btrfs: prepare for variable stripe_len
v2 cover btrfs: prepare for variable stripe_len
[CHANGELOG]
v2:
- Use IS_ALINGED() when possible
Mostly affecting zoned.c
- Refactor is_same_chunk_map()
Not only reduce the length of the parameter list, but also make it easier to expand, with extra unlikely() calls.
- Refactor alloc_scrub_bbio()
Since the @private is always a scrub_stripe pointer, use scrub_stripe as the type instead, so we can grab stripe_len without introducing a new parameter.
All are suggested by Johannes.
This is a preparation patchset, to support a stripe_len that is not 64KiB.
This is for the future variable stripe length RAID56 feature, where RAID56_VSL profile is allowed for data chunks, and the stripe_len can be as small as sectorsize, meanwhile data IO are kept to be aligned to data_size, which is power of 2, and larger than sector size.
Finally, RAID56_VSL will always have a number of devices that nr_data == data_size / sector_size.
So that's why we want a variable stripe_len.
This series does all the preparation, including supporting stripe_len < PAGE_SIZE.
For now, tree-checker will still reject stripe_len != 64K, but that limit will be removed when RAID56_VSL feature is introduced and can be properly tested.
Qu Wenruo (9):
btrfs: reduce the width of num_stripes and sub_stripes
btrfs: load the on-disk btrfs_chunk::stripe_len
btrfs: add btrfs_io_context::stripe_len and load it
btrfs: change scrub to use stripe_len
btrfs: add a new 'stripe_len' parameter to btrfs_stripe_nr_to_offset()
btrfs: convert volumes.c and block-group.c to use map->stripe_len
btrfs: convert raid56.c to use rbio->bioc->stripe_len
btrfs: convert zoned.c to use map->stripe_len
btrfs: introduce extra limits on btrfs_chunk::stripe_len
fs/btrfs/block-group.c | 9 +-- fs/btrfs/raid56.c | 57 ++++++++++++------ fs/btrfs/raid56.h | 7 +-- fs/btrfs/scrub.c | 129 ++++++++++++++++++++++++++-------------- fs/btrfs/tree-checker.c | 13 +++- fs/btrfs/volumes.c | 95 +++++++++++++++++++---------- fs/btrfs/volumes.h | 16 +++-- fs/btrfs/zoned.c | 56 +++++++++-------- 8 files changed, 246 insertions(+), 136 deletions(-)
Signature · 1 line
2.56.0
PATCH v2 1/9 btrfs: reduce the width of num_stripes and sub_stripes
The on-disk btrfs_chunk structure uses u16 for both num_stripes and sub_stripes, so there is no need to use "int" to represent them in btrfs_chunk_map.
Follow the on-disk format to use u16 for those members, including verified_stripes too.
And since verified_stripes is u16, it no longer completely fill the hole between @refs and @start.
Instead use @num_stripes and @sub_stripes to fill the hole, and leave a 6 bytes hole between @verified_stripes and @stripes.
This doesn't save any space for btrfs_chunk_map, but leaves enough space for future expansion.
--- fs/btrfs/volumes.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
fs/btrfs/volumes.h +4 -4
@@ -623,9 +623,9 @@ extern const struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES];struct btrfs_chunk_map {struct rb_node rb_node; -/* For mount time dev extent verification. */ -int verified_stripes;refcount_t refs; +u16 num_stripes; +u16 sub_stripes;u64 start;u64 chunk_len;u64 dev_extent_size;
@@ -638,8 +638,8 @@ struct btrfs_chunk_map {*/ u64 type;u64 on_disk_type; -int num_stripes; -int sub_stripes; +/* For mount time dev extent verification. */ +u16 verified_stripes;struct btrfs_io_stripe stripes[]; };
Signature · 1 line
2.56.0
PATCH v2 2/9 btrfs: load the on-disk btrfs_chunk::stripe_len
In reply to Qu WenruoCurrently btrfs doesn't load btrfs_chunk::stripe_len into btrfs_chunk_map, as tree-checker will reject any chunk whose stripe_len is not 64K.
But for the future we will want stripe length other than 64K.
As a preparation for the variable stripe length in the future, load btrfs_chunk::stripe_len into btrfs_chunk_map::stripe_len and stripe_shift.
For now stripe_len is still 64K, no matter if it's loaded from the on-disk metadata or newly allocated chunk.
As tree-checker still rejects any chunk whose stripe_len is not 64K.
The two new members fit into the existing hole of btrfs_chunk_map, so the size of btrfs_chunk_map stay the same.
--- fs/btrfs/volumes.c | 12 +++++++++++- fs/btrfs/volumes.h | 2 ++ 2 files changed, 13 insertions(+), 1 deletion(-)
fs/btrfs/volumes.c +11 -1
@@ -6166,6 +6166,12 @@ struct btrfs_chunk_map *btrfs_alloc_chunk_map(int num_stripes, gfp_t gfp)refcount_set (&map->refs,1 );RB_CLEAR_NODE (&map->rb_node); +/* +* Initialize the default stripe len, selftests will utilize them +* without an on-disk chunk item. +*/ + map->stripe_len =BTRFS_STRIPE_LEN ; + map->stripe_shift =ilog2 (map->stripe_len);return map; }
@@ -6203,6 +6209,8 @@ static struct btrfs_block_group *create_chunk(struct btrfs_trans_handle *trans, map->on_disk_type = type; map->sub_stripes = ctl->sub_stripes; map->num_stripes = ctl->num_stripes; + map->stripe_len =BTRFS_STRIPE_LEN ; + map->stripe_shift =ilog2 (BTRFS_STRIPE_LEN );set_real_chunk_type (map);for (int i =0 ; i < ctl->ndevs; i++) {
@@ -6381,7 +6389,7 @@ int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,btrfs_set_stack_chunk_length (chunk, bg->length);btrfs_set_stack_chunk_owner (chunk,BTRFS_EXTENT_TREE_OBJECTID ); -btrfs_set_stack_chunk_stripe_len (chunk,BTRFS_STRIPE_LEN ); +btrfs_set_stack_chunk_stripe_len (chunk, map->stripe_len);btrfs_set_stack_chunk_type (chunk, map->on_disk_type);btrfs_set_stack_chunk_num_stripes (chunk, map->num_stripes);btrfs_set_stack_chunk_io_align (chunk,BTRFS_STRIPE_LEN );
@@ -7781,6 +7789,8 @@ static int read_one_chunk(struct btrfs_key *key, struct extent_buffer *leaf, map->chunk_len = length; map->num_stripes = num_stripes; map->on_disk_type = type; + map->stripe_len =btrfs_chunk_stripe_len (leaf, chunk); + map->stripe_shift =ilog2 (map->stripe_len);/* * We can't use the sub_stripes value, as for profiles other than * RAID10, they may have 0 as sub_stripes for filesystems created by
fs/btrfs/volumes.h +2 -0
@@ -640,6 +640,8 @@ struct btrfs_chunk_map {u64 on_disk_type;/* For mount time dev extent verification. */ u16 verified_stripes; +u8 stripe_shift; +u32 stripe_len;struct btrfs_io_stripe stripes[]; };
Signature · 1 line
2.56.0
PATCH v2 3/9 btrfs: add btrfs_io_context::stripe_len and load it
In reply to Qu WenruoThe new member will be utilized by various path that no direct btrfs_chunk_map is available, but a stripe length is still required.
The new member is located inside a hole between @refs and @fs_info, so no size increase for btrfs_io_context.
