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); > + /* increment ip reference counter */ > + kref_get(&ip->kref); > + } while (false); > + up_write(&ip->ip_devs_lock); > + > + kref_put(&ip->kref, free_interposer); > + > +out: > + memalloc_noio_restore(noio_flag); > + mutex_unlock(&interposer_mutex); > + dm_disk_unfreeze(ip_dev->disk); > + > + return ret; > +} > + > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) > +{ > + int ret = 0; > + struct dm_interposer *ip = NULL; > + unsigned int noio_flag = 0; > + > + if (!ip_dev) > + return -EINVAL; > + > + dm_disk_freeze(ip_dev->disk); > + mutex_lock(&interposer_mutex); > + noio_flag = memalloc_noio_save(); > + > + ip = get_interposer(ip_dev->disk); > + if (IS_ERR(ip)) { > + ret = PTR_ERR(ip); > + DMERR("Interposer not found"); > + goto out; > + } > + if (unlikely(ip == NULL)) { > + ret = -ENXIO; > + DMERR("Interposer not found"); > + goto out; > + } > + > + down_write(&ip->ip_devs_lock); > + do { > + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); > + /* the reference counter here cannot be zero */ > + kref_put(&ip->kref, free_interposer); > + > + } while (false); > + up_write(&ip->ip_devs_lock); > + > + /* detach and free interposer if it`s not needed */ s/`/'/ > + kref_put(&ip->kref, free_interposer); > +out: > + memalloc_noio_restore(noio_flag); > + mutex_unlock(&interposer_mutex); > + dm_disk_unfreeze(ip_dev->disk); > + > + return ret; > +} > + > +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) > +{ > + struct mapped_device *md = context; > + > + /* Set acceptor device. */ > + bio->bi_disk = md->disk; > + > + /* Remap disks offset */ > + bio->bi_iter.bi_sector -= node->start; > + > + /* > + * bio should be resubmitted. > + * We can just add bio to bio_list of the current process. > + * current->bio_list must be initialized when this function is called. > + * If call submit_bio_noacct(), the bio will be checked twice. > + */ > + BUG_ON(!current->bio_list); > + bio_list_add(¤t->bio_list[0], bio); > +} > + > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) > +{ > + int ret = 0; > + struct block_device *donor_bdev; > + fmode_t mode = FMODE_READ | FMODE_WRITE; > + > + DMDEBUG("Dm remap install for mapped device %s and donor device %s", > + md->name, donor_device_name); > + > + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); > + if (IS_ERR(donor_bdev)) { > + DMERR("Cannot open device [%s]", donor_device_name); > + return PTR_ERR(donor_bdev); > + } > + > + do { > + sector_t ofs = get_start_sect(donor_bdev); > + sector_t len = bdev_nr_sectors(donor_bdev); > + > + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); > + if (!md->ip_dev) { > + ret = -ENOMEM; > + break; > + } > + > + DMDEBUG("New interposed device 0x%p", md->ip_dev); > + ret = dm_interposer_attach_dev(md->ip_dev); > + if (ret) { > + dm_interposer_free_dev(md->ip_dev); > + > + md->ip_dev = NULL; > + DMERR("Failed to attach dm interposer"); > + break; > + } > + > + DMDEBUG("Attached successfully."); > + } while (false); > + > + blkdev_put(donor_bdev, mode); > + > + return ret; > +} > + > +int dm_remap_uninstall(struct mapped_device *md) > +{ > + int ret = 0; > + > + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); > + > + if (!md->ip_dev) { > + DMERR("Cannot detach dm interposer"); > + return -EINVAL; > + } > + > + ret = dm_interposer_detach_dev(md->ip_dev); > + if (ret) { > + DMERR("Failed to detach dm interposer"); > + return ret; > + } > + > + DMDEBUG("Detached successfully. %llu bios was interposed", > + atomic64_read(&md->ip_dev->ip_cnt)); > + dm_interposer_free_dev(md->ip_dev); > + md->ip_dev = NULL; > + > + return 0; > +} > + > static char *_dm_claim_ptr = "I belong to device-mapper"; > > /* > * Open a table device so we can use it as a map destination. > */ > static int open_table_device(struct table_device *td, dev_t dev, > - struct mapped_device *md) > + struct mapped_device *md, bool non_exclusive) > { > struct block_device *bdev; > - > - int r; > + int ret; > > BUG_ON(td->dm_dev.bdev); > > - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > - if (IS_ERR(bdev)) > - return PTR_ERR(bdev); > + if (non_exclusive) > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); > + else > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > - r = bd_link_disk_holder(bdev, dm_disk(md)); > - if (r) { > - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); > - return r; > + if (IS_ERR(bdev)) { > + ret = PTR_ERR(bdev); > + if (ret != -EBUSY) > + return ret; > + } > + > + if (!non_exclusive) { > + ret = bd_link_disk_holder(bdev, dm_disk(md)); > + if (ret) { > + blkdev_put(bdev, td->dm_dev.mode); > + return ret; > + } > } > > td->dm_dev.bdev = bdev; > @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md > if (!td->dm_dev.bdev) > return; > > - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); > + if (td->dm_dev.mode & FMODE_EXCL) > + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > + > + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); > + > put_dax(td->dm_dev.dax_dev); > td->dm_dev.bdev = NULL; > td->dm_dev.dax_dev = NULL; > } > > static struct table_device *find_table_device(struct list_head *l, dev_t dev, > - fmode_t mode) > + fmode_t mode, bool non_exclusive) > { > struct table_device *td; > > list_for_each_entry(td, l, list) > - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) > + if (td->dm_dev.bdev->bd_dev == dev && > + td->dm_dev.mode == mode && > + td->dm_dev.non_exclusive == non_exclusive) > return td; > > return NULL; > } > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > struct dm_dev **result) > { > int r; > struct table_device *td; > > mutex_lock(&md->table_devices_lock); > - td = find_table_device(&md->table_devices, dev, mode); > + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); > if (!td) { > td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); > if (!td) { > @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > td->dm_dev.mode = mode; > td->dm_dev.bdev = NULL; > > - if ((r = open_table_device(td, dev, md))) { > + r = open_table_device(td, dev, md, non_exclusive); > + if (r) { > mutex_unlock(&md->table_devices_lock); > kfree(td); > return r; > @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) > > might_sleep(); > > + if (md->ip_dev) { > + if (dm_interposer_detach_dev(md->ip_dev)) > + DMERR("Failed to detach dm interposer"); > + > + dm_interposer_free_dev(md->ip_dev); > + md->ip_dev = NULL; > + } > + > spin_lock(&_minor_lock); > idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); > set_bit(DMF_FREEING, &md->flags); > diff --git a/drivers/md/dm.h b/drivers/md/dm.h > index fffe1e289c53..7bf20fb2de74 100644 > --- a/drivers/md/dm.h > +++ b/drivers/md/dm.h > @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); > int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); > int dm_cancel_deferred_remove(struct mapped_device *md); > int dm_request_based(struct mapped_device *md); > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > struct dm_dev **result); > void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); > > diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h > index 61a66fb8ebb3..70002363bfc0 100644 > --- a/include/linux/device-mapper.h > +++ b/include/linux/device-mapper.h > @@ -150,6 +150,7 @@ struct dm_dev { > struct block_device *bdev; > struct dax_device *dax_dev; > fmode_t mode; > + bool non_exclusive; > char name[16]; > }; > > @@ -325,6 +326,12 @@ struct dm_target { > * whether or not its underlying devices have support. > */ > bool discards_supported:1; > + > + /* > + * Set if this target needs to open device without FMODE_EXCL > + * mode. > + */ > + bool non_exclusive:1; > }; > > void *dm_per_bio_data(struct bio *bio, size_t data_size); > diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h > index 4933b6b67b85..08d7dbff80f4 100644 > --- a/include/uapi/linux/dm-ioctl.h > +++ b/include/uapi/linux/dm-ioctl.h > @@ -214,6 +214,15 @@ struct dm_target_msg { > char message[0]; > }; > > +enum { > + REMAP_START_CMD = 1, > + REMAP_FINISH_CMD, > +}; > + > +struct dm_remap_param { > + uint8_t cmd; > + uint8_t params[0]; > +}; > /* > * If you change this make sure you make the corresponding change > * to dm-ioctl.c:lookup_ioctl() > @@ -244,6 +253,7 @@ enum { > DM_DEV_SET_GEOMETRY_CMD, > DM_DEV_ARM_POLL_CMD, > DM_GET_TARGET_VERSION_CMD, > + DM_DEV_REMAP_CMD > }; > > #define DM_IOCTL 0xfd > @@ -259,6 +269,7 @@ enum { > #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) > #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) > #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) > +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) > > #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) > #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) > @@ -272,9 +283,9 @@ enum { > #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) > > #define DM_VERSION_MAJOR 4 > -#define DM_VERSION_MINOR 43 > +#define DM_VERSION_MINOR 44 > #define DM_VERSION_PATCHLEVEL 0 > -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" > +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" > > /* Status bits */ > #define DM_READONLY_FLAG (1 << 0) /* In/Out */ > -- Damien Le Moal Western Digital Research -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear @ 2021-01-29 1:46 ` Damien Le Moal 0 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 1:46 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > Implement a block interposer for device-mapper to attach > to an existing block layer stack. Using the interposer, > we can connect the dm-linear to a device with a mounted > file system. > > changes: > * the new dm_interposer structure contains blk_interposer > to intercept bio from the interposed disk and interval tree > of block devices on this disk. > * the new interval tree for device mapper. > * the dm_submit_bio_interposer_fn() function implements > the bio interception logic. > * the functions dm_interposer_attach_dev() & > dm_interposer_detach_dev() allow to attach and detach devices > to dm_interposer. > * the new parameter 'noexcl' allows to create dm-linear to device > with an already mounted file system. > * the non_exclusive parameter in dm_target structure - it`s a sign > that target device should be opened without FMODE_EXCL mode. > * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to > a regular block device. Same comment about changelog as in the previous patch. > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > --- > drivers/md/dm-core.h | 46 +++- > drivers/md/dm-ioctl.c | 39 ++++ > drivers/md/dm-linear.c | 17 +- > drivers/md/dm-table.c | 12 +- > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > drivers/md/dm.h | 2 +- > include/linux/device-mapper.h | 7 + > include/uapi/linux/dm-ioctl.h | 15 +- > 8 files changed, 493 insertions(+), 28 deletions(-) > > diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h > index 086d293c2b03..0f870b1d4be4 100644 > --- a/drivers/md/dm-core.h > +++ b/drivers/md/dm-core.h > @@ -13,7 +13,7 @@ > #include <linux/ktime.h> > #include <linux/genhd.h> > #include <linux/blk-mq.h> > - whiteline change. > +#include <linux/rbtree.h> > #include <trace/events/block.h> > > #include "dm.h" > @@ -109,6 +109,9 @@ struct mapped_device { > bool init_tio_pdu:1; > > struct srcu_struct io_barrier; > + > + /* interposer device for remap */ > + struct dm_interposed_dev *ip_dev; > }; > > void disable_discard(struct mapped_device *md); > @@ -164,6 +167,47 @@ struct dm_table { > struct dm_md_mempools *mempools; > }; > > +/* > + * Interval tree for device mapper > + */ > +struct dm_rb_range { > + struct rb_node node; > + sector_t start; /* start sector of rb node */ > + sector_t last; /* end sector of rb node */ > + sector_t _subtree_last; /* highest sector in subtree of rb node */ > +}; > + > +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); > +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); > + > +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); > +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); > + > +/* > + * For connecting blk_interposer and dm-targets devices. Is this comment about the callback or the structure ? I think the latter, so it is in the worng place. Please also add a comment for the callback definition explaining what it should be doing. > + */ > +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); > + > +struct dm_interposed_dev { > + struct gendisk *disk; > + struct dm_rb_range node; > + void *context; > + dm_interpose_bio_t dm_interpose_bio; > + > + atomic64_t ip_cnt; /*for debug purpose*/ > +}; > + > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, > + sector_t ofs, sector_t len, > + void *context, > + dm_interpose_bio_t dm_interpose_bio_t); > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); > + > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); > +int dm_remap_uninstall(struct mapped_device *md); > + > static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) > { > return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; > diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c > index 5e306bba4375..2944d442c256 100644 > --- a/drivers/md/dm-ioctl.c > +++ b/drivers/md/dm-ioctl.c > @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para > return r; > } > > +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) > +{ > + char *donor_device_name = (char *)params; > + > + return dm_remap_install(md, donor_device_name); > +} > +static int dev_remap_finish(struct mapped_device *md) > +{ > + return dm_remap_uninstall(md); > +} > + > +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) > +{ > + int ret = 0; > + struct mapped_device *md; > + void *bin_data; > + struct dm_remap_param *remap_param; > + > + md = find_device(param); > + if (!md) > + return -ENXIO; > + > + bin_data = (void *)(param) + param->data_start; > + remap_param = bin_data; > + > + if (remap_param->cmd == REMAP_START_CMD) > + ret = dev_remap_start(md, remap_param->params); > + else if (remap_param->cmd == REMAP_FINISH_CMD) > + ret = dev_remap_finish(md); > + else { > + DMWARN("Invalid remap