make secure erase and write zeroes ioctls interruptible as well
Subsystems (2)
Hi all,
Following discard in the last merge window, this series also makes secure erase and discard interruptible by fatal signals.
The secure erase side is a straight port of the discard support. Unfortunately I don't have a way to test it, so I'm adding the eMMC maintainer as that is where the support originated so maybe they can give it a spin? (just do a blkdiscard -f -s /dev/<dev> and then Ctrl+C)
The write zeroes support is a bit different as it is more complex due to the fallback code and there already is a helper taking flags that we piggy back on. This side has been extensively tested.
Diffstat:
block/blk-lib.c | 267 +++++++++++++++++++++++++++---------------------- block/ioctl.c | 44 +++++++- include/linux/bio.h | 2 include/linux/blkdev.h | 1 4 files changed, 193 insertions(+), 121 deletions(-)
PATCH 1/10 block: move secure erase checks into the ioctl handler
Most bio operations get basic sanity checking in submit_bio and anything more complicated than that is done in the callers. Secure erase is a bit different from that in that a lot of checking is done in blkdev_issue_secure_erase, and the specific errnos for that are returned to userspace. Move the checks that require specific errnos to the ioctl handler instead, and just leave the basic sanity checking in submit_bio
--- block/blk-lib.c | 7 ------- block/ioctl.c | 8 +++++++- 2 files changed, 7 insertions(+), 8 deletions(-)
block/blk-lib.c +0 -7
@@ -299,13 +299,6 @@ int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, max_sectors =UINT_MAX >>SECTOR_SHIFT ; max_sectors &= ~bs_mask; -if (max_sectors ==0 ) -return -EOPNOTSUPP ; -if ((sector | nr_sects) & bs_mask) -return -EINVAL ; -if (bdev_read_only (bdev)) -return -EPERM ; -blk_start_plug (&plug);while (nr_sects) {unsigned int len =min_t (sector_t, nr_sects, max_sectors);
block/ioctl.c +7 -1
@@ -163,6 +163,7 @@ static int blk_ioctl_discard(struct block_device *bdev, blk_mode_t mode,static int blk_ioctl_secure_erase (struct block_device *bdev,blk_mode_t mode,void __user *argp) { +unsigned int bs_mask =bdev_logical_block_size (bdev) -1 ;uint64_t start, len;uint64_t range[2 ];int err;
@@ -171,12 +172,17 @@ static int blk_ioctl_secure_erase(struct block_device *bdev, blk_mode_t mode,return -EBADF ;if (!bdev_max_secure_erase_sectors (bdev))return -EOPNOTSUPP ; +if (bdev_read_only (bdev)) +return -EPERM ;if (copy_from_user (range, argp,sizeof (range)))return -EFAULT ; start = range[0 ]; len = range[1 ]; -if ((start &511 ) || (len &511 )) + +if (!len) +return -EINVAL ; +if ((start | len) & bs_mask)return -EINVAL ;if (start + len >bdev_nr_bytes (bdev))return -EINVAL ;
Signature · 1 line
2.43.0
PATCH 2/10 block: factor out a bio_secure_erase_limit helper
In reply to Christoph HellwigAdd a helper to size the maximum secure erase bio without exceeding the queue limit or overlowing the bi_size member.
--- block/blk-lib.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-)
block/blk-lib.c +9 -7
@@ -285,20 +285,22 @@ int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, }EXPORT_SYMBOL (blkdev_issue_zeroout); +static sector_t bio_secure_erase_limit (struct block_device *bdev) +{ +sector_t bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ; + +return min (bdev_max_secure_erase_sectors (bdev), + (UINT_MAX >>SECTOR_SHIFT ) & ~bs_mask); +} +int blkdev_issue_secure_erase (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp) { -sector_t bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ; -unsigned int max_sectors =bdev_max_secure_erase_sectors (bdev); +unsigned int max_sectors =bio_secure_erase_limit (bdev);struct bio *bio =NULL ;struct blk_plug plug;int ret =0 ; -/* make sure that "len << SECTOR_SHIFT" doesn't overflow */ -if (max_sectors >UINT_MAX >>SECTOR_SHIFT ) - max_sectors =UINT_MAX >>SECTOR_SHIFT ; - max_sectors &= ~bs_mask; -blk_start_plug (&plug);while (nr_sects) {unsigned int len =min_t (sector_t, nr_sects, max_sectors);
Signature · 1 line
2.43.0
PATCH 3/10 block: add a blk_alloc_secure_erase_bio helper
In reply to Christoph HellwigFactor out a helper from blkdev_issue_secure_erase that chews off as much as possible from a secure_erase range and allocates a bio for it.
