iomap: switch __iomap_dio_rw to use iomap_iter
authorChristoph Hellwig <hch@lst.de>
Wed, 11 Aug 2021 01:33:10 +0000 (18:33 -0700)
committerDarrick J. Wong <djwong@kernel.org>
Tue, 17 Aug 2021 04:26:33 +0000 (21:26 -0700)
Switch __iomap_dio_rw to use iomap_iter.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
fs/btrfs/inode.c
fs/iomap/direct-io.c
include/linux/iomap.h

index 0117d86..3b0595e 100644 (file)
@@ -8194,9 +8194,10 @@ static struct btrfs_dio_private *btrfs_create_dio_private(struct bio *dio_bio,
        return dip;
 }
 
-static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
+static blk_qc_t btrfs_submit_direct(const struct iomap_iter *iter,
                struct bio *dio_bio, loff_t file_offset)
 {
+       struct inode *inode = iter->inode;
        const bool write = (btrfs_op(dio_bio) == BTRFS_MAP_WRITE);
        struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
        const bool raid56 = (btrfs_data_alloc_profile(fs_info) &
@@ -8212,7 +8213,7 @@ static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
        int ret;
        blk_status_t status;
        struct btrfs_io_geometry geom;
-       struct btrfs_dio_data *dio_data = iomap->private;
+       struct btrfs_dio_data *dio_data = iter->iomap.private;
        struct extent_map *em = NULL;
 
        dip = btrfs_create_dio_private(dio_bio, inode, file_offset);
index 41ccbfc..4ecd255 100644 (file)
@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 /*
  * Copyright (C) 2010 Red Hat, Inc.
- * Copyright (c) 2016-2018 Christoph Hellwig.
+ * Copyright (c) 2016-2021 Christoph Hellwig.
  */
 #include <linux/module.h>
 #include <linux/compiler.h>
@@ -59,19 +59,17 @@ int iomap_dio_iopoll(struct kiocb *kiocb, bool spin)
 }
 EXPORT_SYMBOL_GPL(iomap_dio_iopoll);
 
-static void iomap_dio_submit_bio(struct iomap_dio *dio, struct iomap *iomap,
-               struct bio *bio, loff_t pos)
+static void iomap_dio_submit_bio(const struct iomap_iter *iter,
+               struct iomap_dio *dio, struct bio *bio, loff_t pos)
 {
        atomic_inc(&dio->ref);
 
        if (dio->iocb->ki_flags & IOCB_HIPRI)
                bio_set_polled(bio, dio->iocb);
 
-       dio->submit.last_queue = bdev_get_queue(iomap->bdev);
+       dio->submit.last_queue = bdev_get_queue(iter->iomap.bdev);
        if (dio->dops && dio->dops->submit_io)
-               dio->submit.cookie = dio->dops->submit_io(
-                               file_inode(dio->iocb->ki_filp),
-                               iomap, bio, pos);
+               dio->submit.cookie = dio->dops->submit_io(iter, bio, pos);
        else
                dio->submit.cookie = submit_bio(bio);
 }
@@ -181,24 +179,23 @@ static void iomap_dio_bio_end_io(struct bio *bio)
        }
 }
 
-static void
-iomap_dio_zero(struct iomap_dio *dio, struct iomap *iomap, loff_t pos,
-               unsigned len)
+static void iomap_dio_zero(const struct iomap_iter *iter, struct iomap_dio *dio,
+               loff_t pos, unsigned len)
 {
        struct page *page = ZERO_PAGE(0);
        int flags = REQ_SYNC | REQ_IDLE;
        struct bio *bio;
 
        bio = bio_alloc(GFP_KERNEL, 1);
-       bio_set_dev(bio, iomap->bdev);
-       bio->bi_iter.bi_sector = iomap_sector(iomap, pos);
+       bio_set_dev(bio, iter->iomap.bdev);
+       bio->bi_iter.bi_sector = iomap_sector(&iter->iomap, pos);
        bio->bi_private = dio;
        bio->bi_end_io = iomap_dio_bio_end_io;
 
        get_page(page);
        __bio_add_page(bio, page, len, 0);
        bio_set_op_attrs(bio, REQ_OP_WRITE, flags);
-       iomap_dio_submit_bio(dio, iomap, bio, pos);
+       iomap_dio_submit_bio(iter, dio, bio, pos);
 }
 
