Patchwork ext4: correctly calculate number of blocks for fiemap

login
register
mail settings
Submitter Surbhi Palande
Date April 1, 2010, 9:21 a.m.
Message ID <1270113678-5901-1-git-send-email-surbhi.palande@canonical.com>
Download mbox | patch
Permalink /patch/49200/
State Accepted
Delegated to: Andy Whitcroft
Headers show

Comments

Surbhi Palande - April 1, 2010, 9:21 a.m.
The following cherry-picked patch fixes the calculation of extents in fiemap.
Without this patch a user of fiemap could be misled into thinking that there
are no extents in a file when actually there are. This patch is necessary for
code such as ureadahead which uses fiemap to find out if the file extents have
increased.

The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
This patch fixes bug 474597 on LP. Do consider merging this in Lucid.

From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
From: Leonard Michlmayr <leonard.michlmayr@gmail.com>
Date: Thu, 4 Mar 2010 17:07:28 -0500
Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap

ext4_fiemap() rounds the length of the requested range down to
blocksize, which is is not the true number of blocks that cover the
requested region.  This problem is especially impressive if the user
requests only the first byte of a file: not a single extent will be
reported.

We fix this by calculating the last block of the region and then
subtract to find the number of blocks in the extents.

Signed-off-by: Leonard Michlmayr <leonard.michlmayr@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
(cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)

BugLink: http://launchpad.net/bugs/474597
Signed-off-by: Surbhi Palande <surbhi.palande@canonical.com>
---
 fs/ext4/extents.c |    9 +++++++--
 1 files changed, 7 insertions(+), 2 deletions(-)
Stefan Bader - April 1, 2010, 9:37 a.m.
Ok, looks good to me now.

