iov_iter: Use I/O direction from kiocb, iomap & request rather than iov_iter
Subsystems (8)
cover iov_iter: Use I/O direction from kiocb, iomap & request rather than iov_iter
Hi Jens, Christoph,
Here are some patches to switch from using the I/O direction indication in the iov_iter struct to using the I/O direction flags to be found in the kiocb struct, the iomap_iter struct and the request struct. The iterator's I/O direction is then only used in some internal checks.
The patches also add direction flags into iov_iter_extract_pages() so that it can perform some checks. New constants are defined rather than using READ and WRITE so that a check can be made that one of them is specified. The problem with the READ constant is that it is zero and is thus the same as no direction being specified - but if we're modifying the buffer contents (ie. reading into it), we need to know to set FOLL_WRITE. Granted this would be the default if unspecified, but it seems better that this case should be explicit.
There are also patches to make 9P and SCSI use iov_iter_extract_pages().
I've pushed the patches here also:
https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/log/?h=iov-extract
David
David Howells (11):
iov_iter: Fix comment refs to iov_iter_get_pages/pages_alloc()
vfs: Set IOCB_WRITE in iocbs that we're going to write from
vfs: Use init_kiocb() to initialise new IOCBs
iov_iter: Use IOCB_WRITE rather than iterator direction
iov_iter: Use IOMAP_WRITE rather than iterator direction
iov_iter: Use op_is_write() rather than iterator direction
cifs: Drop the check using iov_iter_rw()
iov_iter: Drop iov_iter_rw() and fold in last user
iov_iter: Use I/O dir flags with iov_iter_extract_pages() 9p: Pin pages rather than ref'ing if appropriate
scsi: Use extract_iter_to_sg()
block/bio.c | 6 ++ block/blk-map.c | 5 +- block/fops.c | 8 +-- crypto/af_alg.c | 5 +- crypto/algif_hash.c | 3 +- drivers/block/loop.c | 11 ++-- drivers/nvme/target/io-cmd-file.c | 5 +- drivers/target/target_core_file.c | 2 +- drivers/vhost/scsi.c | 79 ++++++++------------------ fs/9p/vfs_addr.c | 2 +- fs/affs/file.c | 4 +- fs/aio.c | 9 ++- fs/btrfs/ioctl.c | 4 +- fs/cachefiles/io.c | 10 ++-- fs/ceph/file.c | 6 +- fs/dax.c | 6 +- fs/direct-io.c | 28 ++++++---- fs/exfat/inode.c | 6 +- fs/ext2/inode.c | 2 +- fs/f2fs/file.c | 10 ++-- fs/fat/inode.c | 4 +- fs/fuse/dax.c | 2 +- fs/fuse/file.c | 8 +-- fs/hfs/inode.c | 2 +- fs/hfsplus/inode.c | 2 +- fs/iomap/direct-io.c | 4 +- fs/jfs/inode.c | 2 +- fs/nfs/direct.c | 2 +- fs/nilfs2/inode.c | 2 +- fs/ntfs3/inode.c | 2 +- fs/ocfs2/aops.c | 2 +- fs/orangefs/inode.c | 2 +- fs/read_write.c | 10 ++-- fs/reiserfs/inode.c | 2 +- fs/seq_file.c | 2 +- fs/smb/client/smbdirect.c | 9 --- fs/splice.c | 2 +- fs/udf/inode.c | 2 +- include/linux/bio.h | 18 +++++- include/linux/fs.h | 16 +++++- include/linux/mm_types.h | 2 +- include/linux/uio.h | 10 ++-- io_uring/rw.c | 10 ++-- lib/iov_iter.c | 14 ++++- lib/scatterlist.c | 12 +++- mm/filemap.c | 2 +- mm/page_io.c | 4 +- net/9p/trans_common.c | 8 +-- net/9p/trans_common.h | 2 +- net/9p/trans_virtio.c | 92 ++++++++++--------------------- 50 files changed, 221 insertions(+), 241 deletions(-)
RFC PATCH 1/11 iov_iter: Fix comment refs to iov_iter_get_pages/pages_alloc()
Fix references to iov_iter_get_pages/pages_alloc() in comments to refer to the *2 interfaces instead.
