[V2] Add flag to files with blocks intentionally past EOF

From: Jiaying Zhang <jiayingz@google.com>

fallocate() may potentially instantiate blocks past EOF, depending
on the flags used when it is called.

e2fsck currently has a test for blocks past i_size, and it
sometimes trips up - noticeably on xfstests 013 which runs fsstress.

This patch from Jiayang does fix it up - it (along with
e2fsprogs updates and other patches recently from Aneesh) has
survived many fsstress runs in a row.


(Eric Sandeen: removed ioctl interface and minor cleanups)

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---


--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Eric,

Thank you very much for bringing the patch up-to-date and improving it!
I have been thinking to re-sync it and send it again but was pulled away
but other things.

The patch looks good. I think it is ok to drop the ioctl support at this
time. We can add it later if there are users need that feature.

Jiaying

On Thu, Jan 21, 2010 at 10:00 AM, Eric Sandeen <sandeen@redhat.com> wrote:
>
> From: Jiaying Zhang <jiayingz@google.com>
>
> fallocate() may potentially instantiate blocks past EOF, depending
> on the flags used when it is called.
>
> e2fsck currently has a test for blocks past i_size, and it
> sometimes trips up - noticeably on xfstests 013 which runs fsstress.
>
> This patch from Jiayang does fix it up - it (along with
> e2fsprogs updates and other patches recently from Aneesh) has
> survived many fsstress runs in a row.
>
>
> (Eric Sandeen: removed ioctl interface and minor cleanups)
>
> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
> ---
>
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 874d169..1f6b936 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -284,6 +284,7 @@ struct flex_groups {
>  #define EXT4_TOPDIR_FL                 0x00020000 /* Top of directory hierarchies*/
>  #define EXT4_HUGE_FILE_FL               0x00040000 /* Set to each huge file */
>  #define EXT4_EXTENTS_FL                        0x00080000 /* Inode uses extents */
> +#define EXT4_EOFBLOCKS_FL              0x00400000 /* Blocks allocated beyond EOF */
>  #define EXT4_RESERVED_FL               0x80000000 /* reserved for ext4 lib */
>
>  #define EXT4_FL_USER_VISIBLE           0x000BDFFF /* User visible flags */
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 765a482..95e94ae 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -3185,7 +3185,7 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
>  {
>        struct ext4_ext_path *path = NULL;
>        struct ext4_extent_header *eh;
> -       struct ext4_extent newex, *ex;
> +       struct ext4_extent newex, *ex, *last_ex;
>        ext4_fsblk_t newblock;
>        int err = 0, depth, ret, cache_type;
>        unsigned int allocated = 0;
> @@ -3366,6 +3366,19 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
>                                        EXT4_STATE_DIO_UNWRITTEN;;
>                }
>        }
> +
> +       if (unlikely(EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL)) {
> +                if (eh->eh_entries) {
> +                       last_ex = EXT_LAST_EXTENT(eh);
> +                       if (iblock + ar.len > le32_to_cpu(last_ex->ee_block)
> +                                           + ext4_ext_get_actual_len(last_ex))
> +                               EXT4_I(inode)->i_flags &= ~EXT4_EOFBLOCKS_FL;
> +               } else {
> +                       WARN_ON(eh->eh_entries == 0);
> +                       ext4_error(inode->i_sb, __func__,
> +                               "inode#%lu, eh->eh_entries = 0!", inode->i_ino);
> +               }
> +       }
>        err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
>        if (err) {
>                /* free data blocks we just allocated */
> @@ -3499,6 +3512,13 @@ static void ext4_falloc_update_inode(struct inode *inode,
>                        i_size_write(inode, new_size);
>                if (new_size > EXT4_I(inode)->i_disksize)
>                        ext4_update_i_disksize(inode, new_size);
> +       } else {
> +               /*
> +                * Mark that we allocate beyond EOF so the subsequent truncate
> +                * can proceed even if the new size is the same as i_size.
> +                */
> +               if (new_size > i_size_read(inode))
> +                       EXT4_I(inode)->i_flags |= EXT4_EOFBLOCKS_FL;
>        }
>
>  }
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index cbf56da..f5802e9 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -4429,6 +4429,8 @@ void ext4_truncate(struct inode *inode)
>        if (!ext4_can_truncate(inode))
>                return;
>
> +       EXT4_I(inode)->i_flags &= ~EXT4_EOFBLOCKS_FL;
> +
>        if (inode->i_size == 0 && !test_opt(inode->i_sb, NO_AUTO_DA_ALLOC))
>                ei->i_state |= EXT4_STATE_DA_ALLOC_CLOSE;
>
> @@ -5284,7 +5286,9 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
>        }
>
>        if (S_ISREG(inode->i_mode) &&
> -           attr->ia_valid & ATTR_SIZE && attr->ia_size < inode->i_size) {
> +           attr->ia_valid & ATTR_SIZE &&
> +           (attr->ia_size < inode->i_size ||
> +            (EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL))) {
>                handle_t *handle;
>
>                handle = ext4_journal_start(inode, 3);
> @@ -5315,6 +5319,9 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
>                                goto err_out;
>                        }
>                }
> +               /* ext4_truncate will clear the flag */
> +               if ((EXT4_I(inode)->i_flags & EXT4_EOFBLOCKS_FL))
> +                       ext4_truncate(inode);
>        }
>
>        rc = inode_setattr(inode, attr);
>
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Thu, Jan 21, 2010 at 12:00:30PM -0600, Eric Sandeen wrote:
> From: Jiaying Zhang <jiayingz@google.com>
> 
> fallocate() may potentially instantiate blocks past EOF, depending
> on the flags used when it is called.
> 
> e2fsck currently has a test for blocks past i_size, and it
> sometimes trips up - noticeably on xfstests 013 which runs fsstress.
> 
> This patch from Jiayang does fix it up - it (along with
> e2fsprogs updates and other patches recently from Aneesh) has
> survived many fsstress runs in a row.
> 
> 
> (Eric Sandeen: removed ioctl interface and minor cleanups)
> 
> Signed-off-by: Eric Sandeen <sandeen@redhat.com>

Applied to the maint branch, although I added back the ability to
truncate the fallocated blocks via chattr.

						- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html