command, %d", remap_param->cmd); > + ret = -EINVAL; > + } > + > + dm_put(md); > + return ret; > +} > + > /* > * The ioctl parameter block consists of two parts, a dm_ioctl struct > * followed by a data buffer. This flag is set if the second part, > @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) > {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, > {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, > {DM_GET_TARGET_VERSION, 0, get_target_version}, > + {DM_DEV_REMAP_CMD, 0, dev_remap}, > }; > > if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) > diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c > index 00774b5d7668..ffb8b5ca4d10 100644 > --- a/drivers/md/dm-linear.c > +++ b/drivers/md/dm-linear.c > @@ -28,12 +28,13 @@ struct linear_c { > */ > static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > { > + fmode_t mode; > struct linear_c *lc; > unsigned long long tmp; > char dummy; > int ret; > > - if (argc != 2) { > + if ((argc < 2) || (argc > 3)) { > ti->error = "Invalid argument count"; > return -EINVAL; > } > @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > } > lc->start = tmp; > > - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); > + ti->non_exclusive = false; > + if (argc > 2) { > + if (strcmp("noexcl", argv[2]) == 0) > + ti->non_exclusive = true; > + else if (strcmp("excl", argv[2]) == 0) > + ti->non_exclusive = false; It already is false. > + else { > + ti->error = "Invalid exclusive option"; > + return -EINVAL; > + } > + } > + > + ret = dm_get_device(ti, argv[0], mode, &lc->dev); Where is mode initialized ? Why remove dm_table_get_mode(ti->table) ? > if (ret) { > ti->error = "Device lookup failed"; > goto bad; I would prefer to see this change to dm-linear in its own patch, following this one, with a clear explanation in the commit message how this change relates to interposer since the explanation for this "exclusive" change is nowhere to be seen. Also please check if there is a file describing dm-linear options under Documentation/ (I can't remember if there is one). If there is one, it will need to be updated too. > diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c > index 4acf2342f7ad..f15bc2171f25 100644 > --- a/drivers/md/dm-table.c > +++ b/drivers/md/dm-table.c > @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, > * device and not to touch the existing bdev field in case > * it is accessed concurrently. > */ > -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, > struct mapped_device *md) > { > int r; > @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > old_dev = dd->dm_dev; > > - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, > + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, > dd->dm_dev->mode | new_mode, &new_dev); > if (r) > return r; > @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > if (!dd) > return -ENOMEM; > > - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { > + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); > + if (r) { > kfree(dd); > return r; > } > @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > list_add(&dd->list, &t->devices); > goto out; > > - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { > - r = upgrade_mode(dd, mode, t->md); > + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && > + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { > + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); > if (r) > return r; > } > diff --git a/drivers/md/dm.c b/drivers/md/dm.c > index 7bac564f3faa..3b871d98b7b6 100644 > --- a/drivers/md/dm.c > +++ b/drivers/md/dm.c > @@ -28,6 +28,7 @@ > #include <linux/refcount.h> > #include <linux/part_stat.h> > #include <linux/blk-crypto.h> > +#include <linux/interval_tree_generic.h> > > #define DM_MSG_PREFIX "core" > > @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; > atomic_t dm_global_event_nr = ATOMIC_INIT(0); > DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); > > +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ Why not dm_interposer_mutex as the name ? And the comment is not very useful: a mutex is always for synchronizing :) > + > void dm_issue_global_event(void) > { > atomic_inc(&dm_global_event_nr); > @@ -162,6 +165,26 @@ struct table_device { > struct dm_dev dm_dev; > }; > > +/* > + * Device mapper`s interposer. > + */ > +struct dm_interposer { > + struct blk_interposer blk_ip; > + struct mapped_device *md; > + > + struct kref kref; > + struct rw_semaphore ip_devs_lock; > + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ > +}; > + > +/* > + * Interval tree for device mapper > + */ > +#define START(node) ((node)->start) > +#define LAST(node) ((node)->last) > +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, > + START, LAST,, dm_rb); > + > /* > * Bio-based DM's mempools' reserved IOs set by the user. > */ > @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) > rcu_read_unlock(); > } > > +static void dm_submit_bio_interposer_fn(struct bio *bio) > +{ > + struct dm_interposer *ip; > + unsigned int noio_flag = 0; > + sector_t start; > + sector_t last; > + struct dm_rb_range *node; > + > + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); > + start = bio->bi_iter.bi_sector; > + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); > + > + noio_flag = memalloc_noio_save(); > + down_read(&ip->ip_devs_lock); > + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); > + while (node) { > + struct dm_interposed_dev *ip_dev = > + container_of(node, struct dm_interposed_dev, node); > + > + atomic64_inc(&ip_dev->ip_cnt); > + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); > + > + node = dm_rb_iter_next(node, start, last); > + } > + up_read(&ip->ip_devs_lock); > + memalloc_noio_restore(noio_flag); > +} > + > +static void free_interposer(struct kref *kref) > +{ > + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); > + > + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); No queue freeze ? > + > + kfree(ip); > +} > + > +static struct dm_interposer *new_interposer(struct gendisk *disk) > +{ > + int ret = 0; > + struct dm_interposer *ip; > + > + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); > + if (!ip) > + return ERR_PTR(-ENOMEM); > + > + kref_init(&ip->kref); > + init_rwsem(&ip->ip_devs_lock); > + ip->ip_devs_root = RB_ROOT_CACHED; > + > + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); No queue freeze ? > + if (ret) { > + DMERR("Failed to attack blk_interposer"); > + kref_put(&ip->kref, free_interposer); > + return ERR_PTR(ret); > + } > + > + return ip; > +} > + > +static struct dm_interposer *get_interposer(struct gendisk *disk) > +{ > + struct dm_interposer *ip; > + > + if (!blk_has_interposer(disk)) > + return NULL; > + > + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { > + DMERR("Disks interposer slot already occupied."); > + return ERR_PTR(-EBUSY); This is weird... If there is an interposer, why not get a ref on that one. That is what the function name suggests at least. > + } > + > + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); > + > + kref_get(&ip->kref); > + return ip; > +} > + > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, > + void *context, dm_interpose_bio_t dm_interpose_bio) > +{ > + sector_t start = ofs; > + sector_t last = ofs + len - 1; > + struct dm_interposed_dev *ip_dev = NULL; > + > + /* Allocate new ip_dev */ > + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); > + if (!ip_dev) > + return NULL; > + > + ip_dev->disk = disk; > + ip_dev->node.start = start; > + ip_dev->node.last = last; > + > + ip_dev->context = context; > + ip_dev->dm_interpose_bio = dm_interpose_bio; > + > + atomic64_set(&ip_dev->ip_cnt, 0); > + > + return ip_dev; > +} > + > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) > +{ > + kfree(ip_dev); > +} Make this inline may be ? > + > +static inline void dm_disk_freeze(struct gendisk *disk) > +{ > + blk_mq_freeze_queue(disk->queue); > + blk_mq_quiesce_queue(disk->queue); I think you can replace this with blk_mq_freeze_queue_wait(). > +} > + > +static inline void dm_disk_unfreeze(struct gendisk *disk) > +{ > + blk_mq_unquiesce_queue(disk->queue); > + blk_mq_unfreeze_queue(disk->queue); > +} > + > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) > +{ > + int ret = 0; > + struct dm_interposer *ip = NULL; > + unsigned int noio_flag = 0; > + > + if (!ip_dev) > + return -EINVAL; > + > + dm_disk_freeze(ip_dev->disk); > + mutex_lock(&interposer_mutex); > + noio_flag = memalloc_noio_save(); > + > + ip = get_interposer(ip_dev->disk); > + if (ip == NULL) > + ip = new_interposer(ip_dev->disk); > + if (IS_ERR(ip)) { > + ret = PTR_ERR(ip); > + goto out; > + } > + > + /* Attach dm_interposed_dev to dm_interposer */ > + down_write(&ip->ip_devs_lock); > + do { > + struct dm_rb_range *node; > + > + /* checking that ip_dev already exists for this region */ > + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); > + if (node) { > + DMERR("Disk part form [%llu] to [%llu] already have interposer", > + node->start, node->last); > + > + ret = -EBUSY; > + break; > + } > + > + /* insert ip_dev to ip tree */ > + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); > + /* increment ip reference counter */ > + kref_get(&ip->kref); > + } while (false); > + up_write(&ip->ip_devs_lock); > + > + kref_put(&ip->kref, free_interposer); > + > +out: > + memalloc_noio_restore(noio_flag); > + mutex_unlock(&interposer_mutex); > + dm_disk_unfreeze(ip_dev->disk); > + > + return ret; > +} > + > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) > +{ > + int ret = 0; > + struct dm_interposer *ip = NULL; > + unsigned int noio_flag = 0; > + > + if (!ip_dev) > + return -EINVAL; > + > + dm_disk_freeze(ip_dev->disk); > + mutex_lock(&interposer_mutex); > + noio_flag = memalloc_noio_save(); > + > + ip = get_interposer(ip_dev->disk); > + if (IS_ERR(ip)) { > + ret = PTR_ERR(ip); > + DMERR("Interposer not found"); > + goto out; > + } > + if (unlikely(ip == NULL)) { > + ret = -ENXIO; > + DMERR("Interposer not found"); > + goto out; > + } > + > + down_write(&ip->ip_devs_lock); > + do { > + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); > + /* the reference counter here cannot be zero */ > + kref_put(&ip->kref, free_interposer); > + > + } while (false); > + up_write(&ip->ip_devs_lock); > + > + /* detach and free interposer if it`s not needed */ s/`/'/ > + kref_put(&ip->kref, free_interposer); > +out: > + memalloc_noio_restore(noio_flag); > + mutex_unlock(&interposer_mutex); > + dm_disk_unfreeze(ip_dev->disk); > + > + return ret; > +} > + > +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) > +{ > + struct mapped_device *md = context; > + > + /* Set acceptor device. */ > + bio->bi_disk = md->disk; > + > + /* Remap disks offset */ > + bio->bi_iter.bi_sector -= node->start; > + > + /* > + * bio should be resubmitted. > + * We can just add bio to bio_list of the current process. > + * current->bio_list must be initialized when this function is called. > + * If call submit_bio_noacct(), the bio will be checked twice. > + */ > + BUG_ON(!current->bio_list); > + bio_list_add(¤t->bio_list[0], bio); > +} > + > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) > +{ > + int ret = 0; > + struct block_device *donor_bdev; > + fmode_t mode = FMODE_READ | FMODE_WRITE; > + > + DMDEBUG("Dm remap install for mapped device %s and donor device %s", > + md->name, donor_device_name); > + > + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); > + if (IS_ERR(donor_bdev)) { > + DMERR("Cannot open device [%s]", donor_device_name); > + return PTR_ERR(donor_bdev); > + } > + > + do { > + sector_t ofs = get_start_sect(donor_bdev); > + sector_t len = bdev_nr_sectors(donor_bdev); > + > + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); > + if (!md->ip_dev) { > + ret = -ENOMEM; > + break; > + } > + > + DMDEBUG("New interposed device 0x%p", md->ip_dev); > + ret = dm_interposer_attach_dev(md->ip_dev); > + if (ret) { > + dm_interposer_free_dev(md->ip_dev); > + > + md->ip_dev = NULL; > + DMERR("Failed to attach dm interposer"); > + break; > + } > + > + DMDEBUG("Attached successfully."); > + } while (false); > + > + blkdev_put(donor_bdev, mode); > + > + return ret; > +} > + > +int dm_remap_uninstall(struct mapped_device *md) > +{ > + int ret = 0; > + > + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); > + > + if (!md->ip_dev) { > + DMERR("Cannot detach dm interposer"); > + return -EINVAL; > + } > + > + ret = dm_interposer_detach_dev(md->ip_dev); > + if (ret) { > + DMERR("Failed to detach dm interposer"); > + return ret; > + } > + > + DMDEBUG("Detached successfully. %llu bios was interposed", > + atomic64_read(&md->ip_dev->ip_cnt)); > + dm_interposer_free_dev(md->ip_dev); > + md->ip_dev = NULL; > + > + return 0; > +} > + > static char *_dm_claim_ptr = "I belong to device-mapper"; > > /* > * Open a table device so we can use it as a map destination. > */ > static int open_table_device(struct table_device *td, dev_t dev, > - struct mapped_device *md) > + struct mapped_device *md, bool non_exclusive) > { > struct block_device *bdev; > - > - int r; > + int ret; > > BUG_ON(td->dm_dev.bdev); > > - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > - if (IS_ERR(bdev)) > - return PTR_ERR(bdev); > + if (non_exclusive) > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); > + else > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > - r = bd_link_disk_holder(bdev, dm_disk(md)); > - if (r) { > - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); > - return r; > + if (IS_ERR(bdev)) { > + ret = PTR_ERR(bdev); > + if (ret != -EBUSY) > + return ret; > + } > + > + if (!non_exclusive) { > + ret = bd_link_disk_holder(bdev, dm_disk(md)); > + if (ret) { > + blkdev_put(bdev, td->dm_dev.mode); > + return ret; > + } > } > > td->dm_dev.bdev = bdev; > @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md > if (!td->dm_dev.bdev) > return; > > - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); > + if (td->dm_dev.mode & FMODE_EXCL) > + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > + > + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); > + > put_dax(td->dm_dev.dax_dev); > td->dm_dev.bdev = NULL; > td->dm_dev.dax_dev = NULL; > } > > static struct table_device *find_table_device(struct list_head *l, dev_t dev, > - fmode_t mode) > + fmode_t mode, bool non_exclusive) > { > struct table_device *td; > > list_for_each_entry(td, l, list) > - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) > + if (td->dm_dev.bdev->bd_dev == dev && > + td->dm_dev.mode == mode && > + td->dm_dev.non_exclusive == non_exclusive) > return td; > > return NULL; > } > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > struct dm_dev **result) > { > int r; > struct table_device *td; > > mutex_lock(&md->table_devices_lock); > - td = find_table_device(&md->table_devices, dev, mode); > + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); > if (!td) { > td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); > if (!td) { > @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > td->dm_dev.mode = mode; > td->dm_dev.bdev = NULL; > > - if ((r = open_table_device(td, dev, md))) { > + r = open_table_device(td, dev, md, non_exclusive); > + if (r) { > mutex_unlock(&md->table_devices_lock); > kfree(td); > return r; > @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) > > might_sleep(); > > + if (md->ip_dev) { > + if (dm_interposer_detach_dev(md->ip_dev)) > + DMERR("Failed to detach dm interposer"); > + > + dm_interposer_free_dev(md->ip_dev); > + md->ip_dev = NULL; > + } > + > spin_lock(&_minor_lock); > idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); > set_bit(DMF_FREEING, &md->flags); > diff --git a/drivers/md/dm.h b/drivers/md/dm.h > index fffe1e289c53..7bf20fb2de74 100644 > --- a/drivers/md/dm.h > +++ b/drivers/md/dm.h > @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); > int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); > int dm_cancel_deferred_remove(struct mapped_device *md); > int dm_request_based(struct mapped_device *md); > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > struct dm_dev **result); > void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); > > diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h > index 61a66fb8ebb3..70002363bfc0 100644 > --- a/include/linux/device-mapper.h > +++ b/include/linux/device-mapper.h > @@ -150,6 +150,7 @@ struct dm_dev { > struct block_device *bdev; > struct dax_device *dax_dev; > fmode_t mode; > + bool non_exclusive; > char name[16]; > }; > > @@ -325,6 +326,12 @@ struct dm_target { > * whether or not its underlying devices have support. > */ > bool discards_supported:1; > + > + /* > + * Set if this target needs to open device without FMODE_EXCL > + * mode. > + */ > + bool non_exclusive:1; > }; > > void *dm_per_bio_data(struct bio *bio, size_t data_size); > diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h > index 4933b6b67b85..08d7dbff80f4 100644 > --- a/include/uapi/linux/dm-ioctl.h > +++ b/include/uapi/linux/dm-ioctl.h > @@ -214,6 +214,15 @@ struct dm_target_msg { > char message[0]; > }; > > +enum { > + REMAP_START_CMD = 1, > + REMAP_FINISH_CMD, > +}; > + > +struct dm_remap_param { > + uint8_t cmd; > + uint8_t params[0]; > +}; > /* > * If you change this make sure you make the corresponding change > * to dm-ioctl.c:lookup_ioctl() > @@ -244,6 +253,7 @@ enum { > DM_DEV_SET_GEOMETRY_CMD, > DM_DEV_ARM_POLL_CMD, > DM_GET_TARGET_VERSION_CMD, > + DM_DEV_REMAP_CMD > }; > > #define DM_IOCTL 0xfd > @@ -259,6 +269,7 @@ enum { > #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) > #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) > #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) > +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) > > #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) > #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) > @@ -272,9 +283,9 @@ enum { > #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) > > #define DM_VERSION_MAJOR 4 > -#define DM_VERSION_MINOR 43 > +#define DM_VERSION_MINOR 44 > #define DM_VERSION_PATCHLEVEL 0 > -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" > +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" > > /* Status bits */ > #define DM_READONLY_FLAG (1 << 0) /* In/Out */ > -- Damien Le Moal Western Digital Research ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear 2021-01-29 1:46 ` Damien Le Moal @ 2021-01-29 16:08 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-29 16:08 UTC (permalink / raw) To: Damien Le Moal Cc: snitzer@redhat.com, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com The 01/29/2021 04:46, Damien Le Moal wrote: > On 2021/01/29 2:23, Sergei Shtepa wrote: > > Implement a block interposer for device-mapper to attach > > to an existing block layer stack. Using the interposer, > > we can connect the dm-linear to a device with a mounted > > file system. > > > > changes: > > * the new dm_interposer structure contains blk_interposer > > to intercept bio from the interposed disk and interval tree > > of block devices on this disk. > > * the new interval tree for device mapper. > > * the dm_submit_bio_interposer_fn() function implements > > the bio interception logic. > > * the functions dm_interposer_attach_dev() & > > dm_interposer_detach_dev() allow to attach and detach devices > > to dm_interposer. > > * the new parameter 'noexcl' allows to create dm-linear to device > > with an already mounted file system. > > * the non_exclusive parameter in dm_target structure - it`s a sign > > that target device should be opened without FMODE_EXCL mode. > > * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to > > a regular block device. > > Same comment about changelog as in the previous patch. > > > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > > --- > > drivers/md/dm-core.h | 46 +++- > > drivers/md/dm-ioctl.c | 39 ++++ > > drivers/md/dm-linear.c | 17 +- > > drivers/md/dm-table.c | 12 +- > > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > > drivers/md/dm.h | 2 +- > > include/linux/device-mapper.h | 7 + > > include/uapi/linux/dm-ioctl.h | 15 +- > > 8 files changed, 493 insertions(+), 28 deletions(-) > > > > diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h > > index 086d293c2b03..0f870b1d4be4 100644 > > --- a/drivers/md/dm-core.h > > +++ b/drivers/md/dm-core.h > > @@ -13,7 +13,7 @@ > > #include <linux/ktime.h> > > #include <linux/genhd.h> > > #include <linux/blk-mq.h> > > - > > whiteline change. > > > +#include <linux/rbtree.h> > > #include <trace/events/block.h> I don't see any problem in the fact that a new include appeared instead of whiteline. It doesn't make sense to split the include section by whiteline. > > > > #include "dm.h" > > @@ -109,6 +109,9 @@ struct mapped_device { > > bool init_tio_pdu:1; > > > > struct srcu_struct io_barrier; > > + > > + /* interposer device for remap */ > > + struct dm_interposed_dev *ip_dev; > > }; > > > > void disable_discard(struct mapped_device *md); > > @@ -164,6 +167,47 @@ struct dm_table { > > struct dm_md_mempools *mempools; > > }; > > > > +/* > > + * Interval tree for device mapper > > + */ > > +struct dm_rb_range { > > + struct rb_node node; > > + sector_t start; /* start sector of rb node */ > > + sector_t last; /* end sector of rb node */ > > + sector_t _subtree_last; /* highest sector in subtree of rb node */ > > +}; > > + > > +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); > > +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); > > + > > +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); > > +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); > > + > > +/* > > + * For connecting blk_interposer and dm-targets devices. > > Is this comment about the callback or the structure ? I think the latter, so it > is in the worng place. Please also add a comment for the callback definition > explaining what it should be doing. Ok. > > > + */ > > +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); > > + > > +struct dm_interposed_dev { > > + struct gendisk *disk; > > + struct dm_rb_range node; > > + void *context; > > + dm_interpose_bio_t dm_interpose_bio; > > + > > + atomic64_t ip_cnt; /*for debug purpose*/ > > +}; > > + > > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, > > + sector_t ofs, sector_t len, > > + void *context, > > + dm_interpose_bio_t dm_interpose_bio_t); > > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); > > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); > > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); > > + > > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); > > +int dm_remap_uninstall(struct mapped_device *md); > > + > > static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) > > { > > return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; > > diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c > > index 5e306bba4375..2944d442c256 100644 > > --- a/drivers/md/dm-ioctl.c > > +++ b/drivers/md/dm-ioctl.c > > @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para > > return r; > > } > > > > +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) > > +{ > > + char *donor_device_name = (char *)params; > > + > > + return dm_remap_install(md, donor_device_name); > > +} > > +static int dev_remap_finish(struct mapped_device *md) > > +{ > > + return dm_remap_uninstall(md); > > +} > > + > > +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) > > +{ > > + int ret = 0; > > + struct mapped_device *md; > > + void *bin_data; > > + struct dm_remap_param *remap_param; > > + > > + md = find_device(param); > > + if (!md) > > + return -ENXIO; > > + > > + bin_data = (void *)(param) + param->data_start; > > + remap_param = bin_data; > > + > > + if (remap_param->cmd == REMAP_START_CMD) > > + ret = dev_remap_start(md, remap_param->params); > > + else if (remap_param->cmd == REMAP_FINISH_CMD) > > + ret = dev_remap_finish(md); > > + else { > > + DMWARN("Invalid remap command, %d", remap_param->cmd); > > + ret = -EINVAL; > > + } > > + > > + dm_put(md); > > + return ret; > > +} > > + > > /* > > * The ioctl parameter block consists of two parts, a dm_ioctl struct > > * followed by a data buffer. This flag is set if the second part, > > @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) > > {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, > > {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, > > {DM_GET_TARGET_VERSION, 0, get_target_version}, > > + {DM_DEV_REMAP_CMD, 0, dev_remap}, > > }; > > > > if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) > > diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c > > index 00774b5d7668..ffb8b5ca4d10 100644 > > --- a/drivers/md/dm-linear.c > > +++ b/drivers/md/dm-linear.c > > @@ -28,12 +28,13 @@ struct linear_c { > > */ > > static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > > { > > + fmode_t mode; > > struct linear_c *lc; > > unsigned long long tmp; > > char dummy; > > int ret; > > > > - if (argc != 2) { > > + if ((argc < 2) || (argc > 3)) { > > ti->error = "Invalid argument count"; > > return -EINVAL; > > } > > @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > > } > > lc->start = tmp; > > > > - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); > > + ti->non_exclusive = false; > > + if (argc > 2) { > > + if (strcmp("noexcl", argv[2]) == 0) > > + ti->non_exclusive = true; > > + else if (strcmp("excl", argv[2]) == 0) > > + ti->non_exclusive = false; > > It already is false. Yes, and even the value of the "excl" parameter is redundant, since it defines the default value. I think this code structure more clearly reflects the meaning of the parameter. > > > + else { > > + ti->error = "Invalid exclusive option"; > > + return -EINVAL; > > + } > > + } > > + > > + ret = dm_get_device(ti, argv[0], mode, &lc->dev); > > Where is mode initialized ? Why remove dm_table_get_mode(ti->table) ? Yes. It`s a bug. In this plaсe should be dm_table_get_mode(). > > > if (ret) { > > ti->error = "Device lookup failed"; > > goto bad; > > I would prefer to see this change to dm-linear in its own patch, following this > one, with a clear explanation in the commit message how this change relates to > interposer since the explanation for this "exclusive" change is nowhere to be > seen. Also please check if there is a file describing dm-linear options under > Documentation/ (I can't remember if there is one). If there is one, it will need > to be updated too. It's a good idea. > > > diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c > > index 4acf2342f7ad..f15bc2171f25 100644 > > --- a/drivers/md/dm-table.c > > +++ b/drivers/md/dm-table.c > > @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, > > * device and not to touch the existing bdev field in case > > * it is accessed concurrently. > > */ > > -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, > > struct mapped_device *md) > > { > > int r; > > @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > > > old_dev = dd->dm_dev; > > > > - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, > > + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, > > dd->dm_dev->mode | new_mode, &new_dev); > > if (r) > > return r; > > @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > > if (!dd) > > return -ENOMEM; > > > > - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { > > + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); > > + if (r) { > > kfree(dd); > > return r; > > } > > @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > > list_add(&dd->list, &t->devices); > > goto out; > > > > - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { > > - r = upgrade_mode(dd, mode, t->md); > > + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && > > + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { > > + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); > > if (r) > > return r; > > } > > diff --git a/drivers/md/dm.c b/drivers/md/dm.c > > index 7bac564f3faa..3b871d98b7b6 100644 > > --- a/drivers/md/dm.c > > +++ b/drivers/md/dm.c > > @@ -28,6 +28,7 @@ > > #include <linux/refcount.h> > > #include <linux/part_stat.h> > > #include <linux/blk-crypto.h> > > +#include <linux/interval_tree_generic.h> > > > > #define DM_MSG_PREFIX "core" > > > > @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; > > atomic_t dm_global_event_nr = ATOMIC_INIT(0); > > DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); > > > > +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ > > Why not dm_interposer_mutex as the name ? And the comment is not very useful: a > mutex is always for synchronizing :) Right. I'll do it. > > > + > > void dm_issue_global_event(void) > > { > > atomic_inc(&dm_global_event_nr); > > @@ -162,6 +165,26 @@ struct table_device { > > struct dm_dev dm_dev; > > }; > > > > +/* > > + * Device mapper`s interposer. > > + */ > > +struct dm_interposer { > > + struct blk_interposer blk_ip; > > + struct mapped_device *md; > > + > > + struct kref kref; > > + struct rw_semaphore ip_devs_lock; > > + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ > > +}; > > + > > +/* > > + * Interval tree for device mapper > > + */ > > +#define START(node) ((node)->start) > > +#define LAST(node) ((node)->last) > > +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, > > + START, LAST,, dm_rb); > > + > > /* > > * Bio-based DM's mempools' reserved IOs set by the user. > > */ > > @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) > > rcu_read_unlock(); > > } > > > > +static void dm_submit_bio_interposer_fn(struct bio *bio) > > +{ > > + struct dm_interposer *ip; > > + unsigned int noio_flag = 0; > > + sector_t start; > > + sector_t last; > > + struct dm_rb_range *node; > > + > > + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); > > + start = bio->bi_iter.bi_sector; > > + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); > > + > > + noio_flag = memalloc_noio_save(); > > + down_read(&ip->ip_devs_lock); > > + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); > > + while (node) { > > + struct dm_interposed_dev *ip_dev = > > + container_of(node, struct dm_interposed_dev, node); > > + > > + atomic64_inc(&ip_dev->ip_cnt); > > + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); > > + > > + node = dm_rb_iter_next(node, start, last); > > + } > > + up_read(&ip->ip_devs_lock); > > + memalloc_noio_restore(noio_flag); > > +} > > + > > +static void free_interposer(struct kref *kref) > > +{ > > + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); > > + > > + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); > > No queue freeze ? Yes. The queue should be already freeze. > > > + > > + kfree(ip); > > +} > > + > > +static struct dm_interposer *new_interposer(struct gendisk *disk) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip; > > + > > + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); > > + if (!ip) > > + return ERR_PTR(-ENOMEM); > > + > > + kref_init(&ip->kref); > > + init_rwsem(&ip->ip_devs_lock); > > + ip->ip_devs_root = RB_ROOT_CACHED; > > + > > + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); > > No queue freeze ? Yes, again. > > > + if (ret) { > > + DMERR("Failed to attack blk_interposer"); > > + kref_put(&ip->kref, free_interposer); > > + return ERR_PTR(ret); > > + } > > + > > + return ip; > > +} > > + > > +static struct dm_interposer *get_interposer(struct gendisk *disk) > > +{ > > + struct dm_interposer *ip; > > + > > + if (!blk_has_interposer(disk)) > > + return NULL; > > + > > + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { > > + DMERR("Disks interposer slot already occupied."); > > + return ERR_PTR(-EBUSY); > > This is weird... If there is an interposer, why not get a ref on that one. That > is what the function name suggests at least. Getting a ref on that just below in this function. But the name "get_dm_interposer" would be better. > > > + } > > + > > + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); > > + > > + kref_get(&ip->kref); > > + return ip; > > +} > > + > > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, > > + void *context, dm_interpose_bio_t dm_interpose_bio) > > +{ > > + sector_t start = ofs; > > + sector_t last = ofs + len - 1; > > + struct dm_interposed_dev *ip_dev = NULL; > > + > > + /* Allocate new ip_dev */ > > + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); > > + if (!ip_dev) > > + return NULL; > > + > > + ip_dev->disk = disk; > > + ip_dev->node.start = start; > > + ip_dev->node.last = last; > > + > > + ip_dev->context = context; > > + ip_dev->dm_interpose_bio = dm_interpose_bio; > > + > > + atomic64_set(&ip_dev->ip_cnt, 0); > > + > > + return ip_dev; > > +} > > + > > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + kfree(ip_dev); > > +} > > Make this inline may be ? Yes. Or even remove this function. > > > + > > +static inline void dm_disk_freeze(struct gendisk *disk) > > +{ > > + blk_mq_freeze_queue(disk->queue); > > + blk_mq_quiesce_queue(disk->queue); > > I think you can replace this with blk_mq_freeze_queue_wait(). I think no. blk_freeze_queue_start() also is required. > > > +} > > + > > +static inline void dm_disk_unfreeze(struct gendisk *disk) > > +{ > > + blk_mq_unquiesce_queue(disk->queue); > > + blk_mq_unfreeze_queue(disk->queue); > > +} > > + > > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip = NULL; > > + unsigned int noio_flag = 0; > > + > > + if (!ip_dev) > > + return -EINVAL; > > + > > + dm_disk_freeze(ip_dev->disk); > > + mutex_lock(&interposer_mutex); > > + noio_flag = memalloc_noio_save(); > > + > > + ip = get_interposer(ip_dev->disk); > > + if (ip == NULL) > > + ip = new_interposer(ip_dev->disk); > > + if (IS_ERR(ip)) { > > + ret = PTR_ERR(ip); > > + goto out; > > + } > > + > > + /* Attach dm_interposed_dev to dm_interposer */ > > + down_write(&ip->ip_devs_lock); > > + do { > > + struct dm_rb_range *node; > > + > > + /* checking that ip_dev already exists for this region */ > > + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); > > + if (node) { > > + DMERR("Disk part form [%llu] to [%llu] already have interposer", > > + node->start, node->last); > > + > > + ret = -EBUSY; > > + break; > > + } > > + > > + /* insert ip_dev to ip tree */ > > + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); > > + /* increment ip reference counter */ > > + kref_get(&ip->kref); > > + } while (false); > > + up_write(&ip->ip_devs_lock); > > + > > + kref_put(&ip->kref, free_interposer); > > + > > +out: > > + memalloc_noio_restore(noio_flag); > > + mutex_unlock(&interposer_mutex); > > + dm_disk_unfreeze(ip_dev->disk); > > + > > + return ret; > > +} > > + > > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip = NULL; > > + unsigned int noio_flag = 0; > > + > > + if (!ip_dev) > > + return -EINVAL; > > + > > + dm_disk_freeze(ip_dev->disk); > > + mutex_lock(&interposer_mutex); > > + noio_flag = memalloc_noio_save(); > > + > > + ip = get_interposer(ip_dev->disk); > > + if (IS_ERR(ip)) { > > + ret = PTR_ERR(ip); > > + DMERR("Interposer not found"); > > + goto out; > > + } > > + if (unlikely(ip == NULL)) { > > + ret = -ENXIO; > > + DMERR("Interposer not found"); > > + goto out; > > + } > > + > > + down_write(&ip->ip_devs_lock); > > + do { > > + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); > > + /* the reference counter here cannot be zero */ > > + kref_put(&ip->kref, free_interposer); > > + > > + } while (false); > > + up_write(&ip->ip_devs_lock); > > + > > + /* detach and free interposer if it`s not needed */ > > s/`/'/ Thanks. It's my problem. > > > + kref_put(&ip->kref, free_interposer); > > +out: > > + memalloc_noio_restore(noio_flag); > > + mutex_unlock(&interposer_mutex); > > + dm_disk_unfreeze(ip_dev->disk); > > + > > + return ret; > > +} > > + > > +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) > > +{ > > + struct mapped_device *md = context; > > + > > + /* Set acceptor device. */ > > + bio->bi_disk = md->disk; > > + > > + /* Remap disks offset */ > > + bio->bi_iter.bi_sector -= node->start; > > + > > + /* > > + * bio should be resubmitted. > > + * We can just add bio to bio_list of the current process. > > + * current->bio_list must be initialized when this function is called. > > + * If call submit_bio_noacct(), the bio will be checked twice. > > + */ > > + BUG_ON(!current->bio_list); > > + bio_list_add(¤t->bio_list[0], bio); > > +} > > + > > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) > > +{ > > + int ret = 0; > > + struct block_device *donor_bdev; > > + fmode_t mode = FMODE_READ | FMODE_WRITE; > > + > > + DMDEBUG("Dm remap install for mapped device %s and donor device %s", > > + md->name, donor_device_name); > > + > > + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); > > + if (IS_ERR(donor_bdev)) { > > + DMERR("Cannot open device [%s]", donor_device_name); > > + return PTR_ERR(donor_bdev); > > + } > > + > > + do { > > + sector_t ofs = get_start_sect(donor_bdev); > > + sector_t len = bdev_nr_sectors(donor_bdev); > > + > > + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); > > + if (!md->ip_dev) { > > + ret = -ENOMEM; > > + break; > > + } > > + > > + DMDEBUG("New interposed device 0x%p", md->ip_dev); > > + ret = dm_interposer_attach_dev(md->ip_dev); > > + if (ret) { > > + dm_interposer_free_dev(md->ip_dev); > > + > > + md->ip_dev = NULL; > > + DMERR("Failed to attach dm interposer"); > > + break; > > + } > > + > > + DMDEBUG("Attached successfully."); > > + } while (false); > > + > > + blkdev_put(donor_bdev, mode); > > + > > + return ret; > > +} > > + > > +int dm_remap_uninstall(struct mapped_device *md) > > +{ > > + int ret = 0; > > + > > + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); > > + > > + if (!md->ip_dev) { > > + DMERR("Cannot detach dm interposer"); > > + return -EINVAL; > > + } > > + > > + ret = dm_interposer_detach_dev(md->ip_dev); > > + if (ret) { > > + DMERR("Failed to detach dm interposer"); > > + return ret; > > + } > > + > > + DMDEBUG("Detached successfully. %llu bios was interposed", > > + atomic64_read(&md->ip_dev->ip_cnt)); > > + dm_interposer_free_dev(md->ip_dev); > > + md->ip_dev = NULL; > > + > > + return 0; > > +} > > + > > static char *_dm_claim_ptr = "I belong to device-mapper"; > > > > /* > > * Open a table device so we can use it as a map destination. > > */ > > static int open_table_device(struct table_device *td, dev_t dev, > > - struct mapped_device *md) > > + struct mapped_device *md, bool non_exclusive) > > { > > struct block_device *bdev; > > - > > - int r; > > + int ret; > > > > BUG_ON(td->dm_dev.bdev); > > > > - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > - if (IS_ERR(bdev)) > > - return PTR_ERR(bdev); > > + if (non_exclusive) > > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); > > + else > > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > > > - r = bd_link_disk_holder(bdev, dm_disk(md)); > > - if (r) { > > - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); > > - return r; > > + if (IS_ERR(bdev)) { > > + ret = PTR_ERR(bdev); > > + if (ret != -EBUSY) > > + return ret; > > + } > > + > > + if (!non_exclusive) { > > + ret = bd_link_disk_holder(bdev, dm_disk(md)); > > + if (ret) { > > + blkdev_put(bdev, td->dm_dev.mode); > > + return ret; > > + } > > } > > > > td->dm_dev.bdev = bdev; > > @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md > > if (!td->dm_dev.bdev) > > return; > > > > - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > > - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); > > + if (td->dm_dev.mode & FMODE_EXCL) > > + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > > + > > + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); > > + > > put_dax(td->dm_dev.dax_dev); > > td->dm_dev.bdev = NULL; > > td->dm_dev.dax_dev = NULL; > > } > > > > static struct table_device *find_table_device(struct list_head *l, dev_t dev, > > - fmode_t mode) > > + fmode_t mode, bool non_exclusive) > > { > > struct table_device *td; > > > > list_for_each_entry(td, l, list) > > - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) > > + if (td->dm_dev.bdev->bd_dev == dev && > > + td->dm_dev.mode == mode && > > + td->dm_dev.non_exclusive == non_exclusive) > > return td; > > > > return NULL; > > } > > > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > > struct dm_dev **result) > > { > > int r; > > struct table_device *td; > > > > mutex_lock(&md->table_devices_lock); > > - td = find_table_device(&md->table_devices, dev, mode); > > + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); > > if (!td) { > > td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); > > if (!td) { > > @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > td->dm_dev.mode = mode; > > td->dm_dev.bdev = NULL; > > > > - if ((r = open_table_device(td, dev, md))) { > > + r = open_table_device(td, dev, md, non_exclusive); > > + if (r) { > > mutex_unlock(&md->table_devices_lock); > > kfree(td); > > return r; > > @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) > > > > might_sleep(); > > > > + if (md->ip_dev) { > > + if (dm_interposer_detach_dev(md->ip_dev)) > > + DMERR("Failed to detach dm interposer"); > > + > > + dm_interposer_free_dev(md->ip_dev); > > + md->ip_dev = NULL; > > + } > > + > > spin_lock(&_minor_lock); > > idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); > > set_bit(DMF_FREEING, &md->flags); > > diff --git a/drivers/md/dm.h b/drivers/md/dm.h > > index fffe1e289c53..7bf20fb2de74 100644 > > --- a/drivers/md/dm.h > > +++ b/drivers/md/dm.h > > @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); > > int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); > > int dm_cancel_deferred_remove(struct mapped_device *md); > > int dm_request_based(struct mapped_device *md); > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > > struct dm_dev **result); > > void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); > > > > diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h > > index 61a66fb8ebb3..70002363bfc0 100644 > > --- a/include/linux/device-mapper.h > > +++ b/include/linux/device-mapper.h > > @@ -150,6 +150,7 @@ struct dm_dev { > > struct block_device *bdev; > > struct dax_device *dax_dev; > > fmode_t mode; > > + bool non_exclusive; > > char name[16]; > > }; > > > > @@ -325,6 +326,12 @@ struct dm_target { > > * whether or not its underlying devices have support. > > */ > > bool discards_supported:1; > > + > > + /* > > + * Set if this target needs to open device without FMODE_EXCL > > + * mode. > > + */ > > + bool non_exclusive:1; > > }; > > > > void *dm_per_bio_data(struct bio *bio, size_t data_size); > > diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h > > index 4933b6b67b85..08d7dbff80f4 100644 > > --- a/include/uapi/linux/dm-ioctl.h > > +++ b/include/uapi/linux/dm-ioctl.h > > @@ -214,6 +214,15 @@ struct dm_target_msg { > > char message[0]; > > }; > > > > +enum { > > + REMAP_START_CMD = 1, > > + REMAP_FINISH_CMD, > > +}; > > + > > +struct dm_remap_param { > > + uint8_t cmd; > > + uint8_t params[0]; > > +}; > > /* > > * If you change this make sure you make the corresponding change > > * to dm-ioctl.c:lookup_ioctl() > > @@ -244,6 +253,7 @@ enum { > > DM_DEV_SET_GEOMETRY_CMD, > > DM_DEV_ARM_POLL_CMD, > > DM_GET_TARGET_VERSION_CMD, > > + DM_DEV_REMAP_CMD > > }; > > > > #define DM_IOCTL 0xfd > > @@ -259,6 +269,7 @@ enum { > > #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) > > #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) > > #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) > > +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) > > > > #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) > > #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) > > @@ -272,9 +283,9 @@ enum { > > #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) > > > > #define DM_VERSION_MAJOR 4 > > -#define DM_VERSION_MINOR 43 > > +#define DM_VERSION_MINOR 44 > > #define DM_VERSION_PATCHLEVEL 0 > > -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" > > +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" > > > > /* Status bits */ > > #define DM_READONLY_FLAG (1 << 0) /* In/Out */ > > > > > -- > Damien Le Moal > Western Digital Research > -- Sergei Shtepa Veeam Software developer. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear @ 2021-01-29 16:08 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-29 16:08 UTC (permalink / raw) To: Damien Le Moal Cc: hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org, Pavel Tide The 01/29/2021 04:46, Damien Le Moal wrote: > On 2021/01/29 2:23, Sergei Shtepa wrote: > > Implement a block interposer for device-mapper to attach > > to an existing block layer stack. Using the interposer, > > we can connect the dm-linear to a device with a mounted > > file system. > > > > changes: > > * the new dm_interposer structure contains blk_interposer > > to intercept bio from the interposed disk and interval tree > > of block devices on this disk. > > * the new interval tree for device mapper. > > * the dm_submit_bio_interposer_fn() function implements > > the bio interception logic. > > * the functions dm_interposer_attach_dev() & > > dm_interposer_detach_dev() allow to attach and detach devices > > to dm_interposer. > > * the new parameter 'noexcl' allows to create dm-linear to device > > with an already mounted file system. > > * the non_exclusive parameter in dm_target structure - it`s a sign > > that target device should be opened without FMODE_EXCL mode. > > * the new ioctl IOCTL_DEV_REMAP allow to attach dm device to > > a regular block device. > > Same comment about changelog as in the previous patch. > > > Signed-off-by: Sergei Shtepa <sergei.shtepa@veeam.com> > > --- > > drivers/md/dm-core.h | 46 +++- > > drivers/md/dm-ioctl.c | 39 ++++ > > drivers/md/dm-linear.c | 17 +- > > drivers/md/dm-table.c | 12 +- > > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > > drivers/md/dm.h | 2 +- > > include/linux/device-mapper.h | 7 + > > include/uapi/linux/dm-ioctl.h | 15 +- > > 8 files changed, 493 insertions(+), 28 deletions(-) > > > > diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h > > index 086d293c2b03..0f870b1d4be4 100644 > > --- a/drivers/md/dm-core.h > > +++ b/drivers/md/dm-core.h > > @@ -13,7 +13,7 @@ > > #include <linux/ktime.h> > > #include <linux/genhd.h> > > #include <linux/blk-mq.h> > > - > > whiteline change. > > > +#include <linux/rbtree.h> > > #include <trace/events/block.h> I don't see any problem in the fact that a new include appeared instead of whiteline. It doesn't make sense to split the include section by whiteline. > > > > #include "dm.h" > > @@ -109,6 +109,9 @@ struct mapped_device { > > bool init_tio_pdu:1; > > > > struct srcu_struct io_barrier; > > + > > + /* interposer device for remap */ > > + struct dm_interposed_dev *ip_dev; > > }; > > > > void disable_discard(struct mapped_device *md); > > @@ -164,6 +167,47 @@ struct dm_table { > > struct dm_md_mempools *mempools; > > }; > > > > +/* > > + * Interval tree for device mapper > > + */ > > +struct dm_rb_range { > > + struct rb_node node; > > + sector_t start; /* start sector of rb node */ > > + sector_t last; /* end sector of rb node */ > > + sector_t _subtree_last; /* highest sector in subtree of rb node */ > > +}; > > + > > +void dm_rb_insert(struct dm_rb_range *node, struct rb_root_cached *root); > > +void dm_rb_remove(struct dm_rb_range *node, struct rb_root_cached *root); > > + > > +struct dm_rb_range *dm_rb_iter_first(struct rb_root_cached *root, sector_t start, sector_t last); > > +struct dm_rb_range *dm_rb_iter_next(struct dm_rb_range *node, sector_t start, sector_t last); > > + > > +/* > > + * For connecting blk_interposer and dm-targets devices. > > Is this comment about the callback or the structure ? I think the latter, so it > is in the worng place. Please also add a comment for the callback definition > explaining what it should be doing. Ok. > > > + */ > > +typedef void (*dm_interpose_bio_t) (void *context, struct dm_rb_range *node, struct bio *bio); > > + > > +struct dm_interposed_dev { > > + struct gendisk *disk; > > + struct dm_rb_range node; > > + void *context; > > + dm_interpose_bio_t dm_interpose_bio; > > + > > + atomic64_t ip_cnt; /*for debug purpose*/ > > +}; > > + > > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, > > + sector_t ofs, sector_t len, > > + void *context, > > + dm_interpose_bio_t dm_interpose_bio_t); > > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev); > > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev); > > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev); > > + > > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name); > > +int dm_remap_uninstall(struct mapped_device *md); > > + > > static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) > > { > > return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; > > diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c > > index 5e306bba4375..2944d442c256 100644 > > --- a/drivers/md/dm-ioctl.c > > +++ b/drivers/md/dm-ioctl.c > > @@ -1649,6 +1649,44 @@ static int target_message(struct file *filp, struct dm_ioctl *param, size_t para > > return r; > > } > > > > +static inline int dev_remap_start(struct mapped_device *md, uint8_t *params) > > +{ > > + char *donor_device_name = (char *)params; > > + > > + return dm_remap_install(md, donor_device_name); > > +} > > +static int dev_remap_finish(struct mapped_device *md) > > +{ > > + return dm_remap_uninstall(md); > > +} > > + > > +static int dev_remap(struct file *filp, struct dm_ioctl *param, size_t param_size) > > +{ > > + int ret = 0; > > + struct mapped_device *md; > > + void *bin_data; > > + struct dm_remap_param *remap_param; > > + > > + md = find_device(param); > > + if (!md) > > + return -ENXIO; > > + > > + bin_data = (void *)(param) + param->data_start; > > + remap_param = bin_data; > > + > > + if (remap_param->cmd == REMAP_START_CMD) > > + ret = dev_remap_start(md, remap_param->params); > > + else if (remap_param->cmd == REMAP_FINISH_CMD) > > + ret = dev_remap_finish(md); > > + else { > > + DMWARN("Invalid remap command, %d", remap_param->cmd); > > + ret = -EINVAL; > > + } > > + > > + dm_put(md); > > + return ret; > > +} > > + > > /* > > * The ioctl parameter block consists of two parts, a dm_ioctl struct > > * followed by a data buffer. This flag is set if the second part, > > @@ -1691,6 +1729,7 @@ static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags) > > {DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}, > > {DM_DEV_ARM_POLL, IOCTL_FLAGS_NO_PARAMS, dev_arm_poll}, > > {DM_GET_TARGET_VERSION, 0, get_target_version}, > > + {DM_DEV_REMAP_CMD, 0, dev_remap}, > > }; > > > > if (unlikely(cmd >= ARRAY_SIZE(_ioctls))) > > diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c > > index 00774b5d7668..ffb8b5ca4d10 100644 > > --- a/drivers/md/dm-linear.c > > +++ b/drivers/md/dm-linear.c > > @@ -28,12 +28,13 @@ struct linear_c { > > */ > > static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > > { > > + fmode_t mode; > > struct linear_c *lc; > > unsigned long long tmp; > > char dummy; > > int ret; > > > > - if (argc != 2) { > > + if ((argc < 2) || (argc > 3)) { > > ti->error = "Invalid argument count"; > > return -EINVAL; > > } > > @@ -51,7 +52,19 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv) > > } > > lc->start = tmp; > > > > - ret = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev); > > + ti->non_exclusive = false; > > + if (argc > 2) { > > + if (strcmp("noexcl", argv[2]) == 0) > > + ti->non_exclusive = true; > > + else if (strcmp("excl", argv[2]) == 0) > > + ti->non_exclusive = false; > > It already is false. Yes, and even the value of the "excl" parameter is redundant, since it defines the default value. I think this code structure more clearly reflects the meaning of the parameter. > > > + else { > > + ti->error = "Invalid exclusive option"; > > + return -EINVAL; > > + } > > + } > > + > > + ret = dm_get_device(ti, argv[0], mode, &lc->dev); > > Where is mode initialized ? Why remove dm_table_get_mode(ti->table) ? Yes. It`s a bug. In this plaсe should be dm_table_get_mode(). > > > if (ret) { > > ti->error = "Device lookup failed"; > > goto bad; > > I would prefer to see this change to dm-linear in its own patch, following this > one, with a clear explanation in the commit message how this change relates to > interposer since the explanation for this "exclusive" change is nowhere to be > seen. Also please check if there is a file describing dm-linear options under > Documentation/ (I can't remember if there is one). If there is one, it will need > to be updated too. It's a good idea. > > > diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c > > index 4acf2342f7ad..f15bc2171f25 100644 > > --- a/drivers/md/dm-table.c > > +++ b/drivers/md/dm-table.c > > @@ -322,7 +322,7 @@ static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev, > > * device and not to touch the existing bdev field in case > > * it is accessed concurrently. > > */ > > -static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > +static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, bool non_exclusive, > > struct mapped_device *md) > > { > > int r; > > @@ -330,7 +330,7 @@ static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode, > > > > old_dev = dd->dm_dev; > > > > - r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, > > + r = dm_get_table_device(md, dd->dm_dev->bdev->bd_dev, non_exclusive, > > dd->dm_dev->mode | new_mode, &new_dev); > > if (r) > > return r; > > @@ -387,7 +387,8 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > > if (!dd) > > return -ENOMEM; > > > > - if ((r = dm_get_table_device(t->md, dev, mode, &dd->dm_dev))) { > > + r = dm_get_table_device(t->md, dev, mode, ti->non_exclusive, &dd->dm_dev); > > + if (r) { > > kfree(dd); > > return r; > > } > > @@ -396,8 +397,9 @@ int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode, > > list_add(&dd->list, &t->devices); > > goto out; > > > > - } else if (dd->dm_dev->mode != (mode | dd->dm_dev->mode)) { > > - r = upgrade_mode(dd, mode, t->md); > > + } else if ((dd->dm_dev->mode != (mode | dd->dm_dev->mode)) && > > + (dd->dm_dev->non_exclusive != ti->non_exclusive)) { > > + r = upgrade_mode(dd, mode, ti->non_exclusive, t->md); > > if (r) > > return r; > > } > > diff --git a/drivers/md/dm.c b/drivers/md/dm.c > > index 7bac564f3faa..3b871d98b7b6 100644 > > --- a/drivers/md/dm.c > > +++ b/drivers/md/dm.c > > @@ -28,6 +28,7 @@ > > #include <linux/refcount.h> > > #include <linux/part_stat.h> > > #include <linux/blk-crypto.h> > > +#include <linux/interval_tree_generic.h> > > > > #define DM_MSG_PREFIX "core" > > > > @@ -56,6 +57,8 @@ static struct workqueue_struct *deferred_remove_workqueue; > > atomic_t dm_global_event_nr = ATOMIC_INIT(0); > > DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq); > > > > +static DEFINE_MUTEX(interposer_mutex); /* synchronizing access to blk_interposer */ > > Why not dm_interposer_mutex as the name ? And the comment is not very useful: a > mutex is always for synchronizing :) Right. I'll do it. > > > + > > void dm_issue_global_event(void) > > { > > atomic_inc(&dm_global_event_nr); > > @@ -162,6 +165,26 @@ struct table_device { > > struct dm_dev dm_dev; > > }; > > > > +/* > > + * Device mapper`s interposer. > > + */ > > +struct dm_interposer { > > + struct blk_interposer blk_ip; > > + struct mapped_device *md; > > + > > + struct kref kref; > > + struct rw_semaphore ip_devs_lock; > > + struct rb_root_cached ip_devs_root; /* dm_interposed_dev tree */ > > +}; > > + > > +/* > > + * Interval tree for device mapper > > + */ > > +#define START(node) ((node)->start) > > +#define LAST(node) ((node)->last) > > +INTERVAL_TREE_DEFINE(struct dm_rb_range, node, sector_t, _subtree_last, > > + START, LAST,, dm_rb); > > + > > /* > > * Bio-based DM's mempools' reserved IOs set by the user. > > */ > > @@ -733,28 +756,340 @@ static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU) > > rcu_read_unlock(); > > } > > > > +static void dm_submit_bio_interposer_fn(struct bio *bio) > > +{ > > + struct dm_interposer *ip; > > + unsigned int noio_flag = 0; > > + sector_t start; > > + sector_t last; > > + struct dm_rb_range *node; > > + > > + ip = container_of(bio->bi_disk->interposer, struct dm_interposer, blk_ip); > > + start = bio->bi_iter.bi_sector; > > + last = start + dm_sector_div_up(bio->bi_iter.bi_size, SECTOR_SIZE); > > + > > + noio_flag = memalloc_noio_save(); > > + down_read(&ip->ip_devs_lock); > > + node = dm_rb_iter_first(&ip->ip_devs_root, start, last); > > + while (node) { > > + struct dm_interposed_dev *ip_dev = > > + container_of(node, struct dm_interposed_dev, node); > > + > > + atomic64_inc(&ip_dev->ip_cnt); > > + ip_dev->dm_interpose_bio(ip_dev->context, node, bio); > > + > > + node = dm_rb_iter_next(node, start, last); > > + } > > + up_read(&ip->ip_devs_lock); > > + memalloc_noio_restore(noio_flag); > > +} > > + > > +static void free_interposer(struct kref *kref) > > +{ > > + struct dm_interposer *ip = container_of(kref, struct dm_interposer, kref); > > + > > + blk_interposer_detach(&ip->blk_ip, dm_submit_bio_interposer_fn); > > No queue freeze ? Yes. The queue should be already freeze. > > > + > > + kfree(ip); > > +} > > + > > +static struct dm_interposer *new_interposer(struct gendisk *disk) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip; > > + > > + ip = kzalloc(sizeof(struct dm_interposer), GFP_NOIO); > > + if (!ip) > > + return ERR_PTR(-ENOMEM); > > + > > + kref_init(&ip->kref); > > + init_rwsem(&ip->ip_devs_lock); > > + ip->ip_devs_root = RB_ROOT_CACHED; > > + > > + ret = blk_interposer_attach(disk, &ip->blk_ip, dm_submit_bio_interposer_fn); > > No queue freeze ? Yes, again. > > > + if (ret) { > > + DMERR("Failed to attack blk_interposer"); > > + kref_put(&ip->kref, free_interposer); > > + return ERR_PTR(ret); > > + } > > + > > + return ip; > > +} > > + > > +static struct dm_interposer *get_interposer(struct gendisk *disk) > > +{ > > + struct dm_interposer *ip; > > + > > + if (!blk_has_interposer(disk)) > > + return NULL; > > + > > + if (disk->interposer->ip_submit_bio != dm_submit_bio_interposer_fn) { > > + DMERR("Disks interposer slot already occupied."); > > + return ERR_PTR(-EBUSY); > > This is weird... If there is an interposer, why not get a ref on that one. That > is what the function name suggests at least. Getting a ref on that just below in this function. But the name "get_dm_interposer" would be better. > > > + } > > + > > + ip = container_of(disk->interposer, struct dm_interposer, blk_ip); > > + > > + kref_get(&ip->kref); > > + return ip; > > +} > > + > > +struct dm_interposed_dev *dm_interposer_new_dev(struct gendisk *disk, sector_t ofs, sector_t len, > > + void *context, dm_interpose_bio_t dm_interpose_bio) > > +{ > > + sector_t start = ofs; > > + sector_t last = ofs + len - 1; > > + struct dm_interposed_dev *ip_dev = NULL; > > + > > + /* Allocate new ip_dev */ > > + ip_dev = kzalloc(sizeof(struct dm_interposed_dev), GFP_KERNEL); > > + if (!ip_dev) > > + return NULL; > > + > > + ip_dev->disk = disk; > > + ip_dev->node.start = start; > > + ip_dev->node.last = last; > > + > > + ip_dev->context = context; > > + ip_dev->dm_interpose_bio = dm_interpose_bio; > > + > > + atomic64_set(&ip_dev->ip_cnt, 0); > > + > > + return ip_dev; > > +} > > + > > +void dm_interposer_free_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + kfree(ip_dev); > > +} > > Make this inline may be ? Yes. Or even remove this function. > > > + > > +static inline void dm_disk_freeze(struct gendisk *disk) > > +{ > > + blk_mq_freeze_queue(disk->queue); > > + blk_mq_quiesce_queue(disk->queue); > > I think you can replace this with blk_mq_freeze_queue_wait(). I think no. blk_freeze_queue_start() also is required. > > > +} > > + > > +static inline void dm_disk_unfreeze(struct gendisk *disk) > > +{ > > + blk_mq_unquiesce_queue(disk->queue); > > + blk_mq_unfreeze_queue(disk->queue); > > +} > > + > > +int dm_interposer_attach_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip = NULL; > > + unsigned int noio_flag = 0; > > + > > + if (!ip_dev) > > + return -EINVAL; > > + > > + dm_disk_freeze(ip_dev->disk); > > + mutex_lock(&interposer_mutex); > > + noio_flag = memalloc_noio_save(); > > + > > + ip = get_interposer(ip_dev->disk); > > + if (ip == NULL) > > + ip = new_interposer(ip_dev->disk); > > + if (IS_ERR(ip)) { > > + ret = PTR_ERR(ip); > > + goto out; > > + } > > + > > + /* Attach dm_interposed_dev to dm_interposer */ > > + down_write(&ip->ip_devs_lock); > > + do { > > + struct dm_rb_range *node; > > + > > + /* checking that ip_dev already exists for this region */ > > + node = dm_rb_iter_first(&ip->ip_devs_root, ip_dev->node.start, ip_dev->node.last); > > + if (node) { > > + DMERR("Disk part form [%llu] to [%llu] already have interposer", > > + node->start, node->last); > > + > > + ret = -EBUSY; > > + break; > > + } > > + > > + /* insert ip_dev to ip tree */ > > + dm_rb_insert(&ip_dev->node, &ip->ip_devs_root); > > + /* increment ip reference counter */ > > + kref_get(&ip->kref); > > + } while (false); > > + up_write(&ip->ip_devs_lock); > > + > > + kref_put(&ip->kref, free_interposer); > > + > > +out: > > + memalloc_noio_restore(noio_flag); > > + mutex_unlock(&interposer_mutex); > > + dm_disk_unfreeze(ip_dev->disk); > > + > > + return ret; > > +} > > + > > +int dm_interposer_detach_dev(struct dm_interposed_dev *ip_dev) > > +{ > > + int ret = 0; > > + struct dm_interposer *ip = NULL; > > + unsigned int noio_flag = 0; > > + > > + if (!ip_dev) > > + return -EINVAL; > > + > > + dm_disk_freeze(ip_dev->disk); > > + mutex_lock(&interposer_mutex); > > + noio_flag = memalloc_noio_save(); > > + > > + ip = get_interposer(ip_dev->disk); > > + if (IS_ERR(ip)) { > > + ret = PTR_ERR(ip); > > + DMERR("Interposer not found"); > > + goto out; > > + } > > + if (unlikely(ip == NULL)) { > > + ret = -ENXIO; > > + DMERR("Interposer not found"); > > + goto out; > > + } > > + > > + down_write(&ip->ip_devs_lock); > > + do { > > + dm_rb_remove(&ip_dev->node, &ip->ip_devs_root); > > + /* the reference counter here cannot be zero */ > > + kref_put(&ip->kref, free_interposer); > > + > > + } while (false); > > + up_write(&ip->ip_devs_lock); > > + > > + /* detach and free interposer if it`s not needed */ > > s/`/'/ Thanks. It's my problem. > > > + kref_put(&ip->kref, free_interposer); > > +out: > > + memalloc_noio_restore(noio_flag); > > + mutex_unlock(&interposer_mutex); > > + dm_disk_unfreeze(ip_dev->disk); > > + > > + return ret; > > +} > > + > > +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) > > +{ > > + struct mapped_device *md = context; > > + > > + /* Set acceptor device. */ > > + bio->bi_disk = md->disk; > > + > > + /* Remap disks offset */ > > + bio->bi_iter.bi_sector -= node->start; > > + > > + /* > > + * bio should be resubmitted. > > + * We can just add bio to bio_list of the current process. > > + * current->bio_list must be initialized when this function is called. > > + * If call submit_bio_noacct(), the bio will be checked twice. > > + */ > > + BUG_ON(!current->bio_list); > > + bio_list_add(¤t->bio_list[0], bio); > > +} > > + > > +int dm_remap_install(struct mapped_device *md, const char *donor_device_name) > > +{ > > + int ret = 0; > > + struct block_device *donor_bdev; > > + fmode_t mode = FMODE_READ | FMODE_WRITE; > > + > > + DMDEBUG("Dm remap install for mapped device %s and donor device %s", > > + md->name, donor_device_name); > > + > > + donor_bdev = blkdev_get_by_path(donor_device_name, mode, "device-mapper remap"); > > + if (IS_ERR(donor_bdev)) { > > + DMERR("Cannot open device [%s]", donor_device_name); > > + return PTR_ERR(donor_bdev); > > + } > > + > > + do { > > + sector_t ofs = get_start_sect(donor_bdev); > > + sector_t len = bdev_nr_sectors(donor_bdev); > > + > > + md->ip_dev = dm_interposer_new_dev(donor_bdev->bd_disk, ofs, len, md, dm_remap_fn); > > + if (!md->ip_dev) { > > + ret = -ENOMEM; > > + break; > > + } > > + > > + DMDEBUG("New interposed device 0x%p", md->ip_dev); > > + ret = dm_interposer_attach_dev(md->ip_dev); > > + if (ret) { > > + dm_interposer_free_dev(md->ip_dev); > > + > > + md->ip_dev = NULL; > > + DMERR("Failed to attach dm interposer"); > > + break; > > + } > > + > > + DMDEBUG("Attached successfully."); > > + } while (false); > > + > > + blkdev_put(donor_bdev, mode); > > + > > + return ret; > > +} > > + > > +int dm_remap_uninstall(struct mapped_device *md) > > +{ > > + int ret = 0; > > + > > + DMDEBUG("Dm remap uninstall for mapped device %s ip_dev=0x%p", md->name, md->ip_dev); > > + > > + if (!md->ip_dev) { > > + DMERR("Cannot detach dm interposer"); > > + return -EINVAL; > > + } > > + > > + ret = dm_interposer_detach_dev(md->ip_dev); > > + if (ret) { > > + DMERR("Failed to detach dm interposer"); > > + return ret; > > + } > > + > > + DMDEBUG("Detached successfully. %llu bios was interposed", > > + atomic64_read(&md->ip_dev->ip_cnt)); > > + dm_interposer_free_dev(md->ip_dev); > > + md->ip_dev = NULL; > > + > > + return 0; > > +} > > + > > static char *_dm_claim_ptr = "I belong to device-mapper"; > > > > /* > > * Open a table device so we can use it as a map destination. > > */ > > static int open_table_device(struct table_device *td, dev_t dev, > > - struct mapped_device *md) > > + struct mapped_device *md, bool non_exclusive) > > { > > struct block_device *bdev; > > - > > - int r; > > + int ret; > > > > BUG_ON(td->dm_dev.bdev); > > > > - bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > - if (IS_ERR(bdev)) > > - return PTR_ERR(bdev); > > + if (non_exclusive) > > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode, NULL); > > + else > > + bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr); > > > > - r = bd_link_disk_holder(bdev, dm_disk(md)); > > - if (r) { > > - blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL); > > - return r; > > + if (IS_ERR(bdev)) { > > + ret = PTR_ERR(bdev); > > + if (ret != -EBUSY) > > + return ret; > > + } > > + > > + if (!non_exclusive) { > > + ret = bd_link_disk_holder(bdev, dm_disk(md)); > > + if (ret) { > > + blkdev_put(bdev, td->dm_dev.mode); > > + return ret; > > + } > > } > > > > td->dm_dev.bdev = bdev; > > @@ -770,33 +1105,38 @@ static void close_table_device(struct table_device *td, struct mapped_device *md > > if (!td->dm_dev.bdev) > > return; > > > > - bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > > - blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL); > > + if (td->dm_dev.mode & FMODE_EXCL) > > + bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md)); > > + > > + blkdev_put(td->dm_dev.bdev, td->dm_dev.mode); > > + > > put_dax(td->dm_dev.dax_dev); > > td->dm_dev.bdev = NULL; > > td->dm_dev.dax_dev = NULL; > > } > > > > static struct table_device *find_table_device(struct list_head *l, dev_t dev, > > - fmode_t mode) > > + fmode_t mode, bool non_exclusive) > > { > > struct table_device *td; > > > > list_for_each_entry(td, l, list) > > - if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode) > > + if (td->dm_dev.bdev->bd_dev == dev && > > + td->dm_dev.mode == mode && > > + td->dm_dev.non_exclusive == non_exclusive) > > return td; > > > > return NULL; > > } > > > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > > struct dm_dev **result) > > { > > int r; > > struct table_device *td; > > > > mutex_lock(&md->table_devices_lock); > > - td = find_table_device(&md->table_devices, dev, mode); > > + td = find_table_device(&md->table_devices, dev, mode, non_exclusive); > > if (!td) { > > td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id); > > if (!td) { > > @@ -807,7 +1147,8 @@ int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > td->dm_dev.mode = mode; > > td->dm_dev.bdev = NULL; > > > > - if ((r = open_table_device(td, dev, md))) { > > + r = open_table_device(td, dev, md, non_exclusive); > > + if (r) { > > mutex_unlock(&md->table_devices_lock); > > kfree(td); > > return r; > > @@ -2182,6 +2523,14 @@ static void __dm_destroy(struct mapped_device *md, bool wait) > > > > might_sleep(); > > > > + if (md->ip_dev) { > > + if (dm_interposer_detach_dev(md->ip_dev)) > > + DMERR("Failed to detach dm interposer"); > > + > > + dm_interposer_free_dev(md->ip_dev); > > + md->ip_dev = NULL; > > + } > > + > > spin_lock(&_minor_lock); > > idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md)))); > > set_bit(DMF_FREEING, &md->flags); > > diff --git a/drivers/md/dm.h b/drivers/md/dm.h > > index fffe1e289c53..7bf20fb2de74 100644 > > --- a/drivers/md/dm.h > > +++ b/drivers/md/dm.h > > @@ -179,7 +179,7 @@ int dm_open_count(struct mapped_device *md); > > int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred); > > int dm_cancel_deferred_remove(struct mapped_device *md); > > int dm_request_based(struct mapped_device *md); > > -int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, > > +int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode, bool non_exclusive, > > struct dm_dev **result); > > void dm_put_table_device(struct mapped_device *md, struct dm_dev *d); > > > > diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h > > index 61a66fb8ebb3..70002363bfc0 100644 > > --- a/include/linux/device-mapper.h > > +++ b/include/linux/device-mapper.h > > @@ -150,6 +150,7 @@ struct dm_dev { > > struct block_device *bdev; > > struct dax_device *dax_dev; > > fmode_t mode; > > + bool non_exclusive; > > char name[16]; > > }; > > > > @@ -325,6 +326,12 @@ struct dm_target { > > * whether or not its underlying devices have support. > > */ > > bool discards_supported:1; > > + > > + /* > > + * Set if this target needs to open device without FMODE_EXCL > > + * mode. > > + */ > > + bool non_exclusive:1; > > }; > > > > void *dm_per_bio_data(struct bio *bio, size_t data_size); > > diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h > > index 4933b6b67b85..08d7dbff80f4 100644 > > --- a/include/uapi/linux/dm-ioctl.h > > +++ b/include/uapi/linux/dm-ioctl.h > > @@ -214,6 +214,15 @@ struct dm_target_msg { > > char message[0]; > > }; > > > > +enum { > > + REMAP_START_CMD = 1, > > + REMAP_FINISH_CMD, > > +}; > > + > > +struct dm_remap_param { > > + uint8_t cmd; > > + uint8_t params[0]; > > +}; > > /* > > * If you change this make sure you make the corresponding change > > * to dm-ioctl.c:lookup_ioctl() > > @@ -244,6 +253,7 @@ enum { > > DM_DEV_SET_GEOMETRY_CMD, > > DM_DEV_ARM_POLL_CMD, > > DM_GET_TARGET_VERSION_CMD, > > + DM_DEV_REMAP_CMD > > }; > > > > #define DM_IOCTL 0xfd > > @@ -259,6 +269,7 @@ enum { > > #define DM_DEV_STATUS _IOWR(DM_IOCTL, DM_DEV_STATUS_CMD, struct dm_ioctl) > > #define DM_DEV_WAIT _IOWR(DM_IOCTL, DM_DEV_WAIT_CMD, struct dm_ioctl) > > #define DM_DEV_ARM_POLL _IOWR(DM_IOCTL, DM_DEV_ARM_POLL_CMD, struct dm_ioctl) > > +#define DM_DEV_REMAP _IOWR(DM_IOCTL, DM_DEV_REMAP_CMD, struct dm_ioctl) > > > > #define DM_TABLE_LOAD _IOWR(DM_IOCTL, DM_TABLE_LOAD_CMD, struct dm_ioctl) > > #define DM_TABLE_CLEAR _IOWR(DM_IOCTL, DM_TABLE_CLEAR_CMD, struct dm_ioctl) > > @@ -272,9 +283,9 @@ enum { > > #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) > > > > #define DM_VERSION_MAJOR 4 > > -#define DM_VERSION_MINOR 43 > > +#define DM_VERSION_MINOR 44 > > #define DM_VERSION_PATCHLEVEL 0 > > -#define DM_VERSION_EXTRA "-ioctl (2020-10-01)" > > +#define DM_VERSION_EXTRA "-ioctl (2020-12-25)" > > > > /* Status bits */ > > #define DM_READONLY_FLAG (1 << 0) /* In/Out */ > > > > > -- > Damien Le Moal > Western Digital Research > -- Sergei Shtepa Veeam Software developer. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-01-28 17:12 ` Sergei Shtepa @ 2021-01-29 0:18 ` Damien Le Moal -1 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 0:18 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > Hi all, > > I`m ready to suggest the blk_interposer again. > blk_interposer allows to intercept bio requests, remap bio to > another devices or add new bios. > > This version has support from device mapper. > > For the dm-linear device creation command, the `noexcl` parameter > has been added, which allows to open block devices without > FMODE_EXCL mode. It allows to create dm-linear device on a block > device with an already mounted file system. > The new ioctl DM_DEV_REMAP allows to enable and disable bio > interception. > > Thus, it is possible to add the dm-device to the block layer stack > without reconfiguring and rebooting. Please add the changelog here instead of adding it in each patch. And keep the changelog for previous versions too (i.e. v1->v2 in this case) so that the changes overall can be tracked. The proper formatting for the title should be [PATCH v3 X/Y] instead of adding v3 in the title itself. With git format-patch, you can use "-v 3" option to format this for you, or --subject-prefix="PATCH v3" option. > > > Sergei Shtepa (2): > block: blk_interposer > [dm] blk_interposer for dm-linear > > block/bio.c | 2 + > block/blk-core.c | 29 +++ > block/blk-mq.c | 13 ++ > block/genhd.c | 82 ++++++++ > drivers/md/dm-core.h | 46 +++- > drivers/md/dm-ioctl.c | 39 ++++ > drivers/md/dm-linear.c | 17 +- > drivers/md/dm-table.c | 12 +- > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > drivers/md/dm.h | 2 +- > include/linux/blk-mq.h | 1 + > include/linux/blk_types.h | 6 +- > include/linux/device-mapper.h | 7 + > include/linux/genhd.h | 19 ++ > include/uapi/linux/dm-ioctl.h | 15 +- > 15 files changed, 643 insertions(+), 30 deletions(-) > -- Damien Le Moal Western Digital Research -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 @ 2021-01-29 0:18 ` Damien Le Moal 0 siblings, 0 replies; 24+ messages in thread From: Damien Le Moal @ 2021-01-29 0:18 UTC (permalink / raw) To: Sergei Shtepa, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org Cc: pavel.tide@veeam.com On 2021/01/29 2:23, Sergei Shtepa wrote: > Hi all, > > I`m ready to suggest the blk_interposer again. > blk_interposer allows to intercept bio requests, remap bio to > another devices or add new bios. > > This version has support from device mapper. > > For the dm-linear device creation command, the `noexcl` parameter > has been added, which allows to open block devices without > FMODE_EXCL mode. It allows to create dm-linear device on a block > device with an already mounted file system. > The new ioctl DM_DEV_REMAP allows to enable and disable bio > interception. > > Thus, it is possible to add the dm-device to the block layer stack > without reconfiguring and rebooting. Please add the changelog here instead of adding it in each patch. And keep the changelog for previous versions too (i.e. v1->v2 in this case) so that the changes overall can be tracked. The proper formatting for the title should be [PATCH v3 X/Y] instead of adding v3 in the title itself. With git format-patch, you can use "-v 3" option to format this for you, or --subject-prefix="PATCH v3" option. > > > Sergei Shtepa (2): > block: blk_interposer > [dm] blk_interposer for dm-linear > > block/bio.c | 2 + > block/blk-core.c | 29 +++ > block/blk-mq.c | 13 ++ > block/genhd.c | 82 ++++++++ > drivers/md/dm-core.h | 46 +++- > drivers/md/dm-ioctl.c | 39 ++++ > drivers/md/dm-linear.c | 17 +- > drivers/md/dm-table.c | 12 +- > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > drivers/md/dm.h | 2 +- > include/linux/blk-mq.h | 1 + > include/linux/blk_types.h | 6 +- > include/linux/device-mapper.h | 7 + > include/linux/genhd.h | 19 ++ > include/uapi/linux/dm-ioctl.h | 15 +- > 15 files changed, 643 insertions(+), 30 deletions(-) > -- Damien Le Moal Western Digital Research ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-01-29 0:18 ` Damien Le Moal (?) @ 2021-01-29 15:39 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-01-29 15:39 UTC (permalink / raw) To: Damien Le Moal Cc: snitzer@redhat.com, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com The 01/29/2021 03:18, Damien Le Moal wrote: > On 2021/01/29 2:23, Sergei Shtepa wrote: > > Hi all, > > > > I`m ready to suggest the blk_interposer again. > > blk_interposer allows to intercept bio requests, remap bio to > > another devices or add new bios. > > > > This version has support from device mapper. > > > > For the dm-linear device creation command, the `noexcl` parameter > > has been added, which allows to open block devices without > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > device with an already mounted file system. > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > interception. > > > > Thus, it is possible to add the dm-device to the block layer stack > > without reconfiguring and rebooting. > > Please add the changelog here instead of adding it in each patch. And keep the > changelog for previous versions too (i.e. v1->v2 in this case) so that the > changes overall can be tracked. > > The proper formatting for the title should be [PATCH v3 X/Y] instead of adding > v3 in the title itself. With git format-patch, you can use "-v 3" option to > format this for you, or --subject-prefix="PATCH v3" option. Thanks. I think I need to work with the style of my patches. > > > > > > > Sergei Shtepa (2): > > block: blk_interposer > > [dm] blk_interposer for dm-linear > > > > block/bio.c | 2 + > > block/blk-core.c | 29 +++ > > block/blk-mq.c | 13 ++ > > block/genhd.c | 82 ++++++++ > > drivers/md/dm-core.h | 46 +++- > > drivers/md/dm-ioctl.c | 39 ++++ > > drivers/md/dm-linear.c | 17 +- > > drivers/md/dm-table.c | 12 +- > > drivers/md/dm.c | 383 ++++++++++++++++++++++++++++++++-- > > drivers/md/dm.h | 2 +- > > include/linux/blk-mq.h | 1 + > > include/linux/blk_types.h | 6 +- > > include/linux/device-mapper.h | 7 + > > include/linux/genhd.h | 19 ++ > > include/uapi/linux/dm-ioctl.h | 15 +- > > 15 files changed, 643 insertions(+), 30 deletions(-) > > > > > -- > Damien Le Moal > Western Digital Research > -- Sergei Shtepa Veeam Software developer. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-01-28 17:12 ` Sergei Shtepa @ 2021-02-01 15:45 ` Bart Van Assche -1 siblings, 0 replies; 24+ messages in thread From: Bart Van Assche @ 2021-02-01 15:45 UTC (permalink / raw) To: Sergei Shtepa, hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: pavel.tide On 1/28/21 9:12 AM, Sergei Shtepa wrote: > I`m ready to suggest the blk_interposer again. > blk_interposer allows to intercept bio requests, remap bio to > another devices or add new bios. > > This version has support from device mapper. > > For the dm-linear device creation command, the `noexcl` parameter > has been added, which allows to open block devices without > FMODE_EXCL mode. It allows to create dm-linear device on a block > device with an already mounted file system. > The new ioctl DM_DEV_REMAP allows to enable and disable bio > interception. > > Thus, it is possible to add the dm-device to the block layer stack > without reconfiguring and rebooting. What functionality does this driver provide that is not yet available in a RAID level 1 (mirroring) driver + a custom dm driver? My understanding is that there are already two RAID level 1 drivers in the kernel tree and that both driver support sending bio's to two different block devices. Thanks, Bart. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 @ 2021-02-01 15:45 ` Bart Van Assche 0 siblings, 0 replies; 24+ messages in thread From: Bart Van Assche @ 2021-02-01 15:45 UTC (permalink / raw) To: Sergei Shtepa, hare, ming.lei, agk, snitzer, dm-devel, linux-block Cc: pavel.tide On 1/28/21 9:12 AM, Sergei Shtepa wrote: > I`m ready to suggest the blk_interposer again. > blk_interposer allows to intercept bio requests, remap bio to > another devices or add new bios. > > This version has support from device mapper. > > For the dm-linear device creation command, the `noexcl` parameter > has been added, which allows to open block devices without > FMODE_EXCL mode. It allows to create dm-linear device on a block > device with an already mounted file system. > The new ioctl DM_DEV_REMAP allows to enable and disable bio > interception. > > Thus, it is possible to add the dm-device to the block layer stack > without reconfiguring and rebooting. What functionality does this driver provide that is not yet available in a RAID level 1 (mirroring) driver + a custom dm driver? My understanding is that there are already two RAID level 1 drivers in the kernel tree and that both driver support sending bio's to two different block devices. Thanks, Bart. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-02-01 15:45 ` Bart Van Assche @ 2021-02-01 18:18 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-02-01 18:18 UTC (permalink / raw) To: Bart Van Assche Cc: snitzer@redhat.com, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com The 02/01/2021 18:45, Bart Van Assche wrote: > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > I`m ready to suggest the blk_interposer again. > > blk_interposer allows to intercept bio requests, remap bio to > > another devices or add new bios. > > > > This version has support from device mapper. > > > > For the dm-linear device creation command, the `noexcl` parameter > > has been added, which allows to open block devices without > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > device with an already mounted file system. > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > interception. > > > > Thus, it is possible to add the dm-device to the block layer stack > > without reconfiguring and rebooting. > > What functionality does this driver provide that is not yet available in > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > is that there are already two RAID level 1 drivers in the kernel tree > and that both driver support sending bio's to two different block devices. > > Thanks, > > Bart. Hi Bart. The proposed patch is not realy aimed at RAID1. Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio requests from the regular block device to the new DM device using the blk_interposer will allow to use device mapper for regular devices. For dm-linear, there is not much benefit from using blk_interposer. This is a good and illustrative example. Later, using blk-interposer, it will be possible to connect the dm-cache "on the fly" without having to reboot and/or reconfigure. My intention is to let users use dm-snap to create snapshots of any device. blk-interposer will allow to add new features to Device Mapper. As per Daniel's advice I want to add a documentation, I'm working on it now. The documentation will also contain a description of new features that blk_interposer will add to Device Mapper Thanks. -- Sergei Shtepa Veeam Software developer. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 @ 2021-02-01 18:18 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-02-01 18:18 UTC (permalink / raw) To: Bart Van Assche Cc: hare@suse.de, ming.lei@redhat.com, agk@redhat.com, snitzer@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org, Pavel Tide The 02/01/2021 18:45, Bart Van Assche wrote: > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > I`m ready to suggest the blk_interposer again. > > blk_interposer allows to intercept bio requests, remap bio to > > another devices or add new bios. > > > > This version has support from device mapper. > > > > For the dm-linear device creation command, the `noexcl` parameter > > has been added, which allows to open block devices without > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > device with an already mounted file system. > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > interception. > > > > Thus, it is possible to add the dm-device to the block layer stack > > without reconfiguring and rebooting. > > What functionality does this driver provide that is not yet available in > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > is that there are already two RAID level 1 drivers in the kernel tree > and that both driver support sending bio's to two different block devices. > > Thanks, > > Bart. Hi Bart. The proposed patch is not realy aimed at RAID1. Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio requests from the regular block device to the new DM device using the blk_interposer will allow to use device mapper for regular devices. For dm-linear, there is not much benefit from using blk_interposer. This is a good and illustrative example. Later, using blk-interposer, it will be possible to connect the dm-cache "on the fly" without having to reboot and/or reconfigure. My intention is to let users use dm-snap to create snapshots of any device. blk-interposer will allow to add new features to Device Mapper. As per Daniel's advice I want to add a documentation, I'm working on it now. The documentation will also contain a description of new features that blk_interposer will add to Device Mapper Thanks. -- Sergei Shtepa Veeam Software developer. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-02-01 18:18 ` Sergei Shtepa @ 2021-02-01 18:50 ` Mike Snitzer -1 siblings, 0 replies; 24+ messages in thread From: Mike Snitzer @ 2021-02-01 18:50 UTC (permalink / raw) To: Sergei Shtepa Cc: Bart Van Assche, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com On Mon, Feb 01 2021 at 1:18pm -0500, Sergei Shtepa <sergei.shtepa@veeam.com> wrote: > The 02/01/2021 18:45, Bart Van Assche wrote: > > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > > I`m ready to suggest the blk_interposer again. > > > blk_interposer allows to intercept bio requests, remap bio to > > > another devices or add new bios. > > > > > > This version has support from device mapper. > > > > > > For the dm-linear device creation command, the `noexcl` parameter > > > has been added, which allows to open block devices without > > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > > device with an already mounted file system. > > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > > interception. > > > > > > Thus, it is possible to add the dm-device to the block layer stack > > > without reconfiguring and rebooting. > > > > What functionality does this driver provide that is not yet available in > > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > > is that there are already two RAID level 1 drivers in the kernel tree > > and that both driver support sending bio's to two different block devices. > > > > Thanks, > > > > Bart. > > Hi Bart. > > The proposed patch is not realy aimed at RAID1. > > Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio > requests from the regular block device to the new DM device using > the blk_interposer will allow to use device mapper for regular devices. > For dm-linear, there is not much benefit from using blk_interposer. > This is a good and illustrative example. Later, using blk-interposer, > it will be possible to connect the dm-cache "on the fly" without having > to reboot and/or reconfigure. > My intention is to let users use dm-snap to create snapshots of any device. > blk-interposer will allow to add new features to Device Mapper. > > As per Daniel's advice I want to add a documentation, I'm working on it now. > The documentation will also contain a description of new features that > blk_interposer will add to Device Mapper More Documentation is fine, but the code needs to be improved and fully formed before you start trying to polish with Documentation -- definitely don't put time to Documentation that is speculative! You'd do well to focus on an implementation that doesn't require an extra clone if interposed device will use DM (DM core already handles cloning all incoming bios). Mike -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH 0/2] block: blk_interposer v3 @ 2021-02-01 18:50 ` Mike Snitzer 0 siblings, 0 replies; 24+ messages in thread From: Mike Snitzer @ 2021-02-01 18:50 UTC (permalink / raw) To: Sergei Shtepa Cc: Bart Van Assche, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org, Pavel Tide On Mon, Feb 01 2021 at 1:18pm -0500, Sergei Shtepa <sergei.shtepa@veeam.com> wrote: > The 02/01/2021 18:45, Bart Van Assche wrote: > > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > > I`m ready to suggest the blk_interposer again. > > > blk_interposer allows to intercept bio requests, remap bio to > > > another devices or add new bios. > > > > > > This version has support from device mapper. > > > > > > For the dm-linear device creation command, the `noexcl` parameter > > > has been added, which allows to open block devices without > > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > > device with an already mounted file system. > > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > > interception. > > > > > > Thus, it is possible to add the dm-device to the block layer stack > > > without reconfiguring and rebooting. > > > > What functionality does this driver provide that is not yet available in > > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > > is that there are already two RAID level 1 drivers in the kernel tree > > and that both driver support sending bio's to two different block devices. > > > > Thanks, > > > > Bart. > > Hi Bart. > > The proposed patch is not realy aimed at RAID1. > > Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio > requests from the regular block device to the new DM device using > the blk_interposer will allow to use device mapper for regular devices. > For dm-linear, there is not much benefit from using blk_interposer. > This is a good and illustrative example. Later, using blk-interposer, > it will be possible to connect the dm-cache "on the fly" without having > to reboot and/or reconfigure. > My intention is to let users use dm-snap to create snapshots of any device. > blk-interposer will allow to add new features to Device Mapper. > > As per Daniel's advice I want to add a documentation, I'm working on it now. > The documentation will also contain a description of new features that > blk_interposer will add to Device Mapper More Documentation is fine, but the code needs to be improved and fully formed before you start trying to polish with Documentation -- definitely don't put time to Documentation that is speculative! You'd do well to focus on an implementation that doesn't require an extra clone if interposed device will use DM (DM core already handles cloning all incoming bios). Mike ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [dm-devel] [PATCH 0/2] block: blk_interposer v3 2021-02-01 18:50 ` Mike Snitzer @ 2021-02-02 1:24 ` Sergei Shtepa -1 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-02-02 1:24 UTC (permalink / raw) To: Mike Snitzer Cc: Bart Van Assche, Pavel Tide, ming.lei@redhat.com, linux-block@vger.kernel.org, dm-devel@redhat.com, agk@redhat.com The 02/01/2021 21:50, Mike Snitzer wrote: > On Mon, Feb 01 2021 at 1:18pm -0500, > Sergei Shtepa <sergei.shtepa@veeam.com> wrote: > > > The 02/01/2021 18:45, Bart Van Assche wrote: > > > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > > > I`m ready to suggest the blk_interposer again. > > > > blk_interposer allows to intercept bio requests, remap bio to > > > > another devices or add new bios. > > > > > > > > This version has support from device mapper. > > > > > > > > For the dm-linear device creation command, the `noexcl` parameter > > > > has been added, which allows to open block devices without > > > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > > > device with an already mounted file system. > > > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > > > interception. > > > > > > > > Thus, it is possible to add the dm-device to the block layer stack > > > > without reconfiguring and rebooting. > > > > > > What functionality does this driver provide that is not yet available in > > > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > > > is that there are already two RAID level 1 drivers in the kernel tree > > > and that both driver support sending bio's to two different block devices. > > > > > > Thanks, > > > > > > Bart. > > > > Hi Bart. > > > > The proposed patch is not realy aimed at RAID1. > > > > Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio > > requests from the regular block device to the new DM device using > > the blk_interposer will allow to use device mapper for regular devices. > > For dm-linear, there is not much benefit from using blk_interposer. > > This is a good and illustrative example. Later, using blk-interposer, > > it will be possible to connect the dm-cache "on the fly" without having > > to reboot and/or reconfigure. > > My intention is to let users use dm-snap to create snapshots of any device. > > blk-interposer will allow to add new features to Device Mapper. > > > > As per Daniel's advice I want to add a documentation, I'm working on it now. > > The documentation will also contain a description of new features that > > blk_interposer will add to Device Mapper > > More Documentation is fine, but the code needs to be improved and fully > formed before you start trying to polish with Documentation -- > definitely don't put time to Documentation that is speculative! > > You'd do well to focus on an implementation that doesn't require an > extra clone if interposed device will use DM (DM core already handles > cloning all incoming bios). > > Mike > Hi Mike. I agree that quality code is more important than documentation. And I think this code has already solved the problem of a bio extra cloning. +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) +{ + struct mapped_device *md = context; + + /* Set acceptor device. */ + bio->bi_disk = md->disk; + + /* Remap disks offset */ + bio->bi_iter.bi_sector -= node->start; + + /* + * bio should be resubmitted. + * We can just add bio to bio_list of the current process. + * current->bio_list must be initialized when this function is called. + * If call submit_bio_noacct(), the bio will be checked twice. + */ + BUG_ON(!current->bio_list); + bio_list_add(¤t->bio_list[0], bio); +} -- Sergei Shtepa Veeam Software developer. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH 0/2] block: blk_interposer v3 @ 2021-02-02 1:24 ` Sergei Shtepa 0 siblings, 0 replies; 24+ messages in thread From: Sergei Shtepa @ 2021-02-02 1:24 UTC (permalink / raw) To: Mike Snitzer Cc: Bart Van Assche, hare@suse.de, ming.lei@redhat.com, agk@redhat.com, dm-devel@redhat.com, linux-block@vger.kernel.org, Pavel Tide The 02/01/2021 21:50, Mike Snitzer wrote: > On Mon, Feb 01 2021 at 1:18pm -0500, > Sergei Shtepa <sergei.shtepa@veeam.com> wrote: > > > The 02/01/2021 18:45, Bart Van Assche wrote: > > > On 1/28/21 9:12 AM, Sergei Shtepa wrote: > > > > I`m ready to suggest the blk_interposer again. > > > > blk_interposer allows to intercept bio requests, remap bio to > > > > another devices or add new bios. > > > > > > > > This version has support from device mapper. > > > > > > > > For the dm-linear device creation command, the `noexcl` parameter > > > > has been added, which allows to open block devices without > > > > FMODE_EXCL mode. It allows to create dm-linear device on a block > > > > device with an already mounted file system. > > > > The new ioctl DM_DEV_REMAP allows to enable and disable bio > > > > interception. > > > > > > > > Thus, it is possible to add the dm-device to the block layer stack > > > > without reconfiguring and rebooting. > > > > > > What functionality does this driver provide that is not yet available in > > > a RAID level 1 (mirroring) driver + a custom dm driver? My understanding > > > is that there are already two RAID level 1 drivers in the kernel tree > > > and that both driver support sending bio's to two different block devices. > > > > > > Thanks, > > > > > > Bart. > > > > Hi Bart. > > > > The proposed patch is not realy aimed at RAID1. > > > > Creating a new dm device in the non-FMODE_EXCL mode and then remaping bio > > requests from the regular block device to the new DM device using > > the blk_interposer will allow to use device mapper for regular devices. > > For dm-linear, there is not much benefit from using blk_interposer. > > This is a good and illustrative example. Later, using blk-interposer, > > it will be possible to connect the dm-cache "on the fly" without having > > to reboot and/or reconfigure. > > My intention is to let users use dm-snap to create snapshots of any device. > > blk-interposer will allow to add new features to Device Mapper. > > > > As per Daniel's advice I want to add a documentation, I'm working on it now. > > The documentation will also contain a description of new features that > > blk_interposer will add to Device Mapper > > More Documentation is fine, but the code needs to be improved and fully > formed before you start trying to polish with Documentation -- > definitely don't put time to Documentation that is speculative! > > You'd do well to focus on an implementation that doesn't require an > extra clone if interposed device will use DM (DM core already handles > cloning all incoming bios). > > Mike > Hi Mike. I agree that quality code is more important than documentation. And I think this code has already solved the problem of a bio extra cloning. +static void dm_remap_fn(void *context, struct dm_rb_range *node, struct bio *bio) +{ + struct mapped_device *md = context; + + /* Set acceptor device. */ + bio->bi_disk = md->disk; + + /* Remap disks offset */ + bio->bi_iter.bi_sector -= node->start; + + /* + * bio should be resubmitted. + * We can just add bio to bio_list of the current process. + * current->bio_list must be initialized when this function is called. + * If call submit_bio_noacct(), the bio will be checked twice. + */ + BUG_ON(!current->bio_list); + bio_list_add(¤t->bio_list[0], bio); +} -- Sergei Shtepa Veeam Software developer. ^ permalink raw reply [flat|nested] 24+ messages in thread
end of thread, other threads:[~2021-02-02 1:25 UTC | newest] Thread overview: 24+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2021-01-28 17:12 [dm-devel] [PATCH 0/2] block: blk_interposer v3 Sergei Shtepa 2021-01-28 17:12 ` Sergei Shtepa 2021-01-28 17:12 ` [dm-devel] [PATCH 1/2] block: blk_interposer Sergei Shtepa 2021-01-28 17:12 ` Sergei Shtepa 2021-01-29 1:21 ` [dm-devel] " Damien Le Moal 2021-01-29 1:21 ` Damien Le Moal 2021-01-29 15:58 ` Sergei Shtepa 2021-01-28 17:12 ` [dm-devel] [PATCH 2/2] [dm] blk_interposer for dm-linear Sergei Shtepa 2021-01-28 17:12 ` Sergei Shtepa 2021-01-29 1:46 ` [dm-devel] " Damien Le Moal 2021-01-29 1:46 ` Damien Le Moal 2021-01-29 16:08 ` Sergei Shtepa 2021-01-29 16:08 ` Sergei Shtepa 2021-01-29 0:18 ` [dm-devel] [PATCH 0/2] block: blk_interposer v3 Damien Le Moal 2021-01-29 0:18 ` Damien Le Moal 2021-01-29 15:39 ` Sergei Shtepa 2021-02-01 15:45 ` Bart Van Assche 2021-02-01 15:45 ` Bart Van Assche 2021-02-01 18:18 ` Sergei Shtepa 2021-02-01 18:18 ` Sergei Shtepa 2021-02-01 18:50 ` Mike Snitzer 2021-02-01 18:50 ` Mike Snitzer 2021-02-02 1:24 ` [dm-devel] " Sergei Shtepa 2021-02-02 1:24 ` Sergei Shtepa
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.