--- fs/btrfs/volumes.c | 1 + fs/btrfs/volumes.h | 1 + 2 files changed, 2 insertions(+)
fs/btrfs/volumes.c +1 -0
@@ -7411,6 +7411,7 @@ int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, ret = -ENOMEM ; goto out; } + bioc->stripe_len = map->stripe_len; bioc->map_type = map->type; bioc->use_rst = io_geom.use_rst;
fs/btrfs/volumes.h +1 -0
@@ -515,6 +515,7 @@ struct btrfs_discard_stripe {*/ struct btrfs_io_context {refcount_t refs; +u32 stripe_len;struct btrfs_fs_info *fs_info;/* Taken from struct btrfs_chunk_map::type. */ u64 map_type;
Signature · 1 line
2.56.0
PATCH v2 4/9 btrfs: change scrub to use stripe_len
In reply to Qu WenruoSeveral points to note:
- scrub_ctx::stripes[] stripe_len updates
We always pre-initialize scrub_ctx::stripes[] at the very beginning of scrub, without holding any btrfs_chunk_map.
So init_scrub_stripe() still uses BTRFS_STRIPE_LEN, and all allocated memory are based on that stripe length.
But before scrubbing each chunk, update_scrub_stripe_len() will be called to update the scrub_stripe::stripe_len.
- Stripe_len should not exceed BTRFS_STRIPE_LEN
Since scrub_stripe's heap memory are all based on BTRFS_STRIPE_LEN, any stripe_len larger than BTRFS_STRIPE_LEN will cause access-beyond-boundary.
Thankfully there is no plan to support such stripe_len > BTRFS_STRIPE_LEN any time soon.
- alloc_scrub_bbio() changes
Previously we ensure BTRFS_STRIPE_LEN >= PAGE_SIZE, so BTRFS_STRIPE_LEN >> PAGE_SHIFT can never be zero.
But now stripe_len can be smaller than PAGE_SIZE. E.g. stripe_len 4K, page size 64K.
So we have to use "round_up(stripe_len, PAGE_SIZE) >> PAGE_SHIFT" instead.
Finally alloc_scrub_bbio() is always called with a scrub_stripe pointer as the @private parameter, change the type from (void *) to (struct scrub_stripe *), so that we can grab the stripe_len without a new parameter.
--- fs/btrfs/scrub.c | 99 ++++++++++++++++++++++++++++++++---------------- 1 file changed, 66 insertions(+), 33 deletions(-)
fs/btrfs/scrub.c +66 -33
@@ -124,7 +124,7 @@ enum { };/* -* Represent one contiguous range with a length of BTRFS_STRIPE_LEN. +* Represent one contiguous range with a length of stripe_len. */ struct scrub_stripe {struct scrub_ctx *sctx;
@@ -139,7 +139,7 @@ struct scrub_stripe {u16 mirror_num; -/* Should be BTRFS_STRIPE_LEN / sectorsize. */ +/* Should be stripe_len / sectorsize. */ u16 nr_sectors;/*
@@ -150,6 +150,8 @@ struct scrub_stripe {u16 nr_meta_extents;atomic_t pending_io; +/* From btrfs_chunk_map::stripe_len. Should not exceed BTRFS_STRIPE_LEN. */ +u32 stripe_len;wait_queue_head_t io_wait;wait_queue_head_t repair_wait;
@@ -347,7 +349,13 @@ static int init_scrub_stripe(struct btrfs_fs_info *fs_info, {memset (stripe,0 ,sizeof (*stripe)); +/* +* Use the max stripe_len for now. +* +* This will be updated for every chunk using its stripe_len. +*/ stripe->nr_sectors =BTRFS_STRIPE_LEN >> fs_info->sectorsize_bits; + stripe->stripe_len =BTRFS_STRIPE_LEN ; stripe->state =0 ;init_waitqueue_head (&stripe->io_wait);
@@ -901,7 +909,7 @@ static void scrub_bio_add_sector(struct btrfs_bio *bbio, struct scrub_stripe *stconst u32 offset = sector_nr << fs_info->sectorsize_bits;int ret; -ASSERT (offset + fs_info->sectorsize <=BTRFS_STRIPE_LEN ); +ASSERT (offset + fs_info->sectorsize <= stripe->stripe_len);if (is_vmalloc_addr (stripe->buffer)) { ret =bio_add_vmalloc (&bbio->bio, stripe->buffer + offset, fs_info->sectorsize);
@@ -914,9 +922,9 @@ static void scrub_bio_add_sector(struct btrfs_bio *bbio, struct scrub_stripe *st }static struct btrfs_bio *alloc_scrub_bbio (struct btrfs_fs_info *fs_info, -blk_opf_t opf, -u64 logical, -btrfs_bio_end_io_t end_io,void *private) +blk_opf_t opf,u64 logical, +btrfs_bio_end_io_t end_io, +struct scrub_stripe *stripe) {struct btrfs_bio *bbio;
@@ -925,9 +933,10 @@ static struct btrfs_bio *alloc_scrub_bbio(struct btrfs_fs_info *fs_info,* different physical addresses, we have to ensure the bbio is large * enough to contain the full stripe. */ - bbio =btrfs_bio_alloc (BTRFS_STRIPE_LEN >>PAGE_SHIFT , opf, + bbio =btrfs_bio_alloc (round_up (stripe->stripe_len,PAGE_SIZE ) >> +PAGE_SHIFT , opf,BTRFS_I (fs_info->btree_inode), - logical, end_io, private); + logical, end_io, stripe); bbio->is_scrub = true; bbio->bio.bi_iter.bi_sector = logical >>SECTOR_SHIFT ;return bbio;
@@ -959,7 +968,8 @@ static void scrub_stripe_submit_repair_read(struct scrub_stripe *stripe,if (!bbio) bbio =alloc_scrub_bbio (fs_info,REQ_OP_READ , - stripe->logical + (i << fs_info->sectorsize_bits), + stripe->logical + + (i << fs_info->sectorsize_bits), scrub_repair_read_endio, stripe);scrub_bio_add_sector (bbio, stripe, i);
@@ -1176,7 +1186,7 @@ static void scrub_stripe_read_repair_worker(struct work_struct *work)const unsigned long old_error_bitmap =scrub_bitmap_read_error (stripe);scrub_stripe_submit_repair_read (stripe, mirror, -BTRFS_STRIPE_LEN , false); + stripe->stripe_len, false);wait_scrub_stripe_io (stripe);scrub_verify_one_stripe (stripe, old_error_bitmap);if (scrub_bitmap_empty_error (stripe))
@@ -1325,7 +1335,8 @@ static void scrub_write_sectors(struct scrub_ctx *sctx, struct scrub_stripe *str }if (!bbio) bbio =alloc_scrub_bbio (fs_info,REQ_OP_WRITE , - stripe->logical + (sector_nr << fs_info->sectorsize_bits), + stripe->logical + + (sector_nr << fs_info->sectorsize_bits), scrub_write_endio, stripe);scrub_bio_add_sector (bbio, stripe, sector_nr); }
@@ -1416,7 +1427,7 @@ static int get_raid56_logic_offset(u64 physical, int num, *offset = last_offset +btrfs_stripe_nr_to_offset (i); - stripe_nr = (u32)(*offset >>BTRFS_STRIPE_LEN_SHIFT ) / data_stripes; + stripe_nr = (u32)(*offset >> map->stripe_shift) / data_stripes;/* Work out the disk rotation on this stripe-set */ rot = stripe_nr % map->num_stripes;
@@ -1591,7 +1602,7 @@ static void fill_one_extent_info(struct btrfs_fs_info *fs_info, u64 extent_flags, u64 extent_gen) {for (u64 cur_logical =max (stripe->logical, extent_start); - cur_logical <min (stripe->logical +BTRFS_STRIPE_LEN , + cur_logical <min (stripe->logical + stripe->stripe_len, extent_start + extent_len); cur_logical += fs_info->sectorsize) { const int nr_sector = (cur_logical - stripe->logical) >>
@@ -1670,15 +1681,15 @@ static int scrub_find_fill_first_stripe(struct btrfs_block_group *bg,* Round down to stripe boundary. * * The extra calculation against bg->start is to handle block groups -* whose logical bytenr is not BTRFS_STRIPE_LEN aligned. +* whose logical bytenr is not stripe_len aligned. */ - stripe->logical =round_down (cur_logical - bg->start,BTRFS_STRIPE_LEN ) + + stripe->logical =round_down (cur_logical - bg->start, stripe->stripe_len) + bg->start; stripe->physical = physical + stripe->logical - logical_start; stripe->dev = dev; stripe->bg = bg; stripe->mirror_num = mirror_num; - stripe_end = stripe->logical +BTRFS_STRIPE_LEN -1 ; + stripe_end = stripe->logical + stripe->stripe_len -1 ;/* Fill the first extent info into stripe->sectors[] array. */ fill_one_extent_info(fs_info, stripe, extent_start, extent_len,