Surbhi Palande wrote:
> The following cherry-picked patch fixes the calculation of extents in fiemap.
> Without this patch a user of fiemap could be misled into thinking that there
> are no extents in a file when actually there are. This patch is necessary for
> code such as ureadahead which uses fiemap to find out if the file extents have
> increased.
> 
> The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
> This patch fixes bug 474597 on LP. Do consider merging this in Lucid.
> 
> From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
> From: Leonard Michlmayr <leonard.michlmayr@gmail.com>
> Date: Thu, 4 Mar 2010 17:07:28 -0500
> Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap
> 
> ext4_fiemap() rounds the length of the requested range down to
> blocksize, which is is not the true number of blocks that cover the
> requested region.  This problem is especially impressive if the user
> requests only the first byte of a file: not a single extent will be
> reported.
> 
> We fix this by calculating the last block of the region and then
> subtract to find the number of blocks in the extents.
> 
> Signed-off-by: Leonard Michlmayr <leonard.michlmayr@gmail.com>
> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
> (cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)
> 
> BugLink: http://launchpad.net/bugs/474597
> Signed-off-by: Surbhi Palande <surbhi.palande@canonical.com>
Acked-by: Stefan Bader <stefan.bader@canonical.com>
> ---
>  fs/ext4/extents.c |    9 +++++++--
>  1 files changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 8b8bae4..f199d20 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -3711,7 +3711,6 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>  		__u64 start, __u64 len)
>  {
>  	ext4_lblk_t start_blk;
> -	ext4_lblk_t len_blks;
>  	int error = 0;
>  
>  	/* fallback to generic here if not in extents fmt */
> @@ -3725,8 +3724,14 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>  	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
>  		error = ext4_xattr_fiemap(inode, fieinfo);
>  	} else {
> +		ext4_lblk_t len_blks;
> +		__u64 last_blk;
> +
>  		start_blk = start >> inode->i_sb->s_blocksize_bits;
> -		len_blks = len >> inode->i_sb->s_blocksize_bits;
> +		last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
> +		if (last_blk >= EXT_MAX_BLOCK)
> +			last_blk = EXT_MAX_BLOCK-1;
> +		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
>  
>  		/*
>  		 * Walk the extent tree gathering extent information.
Brad Figg - April 1, 2010, 2:53 p.m.
On 04/01/2010 02:37 AM, Stefan Bader wrote:
> Ok, looks good to me now.
>
> Surbhi Palande wrote:
>> The following cherry-picked patch fixes the calculation of extents in fiemap.
>> Without this patch a user of fiemap could be misled into thinking that there
>> are no extents in a file when actually there are. This patch is necessary for
>> code such as ureadahead which uses fiemap to find out if the file extents have
>> increased.
>>
>> The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
>> This patch fixes bug 474597 on LP. Do consider merging this in Lucid.
>>
>>  From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
>> From: Leonard Michlmayr<leonard.michlmayr@gmail.com>
>> Date: Thu, 4 Mar 2010 17:07:28 -0500
>> Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap
>>
>> ext4_fiemap() rounds the length of the requested range down to
>> blocksize, which is is not the true number of blocks that cover the
>> requested region.  This problem is especially impressive if the user
>> requests only the first byte of a file: not a single extent will be
>> reported.
>>
>> We fix this by calculating the last block of the region and then
>> subtract to find the number of blocks in the extents.
>>
>> Signed-off-by: Leonard Michlmayr<leonard.michlmayr@gmail.com>
>> Signed-off-by: "Theodore Ts'o"<tytso@mit.edu>
>> (cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)
>>
>> BugLink: http://launchpad.net/bugs/474597
>> Signed-off-by: Surbhi Palande<surbhi.palande@canonical.com>
> Acked-by: Stefan Bader<stefan.bader@canonical.com>
>> ---
>>   fs/ext4/extents.c |    9 +++++++--
>>   1 files changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
>> index 8b8bae4..f199d20 100644
>> --- a/fs/ext4/extents.c
>> +++ b/fs/ext4/extents.c
>> @@ -3711,7 +3711,6 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>>   		__u64 start, __u64 len)
>>   {
>>   	ext4_lblk_t start_blk;
>> -	ext4_lblk_t len_blks;
>>   	int error = 0;
>>
>>   	/* fallback to generic here if not in extents fmt */
>> @@ -3725,8 +3724,14 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>>   	if (fieinfo->fi_flags&  FIEMAP_FLAG_XATTR) {
>>   		error = ext4_xattr_fiemap(inode, fieinfo);
>>   	} else {
>> +		ext4_lblk_t len_blks;
>> +		__u64 last_blk;
>> +
>>   		start_blk = start>>  inode->i_sb->s_blocksize_bits;
>> -		len_blks = len>>  inode->i_sb->s_blocksize_bits;
>> +		last_blk = (start + len - 1)>>  inode->i_sb->s_blocksize_bits;
>> +		if (last_blk>= EXT_MAX_BLOCK)
>> +			last_blk = EXT_MAX_BLOCK-1;
>> +		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
>>
>>   		/*
>>   		 * Walk the extent tree gathering extent information.
>
>

Maybe I'm misreading this bug according to
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/474597/comments/7
ted applied a different patch to fix this issue.

