[PATCH 2/2] xfs_spaceman: Accounting for AGFL blocks
ADRS PICT
adrscube at gmail.com
Wed Jan 28 07:05:22 CST 2015
On Wed, Jan 28, 2015 at 7:27 AM, Dave Chinner <david at fromorbit.com> wrote:
> On Fri, Jan 23, 2015 at 09:04:00AM +0530, Dhruvesh Rathore wrote:
>> Here is the original discussion that Dave wrote for xfs_spaceman:
>>
>> http://oss.sgi.com/archives/xfs/2012-10/msg00363.html
>>
>> These patches were not posted and were just forward ported to a current
>> 3.18-rc7+ for-next XFS tree and were pushed to the fiemapfs branch in
Dave's
>> kernel tree at:
>>
>> git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs.git
>>
>> The original xfs_spaceman tool that Dave wrote is here:
>>
>> http://oss.sgi.com/archives/xfs/2012-10/msg00366.html
>>
>> Dave just updated it to the 3.2.2 code base and pushed it to the
>> spaceman branch in this tree:
>>
>> git://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git
>
> As per last patch, this is series description so it doesn't need to
> be repeated in every patch ;)
>
We will keep that in mind, while sending the updated patches. :)
> From here:
>
>> This xfs_spaceman utility previously failed to account for AGFL blocks.
>> Old output (Before changes).
>>
>> $ sudo xfs_spaceman -c "freesp" /media/xfs
>> from to extents blocks pct
>> 1024 2047 1 1262 0.04
>> 4096 8191 1 5126 0.15
>> 8192 16383 3 35344 1.05
>> 32768 65535 1 43989 1.31
>> 262144 524287 3 1334894 39.78
>> 524288 967428 2 1934840 57.66
>>
>> As you can see the AGFL blocks were unaccounted (4 per AG, and there
>> were 4 AGs in this filesystem).
>>
>> This patch is concerned with adding a new function which will walk the
free
>> extents in AGFL and account for these AGFL blocks, while file system
scanning.
>>
>> New output (After implementing this patch).
>>
>> $ uname -a
>> Linux dhruv-MacBookAir 3.18.0-rc7+ #3 SMP Thu Dec 25 15:29:59 IST 2014
x86_64 x86_64 x86_64 GNU/Linux
>> $ sudo xfs_spaceman -V
>> xfs_spaceman version 3.2.2
>> $ sudo xfs_spaceman -c "freesp" /media/xfs
>> from to extents blocks pct
>> 1 1 16 16 0.00
>> 1024 2047 1 1262 0.04
>> 4096 8191 1 5126 0.15
>> 8192 16383 3 35344 1.05
>> 32768 65535 1 43989 1.31
>> 262144 524287 3 1334894 39.78
>> 524288 967428 2 1934840 57.66
>
> to here, this is a good patch description. :)
>
>> Please do comment.
>
> No need to ask in each patch description - the series description if
> the place for that....
>
We will make a not of this too.
>> +
>> +/*
>> + * Walk the free extents in the AGFL (AG Free List) of each AG, and dump
>
> No need to explain what AGFL means - anyone reading the code already
> knows....
>
>> + * them all into the fieinfo.
>> + *
>> + * With a freshly made filesystem, 4 blocks are reserved immediately
after
>> + * the free space B+tree root blocks (blocks 4 to 7). As they are used
up
>
> The number of blocks is actually dependent on filesystem geometry,
> but again there isn't a need to describe the exact workings of the
> AGFL here.
>
>> + * additional blocks are reserved from the AG and added to the free list
>> + * array. A typical device will have space for 128 elements in the
array.
>> + * The actual size can be determined using XFS_AGFL_SIZE macro. The
array
>> + * is maintained as a circular list and active elements are pointed by
>> + * AGF's agf_flfirst, agf_fllast and agf_flcount values.
>
> Yup, you're describing the code, not /why/ the function exists.
Our bad, we will ensure that the comment, explains why the function is
needed.
>
> /*
> * When we map free space we need to take into account the blocks
> * that are indexed by the AGFL. There aren't found by walking the
> * free space btrees, so we have to walk each AGFL to find them.