--- block/blk-lib.c | 36 +++++++++++++++++++++++------------- include/linux/bio.h | 2 ++ 2 files changed, 25 insertions(+), 13 deletions(-)
block/blk-lib.c +23 -13
@@ -293,26 +293,36 @@ static sector_t bio_secure_erase_limit(struct block_device *bdev) (UINT_MAX >> SECTOR_SHIFT) & ~bs_mask); } +struct bio *blk_alloc_secure_erase_bio (struct block_device *bdev, +sector_t *sector,sector_t *nr_sects,gfp_t gfp) +{ +sector_t bio_sects =min (*nr_sects,bio_secure_erase_limit (bdev)); +struct bio *bio; + +if (!bio_sects) +return NULL ; + bio =bio_alloc (bdev,0 ,REQ_OP_SECURE_ERASE , gfp); +if (!bio) +return NULL ; + bio->bi_iter.bi_sector = *sector; + bio->bi_iter.bi_size = bio_sects <<SECTOR_SHIFT ; + *sector += bio_sects; + *nr_sects -= bio_sects; +cond_resched (); +return bio; +} +int blkdev_issue_secure_erase (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp) { -unsigned int max_sectors =bio_secure_erase_limit (bdev); -struct bio *bio =NULL ; +struct bio *bio =NULL , *next;struct blk_plug plug;int ret =0 ;blk_start_plug (&plug); -while (nr_sects) { -unsigned int len =min_t (sector_t, nr_sects, max_sectors); - - bio =blk_next_bio (bio, bdev,0 ,REQ_OP_SECURE_ERASE , gfp); - bio->bi_iter.bi_sector = sector; - bio->bi_iter.bi_size = len <<SECTOR_SHIFT ; - - sector += len; - nr_sects -= len; -cond_resched (); - } +while ((next =blk_alloc_secure_erase_bio (bdev, §or, &nr_sects, + gfp))) + bio =bio_chain_and_submit (bio, next);if (bio) { ret =submit_bio_wait (bio);bio_put (bio);
include/linux/bio.h +2 -0
@@ -835,5 +835,7 @@ struct bio *bio_chain_and_submit(struct bio *prev, struct bio *new);struct bio *blk_alloc_discard_bio (struct block_device *bdev,sector_t *sector,sector_t *nr_sects,gfp_t gfp_mask); +struct bio *blk_alloc_secure_erase_bio (struct block_device *bdev, +sector_t *sector,sector_t *nr_sects,gfp_t gfp);#endif /* __LINUX_BIO_H */
Signature · 1 line
2.43.0
PATCH 4/10 blk-lib: check for kill signal in ioctl BLKSECDISCARD
In reply to Christoph HellwigSecure Erase can access a significant capacity and take longer than the user expected. A user may change their mind about wanting to run that command and attempt to kill the process and do something else with their device. But since the task is uninterruptable, they have to wait for it to finish, which could be many hours.
Open code blkdev_issue_secure_erase in the BLKSECDISCARD ioctl handler and check for a fatal signal at each iteration so the user doesn't have to wait for their regretted operation to complete naturally.
Heavily based on an earlier patch from Keith Busch.