@@ -1715,10 +1726,15 @@ static int scrub_find_fill_first_stripe(struct btrfs_block_group *bg,ASSERT (stripe->csums);/* -* Our csum bitmap should be large enough, as BTRFS_STRIPE_LEN -* should contain at most 16 sectors. +* Our csum bitmap should be large enough, as stripe_len is no +* larger than BTRFS_STRIPE_LEN, and BTRFS_STRIPE_LEN contains +* at most 16 sectors (4K sector size). +* +* So (stripe_len >> sectorsize_bits) should be at most 16, +* not reaching BITS_PER_LONG. */ -ASSERT (BITS_PER_LONG >=BTRFS_STRIPE_LEN >> fs_info->sectorsize_bits); +ASSERT (BITS_PER_LONG >= (stripe->stripe_len >> + fs_info->sectorsize_bits)); ret = btrfs_lookup_csums_bitmap(csum_root, csum_path, stripe->logical, stripe_end,
@@ -1756,7 +1772,7 @@ static u32 stripe_length(const struct scrub_stripe *stripe) {ASSERT (stripe->bg); -return min (BTRFS_STRIPE_LEN , +return min (stripe->stripe_len, stripe->bg->start + stripe->bg->length - stripe->logical); }
@@ -1766,7 +1782,7 @@ static void scrub_submit_extent_sector_read(struct scrub_stripe *stripe)struct btrfs_bio *bbio =NULL ;unsigned int nr_sectors =stripe_length (stripe) >> fs_info->sectorsize_bits;const unsigned long has_extent =scrub_bitmap_read_has_extent (stripe); -u64 stripe_len =BTRFS_STRIPE_LEN ; +u64 stripe_len = stripe->stripe_len;int mirror = stripe->mirror_num;int i;
@@ -2102,7 +2118,7 @@ static int scrub_raid56_cached_parity(struct scrub_ctx *sctx,/* For RAID56 write there must be an @bioc allocated. */ ASSERT (bioc); rbio =raid56_parity_alloc_scrub_rbio (&bio, bioc, scrub_dev, extent_bitmap, -BTRFS_STRIPE_LEN >> fs_info->sectorsize_bits); + map->stripe_len >> fs_info->sectorsize_bits);btrfs_put_bioc (bioc);if (!rbio) { ret = -ENOMEM ;
@@ -2113,7 +2129,7 @@ static int scrub_raid56_cached_parity(struct scrub_ctx *sctx,struct scrub_stripe *stripe = &sctx->raid56_data_stripes[i];raid56_parity_cache_data_folios (rbio, stripe->buffer, - full_stripe_start + (i <<BTRFS_STRIPE_LEN_SHIFT )); + full_stripe_start + (i << map->stripe_shift)); }raid56_parity_submit_scrub_rbio (rbio);wait_for_completion_io (&io_done);
@@ -2172,7 +2188,7 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx, stripe = &sctx->raid56_data_stripes[i]; rot =div_u64 (full_stripe_start - bg->start, - data_stripes) >>BTRFS_STRIPE_LEN_SHIFT ; + data_stripes) >> map->stripe_shift; stripe_index = (i + rot) % map->num_stripes; physical = map->stripes[stripe_index].physical +btrfs_stripe_nr_to_offset (rot);
@@ -2182,7 +2198,7 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx, ret =scrub_find_fill_first_stripe (bg, &extent_path, &csum_path, map->stripes[stripe_index].dev, physical,1 , full_stripe_start +btrfs_stripe_nr_to_offset (i), -BTRFS_STRIPE_LEN , stripe); + map->stripe_len, stripe);if (ret <0 )return ret;/*
@@ -2270,6 +2286,7 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx,static int scrub_simple_mirror (struct scrub_ctx *sctx,struct btrfs_block_group *bg,u64 logical_start,u64 logical_length, +u32 stripe_len,struct btrfs_device *device,u64 physical,int mirror_num) {
@@ -2317,7 +2334,7 @@ static int scrub_simple_mirror(struct scrub_ctx *sctx,/* queue_scrub_stripe() returned 0, @found_logical must be updated. */ ASSERT (found_logical !=U64_MAX ); - cur_logical = found_logical +BTRFS_STRIPE_LEN ; + cur_logical = found_logical + stripe_len;/* Don't hold CPU for too long time */ cond_resched ();
@@ -2372,6 +2389,7 @@ static int scrub_simple_stripe(struct scrub_ctx *sctx,const u64 orig_logical =simple_stripe_get_logical (map, bg, stripe_index);const u64 orig_physical = map->stripes[stripe_index].physical;const u64 end =btrfs_block_group_end (bg); +const u32 stripe_len = map->stripe_len;const int mirror_num =simple_stripe_mirror_num (map, stripe_index);u64 cur_logical = orig_logical;u64 cur_physical = orig_physical;
@@ -2384,18 +2402,27 @@ static int scrub_simple_stripe(struct scrub_ctx *sctx,* this stripe. */ ret =scrub_simple_mirror (sctx, bg, cur_logical, -BTRFS_STRIPE_LEN , device, cur_physical, + stripe_len, stripe_len, + device, cur_physical, mirror_num);if (ret)return ret;/* Skip to next stripe which belongs to the target device */ cur_logical += logical_increment;/* For physical offset, we just go to next stripe */ - cur_physical +=BTRFS_STRIPE_LEN ; + cur_physical += stripe_len; }return ret; } +static void update_scrub_stripe_len (const struct btrfs_fs_info *fs_info, +const struct btrfs_chunk_map *map, +struct scrub_stripe *stripe) +{ + stripe->stripe_len = map->stripe_len; + stripe->nr_sectors = map->stripe_len >> fs_info->sectorsize_bits; +} + static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx, struct btrfs_block_group *bg, struct btrfs_chunk_map *map,
@@ -2405,6 +2432,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,struct btrfs_fs_info *fs_info = sctx->fs_info;const u64 profile = map->type &BTRFS_BLOCK_GROUP_PROFILE_MASK ;const u64 chunk_logical = bg->start; +const u32 stripe_len = map->stripe_len;int ret;int ret2;u64 physical = map->stripes[stripe_index].physical;
@@ -2447,6 +2475,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx, goto out; sctx->raid56_data_stripes[i].bg = bg; sctx->raid56_data_stripes[i].sctx = sctx; +update_scrub_stripe_len (fs_info, map, &sctx->raid56_data_stripes[i]); } }/*
@@ -2467,7 +2496,8 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,* @stripe_index. */ ret =scrub_simple_mirror (sctx, bg, bg->start, bg->length, - scrub_dev, map->stripes[stripe_index].physical, + stripe_len, scrub_dev, + map->stripes[stripe_index].physical, stripe_index +1 ); offset =0 ; goto out;
@@ -2505,7 +2535,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx, ret =scrub_raid56_parity_stripe (sctx, scrub_dev, bg, map, stripe_logical);spin_lock (&sctx->stat_lock); - sctx->stat.last_physical =min (physical +BTRFS_STRIPE_LEN , + sctx->stat.last_physical =min (physical + stripe_len, physical_end);spin_unlock (&sctx->stat_lock);if (ret)
@@ -2521,13 +2551,13 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,* We can reuse scrub_simple_mirror() here, as the repair part * is still based on @mirror_num. */ - ret =scrub_simple_mirror (sctx, bg, logical,BTRFS_STRIPE_LEN , - scrub_dev, physical,1 ); + ret =scrub_simple_mirror (sctx, bg, logical, stripe_len, + stripe_len, scrub_dev, physical,1 );if (ret <0 ) goto out; next: logical += increment; - physical +=BTRFS_STRIPE_LEN ; + physical += stripe_len;spin_lock (&sctx->stat_lock); sctx->stat.last_physical = physical;spin_unlock (&sctx->stat_lock);
@@ -2587,6 +2617,9 @@ static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,if (map->chunk_len < dev_extent_len) goto out; +for (i =0 ; i <SCRUB_TOTAL_STRIPES ; i++) +update_scrub_stripe_len (fs_info, map, &sctx->stripes[i]); +for (i =0 ; i < map->num_stripes; ++i) {if (map->stripes[i].dev->bdev == scrub_dev->bdev && map->stripes[i].physical == dev_offset) {
Signature · 1 line
2.56.0
Re: [PATCH v2 4/9] btrfs: change scrub to use stripe_len
Looks good,
PATCH v2 5/9 btrfs: add a new 'stripe_len' parameter to btrfs_stripe_nr_to_offset()
In reply to Qu WenruoThat helper is currently using a fixed BTRFS_STRIPE_LEN, change it to accept a 'u32 stripe_len'.
