From: Jens Axboe <axboe@suse.de>
THe direct-io code currently returns -EIO if someone tries to read past the
end of a block device. Change it to shorten the read and to return the
number of bytes which were actually read.
(Written by me, fixed & tested by Jens)
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/fs/direct-io.c | 18 +++++++++++++++---
1 files changed, 15 insertions(+), 3 deletions(-)
diff -puN fs/direct-io.c~dio-handle-eof fs/direct-io.c
--- 25/fs/direct-io.c~dio-handle-eof Wed Oct 20 15:13:10 2004
+++ 25-akpm/fs/direct-io.c Wed Oct 20 15:13:10 2004
@@ -926,6 +926,8 @@ direct_io_worker(int rw, struct kiocb *i
ssize_t ret = 0;
ssize_t ret2;
size_t bytes;
+ size_t bytes_todo;
+ loff_t isize;
dio->bio = NULL;
dio->inode = inode;
@@ -973,16 +975,26 @@ direct_io_worker(int rw, struct kiocb *i
else
dio->pages_in_io = 0;
+ bytes_todo = 0;
for (seg = 0; seg < nr_segs; seg++) {
user_addr = (unsigned long)iov[seg].iov_base;
dio->pages_in_io +=
((user_addr+iov[seg].iov_len +PAGE_SIZE-1)/PAGE_SIZE
- user_addr/PAGE_SIZE);
+ bytes_todo += iov[seg].iov_len;
}
- for (seg = 0; seg < nr_segs; seg++) {
+ isize = i_size_read(inode);
+ if (bytes_todo > (isize - offset))
+ bytes_todo = isize - offset;
+
+ for (seg = 0; seg < nr_segs && bytes_todo; seg++) {
user_addr = (unsigned long)iov[seg].iov_base;
- dio->size += bytes = iov[seg].iov_len;
+ bytes = iov[seg].iov_len;
+ if (bytes > bytes_todo)
+ bytes = bytes_todo;
+ bytes_todo -= bytes;
+ dio->size += bytes;
/* Index into the first page of the first block */
dio->first_block_in_page = (user_addr & ~PAGE_MASK) >> blkbits;
@@ -1003,7 +1015,7 @@ direct_io_worker(int rw, struct kiocb *i
ret = do_direct_IO(dio);
- dio->result += iov[seg].iov_len -
+ dio->result += bytes -
((dio->final_block_in_request - dio->block_in_file) <<
blkbits);
_