Brad
Surbhi Palande - April 1, 2010, 10:44 p.m.
On Thu, 2010-04-01 at 07:53 -0700, Brad Figg wrote:
> On 04/01/2010 02:37 AM, Stefan Bader wrote:
> > Ok, looks good to me now.
> >
> > Surbhi Palande wrote:
> >> The following cherry-picked patch fixes the calculation of extents in fiemap.
> >> Without this patch a user of fiemap could be misled into thinking that there
> >> are no extents in a file when actually there are. This patch is necessary for
> >> code such as ureadahead which uses fiemap to find out if the file extents have
> >> increased.
> >>
> >> The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
> >> This patch fixes bug 474597 on LP. Do consider merging this in Lucid.
> >>
> >>  From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
> >> From: Leonard Michlmayr<leonard.michlmayr@gmail.com>
> >> Date: Thu, 4 Mar 2010 17:07:28 -0500
> >> Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap
> >>
> >> ext4_fiemap() rounds the length of the requested range down to
> >> blocksize, which is is not the true number of blocks that cover the
> >> requested region.  This problem is especially impressive if the user
> >> requests only the first byte of a file: not a single extent will be
> >> reported.
> >>
> >> We fix this by calculating the last block of the region and then
> >> subtract to find the number of blocks in the extents.
> >>
> >> Signed-off-by: Leonard Michlmayr<leonard.michlmayr@gmail.com>
> >> Signed-off-by: "Theodore Ts'o"<tytso@mit.edu>
> >> (cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)
> >>
> >> BugLink: http://launchpad.net/bugs/474597
> >> Signed-off-by: Surbhi Palande<surbhi.palande@canonical.com>
> > Acked-by: Stefan Bader<stefan.bader@canonical.com>
> >> ---
> >>   fs/ext4/extents.c |    9 +++++++--
> >>   1 files changed, 7 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> >> index 8b8bae4..f199d20 100644
> >> --- a/fs/ext4/extents.c
> >> +++ b/fs/ext4/extents.c
> >> @@ -3711,7 +3711,6 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
> >>   		__u64 start, __u64 len)
> >>   {
> >>   	ext4_lblk_t start_blk;
> >> -	ext4_lblk_t len_blks;
> >>   	int error = 0;
> >>
> >>   	/* fallback to generic here if not in extents fmt */
> >> @@ -3725,8 +3724,14 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
> >>   	if (fieinfo->fi_flags&  FIEMAP_FLAG_XATTR) {
> >>   		error = ext4_xattr_fiemap(inode, fieinfo);
> >>   	} else {
> >> +		ext4_lblk_t len_blks;
> >> +		__u64 last_blk;
> >> +
> >>   		start_blk = start>>  inode->i_sb->s_blocksize_bits;
> >> -		len_blks = len>>  inode->i_sb->s_blocksize_bits;
> >> +		last_blk = (start + len - 1)>>  inode->i_sb->s_blocksize_bits;
> >> +		if (last_blk>= EXT_MAX_BLOCK)
> >> +			last_blk = EXT_MAX_BLOCK-1;
> >> +		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
> >>
> >>   		/*
> >>   		 * Walk the extent tree gathering extent information.
> >
> >
> 
> Maybe I'm misreading this bug according to
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/474597/comments/7
> ted applied a different patch to fix this issue.
> 
Brad, thanks for pointing this out. 

The patch that actually got into the linux kernel does seem to have the
additional check (missing from whats on launchpad)
+if (last_blk>= EXT_MAX_BLOCK)
+	last_blk = EXT_MAX_BLOCK-1;
This check seems necessary for the corner case.

I have pointed the difference out on the buglink too.

Warm Regards,
Surbhi.