Also add an ASSERT() to make sure the @stripe_len passed in is power of 2, and no larger than BTRFS_STRIPE_LEN.
There is a special case where we still pass BTRFS_STRIPE_LEN into btrfs_stripe_nr_to_offset(), that's for chunk allocation to determine the minimal device extent size.
In that case, since our future planned map->stripe_len is always power of 2 and no larger than BTRFS_STRIPE_LEN, we are totally fine to use the old BTRFS_STRIPE_LEN.
--- fs/btrfs/block-group.c | 3 ++- fs/btrfs/scrub.c | 32 +++++++++++++++++++------------- fs/btrfs/volumes.c | 40 +++++++++++++++++++++++++++------------- fs/btrfs/volumes.h | 5 +++-- fs/btrfs/zoned.c | 7 +++++-- 5 files changed, 56 insertions(+), 31 deletions(-)
fs/btrfs/block-group.c +2 -1
@@ -2302,7 +2302,8 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,/* For RAID5/6 adjust to a full IO stripe length */ if (map->type &BTRFS_BLOCK_GROUP_RAID56_MASK ) - io_stripe_size =btrfs_stripe_nr_to_offset (nr_data_stripes (map)); + io_stripe_size =btrfs_stripe_nr_to_offset (nr_data_stripes (map), + map->stripe_len); buf =kzalloc_objs (u64, map->num_stripes,GFP_NOFS );if (!buf) {
fs/btrfs/scrub.c +19 -13
@@ -1425,7 +1425,8 @@ static int get_raid56_logic_offset(u64 physical, int num,u32 stripe_index;u32 rot; - *offset = last_offset +btrfs_stripe_nr_to_offset (i); + *offset = last_offset +btrfs_stripe_nr_to_offset (i, + map->stripe_len); stripe_nr = (u32)(*offset >> map->stripe_shift) / data_stripes;
@@ -1439,7 +1440,7 @@ static int get_raid56_logic_offset(u64 physical, int num,if (stripe_index < num) j++; } - *offset = last_offset +btrfs_stripe_nr_to_offset (j); + *offset = last_offset +btrfs_stripe_nr_to_offset (j, map->stripe_len);return 1 ; }
@@ -1924,7 +1925,8 @@ static void submit_initial_group_read(struct scrub_ctx *sctx,ASSERT (first_slot + nr_stripes <=SCRUB_TOTAL_STRIPES );scrub_throttle_dev_io (sctx, sctx->stripes[0 ].dev, -btrfs_stripe_nr_to_offset (nr_stripes)); +btrfs_stripe_nr_to_offset (nr_stripes, + sctx->stripes[0 ].stripe_len));blk_start_plug (&plug);for (int i =0 ; i < nr_stripes; i++) {struct scrub_stripe *stripe = &sctx->stripes[first_slot + i];
@@ -2102,7 +2104,7 @@ static int scrub_raid56_cached_parity(struct scrub_ctx *sctx,struct btrfs_raid_bio *rbio;struct bio bio;const int data_stripes =nr_data_stripes (map); -u64 length =btrfs_stripe_nr_to_offset (data_stripes); +u64 length =btrfs_stripe_nr_to_offset (data_stripes, map->stripe_len);int ret;bio_init (&bio,NULL ,NULL ,0 ,REQ_OP_READ );
@@ -2151,6 +2153,7 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx,BTRFS_PATH_AUTO_RELEASE (csum_path);struct scrub_stripe *stripe;bool all_empty = true; +const u32 stripe_len = map->stripe_len;const int data_stripes =nr_data_stripes (map);unsigned long extent_bitmap =0 ;int ret;
@@ -2191,14 +2194,15 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx, data_stripes) >> map->stripe_shift; stripe_index = (i + rot) % map->num_stripes; physical = map->stripes[stripe_index].physical + -btrfs_stripe_nr_to_offset (rot); +btrfs_stripe_nr_to_offset (rot, stripe_len);scrub_reset_stripe (stripe);set_bit (SCRUB_STRIPE_FLAG_NO_REPORT , &stripe->state); ret =scrub_find_fill_first_stripe (bg, &extent_path, &csum_path, map->stripes[stripe_index].dev, physical,1 , - full_stripe_start +btrfs_stripe_nr_to_offset (i), - map->stripe_len, stripe); + full_stripe_start + +btrfs_stripe_nr_to_offset (i, stripe_len), + stripe_len, stripe);if (ret <0 )return ret;/*
@@ -2207,7 +2211,7 @@ static int scrub_raid56_parity_stripe(struct scrub_ctx *sctx,*/ if (ret >0 ) { stripe->logical = full_stripe_start + -btrfs_stripe_nr_to_offset (i); +btrfs_stripe_nr_to_offset (i, stripe_len); stripe->dev = map->stripes[stripe_index].dev; stripe->mirror_num =1 ;set_bit (SCRUB_STRIPE_FLAG_INITIALIZED , &stripe->state);
@@ -2348,7 +2352,8 @@ static u64 simple_stripe_full_stripe_len(const struct btrfs_chunk_map *map)ASSERT (map->type & (BTRFS_BLOCK_GROUP_RAID0 |BTRFS_BLOCK_GROUP_RAID10 )); -return btrfs_stripe_nr_to_offset (map->num_stripes / map->sub_stripes); +return btrfs_stripe_nr_to_offset (map->num_stripes / map->sub_stripes, + map->stripe_len); }/* Get the logical bytenr for the stripe */
@@ -2364,8 +2369,8 @@ static u64 simple_stripe_get_logical(struct btrfs_chunk_map *map,* (stripe_index / sub_stripes) gives how many data stripes we need to * skip. */ -return btrfs_stripe_nr_to_offset (stripe_index / map->sub_stripes) + - bg->start; +return btrfs_stripe_nr_to_offset (stripe_index / map->sub_stripes, + map->stripe_len) + bg->start; }/* Get the mirror number for the stripe */
@@ -2504,7 +2509,8 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx, }if (profile & (BTRFS_BLOCK_GROUP_RAID0 |BTRFS_BLOCK_GROUP_RAID10 )) { ret =scrub_simple_stripe (sctx, bg, map, scrub_dev, stripe_index); - offset =btrfs_stripe_nr_to_offset (stripe_index / map->sub_stripes); + offset =btrfs_stripe_nr_to_offset (stripe_index / map->sub_stripes, + stripe_len); goto out; }
@@ -2519,7 +2525,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,/* Initialize @offset in case we need to go to out: label */ get_raid56_logic_offset (physical, stripe_index, map, &offset,NULL ); - increment =btrfs_stripe_nr_to_offset (nr_data_stripes (map)); + increment =btrfs_stripe_nr_to_offset (nr_data_stripes (map), stripe_len);/* * Due to the rotation, for RAID56 it's better to iterate each stripe
fs/btrfs/volumes.c +27 -13
@@ -5823,7 +5823,8 @@ static void init_alloc_chunk_ctl_policy_regular(/* We don't want a chunk larger than 10% of writable space */ ctl->max_chunk_size =min (mult_perc (fs_devices->total_rw_bytes,10 ), ctl->max_chunk_size); - ctl->dev_extent_min =btrfs_stripe_nr_to_offset (ctl->dev_stripes); + ctl->dev_extent_min =btrfs_stripe_nr_to_offset (ctl->dev_stripes, +BTRFS_STRIPE_LEN ); } static void init_alloc_chunk_ctl_policy_zoned(
@@ -6587,7 +6588,8 @@ unsigned long btrfs_full_stripe_len(struct btrfs_fs_info *fs_info,if (!WARN_ON (IS_ERR (map))) {if (map->type &BTRFS_BLOCK_GROUP_RAID56_MASK ) - len =btrfs_stripe_nr_to_offset (nr_data_stripes (map)); + len =btrfs_stripe_nr_to_offset (nr_data_stripes (map), + map->stripe_len);btrfs_free_chunk_map (map); }return len;
