xfs
[Top] [All Lists]

Re: [PATCH] xfs: lseek SEEK_DATA/SEEK_HOLE consolidation

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH] xfs: lseek SEEK_DATA/SEEK_HOLE consolidation
From: Jeff Liu <jeff.liu@xxxxxxxxxx>
Date: Mon, 24 Jun 2013 11:28:00 +0800
Cc: "xfs@xxxxxxxxxxx" <xfs@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20130623231320.GC29376@dastard>
References: <51C454DE.2010008@xxxxxxxxxx> <20130623231320.GC29376@dastard>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:11.0) Gecko/20120410 Thunderbird/11.0.1
Hi Dave,
On 06/24/2013 07:13 AM, Dave Chinner wrote:

> On Fri, Jun 21, 2013 at 09:27:58PM +0800, Jeff Liu wrote:
>> From: Jie Liu <jeff.liu@xxxxxxxxxx>
>>
>> Consolidate lseek(2) SEEK_DATA/SEEK_HOLE according to the
>> implementation of VFS lseek_execute():
>> - if end up with a negative offset, return EINVAL if file
>>   is not huge.
>> - if end up with an offset larger than s_maxbytes, return
>>   EINVAL as well.
>> - reset file version to 0 if end up with an offset that is
>>   not equal to the current file offset.
>>
>> Signed-off-by: Jie Liu <jeff.liu@xxxxxxxxxx>
>>
>> ---
>>  fs/xfs/xfs_file.c |   26 ++++++++++++++++++++++++--
>>  1 file changed, 24 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
>> index a5f2042..dc42751 100644
>> --- a/fs/xfs/xfs_file.c
>> +++ b/fs/xfs/xfs_file.c
>> @@ -1270,8 +1270,19 @@ xfs_seek_data(
>>      }
>>  
>>  out:
>> -    if (offset != file->f_pos)
>> +    if (offset < 0 && !(file->f_mode & FMODE_UNSIGNED_OFFSET)) {
>> +            error = EINVAL;
>> +            goto out_unlock;
>> +    }
>> +    if (offset > inode->i_sb->s_maxbytes) {
>> +            error = EINVAL;
>> +            goto out_unlock;
>> +    }
>> +
>> +    if (offset != file->f_pos) {
>>              file->f_pos = offset;
>> +            file->f_version = 0;
>> +    }
> 
> Hi Jeff, why are you copy-n-pasting this code from lseek_execute()
> rather than making lseek_execute() an exported function and calling
> that directly? 

I found other file systems are implemented in this way.  But I should
consider how to make it better rather than following others in this
situation.

> 
>>  
>>  out_unlock:
>>      xfs_iunlock_map_shared(ip, lock);
>> @@ -1372,6 +1383,15 @@ xfs_seek_hole(
>>      }
>>  
>>  out:
>> +    if (offset < 0 && !(file->f_mode & FMODE_UNSIGNED_OFFSET)) {
>> +            error = EINVAL;
>> +            goto out_unlock;
>> +    }
>> +    if (offset > inode->i_sb->s_maxbytes) {
>> +            error = EINVAL;
>> +            goto out_unlock;
>> +    }
> 
> These checks belong after we truncated offset to isize, don't they?

They do.

> And that would make both of these functions simply require a call to
> lseek_execute(), yes?

Yep, I'll post a patch set to export this call and propagate it to other
file systems as well.

Thanks,
-Jeff

<Prev in Thread] Current Thread [Next in Thread>