> Brad
Andy Whitcroft - April 8, 2010, 5:57 p.m.
On Fri, Apr 02, 2010 at 01:44:41AM +0300, Surbhi Palande wrote:
> On Thu, 2010-04-01 at 07:53 -0700, Brad Figg wrote:
> > On 04/01/2010 02:37 AM, Stefan Bader wrote:
> > > Ok, looks good to me now.
> > >
> > > Surbhi Palande wrote:
> > >> The following cherry-picked patch fixes the calculation of extents in fiemap.
> > >> Without this patch a user of fiemap could be misled into thinking that there
> > >> are no extents in a file when actually there are. This patch is necessary for
> > >> code such as ureadahead which uses fiemap to find out if the file extents have
> > >> increased.
> > >>
> > >> The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
> > >> This patch fixes bug 474597 on LP. Do consider merging this in Lucid.
> > >>
> > >>  From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
> > >> From: Leonard Michlmayr<leonard.michlmayr@gmail.com>
> > >> Date: Thu, 4 Mar 2010 17:07:28 -0500
> > >> Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap
> > >>
> > >> ext4_fiemap() rounds the length of the requested range down to
> > >> blocksize, which is is not the true number of blocks that cover the
> > >> requested region.  This problem is especially impressive if the user
> > >> requests only the first byte of a file: not a single extent will be
> > >> reported.
> > >>
> > >> We fix this by calculating the last block of the region and then
> > >> subtract to find the number of blocks in the extents.
> > >>
> > >> Signed-off-by: Leonard Michlmayr<leonard.michlmayr@gmail.com>
> > >> Signed-off-by: "Theodore Ts'o"<tytso@mit.edu>
> > >> (cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)
> > >>
> > >> BugLink: http://launchpad.net/bugs/474597
> > >> Signed-off-by: Surbhi Palande<surbhi.palande@canonical.com>
> > > Acked-by: Stefan Bader<stefan.bader@canonical.com>
> > >> ---
> > >>   fs/ext4/extents.c |    9 +++++++--
> > >>   1 files changed, 7 insertions(+), 2 deletions(-)
> > >>
> > >> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> > >> index 8b8bae4..f199d20 100644
> > >> --- a/fs/ext4/extents.c
> > >> +++ b/fs/ext4/extents.c
> > >> @@ -3711,7 +3711,6 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
> > >>   		__u64 start, __u64 len)
> > >>   {
> > >>   	ext4_lblk_t start_blk;
> > >> -	ext4_lblk_t len_blks;
> > >>   	int error = 0;
> > >>
> > >>   	/* fallback to generic here if not in extents fmt */
> > >> @@ -3725,8 +3724,14 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
> > >>   	if (fieinfo->fi_flags&  FIEMAP_FLAG_XATTR) {
> > >>   		error = ext4_xattr_fiemap(inode, fieinfo);
> > >>   	} else {
> > >> +		ext4_lblk_t len_blks;
> > >> +		__u64 last_blk;
> > >> +
> > >>   		start_blk = start>>  inode->i_sb->s_blocksize_bits;
> > >> -		len_blks = len>>  inode->i_sb->s_blocksize_bits;
> > >> +		last_blk = (start + len - 1)>>  inode->i_sb->s_blocksize_bits;
> > >> +		if (last_blk>= EXT_MAX_BLOCK)
> > >> +			last_blk = EXT_MAX_BLOCK-1;
> > >> +		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
> > >>
> > >>   		/*
> > >>   		 * Walk the extent tree gathering extent information.
> > >
> > >
> > 
> > Maybe I'm misreading this bug according to
> > https://bugs.launchpad.net/ubuntu/+source/linux/+bug/474597/comments/7
> > ted applied a different patch to fix this issue.
> > 
> Brad, thanks for pointing this out. 
> 
> The patch that actually got into the linux kernel does seem to have the
> additional check (missing from whats on launchpad)
> +if (last_blk>= EXT_MAX_BLOCK)
> +	last_blk = EXT_MAX_BLOCK-1;
> This check seems necessary for the corner case.
> 
> I have pointed the difference out on the buglink too.

That implies that there is a better patch out there.  Am I expecting
that patch to be posted instead here?

-apw
Surbhi Palande - April 14, 2010, 8:19 p.m.
Hi Folks,

I was not clear enough about this patch before.

The patch that I picked was cherry picked from upstream and is the 
correct patch. The one on launchpad was obsolete and I have not picked 
that one. Please do consider this patch (cherry picked from upstream) 
for Lucid.

Thanks!

Warm Regards,
Surbhi.


