[patch] fix O_DIRECT read of last block in a sparse file

[patch] fix O_DIRECT read of last block in a sparse file

[Jeff Moyer]

Hi,

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.

Thanks,

Jeff

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>

--- linux-2.6.16-rc1-mm1/fs/direct-io.c.orig	2006-01-19 15:45:50.000000000 -0500
+++ linux-2.6.16-rc1-mm1/fs/direct-io.c	2006-01-19 16:09:33.000000000 -0500
@@ -857,6 +857,7 @@ do_holes:
 			/* Handle holes */
 			if (!buffer_mapped(map_bh)) {
 				char *kaddr;
+				loff_t i_size = i_size_read(dio->inode);
 
 				/* AKPM: eargh, -ENOTBLK is a hack */
 				if (dio->rw == WRITE) {
@@ -864,8 +865,10 @@ do_holes:
 					return -ENOTBLK;
 				}
 
-				if (dio->block_in_file >=
-					i_size_read(dio->inode)>>blkbits) {
+				/* Be sure to account for a partial block as
+				 * the last block in the file */
+				if (dio->block_in_file >= i_size >> blkbits &&
+					!(i_size & ~((1<<blkbits)-1))) {
 					/* We hit eof */
 					page_cache_release(page);
 					goto out;

Re: [patch] fix O_DIRECT read of last block in a sparse file

[Andrew Morton]

Jeff Moyer <jmoyer@redhat.com> wrote:
>
> 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.

OK.  We do have some helper functions to make the math a little clearer. 
How does this look?

--- devel/fs/direct-io.c~fix-o_direct-read-of-last-block-in-a-sparse-file	2006-01-23 15:42:31.000000000 -0800
+++ devel-akpm/fs/direct-io.c	2006-01-23 15:51:14.000000000 -0800
@@ -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;
_

Re: [patch] fix O_DIRECT read of last block in a sparse file

[Jeff Moyer]

==> Regarding Re: [patch] fix O_DIRECT read of last block in a sparse file; Andrew Morton <akpm@osdl.org> adds:

akpm> Jeff Moyer <jmoyer@redhat.com> wrote:
>> 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.

akpm> OK.  We do have some helper functions to make the math a little
akpm> clearer. How does this look?

That's much cleaner, thanks!

-Jeff