@@ -6784,6 +6786,7 @@ struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,u64 stripe_end_offset;u64 stripe_offset;u32 stripe_index; +u32 stripe_len;u32 factor =0 ;u32 sub_stripes =0 ;u32 stripes_per_dev =0 ;
@@ -6820,6 +6823,7 @@ struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info, goto out_free_map; } + stripe_len = map->stripe_len; offset = logical - map->start; length =min_t (u64, map->start + map->chunk_len - logical, length); *length_ret = length;
@@ -6831,12 +6835,13 @@ struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info, stripe_nr = offset >>BTRFS_STRIPE_LEN_SHIFT ;/* stripe_offset is the offset of this block in its stripe */ - stripe_offset = offset -btrfs_stripe_nr_to_offset (stripe_nr); + stripe_offset = offset -btrfs_stripe_nr_to_offset (stripe_nr, + stripe_len); stripe_nr_end =round_up (offset + length,BTRFS_STRIPE_LEN ) >>BTRFS_STRIPE_LEN_SHIFT ; stripe_cnt = stripe_nr_end - stripe_nr; - stripe_end_offset =btrfs_stripe_nr_to_offset (stripe_nr_end) - + stripe_end_offset =btrfs_stripe_nr_to_offset (stripe_nr_end, stripe_len) - (offset + length);/* * after this, stripe_nr is the number of stripes on this
@@ -6879,12 +6884,14 @@ struct btrfs_discard_stripe *btrfs_map_discard(struct btrfs_fs_info *fs_info,for (i =0 ; i < *num_stripes; i++) { stripes[i].physical = map->stripes[stripe_index].physical + - stripe_offset +btrfs_stripe_nr_to_offset (stripe_nr); + stripe_offset +btrfs_stripe_nr_to_offset (stripe_nr, + stripe_len); stripes[i].dev = map->stripes[stripe_index].dev;if (map->type & (BTRFS_BLOCK_GROUP_RAID0 |BTRFS_BLOCK_GROUP_RAID10 )) { - stripes[i].length =btrfs_stripe_nr_to_offset (stripes_per_dev); + stripes[i].length =btrfs_stripe_nr_to_offset (stripes_per_dev, + stripe_len);if (i / sub_stripes < remaining_stripes) stripes[i].length +=BTRFS_STRIPE_LEN ;
@@ -7022,6 +7029,8 @@ static void handle_ops_on_dev_replace(struct btrfs_io_context *bioc,static u64 btrfs_max_io_len (struct btrfs_chunk_map *map,u64 offset,struct btrfs_io_geometry *io_geom) { +const u32 stripe_len = map->stripe_len; +/* * Stripe_nr is the stripe where this block falls. stripe_offset is * the offset of this block in its stripe.
@@ -7033,7 +7042,8 @@ static u64 btrfs_max_io_len(struct btrfs_chunk_map *map, u64 offset,if (map->type &BTRFS_BLOCK_GROUP_RAID56_MASK ) {unsigned long full_stripe_len = -btrfs_stripe_nr_to_offset (nr_data_stripes (map)); +btrfs_stripe_nr_to_offset (nr_data_stripes (map), + stripe_len);/* * For full stripe start, we use previously calculated
@@ -7045,7 +7055,8 @@ static u64 btrfs_max_io_len(struct btrfs_chunk_map *map, u64 offset,* not ensured to be power of 2. */ io_geom->raid56_full_stripe_start =btrfs_stripe_nr_to_offset ( -rounddown (io_geom->stripe_nr,nr_data_stripes (map))); +rounddown (io_geom->stripe_nr,nr_data_stripes (map)), + stripe_len); ASSERT(io_geom->raid56_full_stripe_start + full_stripe_len > offset, "raid56_full_stripe_start=%llu full_stripe_len=%lu offset=%llu",
@@ -7084,7 +7095,8 @@ static int set_io_stripe(struct btrfs_fs_info *fs_info, u64 logical, dst->physical = map->stripes[io_geom->stripe_index].physical + io_geom->stripe_offset + -btrfs_stripe_nr_to_offset (io_geom->stripe_nr); +btrfs_stripe_nr_to_offset (io_geom->stripe_nr, + map->stripe_len);return 0 ; }
@@ -7207,7 +7219,7 @@ static void map_blocks_raid56_write(struct btrfs_chunk_map *map,/* Return the length to the full stripe end. */ *length =min (logical + *length, io_geom->raid56_full_stripe_start + map->start + -btrfs_stripe_nr_to_offset (data_stripes)) - +btrfs_stripe_nr_to_offset (data_stripes, map->stripe_len)) - logical; io_geom->stripe_index =0 ; io_geom->stripe_offset =0 ;
@@ -7434,7 +7446,8 @@ int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,*/ bioc->full_stripe_logical = map->start +btrfs_stripe_nr_to_offset (io_geom.stripe_nr * -nr_data_stripes (map)); +nr_data_stripes (map), + map->stripe_len);for (int i =0 ; i < io_geom.num_stripes; i++) {struct btrfs_io_stripe *dst = &bioc->stripes[i];u32 stripe_index;
@@ -7444,7 +7457,8 @@ int btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op, dst->physical = map->stripes[stripe_index].physical + io_geom.stripe_offset + -btrfs_stripe_nr_to_offset (io_geom.stripe_nr); +btrfs_stripe_nr_to_offset (io_geom.stripe_nr, + map->stripe_len); } }else {/*
@@ -9070,7 +9084,7 @@ static void map_raid56_repair_block(struct btrfs_io_context *bioc,for (i =0 ; i < data_stripes; i++) {u64 stripe_start = bioc->full_stripe_logical + -btrfs_stripe_nr_to_offset (i); +btrfs_stripe_nr_to_offset (i, bioc->stripe_len);if (logical >= stripe_start && logical < stripe_start +BTRFS_STRIPE_LEN )
fs/btrfs/volumes.h +3 -2
@@ -722,9 +722,10 @@ static inline unsigned long btrfs_chunk_item_size(int num_stripes)* @stripe_nr is u32, with left shift it can overflow u32 for chunks larger * than 4G. This does the proper type cast to avoid overflow. */ -static inline u64 btrfs_stripe_nr_to_offset (u32 stripe_nr) +static inline u64 btrfs_stripe_nr_to_offset (u32 stripe_nr,u32 stripe_len) { -return (u64 )stripe_nr <<BTRFS_STRIPE_LEN_SHIFT ; +ASSERT (is_power_of_2 (stripe_len) && stripe_len <=BTRFS_STRIPE_LEN ); +return (u64 )stripe_nr <<ilog2 (stripe_len); }void btrfs_get_bioc (struct btrfs_io_context *bioc);
fs/btrfs/zoned.c +5 -2
@@ -1630,7 +1630,9 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,if (zone_info[i].alloc_offset ==WP_CONVENTIONAL ) { has_conventional = true; - zone_info[i].alloc_offset =btrfs_stripe_nr_to_offset (stripe_nr); + zone_info[i].alloc_offset = +btrfs_stripe_nr_to_offset (stripe_nr, + map->stripe_len);if (stripe_index > i) zone_info[i].alloc_offset +=BTRFS_STRIPE_LEN ;
@@ -1772,7 +1774,8 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,if (raid0_allocs[idx] ==WP_CONVENTIONAL ) { has_conventional = true; - raid0_allocs[idx] =btrfs_stripe_nr_to_offset (stripe_nr); + raid0_allocs[idx] =btrfs_stripe_nr_to_offset (stripe_nr, + map->stripe_len);if (stripe_index > idx) raid0_allocs[idx] +=BTRFS_STRIPE_LEN ;
Signature · 1 line
2.56.0
Re: [PATCH v2 5/9] btrfs: add a new 'stripe_len' parameter to btrfs_stripe_nr_to_offset()
Looks good,
PATCH v2 6/9 btrfs: convert volumes.c and block-group.c to use map->stripe_len
In reply to Qu WenruoThere are still some BTRFS_STRIPE_LEN usage remaining, but they are all for chunk allocation.