 /*
@@ -206,8 +203,8 @@ iomap_dio_zero(struct iomap_dio *dio, struct iomap *iomap, loff_t pos,
  * mapping, and whether or not we want FUA.  Note that we can end up
  * clearing the WRITE_FUA flag in the dio request.
  */
-static inline unsigned int
-iomap_dio_bio_opflags(struct iomap_dio *dio, struct iomap *iomap, bool use_fua)
+static inline unsigned int iomap_dio_bio_opflags(struct iomap_dio *dio,
+               const struct iomap *iomap, bool use_fua)
 {
        unsigned int opflags = REQ_SYNC | REQ_IDLE;
 
@@ -229,13 +226,16 @@ iomap_dio_bio_opflags(struct iomap_dio *dio, struct iomap *iomap, bool use_fua)
        return opflags;
 }
 
-static loff_t
-iomap_dio_bio_actor(struct inode *inode, loff_t pos, loff_t length,
-               struct iomap_dio *dio, struct iomap *iomap)
+static loff_t iomap_dio_bio_iter(const struct iomap_iter *iter,
+               struct iomap_dio *dio)
 {
+       const struct iomap *iomap = &iter->iomap;
+       struct inode *inode = iter->inode;
        unsigned int blkbits = blksize_bits(bdev_logical_block_size(iomap->bdev));
        unsigned int fs_block_size = i_blocksize(inode), pad;
        unsigned int align = iov_iter_alignment(dio->submit.iter);
+       loff_t length = iomap_length(iter);
+       loff_t pos = iter->pos;
        unsigned int bio_opf;
        struct bio *bio;
        bool need_zeroout = false;
@@ -286,7 +286,7 @@ iomap_dio_bio_actor(struct inode *inode, loff_t pos, loff_t length,
                /* zero out from the start of the block to the write offset */
                pad = pos & (fs_block_size - 1);
                if (pad)
-                       iomap_dio_zero(dio, iomap, pos - pad, pad);
+                       iomap_dio_zero(iter, dio, pos - pad, pad);
        }
 
        /*
@@ -339,7 +339,7 @@ iomap_dio_bio_actor(struct inode *inode, loff_t pos, loff_t length,
 
                nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter,
                                                 BIO_MAX_VECS);
-               iomap_dio_submit_bio(dio, iomap, bio, pos);
+               iomap_dio_submit_bio(iter, dio, bio, pos);
                pos += n;
        } while (nr_pages);
 
@@ -355,7 +355,7 @@ zero_tail:
                /* zero out from the end of the write to the end of the block */
                pad = pos & (fs_block_size - 1);
                if (pad)
-                       iomap_dio_zero(dio, iomap, pos, fs_block_size - pad);
+                       iomap_dio_zero(iter, dio, pos, fs_block_size - pad);
        }
 out:
        /* Undo iter limitation to current extent */
@@ -365,35 +365,38 @@ out:
        return ret;
 }
 
-static loff_t
-iomap_dio_hole_actor(loff_t length, struct iomap_dio *dio)
+static loff_t iomap_dio_hole_iter(const struct iomap_iter *iter,
+               struct iomap_dio *dio)
 {
-       length = iov_iter_zero(length, dio->submit.iter);
+       loff_t length = iov_iter_zero(iomap_length(iter), dio->submit.iter);
+
        dio->size += length;
        return length;
 }
 
-static loff_t
-iomap_dio_inline_actor(struct inode *inode, loff_t pos, loff_t length,
-               struct iomap_dio *dio, struct iomap *iomap)
+static loff_t iomap_dio_inline_iter(const struct iomap_iter *iomi,
+               struct iomap_dio *dio)
 {
+       const struct iomap *iomap = &iomi->iomap;
        struct iov_iter *iter = dio->submit.iter;
-       void *inline_data = iomap_inline_data(iomap, pos);
+       void *inline_data = iomap_inline_data(iomap, iomi->pos);
+       loff_t length = iomap_length(iomi);
+       loff_t pos = iomi->pos;
        size_t copied;
 
        if (WARN_ON_ONCE(!iomap_inline_data_valid(iomap)))
                return -EIO;
 
        if (dio->flags & IOMAP_DIO_WRITE) {
-               loff_t size = inode->i_size;
+               loff_t size = iomi->inode->i_size;
 
                if (pos > size)
                        memset(iomap_inline_data(iomap, size), 0, pos - size);
                copied = copy_from_iter(inline_data, length, iter);
                if (copied) {
                        if (pos + copied > size)
-                               i_size_write(inode, pos + copied);
-                       mark_inode_dirty(inode);
+                               i_size_write(iomi->inode, pos + copied);
+                       mark_inode_dirty(iomi->inode);
                }
        } else {
                copied = copy_to_iter(inline_data, length, iter);
@@ -402,30 +405,27 @@ iomap_dio_inline_actor(struct inode *inode, loff_t pos, loff_t length,
        return copied;
 }
 