--- block/ioctl.c | 34 +++++++++++++++++++++++++++++++--- 1 file changed, 31 insertions(+), 3 deletions(-)
block/ioctl.c +31 -3
@@ -164,6 +164,9 @@ static int blk_ioctl_secure_erase(struct block_device *bdev, blk_mode_t mode, void __user *argp) {unsigned int bs_mask =bdev_logical_block_size (bdev) -1 ; +struct bio *prev =NULL , *bio; +sector_t sector, nr_sects; +struct blk_plug plug;uint64_t start, len;uint64_t range[2 ];int err;
@@ -187,11 +190,36 @@ static int blk_ioctl_secure_erase(struct block_device *bdev, blk_mode_t mode,if (start + len >bdev_nr_bytes (bdev))return -EINVAL ; + sector = start >>SECTOR_SHIFT ; + nr_sects = len >>SECTOR_SHIFT ; +filemap_invalidate_lock (bdev->bd_mapping); err =truncate_bdev_range (bdev, mode, start, start + len -1 ); -if (!err) - err =blkdev_issue_secure_erase (bdev, start >>9 , len >>9 , -GFP_KERNEL ); +if (err) + goto out_unlock; + +blk_start_plug (&plug); +while (1 ) { +if (fatal_signal_pending (current)) { +if (prev) +bio_await_chain (prev); + err = -EINTR ; + goto out_unplug; + } + bio =blk_alloc_secure_erase_bio (bdev, §or, &nr_sects, +GFP_KERNEL ); +if (!bio) +break ; + prev =bio_chain_and_submit (prev, bio); + } +if (prev) { + err =submit_bio_wait (prev); +bio_put (prev); + } + +out_unplug: +blk_finish_plug (&plug); +out_unlock:filemap_invalidate_unlock (bdev->bd_mapping);return err; }
Signature · 1 line
2.43.0
PATCH 5/10 block: factor out a blk_write_zeroes_limit helper
In reply to Christoph HellwigContrary to the comment in __blkdev_issue_write_zeroes, nothing here checks for a potential bi_size overflow. Add a helper mirroring the secure erase code for the check.
--- block/blk-lib.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-)
block/blk-lib.c +11 -7
@@ -103,24 +103,28 @@ int blkdev_issue_discard(struct block_device *bdev, sector_t sector, }EXPORT_SYMBOL (blkdev_issue_discard); +static sector_t bio_write_zeroes_limit (struct block_device *bdev) +{ +sector_t bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ; + +return min (bdev_write_zeroes_sectors (bdev), + (UINT_MAX >>SECTOR_SHIFT ) & ~bs_mask); +} +static int __blkdev_issue_write_zeroes (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp_mask,struct bio **biop,unsigned flags) {struct bio *bio = *biop; -unsigned int max_sectors;if (bdev_read_only (bdev))return -EPERM ; - -/* Ensure that max_sectors doesn't overflow bi_size */ - max_sectors =bdev_write_zeroes_sectors (bdev); - -if (max_sectors ==0 ) +if (!bdev_write_zeroes_sectors (bdev))return -EOPNOTSUPP ;while (nr_sects) { -unsigned int len =min_t (sector_t, nr_sects, max_sectors); +unsigned int len =min_t (sector_t, nr_sects, +bio_write_zeroes_limit (bdev)); bio =blk_next_bio (bio, bdev,0 ,REQ_OP_WRITE_ZEROES , gfp_mask); bio->bi_iter.bi_sector = sector;
Signature · 1 line
2.43.0
PATCH 6/10 block: remove the LBA alignment check in __blkdev_issue_zeroout
In reply to Christoph Hellwig__blkdev_issue_zeroout is a purely kernel internal API and thus can rely on the block layer sector alignment checks.
--- block/blk-lib.c | 5 ----- 1 file changed, 5 deletions(-)
block/blk-lib.c +0 -5
@@ -209,11 +209,6 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, unsigned flags) {int ret; -sector_t bs_mask; - - bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ; -if ((sector | nr_sects) & bs_mask) -return -EINVAL ; ret =__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp_mask, biop, flags);
Signature · 1 line
2.43.0
PATCH 7/10 block: move read-only and supported checks into (__)blkdev_issue_zeroout
In reply to Christoph HellwigMove these checks out of the lower level helpers and into the higher level ones to prepare for refactoring.