There is one more refactor for is_same_chunk_map(), as there are more members to compare, the parameter list can go longer and longer.
Instead of passing a new @stripe_len member, pass extent_buffer and btrfs_chunk pointers, and grab the needed member inside that function.
And doing "if (unlikely()) return false" for each member check.
Although this make is_same_chunk_map() more complex, considering it's only called for system chunks, it should not bring too much overhead.
--- fs/btrfs/block-group.c | 6 ++--- fs/btrfs/volumes.c | 52 ++++++++++++++++++++++++------------------ 2 files changed, 33 insertions(+), 25 deletions(-)
fs/btrfs/block-group.c +3 -3
@@ -2297,7 +2297,7 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,return -EIO ; data_stripe_length = map->dev_extent_size; - io_stripe_size =BTRFS_STRIPE_LEN ; + io_stripe_size = map->stripe_len; chunk_start = map->start;/* For RAID5/6 adjust to a full IO stripe length */
@@ -2322,9 +2322,9 @@ int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,continue ; stripe_nr = (physical - map->stripes[i].physical) >> -BTRFS_STRIPE_LEN_SHIFT ; + map->stripe_shift; offset = (physical - map->stripes[i].physical) & -BTRFS_STRIPE_LEN_MASK ; + (map->stripe_len -1 );if (map->type & (BTRFS_BLOCK_GROUP_RAID0 |BTRFS_BLOCK_GROUP_RAID10 ))
fs/btrfs/volumes.c +30 -22
142 lines not on this page · raw message
Signature · 1 line
2.56.0
Re: [PATCH v2 6/9] btrfs: convert volumes.c and block-group.c to use map->stripe_len
Looks good,
PATCH v2 7/9 btrfs: convert raid56.c to use rbio->bioc->stripe_len
In reply to Qu WenruoSo that raid56 can support stripe length other than BTRFS_STRIPE_LEN.
There are several extra things to note:
- btrfs_raid_bio::stripe_npages is removed
With stripe_len < BTRFS_STRIPE_LEN, there is no guarantee that a page won't cross multiple stripes.
E.g. with 4K stripe_len and sectorsize, 2 data stripes, with 64K page size, one page will cover the full stripe.
So stripe_npages no longer makes sense, just remove it.
- alloc_rbio_data_pages()/alloc_rbio_parity_pages() fallback For the same above reason, a page can contain both data and parity blocks.
In that case we can no longer only allocate pages for data/parity.
Add a fallback that if stripe_len is smaller than page size, always allocate all the pages.
This is not the best solution, but should be good enough for now until the variable stripe_len is properly tested.
- steal_rbio() fallback
If stripe_len is smaller than PAGE_SIZE, it's hard to only steal the data pages.
Again skip the stealing if stripe_len is smaller than PAGE_SIZE.
--- fs/btrfs/raid56.c | 57 +++++++++++++++++++++++++++++++---------------- fs/btrfs/raid56.h | 7 +----- 2 files changed, 39 insertions(+), 25 deletions(-)
fs/btrfs/raid56.c +38 -19
@@ -395,8 +395,9 @@ static bool is_data_stripe_page(struct btrfs_raid_bio *rbio, int page_nr)const int sector_nr =page_nr_to_sector_nr (rbio, page_nr);/* -* We have ensured PAGE_SIZE is aligned with sectorsize, thus -* we won't have a page which is half data half parity. +* The caller has ensured stripe_len >= PAGE_SIZE, and PAGE_SIZE is +* aligned with sectorsize, thus we won't have a page which is half +* data half parity. * * Thus if the first sector of the page belongs to data stripes, then * the full page belongs to data stripes.
@@ -417,6 +418,9 @@ static void steal_rbio(struct btrfs_raid_bio *src, struct btrfs_raid_bio *dest)if (!test_bit (RBIO_CACHE_READY_BIT , &src->flags))return ; +/* Data and P/Q can be mixed into a page, skip stealing. */ +if (src->bioc->stripe_len <PAGE_SIZE ) +return ;for (i =0 ; i < dest->nr_pages; i++) {struct page *p = src->stripe_pages[i];
@@ -624,12 +628,13 @@ static void cache_rbio(struct btrfs_raid_bio *rbio)static int rbio_is_full (struct btrfs_raid_bio *rbio) {unsigned long size = rbio->bio_list_bytes; +const u32 stripe_len = rbio->bioc->stripe_len;int ret =1 ;spin_lock (&rbio->bio_list_lock); -if (size != rbio->nr_data *BTRFS_STRIPE_LEN ) +if (size != rbio->nr_data * stripe_len) ret =0 ; -BUG_ON (size > rbio->nr_data *BTRFS_STRIPE_LEN ); +BUG_ON (size > rbio->nr_data * stripe_len);spin_unlock (&rbio->bio_list_lock);return ret;
@@ -1041,11 +1046,12 @@ static phys_addr_t sector_paddr_in_rbio(struct btrfs_raid_bio *rbio,static struct btrfs_raid_bio *alloc_rbio (struct btrfs_fs_info *fs_info,struct btrfs_io_context *bioc) { +const u32 stripe_len = bioc->stripe_len;const unsigned int real_stripes = bioc->num_stripes - bioc->replace_nr_stripes; -const unsigned int stripe_npages =BTRFS_STRIPE_LEN >>PAGE_SHIFT ; -const unsigned int num_pages = stripe_npages * real_stripes; +const unsigned int num_pages =round_up (stripe_len * real_stripes,PAGE_SIZE ) >> +PAGE_SHIFT ;const unsigned int stripe_nsectors = -BTRFS_STRIPE_LEN >> fs_info->sectorsize_bits; + stripe_len >> fs_info->sectorsize_bits;const unsigned int num_sectors = stripe_nsectors * real_stripes;const unsigned int step =min (fs_info->sectorsize,PAGE_SIZE );const unsigned int sector_nsteps = fs_info->sectorsize / step;