--- fs/ceph/file.c | 4 ++-- include/linux/mm_types.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-)
fs/ceph/file.c +2 -2
@@ -75,7 +75,7 @@ static __le32 ceph_flags_sys2wire(u32 flags)*/ /* -* How many pages to get in one call to iov_iter_get_pages(). This +* How many pages to get in one call to iov_iter_get_pages2(). This * determines the size of the on-stack array used as a buffer. */ #define ITER_GET_BVECS_PAGES 64
@@ -115,7 +115,7 @@ static ssize_t __iter_get_bvecs(struct iov_iter *iter, size_t maxsize, }/* -* iov_iter_get_pages() only considers one iov_iter segment, no matter +* iov_iter_get_pages2() only considers one iov_iter segment, no matter * what maxsize or maxpages are given. For ITER_BVEC that is a single * page. *
include/linux/mm_types.h +1 -1
@@ -1249,7 +1249,7 @@ enum {/* * FOLL_LONGTERM indicates that the page will be held for an indefinite * time period _often_ under userspace control. This is in contrast to -* iov_iter_get_pages(), whose usages are transient. +* iov_iter_get_pages2(), whose usages are transient. */ FOLL_LONGTERM = 1 << 8,/* split huge pmd before returning */
RFC PATCH 2/11 vfs: Set IOCB_WRITE in iocbs that we're going to write from
In reply to David HowellsIOCB_WRITE is set by aio, io_uring and cachefiles before submitting a write operation to the VFS, but it isn't set by, say, the write() system call.
Fix this by adding an extra argument to init_sync_kiocb() to indicate the direction and setting that to READ or WRITE, which will cause IOCB_WRITE to be set as appropriate.
Whilst we're at it, rename init_sync_kiocb() to init_kiocb().
This will allow drivers to use IOCB_WRITE instead of the iterator data source to determine the I/O direction.
--- fs/btrfs/ioctl.c | 4 ++-- fs/read_write.c | 10 +++++----- fs/seq_file.c | 2 +- fs/splice.c | 2 +- include/linux/fs.h | 6 +++++- mm/filemap.c | 2 +- mm/page_io.c | 4 ++-- 7 files changed, 17 insertions(+), 13 deletions(-)
fs/btrfs/ioctl.c +2 -2
@@ -4422,7 +4422,7 @@ static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,if (ret <0 ) goto out_iov; -init_sync_kiocb (&kiocb, file); +init_kiocb (&kiocb, file,READ ); kiocb.ki_pos = pos; ret =btrfs_encoded_read (&kiocb, &iter, &args);
@@ -4523,7 +4523,7 @@ static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, boolif (ret <0 ) goto out_end_write; -init_sync_kiocb (&kiocb, file); +init_kiocb (&kiocb, file,WRITE ); ret =kiocb_set_rw_flags (&kiocb,0 );if (ret) goto out_end_write;
fs/read_write.c +5 -5
@@ -382,7 +382,7 @@ static ssize_t new_sync_read(struct file *filp, char __user *buf, size_t len, lostruct iov_iter iter;ssize_t ret; -init_sync_kiocb (&kiocb, filp); +init_kiocb (&kiocb, filp,READ ); kiocb.ki_pos = (ppos ? *ppos :0 );iov_iter_ubuf (&iter,ITER_DEST , buf, len);
@@ -422,7 +422,7 @@ ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos)if (unlikely (!file->f_op->read_iter || file->f_op->read))return warn_unsupported (file,"read" ); -init_sync_kiocb (&kiocb, file); +init_kiocb (&kiocb, file,READ ); kiocb.ki_pos = pos ? *pos :0 ;iov_iter_kvec (&iter,ITER_DEST , &iov,1 , iov.iov_len); ret = file->f_op->read_iter (&kiocb, &iter);