--- block/blk-lib.c | 51 ++++++++++++++++++++++--------------------------- 1 file changed, 23 insertions(+), 28 deletions(-)
block/blk-lib.c +23 -28
@@ -111,17 +111,12 @@ static sector_t bio_write_zeroes_limit(struct block_device *bdev) (UINT_MAX >> SECTOR_SHIFT) & ~bs_mask); } -static int __blkdev_issue_write_zeroes (struct block_device *bdev, +static void __blkdev_issue_write_zeroes (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp_mask,struct bio **biop,unsigned flags) {struct bio *bio = *biop; -if (bdev_read_only (bdev)) -return -EPERM ; -if (!bdev_write_zeroes_sectors (bdev)) -return -EOPNOTSUPP ; -while (nr_sects) {unsigned int len =min_t (sector_t, nr_sects,bio_write_zeroes_limit (bdev));
@@ -138,7 +133,6 @@ static int __blkdev_issue_write_zeroes(struct block_device *bdev, } *biop = bio; -return 0 ; }/*
@@ -154,7 +148,7 @@ static unsigned int __blkdev_sectors_to_bio_pages(sector_t nr_sects)return min (pages, (sector_t )BIO_MAX_VECS ); } -static int __blkdev_issue_zero_pages (struct block_device *bdev, +static void __blkdev_issue_zero_pages (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp_mask,struct bio **biop) {
@@ -162,9 +156,6 @@ static int __blkdev_issue_zero_pages(struct block_device *bdev,int bi_size =0 ;unsigned int sz; -if (bdev_read_only (bdev)) -return -EPERM ; -while (nr_sects !=0 ) { bio =blk_next_bio (bio, bdev,__blkdev_sectors_to_bio_pages (nr_sects),REQ_OP_WRITE , gfp_mask);
@@ -182,7 +173,6 @@ static int __blkdev_issue_zero_pages(struct block_device *bdev, } *biop = bio; -return 0 ; }/**
@@ -208,15 +198,19 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { - int ret; - - ret =__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp_mask, - biop, flags); -if (ret != -EOPNOTSUPP || (flags &BLKDEV_ZERO_NOFALLBACK )) -return ret; + if (bdev_read_only(bdev)) + return -EPERM; -return __blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, - biop); +if (bdev_write_zeroes_sectors (bdev)) { +__blkdev_issue_write_zeroes (bdev, sector, nr_sects, + gfp_mask, biop, flags); + }else { +if (flags &BLKDEV_ZERO_NOFALLBACK ) +return -EOPNOTSUPP ; +__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, + biop); + } +return 0 ; }EXPORT_SYMBOL (__blkdev_issue_zeroout);
@@ -245,21 +239,22 @@ int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ;if ((sector | nr_sects) & bs_mask)return -EINVAL ; +if (bdev_read_only (bdev)) +return -EPERM ; +if ((flags &BLKDEV_ZERO_NOFALLBACK ) && !try_write_zeroes) +return -EOPNOTSUPP ; retry: bio =NULL ;blk_start_plug (&plug);if (try_write_zeroes) { - ret =__blkdev_issue_write_zeroes (bdev, sector, nr_sects, - gfp_mask, &bio, flags); - }else if (!(flags &BLKDEV_ZERO_NOFALLBACK )) { - ret =__blkdev_issue_zero_pages (bdev, sector, nr_sects, - gfp_mask, &bio); +__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp_mask, + &bio, flags); }else { -/* No zeroing offload support */ - ret = -EOPNOTSUPP ; +__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, + &bio); } -if (ret ==0 && bio) { +if (bio) { ret =submit_bio_wait (bio);bio_put (bio); }
Signature · 1 line
2.43.0
PATCH 8/10 block: refacto blkdev_issue_zeroout
In reply to Christoph HellwigSplit out two well-defined helpers for hardware supported Write Zeroes and manually writing zeroes using the Write command.
--- block/blk-lib.c | 94 +++++++++++++++++++++++++++++-------------------- 1 file changed, 55 insertions(+), 39 deletions(-)
block/blk-lib.c +55 -39
@@ -135,6 +135,32 @@ static void __blkdev_issue_write_zeroes(struct block_device *bdev, *biop = bio; } +static int blkdev_issue_write_zeroes (struct block_device *bdev,sector_t sector, +sector_t nr_sects,gfp_t gfp,unsigned flags) +{ +struct bio *bio =NULL ; +struct blk_plug plug; +int ret =0 ; + +blk_start_plug (&plug); +__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp, &bio, flags); +if (bio) { + ret =submit_bio_wait (bio); +bio_put (bio); + } +blk_finish_plug (&plug); + +/* +* For some devices there is no non-destructive way to verify whether +* WRITE ZEROES is actually supported. These will clear the capability +* on an I/O error, in which case we'll turn any error into +* "not supported" here. +*/ +if (ret && !bdev_write_zeroes_sectors (bdev)) +return -EOPNOTSUPP ; +return ret; +} +/* * Convert a number of 512B sectors to a number of pages. * The result is limited to a number of pages that can fit into a BIO.