@@ -1105,7 +1111,6 @@ static struct btrfs_raid_bio *alloc_rbio(struct btrfs_fs_info *fs_info, rbio->nr_pages = num_pages; rbio->nr_sectors = num_sectors; rbio->real_stripes = real_stripes; - rbio->stripe_npages = stripe_npages; rbio->stripe_nsectors = stripe_nsectors; rbio->sector_nsteps = sector_nsteps;refcount_set (&rbio->refs,1 );
@@ -1134,9 +1139,17 @@ static int alloc_rbio_pages(struct btrfs_raid_bio *rbio)/* only allocate pages for p/q stripes */ static int alloc_rbio_parity_pages (struct btrfs_raid_bio *rbio) { -const int data_pages = rbio->nr_data * rbio->stripe_npages; +int data_pages;int ret; +/* +* Page size larger than stripe size, we can not only allocate +* pages for P/Q stripes. +*/ +if (rbio->bioc->stripe_len <PAGE_SIZE ) +return alloc_rbio_pages (rbio); + + data_pages = rbio->nr_data * (rbio->bioc->stripe_len >>PAGE_SHIFT ); ret =btrfs_alloc_page_array (rbio->nr_pages - data_pages, rbio->stripe_pages + data_pages,GFP_NOFS );if (ret <0 )
@@ -1274,7 +1287,7 @@ static int rbio_add_io_paddrs(struct btrfs_raid_bio *rbio, struct bio_list *bio_/* put a new bio on the list */ bio =bio_alloc (stripe->dev->bdev, -max (BTRFS_STRIPE_LEN >>PAGE_SHIFT ,1 ), +max (rbio->bioc->stripe_len >>PAGE_SHIFT ,1 ), op,GFP_NOFS ); bio->bi_iter.bi_sector = disk_start >>SECTOR_SHIFT ; bio->bi_private = rbio;
@@ -1715,9 +1728,13 @@ static void submit_read_wait_bio_list(struct btrfs_raid_bio *rbio,static int alloc_rbio_data_pages (struct btrfs_raid_bio *rbio) { -const int data_pages = rbio->nr_data * rbio->stripe_npages; +int data_pages;int ret; +if (rbio->bioc->stripe_len <PAGE_SIZE ) +return alloc_rbio_pages (rbio); + + data_pages = rbio->nr_data * (rbio->bioc->stripe_len >>PAGE_SHIFT ); ret =btrfs_alloc_page_array (data_pages, rbio->stripe_pages,GFP_NOFS );if (ret <0 )return ret;
@@ -1804,7 +1821,7 @@ static void rbio_add_bio(struct btrfs_raid_bio *rbio, struct bio *orig_bio)ASSERT_RBIO_LOGICAL (orig_logical >= full_stripe_start && orig_logical + orig_len <= full_stripe_start + - rbio->nr_data *BTRFS_STRIPE_LEN , + rbio->nr_data * rbio->bioc->stripe_len, rbio, orig_logical);bio_list_add (&rbio->bio_list, orig_bio);
@@ -3012,6 +3029,7 @@ void raid56_parity_cache_data_folios(struct btrfs_raid_bio *rbio,struct btrfs_fs_info *fs_info = rbio->bioc->fs_info;const u64 offset_in_full_stripe = data_logical - rbio->bioc->full_stripe_logical; +const u32 stripe_len = rbio->bioc->stripe_len;int ret;/*
@@ -3027,19 +3045,20 @@ void raid56_parity_cache_data_folios(struct btrfs_raid_bio *rbio,return ;/* data_logical must be at stripe boundary and inside the full stripe. */ -ASSERT (IS_ALIGNED (offset_in_full_stripe,BTRFS_STRIPE_LEN )); -ASSERT (offset_in_full_stripe < (rbio->nr_data <<BTRFS_STRIPE_LEN_SHIFT )); +ASSERT (IS_ALIGNED (offset_in_full_stripe, stripe_len)); +ASSERT (offset_in_full_stripe < (rbio->nr_data <<ilog2 (stripe_len)));for (unsigned int cur_off = offset_in_full_stripe; - cur_off < offset_in_full_stripe +BTRFS_STRIPE_LEN ; + cur_off < offset_in_full_stripe + stripe_len; cur_off +=PAGE_SIZE ) {const unsigned int pindex = cur_off >>PAGE_SHIFT ; +const u32 cur_len =min (stripe_len,PAGE_SIZE ); -ASSERT (cur_off - offset_in_full_stripe +PAGE_SIZE <=BTRFS_STRIPE_LEN ); -memcpy_to_page (rbio->stripe_pages[pindex],0 , - vaddr + cur_off - offset_in_full_stripe,PAGE_SIZE ); +ASSERT (cur_off - offset_in_full_stripe + cur_len <= stripe_len); +memcpy_to_page (rbio->stripe_pages[pindex],offset_in_page (cur_off), + vaddr + cur_off - offset_in_full_stripe, cur_len); }bitmap_set (rbio->stripe_uptodate_bitmap, offset_in_full_stripe >> fs_info->sectorsize_bits, -BTRFS_STRIPE_LEN >> fs_info->sectorsize_bits); + stripe_len >> fs_info->sectorsize_bits); }
fs/btrfs/raid56.h +1 -6
@@ -60,8 +60,6 @@ enum btrfs_rbio_ops {* stripe_pages[] array stores all the pages covering the full stripe, including * data and P/Q pages. * stripe_pages[0] is the first page of the first data stripe. -* stripe_pages[BTRFS_STRIPE_LEN / PAGE_SIZE] is the first page of the second -* data stripe. * * Some pointers inside stripe_pages[] can be NULL, e.g. for a full stripe write * (the bio covers all data stripes) there is no need to allocate pages for
@@ -81,7 +79,7 @@ enum btrfs_rbio_ops {* * - sector_nr * Starts from 0 (representing the first sector of the stripe), ends -* at BTRFS_STRIPE_LEN / sectorsize - 1. +* at stripe_len / sectorsize - 1. * * - step_nr * A step is min(sector_size, PAGE_SIZE).
@@ -153,9 +151,6 @@ struct btrfs_raid_bio {/* Number of all stripes (including P/Q) */ u8 real_stripes; -/* How many pages there are for each stripe */ -u8 stripe_npages; -/* How many sectors there are for each stripe */ u8 stripe_nsectors;
Signature · 1 line
2.56.0
Re: [PATCH v2 7/9] btrfs: convert raid56.c to use rbio->bioc->stripe_len
Looks good,
PATCH v2 8/9 btrfs: convert zoned.c to use map->stripe_len
In reply to Qu WenruoThere are still several call sites using BTRFS_STRIPE_LEN, but that's when doing the basic zoned checks.
For other call sites with a btrfs_chunk_map, use map->stripe_len instead of the hardcoded BTRFS_STRIPE_LEN.
