Implement ->direct_IO() method in aops. The ->direct_IO() method combines
the existing fuse_direct_read/fuse_direct_write with the generic
do_loop_readv/writev to neatly re-use the existing direct_io framework.
Reaching ->direct_IO() in the read path via generic_file_aio_read ensures
proper synchronization with page cache with its existing framework.
Reaching ->direct_IO() in the write path via fuse_file_aio_write is made
to come via generic_file_direct_write() which makes it play nice with
the page cache w.r.t other mmap pages etc.
On files marked 'direct_io' by the filesystem server, IO always follows
the fuse_direct_read/write path. There is no effect of fcntl(O_DIRECT)
and it always succeeds.
On files not marked with 'direct_io' by the filesystem server, the IO
path depends on O_DIRECT flag by the application. This can be passed
at the time of open() as well as via fcntl().
Note that asynchronous O_DIRECT iocb jobs are completed synchronously
always (this has been the case with FUSE even before this patch)
Signed-off-by: Anand Avati <avati@xxxxxxxxxxx>
---
Sending out this patch for early comments. This patch is STILL UNTESTED.
Intention is to get comments and feedback on a couple of points
0. General approach taken by the patch. Are there any obvious considerations
left out?
1. including "../read_write.h" and usage of its do_loop_read_write. There is
a real good use of it inside ->direct_IO() but it looks like an ugly way
to include the header for the prototypes. Should I be writing my own helper
or extern'ing the prototype definition instead of #include or move the
prototypes (or read_write.h) to include/linux/ ? Or just leave it like this?
fs/fuse/dir.c | 3 ---
fs/fuse/file.c | 43 ++++++++++++++++++++++++++++++++++++++-----
2 files changed, 38 insertions(+), 8 deletions(-)
diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
index d5016071..4271f04 100644
--- a/fs/fuse/dir.c
+++ b/fs/fuse/dir.c
@@ -387,9 +387,6 @@ static int fuse_create_open(struct inode *dir, struct dentry *entry, int mode,
if (fc->no_create)
return -ENOSYS;
- if (flags & O_DIRECT)
- return -EINVAL;
-
forget = fuse_alloc_forget();
if (!forget)
return -ENOMEM;
diff --git a/fs/fuse/file.c b/fs/fuse/file.c
index 82a6646..205696a 100644
--- a/fs/fuse/file.c
+++ b/fs/fuse/file.c
@@ -14,6 +14,7 @@
#include <linux/sched.h>
#include <linux/module.h>
#include <linux/compat.h>
+#include "../read_write.h" /* for io_fn_t and do_loop_readv_writev */
static const struct file_operations fuse_direct_io_file_operations;
@@ -193,10 +194,6 @@ int fuse_open_common(struct inode *inode, struct file *file, bool isdir)
struct fuse_conn *fc = get_fuse_conn(inode);
int err;
- /* VFS checks this, but only _after_ ->open() */
- if (file->f_flags & O_DIRECT)
- return -EINVAL;
-
err = generic_file_open(inode, file);
if (err)
return err;
@@ -979,6 +976,7 @@ static ssize_t fuse_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
size_t count = 0;
+ size_t ocount = 0;
ssize_t written = 0;
struct inode *inode = mapping->host;
ssize_t err;
@@ -986,10 +984,13 @@ static ssize_t fuse_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
WARN_ON(iocb->ki_pos != pos);
- err = generic_segment_checks(iov, &nr_segs, &count, VERIFY_READ);
+ ocount = 0;
+ err = generic_segment_checks(iov, &nr_segs, &ocount, VERIFY_READ);
if (err)
return err;
+ count = ocount;
+
mutex_lock(&inode->i_mutex);
vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
@@ -1009,6 +1010,13 @@ static ssize_t fuse_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
file_update_time(file);
+ if (file->f_flags & O_DIRECT) {
+ written = generic_file_direct_write(iocb, iov, &nr_segs,
+ pos, &iocb->ki_pos,
+ count, ocount);
+ goto out;
+ }
+
iov_iter_init(&i, iov, nr_segs, count, 0);
written = fuse_perform_write(file, mapping, &i, pos);
if (written >= 0)
@@ -2132,6 +2140,30 @@ int fuse_notify_poll_wakeup(struct fuse_conn *fc,
return 0;
}
+static ssize_t
+fuse_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov,
+ loff_t offset, unsigned long nr_segs)
+{
+ ssize_t ret = 0;
+ struct file *file = NULL;
+ loff_t pos = 0;
+ io_fn_t fn = NULL;
+ struct iov *iovp = NULL;
+
+ file = iocb->ki_filp;
+ pos = offset;
+ iovp = (struct iovec *) iov; /* non const */
+
+ if (rw == WRITE)
+ fn = (io_fn_t)fuse_direct_write;
+ else
+ fn = fuse_direct_read;
+
+ ret = do_loop_readv_writev(file, iovp, nr_segs, &pos, fn);
+
+ return ret;
+}
+
static const struct file_operations fuse_file_operations = {
.llseek = fuse_file_llseek,
.read = do_sync_read,
@@ -2177,6 +2209,7 @@ static const struct address_space_operations fuse_file_aops = {
.readpages = fuse_readpages,
.set_page_dirty = __set_page_dirty_nobuffers,
.bmap = fuse_bmap,
+ .direct_IO = fuse_direct_IO,
};
void fuse_init_file_inode(struct inode *inode)
--
1.7.4.4
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
