2010-06-11 06:03:30

by Tao Ma

[permalink] [raw]
Subject: [PATCH] xfs: Make fiemap works with sparse file.

In xfs_vn_fiemap, we set bvm_count to fi_extent_max + 1 and want
to return fi_extent_max extents, but actually it won't work for
a sparse file. The reason is that in xfs_getbmap we will
calculate holes and set it in 'out', while out is malloced by
bmv_count(fi_extent_max+1) which didn't consider holes. So in the
worst case, if 'out' vector looks like
[hole, extent, hole, extent, hole, ... hole, extent, hole],
we will only return half of fi_extent_max extents.

So in xfs_vn_fiemap, we should consider this worst case. If the
user wants fi_extent_max extents, we need a 'out' with size of
2 *fi_extent_max + 1.

Cc: Alex Elder <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Cc: Dave Chinner <[email protected]>
Signed-off-by: Tao Ma <[email protected]>
---
fs/xfs/linux-2.6/xfs_iops.c | 16 ++++++++++++++--
1 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
index 9c8019c..1db92e3 100644
--- a/fs/xfs/linux-2.6/xfs_iops.c
+++ b/fs/xfs/linux-2.6/xfs_iops.c
@@ -672,9 +672,21 @@ xfs_vn_fiemap(
else
bm.bmv_length = BTOBB(length);

- /* We add one because in getbmap world count includes the header */
+ /*
+ * It is a bit tricky for us to calculate the bmv_count from
+ * fi_extent_max.
+ * If we support to return fi_extent_max extents to the user,
+ * we need at most 2 * fi_extent_max + 1 for bmv_count since
+ * in xfs_getbmap we will calculate holes while fi_extent_max
+ * don't have them. So in the worst case, bmv can looks like
+ * [hole, extent, hole, extent, hole, ... hole, extent, hole].
+ * So there will be 2 *fi_extent_max + 1.
+ * What's more, in getbmap world count have to include the
+ * header, so we need another bmv. So the total number will
+ * be 2 * fieinfo->fi_extents_max + 2.
+ */
bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
- fieinfo->fi_extents_max + 1;
+ 2 * fieinfo->fi_extents_max + 2;
bm.bmv_count = min_t(__s32, bm.bmv_count,
(PAGE_SIZE * 16 / sizeof(struct getbmapx)));
bm.bmv_iflags = BMV_IF_PREALLOC;
--
1.5.5


2010-06-11 15:53:52

by Eric Sandeen

[permalink] [raw]
Subject: Re: [PATCH] xfs: Make fiemap works with sparse file.

Tao Ma wrote:
> In xfs_vn_fiemap, we set bvm_count to fi_extent_max + 1 and want
> to return fi_extent_max extents, but actually it won't work for
> a sparse file. The reason is that in xfs_getbmap we will
> calculate holes and set it in 'out', while out is malloced by
> bmv_count(fi_extent_max+1) which didn't consider holes. So in the
> worst case, if 'out' vector looks like
> [hole, extent, hole, extent, hole, ... hole, extent, hole],
> we will only return half of fi_extent_max extents.
>
> So in xfs_vn_fiemap, we should consider this worst case. If the
> user wants fi_extent_max extents, we need a 'out' with size of
> 2 *fi_extent_max + 1.

This all seems right to me, though your commit message above (+1)
doesn't match the comment and code in the patch (+2)

-Eric

> Cc: Alex Elder <[email protected]>
> Cc: Christoph Hellwig <[email protected]>
> Cc: Dave Chinner <[email protected]>
> Signed-off-by: Tao Ma <[email protected]>
> ---
> fs/xfs/linux-2.6/xfs_iops.c | 16 ++++++++++++++--
> 1 files changed, 14 insertions(+), 2 deletions(-)
>
> diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
> index 9c8019c..1db92e3 100644
> --- a/fs/xfs/linux-2.6/xfs_iops.c
> +++ b/fs/xfs/linux-2.6/xfs_iops.c
> @@ -672,9 +672,21 @@ xfs_vn_fiemap(
> else
> bm.bmv_length = BTOBB(length);
>
> - /* We add one because in getbmap world count includes the header */
> + /*
> + * It is a bit tricky for us to calculate the bmv_count from
> + * fi_extent_max.
> + * If we support to return fi_extent_max extents to the user,
> + * we need at most 2 * fi_extent_max + 1 for bmv_count since
> + * in xfs_getbmap we will calculate holes while fi_extent_max
> + * don't have them. So in the worst case, bmv can looks like
> + * [hole, extent, hole, extent, hole, ... hole, extent, hole].
> + * So there will be 2 *fi_extent_max + 1.
> + * What's more, in getbmap world count have to include the
> + * header, so we need another bmv. So the total number will
> + * be 2 * fieinfo->fi_extents_max + 2.
> + */
> bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
> - fieinfo->fi_extents_max + 1;
> + 2 * fieinfo->fi_extents_max + 2;
> bm.bmv_count = min_t(__s32, bm.bmv_count,
> (PAGE_SIZE * 16 / sizeof(struct getbmapx)));
> bm.bmv_iflags = BMV_IF_PREALLOC;

2010-06-11 23:39:11

by Tao Ma

[permalink] [raw]
Subject: Re: [PATCH] xfs: Make fiemap works with sparse file.