@@ -175,6 +201,27 @@ static void __blkdev_issue_zero_pages(struct block_device *bdev, *biop = bio; } +static int blkdev_issue_zero_pages (struct block_device *bdev,sector_t sector, +sector_t nr_sects,gfp_t gfp,unsigned flags) +{ +struct bio *bio =NULL ; +struct blk_plug plug; +int ret =0 ; + +if (flags &BLKDEV_ZERO_NOFALLBACK ) +return -EOPNOTSUPP ; + +blk_start_plug (&plug); +__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp, &bio); +if (bio) { + ret =submit_bio_wait (bio); +bio_put (bio); + } +blk_finish_plug (&plug); + +return ret; +} +/** * __blkdev_issue_zeroout - generate number of zero filed write bios * @bdev: blockdev to issue
@@ -230,52 +277,21 @@ EXPORT_SYMBOL(__blkdev_issue_zeroout);int blkdev_issue_zeroout (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp_mask,unsigned flags) { -int ret =0 ; -sector_t bs_mask; -struct bio *bio; -struct blk_plug plug; -bool try_write_zeroes = !!bdev_write_zeroes_sectors (bdev); +int ret; - bs_mask = (bdev_logical_block_size (bdev) >>9 ) -1 ; -if ((sector | nr_sects) & bs_mask) +if ((sector | nr_sects) & ((bdev_logical_block_size (bdev) >>9 ) -1 ))return -EINVAL ;if (bdev_read_only (bdev))return -EPERM ; -if ((flags &BLKDEV_ZERO_NOFALLBACK ) && !try_write_zeroes) -return -EOPNOTSUPP ; -retry: - bio =NULL ; -blk_start_plug (&plug); -if (try_write_zeroes) { -__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp_mask, - &bio, flags); - }else { -__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, - &bio); - } -if (bio) { - ret =submit_bio_wait (bio); -bio_put (bio); - } -blk_finish_plug (&plug); -if (ret && try_write_zeroes) { -if (!(flags &BLKDEV_ZERO_NOFALLBACK )) { - try_write_zeroes = false; - goto retry; - } -if (!bdev_write_zeroes_sectors (bdev)) { -/* -* Zeroing offload support was indicated, but the -* device reported ILLEGAL REQUEST (for some devices -* there is no non-destructive way to verify whether -* WRITE ZEROES is actually supported). -*/ - ret = -EOPNOTSUPP ; - } +if (bdev_write_zeroes_sectors (bdev)) { + ret =blkdev_issue_write_zeroes (bdev, sector, nr_sects, + gfp_mask, flags); +if (!ret) +return ret; } -return ret; +return blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, flags); }EXPORT_SYMBOL (blkdev_issue_zeroout);
Signature · 1 line
2.43.0
PATCH 9/10 block: limit the Write Zeroes to manually writing zeroes fallback
In reply to Christoph HellwigOnly fall back from hardware Write Zeroes failures when blkdev_issue_write_zeroes returns -EOPNOTSUPP;
Note that blkdev_issue_write_zeroes turns any failure into -EOPNOTSUPP when the write zeroes queue limit has been cleared to 0, so this still catches all I/O errors where the driver detected missing support for the hardware acceleration.
--- block/blk-lib.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
block/blk-lib.c +1 -1
@@ -287,7 +287,7 @@ int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,if (bdev_write_zeroes_sectors (bdev)) { ret =blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp_mask, flags); -if (!ret) +if (ret != -EOPNOTSUPP )return ret; }
Signature · 1 line
2.43.0
PATCH 10/10 blk-lib: check for kill signal in ioctl BLKZEROOUT
In reply to Christoph HellwigZeroout can access a significant capacity and take longer than the user expected. A user may change their mind about wanting to run that command and attempt to kill the process and do something else with their device. But since the task is uninterruptable, they have to wait for it to finish, which could be many hours.
Add a new BLKDEV_ZERO_KILLABLE flag for blkdev_issue_zeroout that checks for a fatal signal at each iteration so the user doesn't have to wait for their regretted operation to complete naturally.
Heavily based on an earlier patch from Keith Busch.
