xfs
[Top] [All Lists]

Re: [PATCH v6] ext4: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate

To: "Theodore Ts'o" <tytso@xxxxxxx>
Subject: Re: [PATCH v6] ext4: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
From: Namjae Jeon <linkinjeon@xxxxxxxxx>
Date: Mon, 24 Feb 2014 10:22:10 +0900
Cc: Namjae Jeon <linkinjeon@xxxxxxxxx>, viro@xxxxxxxxxxxxxxxxxx, david@xxxxxxxxxxxxx, bpm@xxxxxxx, adilger.kernel@xxxxxxxxx, jack@xxxxxxx, mtk.manpages@xxxxxxxxx, lczerner@xxxxxxxxxx, linux-fsdevel@xxxxxxxxxxxxxxx, xfs@xxxxxxxxxxx, linux-ext4@xxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, Namjae Jeon <namjae.jeon@xxxxxxxxxxx>, Ashish Sangwan <a.sangwan@xxxxxxxxxxx>
Delivered-to: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type; bh=sj00hw7QCMau10S7wg/S9tWvvKmC5+LN9Z6RfpHK8GU=; b=hsk8BZsqbgp5HJ7mDle6rXPbsq9t36SG59qkkFbUhDjlzc1rHfLyZYzWaSoEeSAYmS SewlcYk5e6fhfbABtrXLcuzvr30mCCWqL1/6Ppf/UKyGvF4Ky1sFQqfxP4mikcRr92uP lVxsqX/+rupYCNl8Rdnw6sMH1YYNDBuNplPFiqgy612hdndjmT/IaM96TVfQRtHz2ejq H1el3BBQvrsdHtC8Ps2O2L2oPBtOQR+6NOThMhTy4PmP5iVKmLjQ5zSFZCZZXKrL6LIB TP6fLzqOqUmviMJrtQWUDRfbJL0QTzngB9xU+IsdN8sdObZOYr9OX2Hr/Fw0eIKdjIhr UeXg==
In-reply-to: <20140222170930.GE26637@xxxxxxxxx>
References: <1392908861-3563-1-git-send-email-linkinjeon@xxxxxxxxx> <20140222170930.GE26637@xxxxxxxxx>
2014-02-23 2:09 GMT+09:00, Theodore Ts'o <tytso@xxxxxxx>:
> On Fri, Feb 21, 2014 at 12:07:41AM +0900, Namjae Jeon wrote:
>> From: Namjae Jeon <namjae.jeon@xxxxxxxxxxx>
>>
>> This patch implements fallocate's FALLOC_FL_COLLAPSE_RANGE for Ext4.
>>
>> The semantics of this flag are following:
>> 1) It collapses the range lying between offset and length by removing any
>> data
>>    blocks which are present in this range and than updates all the
>> logical
>>    offsets of extents beyond "offset + len" to nullify the hole created
>> by
>>    removing blocks. In short, it does not leave a hole.
>> 2) It should be used exclusively. No other fallocate flag in combination.
>> 3) Offset and length supplied to fallocate should be fs block size
>> aligned
>>    in case of xfs and ext4.
>> 4) Collaspe range does not work beyond i_size.
>>
>> Signed-off-by: Namjae Jeon <namjae.jeon@xxxxxxxxxxx>
>> Signed-off-by: Ashish Sangwan <a.sangwan@xxxxxxxxxxx>
>> Tested-by: Dongsu Park <dongsu.park@xxxxxxxxxxxxxxxx>
>
> In terms of how to get this upstream, it looks like if we do something
> like this, we can let this patch go via the ext4 tree and we don't
> need to worry about whether the vfs level changes have gone in our
> not.
Okay, Dave already answered.
>
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index ad13359..d7a78ed 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -46,6 +46,10 @@
>
>  #include <trace/events/ext4.h>
>
> +#ifndef FALLOC_FL_COLLAPSE_RANGE
> +#define FALLOC_FL_COLLAPSE_RANGE     0x08
> +#endif
> +
>  /*
>   * used by extent splitting.
>   */
>
>
>> +    ret = ext4_es_remove_extent(inode, punch_start,
>> +                                EXT_MAX_BLOCKS - punch_start - 1);
>> +    if (ret) {
>> +            up_write(&EXT4_I(inode)->i_data_sem);
>> +            goto out_stop;
>> +    }
>
> Doing this at first is probably a bad idea; you should do this at the
> end, and then completely invalidate the es cache for that inode.  That
> way, the right thing happens if you get an error in the middle
> releasing the boxes and shifting the extents:
Okay, I see.

>
>> +
>> +    ret = ext4_ext_remove_space(inode, punch_start, punch_stop - 1);
>> +    if (ret) {
>> +            up_write(&EXT4_I(inode)->i_data_sem);
>> +            goto out_stop;
>> +    }
>> +
>> +    ret = ext4_ext_shift_extents(inode, handle, punch_stop,
>> +                                 punch_stop - punch_start);
>> +    if (ret) {
>> +            up_write(&EXT4_I(inode)->i_data_sem);
>> +            goto out_stop;
>> +    }
>
Hi Ted.
> The fact that you are doing these two as two separate steps is
> dangerous; what if you've already released the blocks in
> ext4_ext_remove_space(), and ext4_ext_shift_extents() fails in the
> middle of the processing?  That will leave the file system
> inconsistent, which would be bad.
If there is error in the middle of extent shifting, the hole will
present between the last shifted extent and the extent at which error
happen so this will be consistent state.
IMHO even if there is error in between the shift, filesystem will be
in consistent state.
Am I missing something?
>
> Making sure that the right happens if there is a failure in the middle
> of the operation is Really Important....
>
>                                       - Ted
>

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