Eric Sandeen wrote:
> Tao Ma wrote:
>
>> In xfs_vn_fiemap, we set bvm_count to fi_extent_max + 1 and want
>> to return fi_extent_max extents, but actually it won't work for
>> a sparse file. The reason is that in xfs_getbmap we will
>> calculate holes and set it in 'out', while out is malloced by
>> bmv_count(fi_extent_max+1) which didn't consider holes. So in the
>> worst case, if 'out' vector looks like
>> [hole, extent, hole, extent, hole, ... hole, extent, hole],
>> we will only return half of fi_extent_max extents.
>>
>> So in xfs_vn_fiemap, we should consider this worst case. If the
>> user wants fi_extent_max extents, we need a 'out' with size of
>> 2 *fi_extent_max + 1.
>>
>
> This all seems right to me, though your commit message above (+1)
> doesn't match the comment and code in the patch (+2)
>
oh, yes, I will change the commit log and send a v2.
and can I add your ack for this patch?

Regards,
Tao
> -Eric
>
>
>> Cc: Alex Elder <[email protected]>
>> Cc: Christoph Hellwig <[email protected]>
>> Cc: Dave Chinner <[email protected]>
>> Signed-off-by: Tao Ma <[email protected]>
>> ---
>> fs/xfs/linux-2.6/xfs_iops.c | 16 ++++++++++++++--
>> 1 files changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
>> index 9c8019c..1db92e3 100644
>> --- a/fs/xfs/linux-2.6/xfs_iops.c
>> +++ b/fs/xfs/linux-2.6/xfs_iops.c
>> @@ -672,9 +672,21 @@ xfs_vn_fiemap(
>> else
>> bm.bmv_length = BTOBB(length);
>>
>> - /* We add one because in getbmap world count includes the header */
>> + /*
>> + * It is a bit tricky for us to calculate the bmv_count from
>> + * fi_extent_max.
>> + * If we support to return fi_extent_max extents to the user,
>> + * we need at most 2 * fi_extent_max + 1 for bmv_count since
>> + * in xfs_getbmap we will calculate holes while fi_extent_max
>> + * don't have them. So in the worst case, bmv can looks like
>> + * [hole, extent, hole, extent, hole, ... hole, extent, hole].
>> + * So there will be 2 *fi_extent_max + 1.
>> + * What's more, in getbmap world count have to include the
>> + * header, so we need another bmv. So the total number will
>> + * be 2 * fieinfo->fi_extents_max + 2.
>> + */
>> bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
>> - fieinfo->fi_extents_max + 1;
>> + 2 * fieinfo->fi_extents_max + 2;
>> bm.bmv_count = min_t(__s32, bm.bmv_count,
>> (PAGE_SIZE * 16 / sizeof(struct getbmapx)));
>> bm.bmv_iflags = BMV_IF_PREALLOC;
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>

2010-06-11 23:54:17

by Eric Sandeen

[permalink] [raw]
Subject: Re: [PATCH] xfs: Make fiemap works with sparse file.

Tao Ma wrote:
> Eric Sandeen wrote:
>> Tao Ma wrote:
>>
>>> In xfs_vn_fiemap, we set bvm_count to fi_extent_max + 1 and want
>>> to return fi_extent_max extents, but actually it won't work for
>>> a sparse file. The reason is that in xfs_getbmap we will
>>> calculate holes and set it in 'out', while out is malloced by
>>> bmv_count(fi_extent_max+1) which didn't consider holes. So in the
>>> worst case, if 'out' vector looks like
>>> [hole, extent, hole, extent, hole, ... hole, extent, hole],
>>> we will only return half of fi_extent_max extents.
>>>
>>> So in xfs_vn_fiemap, we should consider this worst case. If the
>>> user wants fi_extent_max extents, we need a 'out' with size of
>>> 2 *fi_extent_max + 1.
>>>
>>
>> This all seems right to me, though your commit message above (+1)
>> doesn't match the comment and code in the patch (+2)
>>
> oh, yes, I will change the commit log and send a v2.
> and can I add your ack for this patch?

Sure, it seems right to me.

Thanks,
-Eric

> Regards,
> Tao
>> -Eric
>>
>>
>>> Cc: Alex Elder <[email protected]>
>>> Cc: Christoph Hellwig <[email protected]>
>>> Cc: Dave Chinner <[email protected]>
>>> Signed-off-by: Tao Ma <[email protected]>
>>> ---
>>> fs/xfs/linux-2.6/xfs_iops.c | 16 ++++++++++++++--
>>> 1 files changed, 14 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
>>> index 9c8019c..1db92e3 100644
>>> --- a/fs/xfs/linux-2.6/xfs_iops.c
>>> +++ b/fs/xfs/linux-2.6/xfs_iops.c
>>> @@ -672,9 +672,21 @@ xfs_vn_fiemap(
>>> else
>>> bm.bmv_length = BTOBB(length);
>>>
>>> - /* We add one because in getbmap world count includes the header */
>>> + /*
>>> + * It is a bit tricky for us to calculate the bmv_count from
>>> + * fi_extent_max.
>>> + * If we support to return fi_extent_max extents to the user,
>>> + * we need at most 2 * fi_extent_max + 1 for bmv_count since
>>> + * in xfs_getbmap we will calculate holes while fi_extent_max
>>> + * don't have them. So in the worst case, bmv can looks like
>>> + * [hole, extent, hole, extent, hole, ... hole, extent, hole].
>>> + * So there will be 2 *fi_extent_max + 1.
>>> + * What's more, in getbmap world count have to include the
>>> + * header, so we need another bmv. So the total number will
>>> + * be 2 * fieinfo->fi_extents_max + 2.
>>> + */
>>> bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
>>> - fieinfo->fi_extents_max + 1;
>>> + 2 * fieinfo->fi_extents_max + 2;
>>> bm.bmv_count = min_t(__s32, bm.bmv_count,
>>> (PAGE_SIZE * 16 / sizeof(struct getbmapx)));
>>> bm.bmv_iflags = BMV_IF_PREALLOC;
>>>
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe
>> linux-kernel" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://www.tux.org/lkml/
>>
>