@@ -484,7 +484,7 @@ static ssize_t new_sync_write(struct file *filp, const char __user *buf, size_tstruct iov_iter iter;ssize_t ret; -init_sync_kiocb (&kiocb, filp); +init_kiocb (&kiocb, filp,WRITE ); kiocb.ki_pos = (ppos ? *ppos :0 );iov_iter_ubuf (&iter,ITER_SOURCE , (void __user *)buf, len);
@@ -512,7 +512,7 @@ ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *poif (unlikely (!file->f_op->write_iter || file->f_op->write))return warn_unsupported (file,"write" ); -init_sync_kiocb (&kiocb, file); +init_kiocb (&kiocb, file,WRITE ); kiocb.ki_pos = pos ? *pos :0 ; ret = file->f_op->write_iter (&kiocb, from);if (ret >0 ) {
@@ -723,7 +723,7 @@ static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter,struct kiocb kiocb;ssize_t ret; -init_sync_kiocb (&kiocb, filp); +init_kiocb (&kiocb, filp, type); ret =kiocb_set_rw_flags (&kiocb, flags);if (ret)return ret;
fs/seq_file.c +1 -1
@@ -155,7 +155,7 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos)struct iov_iter iter;ssize_t ret; -init_sync_kiocb (&kiocb, file); +init_kiocb (&kiocb, file,READ );iov_iter_init (&iter,ITER_DEST , &iov,1 , size); kiocb.ki_pos = *ppos;
fs/splice.c +1 -1
@@ -362,7 +362,7 @@ ssize_t copy_splice_read(struct file *in, loff_t *ppos,/* Do the I/O */ iov_iter_bvec (&to,ITER_DEST , bv, npages, len); -init_sync_kiocb (&kiocb, in); +init_kiocb (&kiocb, in,READ ); kiocb.ki_pos = *ppos; ret =call_read_iter (in, &kiocb, &to);
include/linux/fs.h +5 -1
@@ -2017,13 +2017,17 @@ static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap, !vfsgid_valid (i_gid_into_vfsgid (idmap, inode)); } -static inline void init_sync_kiocb (struct kiocb *kiocb,struct file *filp) +static inline void init_kiocb (struct kiocb *kiocb,struct file *filp, +unsigned int rw) { *kiocb = (struct kiocb ) { .ki_filp = filp, .ki_flags = filp->f_iocb_flags, .ki_ioprio =get_current_ioprio (), }; + +if (rw ==WRITE ) + kiocb->ki_flags |=IOCB_WRITE ; } static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
mm/filemap.c +1 -1
@@ -2905,7 +2905,7 @@ ssize_t filemap_splice_read(struct file *in, loff_t *ppos,if (unlikely (*ppos >= in->f_mapping->host->i_sb->s_maxbytes))return 0 ; -init_sync_kiocb (&iocb, in); +init_kiocb (&iocb, in,READ ); iocb.ki_pos = *ppos;/* Work out how much data we can actually add into the pipe */
mm/page_io.c +2 -2
@@ -312,7 +312,7 @@ static void swap_writepage_fs(struct page *page, struct writeback_control *wbc) }if (!sio) { sio =mempool_alloc (sio_pool,GFP_NOIO ); -init_sync_kiocb (&sio->iocb, swap_file); +init_kiocb (&sio->iocb, swap_file,WRITE ); sio->iocb.ki_complete = sio_write_complete; sio->iocb.ki_pos = pos; sio->pages =0 ;
@@ -443,7 +443,7 @@ static void swap_readpage_fs(struct page *page, }if (!sio) { sio =mempool_alloc (sio_pool,GFP_KERNEL ); -init_sync_kiocb (&sio->iocb, sis->swap_file); +init_kiocb (&sio->iocb, sis->swap_file,READ ); sio->iocb.ki_pos = pos; sio->iocb.ki_complete = sio_read_complete; sio->pages =0 ;