And since we're here, replace the following pattern
if (val & BTRFS_STRIPE_LEN_MASK)
with
if (!IS_ALIGNED(val, stripe_len))
--- fs/btrfs/zoned.c | 55 ++++++++++++++++++++++++++---------------------- 1 file changed, 30 insertions(+), 25 deletions(-)
fs/btrfs/zoned.c +30 -25
@@ -1576,6 +1576,8 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg, u64 last_alloc) {struct btrfs_fs_info *fs_info = bg->fs_info; +const u32 stripe_len = map->stripe_len; +const u8 stripe_shift = map->stripe_shift;u64 stripe_nr =0 , stripe_offset =0 ;u64 prev_offset =0 ;u32 stripe_index =0 ;
@@ -1599,18 +1601,18 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg, zone_info[i].alloc_offset == WP_CONVENTIONAL) continue; - stripe_nr = zone_info[i].alloc_offset >>BTRFS_STRIPE_LEN_SHIFT ; - stripe_offset = zone_info[i].alloc_offset &BTRFS_STRIPE_LEN_MASK ; + stripe_nr = zone_info[i].alloc_offset >> stripe_shift; + stripe_offset = zone_info[i].alloc_offset & (stripe_len -1 );if (stripe_offset ==0 && stripe_nr >0 ) { stripe_nr--; - stripe_offset =BTRFS_STRIPE_LEN ; + stripe_offset = stripe_len; } - alloc = ((stripe_nr * map->num_stripes + i) <<BTRFS_STRIPE_LEN_SHIFT ) + + alloc = ((stripe_nr * map->num_stripes + i) << stripe_shift) + stripe_offset; last_alloc =max (last_alloc, alloc);/* Partially written stripe found. It should be last. */ -if (zone_info[i].alloc_offset &BTRFS_STRIPE_LEN_MASK ) +if (!IS_ALIGNED (zone_info[i].alloc_offset, stripe_len))break ; } stripe_nr =0 ;
@@ -1619,8 +1621,8 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,if (last_alloc) {u32 factor = map->num_stripes; - stripe_nr = last_alloc >>BTRFS_STRIPE_LEN_SHIFT ; - stripe_offset = last_alloc &BTRFS_STRIPE_LEN_MASK ; + stripe_nr = last_alloc >> stripe_shift; + stripe_offset = last_alloc & (stripe_len -1 ); stripe_nr =div_u64_rem (stripe_nr, factor, &stripe_index); }
@@ -1631,11 +1633,10 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,if (zone_info[i].alloc_offset ==WP_CONVENTIONAL ) { has_conventional = true; zone_info[i].alloc_offset = -btrfs_stripe_nr_to_offset (stripe_nr, - map->stripe_len); +btrfs_stripe_nr_to_offset (stripe_nr, stripe_len);if (stripe_index > i) - zone_info[i].alloc_offset +=BTRFS_STRIPE_LEN ; + zone_info[i].alloc_offset += stripe_len;else if (stripe_index == i) zone_info[i].alloc_offset += stripe_offset; }
@@ -1650,7 +1651,8 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg, }if (unlikely (has_partial && - (zone_info[i].alloc_offset &BTRFS_STRIPE_LEN_MASK ))) { + !IS_ALIGNED (zone_info[i].alloc_offset, + stripe_len))) {btrfs_err (fs_info,"zoned: multiple partial written stripe found in block group %llu" , bg->start);
@@ -1659,7 +1661,7 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg, } prev_offset = zone_info[i].alloc_offset; -if ((zone_info[i].alloc_offset &BTRFS_STRIPE_LEN_MASK ) !=0 ) +if (!IS_ALIGNED (zone_info[i].alloc_offset, stripe_len)) has_partial = true;if (test_bit (0 , active) !=test_bit (i, active)) {
@@ -1675,7 +1677,7 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,/* Check if all devices stay in the same stripe row. */ if (unlikely (zone_info[0 ].alloc_offset - - zone_info[map->num_stripes -1 ].alloc_offset >BTRFS_STRIPE_LEN )) { + zone_info[map->num_stripes -1 ].alloc_offset > stripe_len)) {btrfs_err (fs_info,"zoned: stripe gap too large in block group %llu" , bg->start);return -EIO ; }
@@ -1699,6 +1701,8 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,u64 AUTO_KFREE (raid0_allocs );u64 stripe_nr =0 , stripe_offset =0 ;u32 stripe_index =0 ; +const u32 stripe_len = map->stripe_len; +const u8 stripe_shift = map->stripe_shift;bool has_partial = false, has_conventional = false;u64 prev_offset =0 ;
@@ -1746,16 +1750,16 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,return -EIO ; } - stripe_nr = alloc >>BTRFS_STRIPE_LEN_SHIFT ; - stripe_offset = alloc &BTRFS_STRIPE_LEN_MASK ; + stripe_nr = alloc >> stripe_shift; + stripe_offset = alloc & (stripe_len -1 );if (stripe_offset ==0 && stripe_nr >0 ) { stripe_nr--; - stripe_offset =BTRFS_STRIPE_LEN ; + stripe_offset = stripe_len; } alloc = ((stripe_nr * (map->num_stripes / map->sub_stripes) + - (i / map->sub_stripes)) << -BTRFS_STRIPE_LEN_SHIFT ) + stripe_offset; + (i / map->sub_stripes)) << stripe_shift) + + stripe_offset; last_alloc =max (last_alloc, alloc); } stripe_nr =0 ;
@@ -1764,8 +1768,8 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,if (last_alloc) {u32 factor = map->num_stripes / map->sub_stripes; - stripe_nr = last_alloc >>BTRFS_STRIPE_LEN_SHIFT ; - stripe_offset = last_alloc &BTRFS_STRIPE_LEN_MASK ; + stripe_nr = last_alloc >> stripe_shift; + stripe_offset = last_alloc & (stripe_len -1 ); stripe_nr =div_u64_rem (stripe_nr, factor, &stripe_index); }
@@ -1775,10 +1779,10 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,if (raid0_allocs[idx] ==WP_CONVENTIONAL ) { has_conventional = true; raid0_allocs[idx] =btrfs_stripe_nr_to_offset (stripe_nr, - map->stripe_len); + stripe_len);if (stripe_index > idx) - raid0_allocs[idx] +=BTRFS_STRIPE_LEN ; + raid0_allocs[idx] += stripe_len;else if (stripe_index == idx) raid0_allocs[idx] += stripe_offset; }
@@ -1794,7 +1798,8 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg, }if (unlikely (has_partial && - (raid0_allocs[idx] &BTRFS_STRIPE_LEN_MASK ))) { + !IS_ALIGNED (raid0_allocs[idx], + stripe_len))) {btrfs_err (fs_info,"zoned: multiple partial written stripe found in block group %llu" , bg->start);
@@ -1803,7 +1808,7 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg, } prev_offset = raid0_allocs[idx]; -if ((raid0_allocs[idx] &BTRFS_STRIPE_LEN_MASK ) !=0 ) +if (!IS_ALIGNED (raid0_allocs[idx], stripe_len)) has_partial = true; }
@@ -1826,7 +1831,7 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,/* Check if all devices stay in the same stripe row. */ if (unlikely (zone_info[0 ].alloc_offset - - zone_info[map->num_stripes -1 ].alloc_offset >BTRFS_STRIPE_LEN )) { + zone_info[map->num_stripes -1 ].alloc_offset > stripe_len)) {btrfs_err (fs_info,"zoned: stripe gap too large in block group %llu" , bg->start);return -EIO ;
Signature · 1 line
2.56.0
Re: [PATCH v2 8/9] btrfs: convert zoned.c to use map->stripe_len
Looks good,
PATCH v2 9/9 btrfs: introduce extra limits on btrfs_chunk::stripe_len
In reply to Qu WenruoThe new limits are:
- Stripe_len must be no smaller than sectorsize
As sectorsize is still the minimal unit.
- Stripe_len must be no larger than BTRFS_STRIPE_LEN
We still have a lot of things requiring on BTRFS_STRIPE_LEN alignment, e.g. dev extent and chunk allocation.
A stripe_len larger than BTRFS_STRIPE_LEN will cause problems for the existing chunk allocator.
--- fs/btrfs/tree-checker.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-)
fs/btrfs/tree-checker.c +12 -1
@@ -1044,7 +1044,18 @@ int btrfs_check_chunk_valid(const struct btrfs_fs_info *fs_info, logical, length);return -EUCLEAN ; } -if (unlikely (!is_power_of_2 (stripe_len) || stripe_len !=BTRFS_STRIPE_LEN )) { + +/* +* For now the "stripe_len != BTRFS_STRIPE_LEN" will cover all cases, +* but the power_of_2 and stripe_len >= sectorsize and +* stripe_len <= BTRFS_STRIPE_LEN are the requirements for future +* stripe_len support. +* +* In the future, "stripe_len != BTRFS_STRIPE_LEN" will be removed. +*/ +if (unlikely (stripe_len !=BTRFS_STRIPE_LEN || + !is_power_of_2 (stripe_len) || stripe_len < sectorsize || + stripe_len >BTRFS_STRIPE_LEN )) {chunk_err (fs_info, leaf, chunk, logical,"invalid chunk stripe length: %llu" , stripe_len);
Signature · 1 line
2.56.0