[PATCH] fix O_DIRECT read of last block in a sparse file
Jeff Moyer [Fri, 3 Feb 2006 11:04:27 +0000 (03:04 -0800)]
Currently, if you open a file O_DIRECT, truncate it to a size that is not a
multiple of the disk block size, and then try to read the last block in the
file, the read will return 0.  The problem is in do_direct_IO, here:

        /* Handle holes */
        if (!buffer_mapped(map_bh)) {
                char *kaddr;

...

                if (dio->block_in_file >=
                        i_size_read(dio->inode)>>blkbits) {
                        /* We hit eof */
                        page_cache_release(page);
                        goto out;
                }

We shift off any remaining bytes in the final block of the I/O, resulting
in a 0-sized read.  I've attached a patch that fixes this.  I'm not happy
about how ugly the math is getting, so suggestions are more than welcome.

I've tested this with a simple program that performs the steps outlined for
reproducing the problem above.  Without the patch, we get a 0-sized result
from read.  With the patch, we get the correct return value from the short
read.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Cc: Mingming Cao <cmm@us.ibm.com>
Cc: Joel Becker <Joel.Becker@oracle.com>
Cc: "Chen, Kenneth W" <kenneth.w.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/direct-io.c

index 30dbbd1..848044a 100644 (file)
@@ -857,6 +857,7 @@ do_holes:
                        /* Handle holes */
                        if (!buffer_mapped(map_bh)) {
                                char *kaddr;
+                               loff_t i_size_aligned;
 
                                /* AKPM: eargh, -ENOTBLK is a hack */
                                if (dio->rw == WRITE) {
@@ -864,8 +865,14 @@ do_holes:
                                        return -ENOTBLK;
                                }
 
+                               /*
+                                * Be sure to account for a partial block as the
+                                * last block in the file
+                                */
+                               i_size_aligned = ALIGN(i_size_read(dio->inode),
+                                                       1 << blkbits);
                                if (dio->block_in_file >=
-                                       i_size_read(dio->inode)>>blkbits) {
+                                               i_size_aligned >> blkbits) {
                                        /* We hit eof */
                                        page_cache_release(page);
                                        goto out;