| Message ID | 20250409111055.3640328-4-hch@lst.de (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [1/8] btrfs: remove the alignment checks in end_bbio_data_read | expand |
在 2025/4/9 20:40, Christoph Hellwig 写道: > This avoid stepping into bvec internals or complicated folio to > page offset calculations. > > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > fs/btrfs/bio.c | 9 ++++----- > fs/btrfs/bio.h | 3 +-- > fs/btrfs/disk-io.c | 7 ++++--- > 3 files changed, 9 insertions(+), 10 deletions(-) > > diff --git a/fs/btrfs/bio.c b/fs/btrfs/bio.c > index 8c2eee1f1878..10f31ee1e8c0 100644 > --- a/fs/btrfs/bio.c > +++ b/fs/btrfs/bio.c > @@ -192,7 +192,7 @@ static void btrfs_end_repair_bio(struct btrfs_bio *repair_bbio, > btrfs_repair_io_failure(fs_info, btrfs_ino(inode), > repair_bbio->file_offset, fs_info->sectorsize, > repair_bbio->saved_iter.bi_sector << SECTOR_SHIFT, > - page_folio(bv->bv_page), bv->bv_offset, mirror); > + bvec_phys(bv), mirror); Just one concern, this is highmem safe? If it's highmem safe, I'm pretty happy to go the single physical address way instead of folio/page + offset. That's way more convenient to handle block size < page size cases. Thanks, Qu
On Thu, Apr 10, 2025 at 07:49:30AM +0930, Qu Wenruo wrote: >> btrfs_repair_io_failure(fs_info, btrfs_ino(inode), >> repair_bbio->file_offset, fs_info->sectorsize, >> repair_bbio->saved_iter.bi_sector << SECTOR_SHIFT, >> - page_folio(bv->bv_page), bv->bv_offset, mirror); >> + bvec_phys(bv), mirror); > > Just one concern, this is highmem safe? > > If it's highmem safe, I'm pretty happy to go the single physical address > way instead of folio/page + offset. > > That's way more convenient to handle block size < page size cases. Physicall addresses and thus bvec_phys work perfectly fine for highmem. That's why I did that here instead of the more obvious virtual address.
On 09.04.25 13:11, Christoph Hellwig wrote: > bio_init(&bio, smap.dev->bdev, &bvec, 1, REQ_OP_WRITE | REQ_SYNC); > bio.bi_iter.bi_sector = smap.physical >> SECTOR_SHIFT; > - ret = bio_add_folio(&bio, folio, length, folio_offset); > - ASSERT(ret); > + __bio_add_page(&bio, phys_to_page(paddr), length, offset_in_page(paddr)); Why are we going back to using a pages and __bio_add_page() here? Can we lift phys_to_folio() from s390 into asm-generic/memory_model.h?
On Thu, Apr 10, 2025 at 06:06:08AM +0000, Johannes Thumshirn wrote: > On 09.04.25 13:11, Christoph Hellwig wrote: > > bio_init(&bio, smap.dev->bdev, &bvec, 1, REQ_OP_WRITE | REQ_SYNC); > > bio.bi_iter.bi_sector = smap.physical >> SECTOR_SHIFT; > > - ret = bio_add_folio(&bio, folio, length, folio_offset); > > - ASSERT(ret); > > + __bio_add_page(&bio, phys_to_page(paddr), length, offset_in_page(paddr)); > > Why are we going back to using a pages and __bio_add_page() here? Because it is the most direct interface right now, bio_add_folio is just a pointless wrapper. I plan to add a bio_add_phys in a bit, but don't want this to depend on it for now. > Can we > lift phys_to_folio() from s390 into asm-generic/memory_model.h? Yes, volunteers please. Also we shouldn't depend on that for now.
diff --git a/fs/btrfs/bio.c b/fs/btrfs/bio.c index 8c2eee1f1878..10f31ee1e8c0 100644 --- a/fs/btrfs/bio.c +++ b/fs/btrfs/bio.c @@ -192,7 +192,7 @@ static void btrfs_end_repair_bio(struct btrfs_bio *repair_bbio, btrfs_repair_io_failure(fs_info, btrfs_ino(inode), repair_bbio->file_offset, fs_info->sectorsize, repair_bbio->saved_iter.bi_sector << SECTOR_SHIFT, - page_folio(bv->bv_page), bv->bv_offset, mirror); + bvec_phys(bv), mirror); } while (mirror != fbio->bbio->mirror_num); done: @@ -803,8 +803,8 @@ void btrfs_submit_bbio(struct btrfs_bio *bbio, int mirror_num) * freeing the bio. */ int btrfs_repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start, - u64 length, u64 logical, struct folio *folio, - unsigned int folio_offset, int mirror_num) + u64 length, u64 logical, phys_addr_t paddr, + int mirror_num) { struct btrfs_io_stripe smap = { 0 }; struct bio_vec bvec; @@ -835,8 +835,7 @@ int btrfs_repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start, bio_init(&bio, smap.dev->bdev, &bvec, 1, REQ_OP_WRITE | REQ_SYNC); bio.bi_iter.bi_sector = smap.physical >> SECTOR_SHIFT; - ret = bio_add_folio(&bio, folio, length, folio_offset); - ASSERT(ret); + __bio_add_page(&bio, phys_to_page(paddr), length, offset_in_page(paddr)); ret = submit_bio_wait(&bio); if (ret) { /* try to remap that extent elsewhere? */ diff --git a/fs/btrfs/bio.h b/fs/btrfs/bio.h index e2fe16074ad6..dc2eb43b7097 100644 --- a/fs/btrfs/bio.h +++ b/fs/btrfs/bio.h @@ -110,7 +110,6 @@ void btrfs_bio_end_io(struct btrfs_bio *bbio, blk_status_t status); void btrfs_submit_bbio(struct btrfs_bio *bbio, int mirror_num); void btrfs_submit_repair_write(struct btrfs_bio *bbio, int mirror_num, bool dev_replace); int btrfs_repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start, - u64 length, u64 logical, struct folio *folio, - unsigned int folio_offset, int mirror_num); + u64 length, u64 logical, phys_addr_t paddr, int mirror_num); #endif diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 3dd555db3d32..7cd1a7a775ed 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -193,10 +193,11 @@ static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, u64 end = min_t(u64, eb->start + eb->len, folio_pos(folio) + eb->folio_size); u32 len = end - start; + phys_addr_t paddr = PFN_PHYS(folio_pfn(folio)) + + offset_in_folio(folio, start); - ret = btrfs_repair_io_failure(fs_info, 0, start, len, - start, folio, offset_in_folio(folio, start), - mirror_num); + ret = btrfs_repair_io_failure(fs_info, 0, start, len, start, + paddr, mirror_num); if (ret) break; }
This avoid stepping into bvec internals or complicated folio to page offset calculations. Signed-off-by: Christoph Hellwig <hch@lst.de> --- fs/btrfs/bio.c | 9 ++++----- fs/btrfs/bio.h | 3 +-- fs/btrfs/disk-io.c | 7 ++++--- 3 files changed, 9 insertions(+), 10 deletions(-)