If a file is fallocated on a hole, map->m_lblk + map->m_len may be greater
than ee_block + ee_len.
Signed-off-by: Yongqiang Yang <[email protected]>
---
fs/ext4/extents.c | 1 -
1 files changed, 0 insertions(+), 1 deletions(-)
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 6f0300e..29bb629 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -2943,7 +2943,6 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
/* Pre-conditions */
BUG_ON(!ext4_ext_is_uninitialized(ex));
BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));
- BUG_ON(map->m_lblk + map->m_len > ee_block + ee_len);
/*
* Attempt to transfer newly initialized blocks from the currently
--
1.7.5.1
We need to zero out part of a page which beyond EOF before setting uptodate,
otherwise, mapread or write will see non-zero data beyond EOF.
Signed-off-by: Yongqiang Yang <[email protected]>
---
fs/ext4/page-io.c | 12 ++++++++++++
1 files changed, 12 insertions(+), 0 deletions(-)
diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index 235b79d..9e145b8 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -385,6 +385,18 @@ int ext4_bio_write_page(struct ext4_io_submit *io,
block_end = block_start + blocksize;
if (block_start >= len) {
+ /*
+ * Comments copied from block_write_full_page_endio:
+ *
+ * The page straddles i_size. It must be zeroed out on
+ * each and every writepage invocation because it may
+ * be mmapped. "A file is mapped in multiples of the
+ * page size. For a file that is not a multiple of
+ * the page size, the remaining memory is zeroed when
+ * mapped, and writes to that region are not written
+ * out to the file."
+ */
+ zero_user_segment(page, block_start, block_end);
clear_buffer_dirty(bh);
set_buffer_uptodate(bh);
continue;
--
1.7.5.1
On Tue, Dec 6, 2011 at 21:21, Yongqiang Yang <[email protected]> wrote:
> If a file is fallocated on a hole, map->m_lblk + map->m_len may be greater
> than ee_block + ee_len.
Could you please detail a scenario that leads to this check being invalid?
As I'm to blame for the faulty BUG_ON I'd like to use this as an opportunity
to get properly edified.
> Signed-off-by: Yongqiang Yang <[email protected]>
> ---
> fs/ext4/extents.c | 1 -
> 1 files changed, 0 insertions(+), 1 deletions(-)
>
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 6f0300e..29bb629 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -2943,7 +2943,6 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
> /* Pre-conditions */
> BUG_ON(!ext4_ext_is_uninitialized(ex));
> BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));
> - BUG_ON(map->m_lblk + map->m_len > ee_block + ee_len);
For a bit I thought this would break the fast path logic, however it gets
protected by the checks marked /*L1*/ and /*L2*/ since m_lblk == ee_block
(L1) and m_len < ee_len (L2).
Regards - Eric
>
> /*
> * Attempt to transfer newly initialized blocks from the currently
> --
> 1.7.5.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
[Yet another resend, without HTML and _with_ the list CC'ed. Apologies
for the spam]
On Fri, Dec 9, 2011 at 19:36, Yongqiang Yang <[email protected]> wrote:
>
>
> On Saturday, December 10, 2011, Eric Gouriou <[email protected]> wrote:
>> On Tue, Dec 6, 2011 at 21:21, Yongqiang Yang <[email protected]>
>> wrote:
>>> If a file is fallocated on a hole, map->m_lblk + map->m_len may be
>>> greater
>>> than ee_block + ee_len.
>>
>> Could you please detail a scenario that leads to this check being invalid?
>> As I'm to blame for the faulty BUG_ON I'd like to use this as an
>> opportunity
>> to get properly edified.
> It's easy to reproduce, I think. Just need to write beyond fallocated
> blocks, write back would submit a request to ext4_map_blocks with blocks
> part of which are fallocated while other part of which are not, then the
> bug-on would happen.
Thanks, this makes sense.
>
> I met the bug_on during fsx and after the patch applied it works normal.
>
> I think Allison also tested with this patch.
>
> Yongqiang..
>
>>
>>> Signed-off-by: Yongqiang Yang <[email protected]>
Reviewed-by: Eric Gouriou <[email protected]>
Thanks - Eric
>>> ---
>>> fs/ext4/extents.c | 1 -
>>> 1 files changed, 0 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
>>> index 6f0300e..29bb629 100644
>>> --- a/fs/ext4/extents.c
>>> +++ b/fs/ext4/extents.c
>>> @@ -2943,7 +2943,6 @@ static int ext4_ext_convert_to_initialized(handle_t
>>> *handle,
>>> /* Pre-conditions */
>>> BUG_ON(!ext4_ext_is_uninitialized(ex));
>>> BUG_ON(!in_range(map->m_lblk, ee_block, ee_len));
>>> - BUG_ON(map->m_lblk + map->m_len > ee_block + ee_len);
>>
>> For a bit I thought this would break the fast path logic, however it gets
>> protected by the checks marked /*L1*/ and /*L2*/ since m_lblk == ee_block
>> (L1) and m_len < ee_len (L2).
>>
>> Regards - Eric
>>
>>>
>>> /*
>>> * Attempt to transfer newly initialized blocks from the currently
>>> --
>>> 1.7.5.1
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
> --
> Best Wishes
> Yongqiang Yang
>
On Wed, Dec 07, 2011 at 01:21:26PM +0800, Yongqiang Yang wrote:
> If a file is fallocated on a hole, map->m_lblk + map->m_len may be greater
> than ee_block + ee_len.
>
> Signed-off-by: Yongqiang Yang <[email protected]>
Thanks, applied.
- Ted
On Wed, Dec 07, 2011 at 01:21:27PM +0800, Yongqiang Yang wrote:
> We need to zero out part of a page which beyond EOF before setting uptodate,
> otherwise, mapread or write will see non-zero data beyond EOF.
>
> Signed-off-by: Yongqiang Yang <[email protected]>
Thanks, applied.
- Ted