-static loff_t
-iomap_dio_actor(struct inode *inode, loff_t pos, loff_t length,
-               void *data, struct iomap *iomap, struct iomap *srcmap)
+static loff_t iomap_dio_iter(const struct iomap_iter *iter,
+               struct iomap_dio *dio)
 {
-       struct iomap_dio *dio = data;
-
-       switch (iomap->type) {
+       switch (iter->iomap.type) {
        case IOMAP_HOLE:
                if (WARN_ON_ONCE(dio->flags & IOMAP_DIO_WRITE))
                        return -EIO;
-               return iomap_dio_hole_actor(length, dio);
+               return iomap_dio_hole_iter(iter, dio);
        case IOMAP_UNWRITTEN:
                if (!(dio->flags & IOMAP_DIO_WRITE))
-                       return iomap_dio_hole_actor(length, dio);
-               return iomap_dio_bio_actor(inode, pos, length, dio, iomap);
+                       return iomap_dio_hole_iter(iter, dio);
+               return iomap_dio_bio_iter(iter, dio);
        case IOMAP_MAPPED:
-               return iomap_dio_bio_actor(inode, pos, length, dio, iomap);
+               return iomap_dio_bio_iter(iter, dio);
        case IOMAP_INLINE:
-               return iomap_dio_inline_actor(inode, pos, length, dio, iomap);
+               return iomap_dio_inline_iter(iter, dio);
        case IOMAP_DELALLOC:
                /*
                 * DIO is not serialised against mmap() access at all, and so
                 * if the page_mkwrite occurs between the writeback and the
-                * iomap_apply() call in the DIO path, then it will see the
+                * iomap_iter() call in the DIO path, then it will see the
                 * DELALLOC block that the page-mkwrite allocated.
                 */
                pr_warn_ratelimited("Direct I/O collision with buffered writes! File: %pD4 Comm: %.20s\n",
@@ -456,16 +456,19 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 {
        struct address_space *mapping = iocb->ki_filp->f_mapping;
        struct inode *inode = file_inode(iocb->ki_filp);
-       size_t count = iov_iter_count(iter);
-       loff_t pos = iocb->ki_pos;
-       loff_t end = iocb->ki_pos + count - 1, ret = 0;
+       struct iomap_iter iomi = {
+               .inode          = inode,
+               .pos            = iocb->ki_pos,
+               .len            = iov_iter_count(iter),
+               .flags          = IOMAP_DIRECT,
+       };
+       loff_t end = iomi.pos + iomi.len - 1, ret = 0;
        bool wait_for_completion =
                is_sync_kiocb(iocb) || (dio_flags & IOMAP_DIO_FORCE_WAIT);
-       unsigned int iomap_flags = IOMAP_DIRECT;
        struct blk_plug plug;
        struct iomap_dio *dio;
 
-       if (!count)
+       if (!iomi.len)
                return NULL;
 
        dio = kmalloc(sizeof(*dio), GFP_KERNEL);
@@ -486,29 +489,30 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
        dio->submit.last_queue = NULL;
 
        if (iov_iter_rw(iter) == READ) {
-               if (pos >= dio->i_size)
+               if (iomi.pos >= dio->i_size)
                        goto out_free_dio;
 
                if (iocb->ki_flags & IOCB_NOWAIT) {
-                       if (filemap_range_needs_writeback(mapping, pos, end)) {
+                       if (filemap_range_needs_writeback(mapping, iomi.pos,
+                                       end)) {
                                ret = -EAGAIN;
                                goto out_free_dio;
                        }
-                       iomap_flags |= IOMAP_NOWAIT;
+                       iomi.flags |= IOMAP_NOWAIT;
                }
 