--- block/blk-lib.c | 66 +++++++++++++++++++++++++++--------------- block/ioctl.c | 2 +- include/linux/blkdev.h | 1 + 3 files changed, 45 insertions(+), 24 deletions(-)
block/blk-lib.c +43 -23
@@ -115,24 +115,27 @@ static void __blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { - struct bio *bio = *biop; - while (nr_sects) { unsigned int len = min_t(sector_t, nr_sects,bio_write_zeroes_limit (bdev)); +struct bio *bio; + +if ((flags &BLKDEV_ZERO_KILLABLE ) && +fatal_signal_pending (current)) +break ; - bio =blk_next_bio (bio, bdev,0 ,REQ_OP_WRITE_ZEROES , gfp_mask); + bio =bio_alloc (bdev,0 ,REQ_OP_WRITE_ZEROES , gfp_mask); bio->bi_iter.bi_sector = sector;if (flags &BLKDEV_ZERO_NOUNMAP ) bio->bi_opf |=REQ_NOUNMAP ; bio->bi_iter.bi_size = len <<SECTOR_SHIFT ; + *biop =bio_chain_and_submit (*biop, bio); + nr_sects -= len; sector += len;cond_resched (); } - - *biop = bio; } static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector,
@@ -145,6 +148,12 @@ static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector,blk_start_plug (&plug);__blkdev_issue_write_zeroes (bdev, sector, nr_sects, gfp, &bio, flags);if (bio) { +if ((flags &BLKDEV_ZERO_KILLABLE ) && +fatal_signal_pending (current)) { +bio_await_chain (bio); +blk_finish_plug (&plug); +return -EINTR ; + } ret =submit_bio_wait (bio);bio_put (bio); }
@@ -176,29 +185,34 @@ static unsigned int __blkdev_sectors_to_bio_pages(sector_t nr_sects)static void __blkdev_issue_zero_pages (struct block_device *bdev,sector_t sector,sector_t nr_sects,gfp_t gfp_mask, -struct bio **biop) +struct bio **biop,unsigned int flags) { -struct bio *bio = *biop; -int bi_size =0 ; -unsigned int sz; +while (nr_sects) { +unsigned int nr_vecs =__blkdev_sectors_to_bio_pages (nr_sects); +struct bio *bio; -while (nr_sects !=0 ) { - bio =blk_next_bio (bio, bdev,__blkdev_sectors_to_bio_pages (nr_sects), -REQ_OP_WRITE , gfp_mask); + bio =bio_alloc (bdev, nr_vecs,REQ_OP_WRITE , gfp_mask); bio->bi_iter.bi_sector = sector; -while (nr_sects !=0 ) { - sz =min ((sector_t )PAGE_SIZE , nr_sects <<9 ); - bi_size =bio_add_page (bio,ZERO_PAGE (0 ), sz,0 ); - nr_sects -= bi_size >>9 ; - sector += bi_size >>9 ; -if (bi_size < sz) +if ((flags &BLKDEV_ZERO_KILLABLE ) && +fatal_signal_pending (current)) +break ; + +do { +unsigned int len, added; + + len =min_t (sector_t, +PAGE_SIZE , nr_sects <<SECTOR_SHIFT ); + added =bio_add_page (bio,ZERO_PAGE (0 ), len,0 ); +if (added < len)break ; - } + nr_sects -= added >>SECTOR_SHIFT ; + sector += added >>SECTOR_SHIFT ; + }while (nr_sects); + + *biop =bio_chain_and_submit (*biop, bio);cond_resched (); } - - *biop = bio; } static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector,
@@ -212,8 +226,14 @@ static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector,return -EOPNOTSUPP ;blk_start_plug (&plug); -__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp, &bio); +__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp, &bio, flags);if (bio) { +if ((flags &BLKDEV_ZERO_KILLABLE ) && +fatal_signal_pending (current)) { +bio_await_chain (bio); +blk_finish_plug (&plug); +return -EINTR ; + } ret =submit_bio_wait (bio);bio_put (bio); }
@@ -255,7 +275,7 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,if (flags &BLKDEV_ZERO_NOFALLBACK )return -EOPNOTSUPP ;__blkdev_issue_zero_pages (bdev, sector, nr_sects, gfp_mask, - biop); + biop, flags); }return 0 ; }
block/ioctl.c +1 -1
@@ -258,7 +258,7 @@ static int blk_ioctl_zeroout(struct block_device *bdev, blk_mode_t mode, goto fail; err =blkdev_issue_zeroout (bdev, start >>9 , len >>9 ,GFP_KERNEL , -BLKDEV_ZERO_NOUNMAP ); +BLKDEV_ZERO_NOUNMAP |BLKDEV_ZERO_KILLABLE ); fail:filemap_invalidate_unlock (bdev->bd_mapping);
include/linux/blkdev.h +1 -0
@@ -1100,6 +1100,7 @@ int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector,#define BLKDEV_ZERO_NOUNMAP (1 << 0)/* do not free blocks */ #define BLKDEV_ZERO_NOFALLBACK (1 << 1)/* don't write explicit zeroes */ +#define BLKDEV_ZERO_KILLABLE (1 << 2)/* interruptible by fatal signals */ extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
Signature · 1 line
2.43.0
Martin K. Petersen
Reviewed-byChristoph,
Quoted 9 lines
Following discard in the last merge window, this series also makes secure erase and discard interruptible by fatal signals.