On 04/01/2010 03:44 PM, Surbhi Palande wrote:
> On Thu, 2010-04-01 at 07:53 -0700, Brad Figg wrote:
>> On 04/01/2010 02:37 AM, Stefan Bader wrote:
>>> Ok, looks good to me now.
>>>
>>> Surbhi Palande wrote:
>>>> The following cherry-picked patch fixes the calculation of extents in fiemap.
>>>> Without this patch a user of fiemap could be misled into thinking that there
>>>> are no extents in a file when actually there are. This patch is necessary for
>>>> code such as ureadahead which uses fiemap to find out if the file extents have
>>>> increased.
>>>>
>>>> The commit id in the upstream code is: aca92ff6f57c000d1b4523e383c8bd6b8269b8b1
>>>> This patch fixes bug 474597 on LP. Do consider merging this in Lucid.
>>>>
>>>>   From dfaa33f0502a8bf9dff0e596783a40ce330c7a9b Mon Sep 17 00:00:00 2001
>>>> From: Leonard Michlmayr<leonard.michlmayr@gmail.com>
>>>> Date: Thu, 4 Mar 2010 17:07:28 -0500
>>>> Subject: [PATCH] ext4: correctly calculate number of blocks for fiemap
>>>>
>>>> ext4_fiemap() rounds the length of the requested range down to
>>>> blocksize, which is is not the true number of blocks that cover the
>>>> requested region.  This problem is especially impressive if the user
>>>> requests only the first byte of a file: not a single extent will be
>>>> reported.
>>>>
>>>> We fix this by calculating the last block of the region and then
>>>> subtract to find the number of blocks in the extents.
>>>>
>>>> Signed-off-by: Leonard Michlmayr<leonard.michlmayr@gmail.com>
>>>> Signed-off-by: "Theodore Ts'o"<tytso@mit.edu>
>>>> (cherry picked from commit aca92ff6f57c000d1b4523e383c8bd6b8269b8b1 upstream)
>>>>
>>>> BugLink: http://launchpad.net/bugs/474597
>>>> Signed-off-by: Surbhi Palande<surbhi.palande@canonical.com>
>>> Acked-by: Stefan Bader<stefan.bader@canonical.com>
>>>> ---
>>>>    fs/ext4/extents.c |    9 +++++++--
>>>>    1 files changed, 7 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
>>>> index 8b8bae4..f199d20 100644
>>>> --- a/fs/ext4/extents.c
>>>> +++ b/fs/ext4/extents.c
>>>> @@ -3711,7 +3711,6 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>>>>    		__u64 start, __u64 len)
>>>>    {
>>>>    	ext4_lblk_t start_blk;
>>>> -	ext4_lblk_t len_blks;
>>>>    	int error = 0;
>>>>
>>>>    	/* fallback to generic here if not in extents fmt */
>>>> @@ -3725,8 +3724,14 @@ int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
>>>>    	if (fieinfo->fi_flags&   FIEMAP_FLAG_XATTR) {
>>>>    		error = ext4_xattr_fiemap(inode, fieinfo);
>>>>    	} else {
>>>> +		ext4_lblk_t len_blks;
>>>> +		__u64 last_blk;
>>>> +
>>>>    		start_blk = start>>   inode->i_sb->s_blocksize_bits;
>>>> -		len_blks = len>>   inode->i_sb->s_blocksize_bits;
>>>> +		last_blk = (start + len - 1)>>   inode->i_sb->s_blocksize_bits;
>>>> +		if (last_blk>= EXT_MAX_BLOCK)
>>>> +			last_blk = EXT_MAX_BLOCK-1;
>>>> +		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
>>>>
>>>>    		/*
>>>>    		 * Walk the extent tree gathering extent information.
>>>
>>>
>>
>> Maybe I'm misreading this bug according to
>> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/474597/comments/7
>> ted applied a different patch to fix this issue.
>>
> Brad, thanks for pointing this out.
>
> The patch that actually got into the linux kernel does seem to have the
> additional check (missing from whats on launchpad)
> +if (last_blk>= EXT_MAX_BLOCK)
> +	last_blk = EXT_MAX_BLOCK-1;
> This check seems necessary for the corner case.
>
> I have pointed the difference out on the buglink too.
>
> Warm Regards,
> Surbhi.
>
>> Brad
>
>
>
Andy Whitcroft - April 22, 2010, 5:59 p.m.
Surbhi, the patch will now hit after release so the bug will need the
SRU justification filling in and the right people subscribing etc.
Thanks!

Applied to Lucid.

-apw

Patch

diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 8b8bae4..f199d20 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -3711,7 +3711,6 @@  int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		__u64 start, __u64 len)
 {
 	ext4_lblk_t start_blk;
-	ext4_lblk_t len_blks;
 	int error = 0;
 
 	/* fallback to generic here if not in extents fmt */
@@ -3725,8 +3724,14 @@  int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
 		error = ext4_xattr_fiemap(inode, fieinfo);
 	} else {
+		ext4_lblk_t len_blks;
+		__u64 last_blk;
+
 		start_blk = start >> inode->i_sb->s_blocksize_bits;
-		len_blks = len >> inode->i_sb->s_blocksize_bits;
+		last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
+		if (last_blk >= EXT_MAX_BLOCK)
+			last_blk = EXT_MAX_BLOCK-1;
+		len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
 
 		/*
 		 * Walk the extent tree gathering extent information.