                if (iter_is_iovec(iter))
                        dio->flags |= IOMAP_DIO_DIRTY;
        } else {
-               iomap_flags |= IOMAP_WRITE;
+               iomi.flags |= IOMAP_WRITE;
                dio->flags |= IOMAP_DIO_WRITE;
 
                if (iocb->ki_flags & IOCB_NOWAIT) {
-                       if (filemap_range_has_page(mapping, pos, end)) {
+                       if (filemap_range_has_page(mapping, iomi.pos, end)) {
                                ret = -EAGAIN;
                                goto out_free_dio;
                        }
-                       iomap_flags |= IOMAP_NOWAIT;
+                       iomi.flags |= IOMAP_NOWAIT;
                }
 
                /* for data sync or sync, we need sync completion processing */
@@ -527,12 +531,13 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 
        if (dio_flags & IOMAP_DIO_OVERWRITE_ONLY) {
                ret = -EAGAIN;
-               if (pos >= dio->i_size || pos + count > dio->i_size)
+               if (iomi.pos >= dio->i_size ||
+                   iomi.pos + iomi.len > dio->i_size)
                        goto out_free_dio;
-               iomap_flags |= IOMAP_OVERWRITE_ONLY;
+               iomi.flags |= IOMAP_OVERWRITE_ONLY;
        }
 
-       ret = filemap_write_and_wait_range(mapping, pos, end);
+       ret = filemap_write_and_wait_range(mapping, iomi.pos, end);
        if (ret)
                goto out_free_dio;
 
@@ -542,9 +547,10 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
                 * If this invalidation fails, let the caller fall back to
                 * buffered I/O.
                 */
-               if (invalidate_inode_pages2_range(mapping, pos >> PAGE_SHIFT,
-                               end >> PAGE_SHIFT)) {
-                       trace_iomap_dio_invalidate_fail(inode, pos, count);
+               if (invalidate_inode_pages2_range(mapping,
+                               iomi.pos >> PAGE_SHIFT, end >> PAGE_SHIFT)) {
+                       trace_iomap_dio_invalidate_fail(inode, iomi.pos,
+                                                       iomi.len);
                        ret = -ENOTBLK;
                        goto out_free_dio;
                }
@@ -559,31 +565,23 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
        inode_dio_begin(inode);
 
        blk_start_plug(&plug);
-       do {
-               ret = iomap_apply(inode, pos, count, iomap_flags, ops, dio,
-                               iomap_dio_actor);
-               if (ret <= 0) {
-                       /* magic error code to fall back to buffered I/O */
-                       if (ret == -ENOTBLK) {
-                               wait_for_completion = true;
-                               ret = 0;
-                       }
-                       break;
-               }
-               pos += ret;
-
-               if (iov_iter_rw(iter) == READ && pos >= dio->i_size) {
-                       /*
-                        * We only report that we've read data up to i_size.
-                        * Revert iter to a state corresponding to that as
-                        * some callers (such as splice code) rely on it.
-                        */
-                       iov_iter_revert(iter, pos - dio->i_size);
-                       break;
-               }
-       } while ((count = iov_iter_count(iter)) > 0);
+       while ((ret = iomap_iter(&iomi, ops)) > 0)
+               iomi.processed = iomap_dio_iter(&iomi, dio);
        blk_finish_plug(&plug);
 
+       /*
+        * We only report that we've read data up to i_size.
+        * Revert iter to a state corresponding to that as some callers (such
+        * as the splice code) rely on it.
+        */
+       if (iov_iter_rw(iter) == READ && iomi.pos >= dio->i_size)
+               iov_iter_revert(iter, iomi.pos - dio->i_size);
+
+       /* magic error code to fall back to buffered I/O */
+       if (ret == -ENOTBLK) {
+               wait_for_completion = true;
+               ret = 0;
+       }
        if (ret < 0)
                iomap_dio_set_error(dio, ret);
 
index aac4176..66e04ae 100644 (file)
@@ -322,8 +322,8 @@ int iomap_writepages(struct address_space *mapping,
 struct iomap_dio_ops {
        int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
                      unsigned flags);
-       blk_qc_t (*submit_io)(struct inode *inode, struct iomap *iomap,
-                       struct bio *bio, loff_t file_offset);
+       blk_qc_t (*submit_io)(const struct iomap_iter *iter, struct bio *bio,
+                             loff_t file_offset);
 };
 
 /*