The secure erase side is a straight port of the discard support. Unfortunately I don't have a way to test it, so I'm adding the eMMC maintainer as that is where the support originated so maybe they can give it a spin? (just do a blkdiscard -f -s /dev/<dev> and then Ctrl+C)
Quoted 3 lines
The write zeroes support is a bit different as it is more complex due to the fallback code and there already is a helper taking flags that we piggy back on. This side has been extensively tested.
Looks good to me.
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering
Jens Axboe
maintainerQuoted 9 lines · Christoph Hellwig
Hi all,
Following discard in the last merge window, this series also makes secure erase and discard interruptible by fatal signals.
The secure erase side is a straight port of the discard support. Unfortunately I don't have a way to test it, so I'm adding the eMMC maintainer as that is where the support originated so maybe they can give it a spin? (just do a blkdiscard -f -s /dev/<dev> and then Ctrl+C)
In lieu of that, qemu does support mmc it looks like?
Wanted to get this queued up, but would probably be best to have that tested first.
Signature · 1 line
Jens Axboe
Quoted 1 line · Jens Axboe
In lieu of that, qemu does support mmc it looks like?
Hmm, I can take a look.
Quoted 2 lines
Wanted to get this queued up, but would probably be best to have that tested first.
The write zeroes patches do not depend on the secure erase patches, and they are what people really care about. Maybe just skip the secure erase patches and apply the rest for now?
Jens Axboe
maintainerQuoted 4 lines · On Fri, Jul 5, 2024 at 12:50?AM Christoph Hellwig <[email protected]> wrote:
Quoted 1 line · Jens Axboe
In lieu of that, qemu does support mmc it looks like?
Quoted 8 lines
Hmm, I can take a look.
Quoted 2 lines
Wanted to get this queued up, but would probably be best to have that tested first.
The write zeroes patches do not depend on the secure erase patches, and they are what people really care about. Maybe just skip the secure erase patches and apply the rest for now?
Done - please just resend the secure erase bits when the mmc side is happy.
Signature · 1 line
Jens Axboe
Quoted 1 line · Jens Axboe
In lieu of that, qemu does support mmc it looks like?
qemu does support mmc, but not the secure erase command. virtio also has secure erase support, but that isn't implemented by qemu either.
Re: (subset) make secure erase and write zeroes ioctls interruptible as well
In reply to Christoph HellwigQuoted 9 lines · Christoph Hellwig
Following discard in the last merge window, this series also makes secure erase and discard interruptible by fatal signals.
The secure erase side is a straight port of the discard support. Unfortunately I don't have a way to test it, so I'm adding the eMMC maintainer as that is where the support originated so maybe they can give it a spin? (just do a blkdiscard -f -s /dev/<dev> and then Ctrl+C)
[...]
Applied, thanks!
[05/10] block: factor out a blk_write_zeroes_limit helper
commit: 73a768d5f95533574bb8ace34eb683a88c40509e
[06/10] block: remove the LBA alignment check in __blkdev_issue_zeroout
commit: ff760a8f0d09f4ba7574ae2ca8be987854f5246d
[07/10] block: move read-only and supported checks into (__)blkdev_issue_zeroout
commit: f6eacb26541ad1eabc40d7e9f5cd86bae7dc0b46
[08/10] block: refacto blkdev_issue_zeroout
commit: 99800ced26b9d87a918aa9824881bdb90a3c1b03
[09/10] block: limit the Write Zeroes to manually writing zeroes fallback
commit: 39722a2f2bcd82bdecc226711412d88b54fcb05b
[10/10] blk-lib: check for kill signal in ioctl BLKZEROOUT
commit: bf86bcdb40123ee99669ee91b67e023669433a1aBest regards,
Signature · 1 line
Jens Axboe