> */
>
>> + */
>> +static int
>> +xfs_alloc_agfl_freespace_map(
>> + struct xfs_buf **agbp, /* buffer for a.g. freelist header
*/
>> + struct xfs_btree_cur *cur,
>> + struct fiemap_extent_info *fieinfo,
>> + xfs_agblock_t sagbno,
>> + xfs_agblock_t eagbno)
>
> No need to pass the freespace btree cursor into this function,
> nor the agf buffer. What is required is the struct xfs_mount,
> the current agno and the agf structure. i.e.
> xfs_alloc_agfl_freespace_map(
> struct xfs_mount *mp,
> struct xfs_agf *agf,
> struct fiemap_extent_info *fieinfo,
> xfs_agnumber_t agno,
> xfs_agblock_t sagbno,
> xfs_agblock_t eagbno)
>
>> +{
>> + xfs_agf_t *agf; /* a.g. freespace structure */
>> + xfs_buf_t *agflbp; /* buffer for a.g. freelist
structure */
>> + __be32 *agfl_bno; /* Reference to freelist block */
>> + int j;
> ^
>> + int index;
> ^
>> + int error = 0;
> ^
>
> Stray whitespace.
>
>> +
>> + agf = XFS_BUF_TO_AGF(*agbp);
>> + error = xfs_alloc_read_agfl(cur->bc_mp, NULL,
be32_to_cpu(agf->agf_seqno),
>> + &agflbp);
>
> error = xfs_alloc_read_agfl(mp, NULL, agno, &agflbp);
>
>> + if (error)
>> + return error;
>> +
>> + agfl_bno = XFS_BUF_TO_AGFL_BNO(cur->bc_mp, agflbp);
>> +
>> + index = be32_to_cpu(agf->agf_flfirst);
>> +
>> + for(j=1 ; j<=be32_to_cpu(agf->agf_flcount); j++)
>> + {
>
> formatting, and no need for j and index parameters.
>
> for (i = be32_to_cpu(agf->agf_flfirst);
> i < be32_to_cpu(agf->agf_fllast);) {
> ....
> if (++i == XFS_AGFL_SIZE(mp))
> i = 0;
> }
>
>> + xfs_agblock_t fbno;
>> + xfs_extlen_t flen;
>> + xfs_daddr_t dbno;
>> + xfs_fileoff_t dlen;
>> + int flags = 0;
>
> all the indenting in the function needs fixing.
>
>> +
>> + /* Relative AG block number */
>> + fbno = be32_to_cpu(agfl_bno[index]);
>> + /* Each entry in the AGFL is a single entry i.e length is 1 block
*/
>> + flen = 1;
>
> These can all be folded into the code below - no need for
> intermediate variables...
>
>> +
>> + /*
>> + * AGFL is maintained as a circular list. Following is needed
>> + * to handle array wrapping correctly.
>> + */
>> + if( index == ( (XFS_AGFL_SIZE(cur->bc_mp))-1 ) )
>> + index = 0; /* First index of AGFL */
>> + else
>> + index++; /* Next index in AGFL */
>
> This is handled by the above loop construct so can be ignored.
>> +
>> + /*
>> + * Use daddr format for all range/len calculations as that is
>> + * the format the range/len variables are supplied in by
>> + * userspace.
>> + */
>> + dbno = XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_private.a.agno, fbno);
>> + dlen = XFS_FSB_TO_BB(cur->bc_mp, flen);
>> + error = -fiemap_fill_next_extent(fieinfo, BBTOB(dbno),
>> + BBTOB(dbno), BBTOB(dlen), flags);
>
> This takes bytes as it's values, and we have AG block numbers and
> filesystem block lengths. so:
>
>
> and the entire loop becomes:
>
> for (i = be32_to_cpu(agf->agf_flfirst);
> i < be32_to_cpu(agf->agf_fllast);) {
> /*
> * Use daddr format for all range/len calculations as
that is
> * the format the range/len variables are supplied in by
> * userspace.
> */
> dbno = XFS_AGB_TO_DADDR(mp, agno,
be32_to_cpu(agfl_bno[i]));
> error = fiemap_fill_next_extent(fieinfo,
> BBTOB(dbno), BBTOB(dbno),
> XFS_FSB_TO_B(mp, 1),
flags);
> if (error)
> break;
> if (++i == XFS_AGFL_SIZE(mp))
> i = 0;
> }
> xfs_buf_relse(agflbp);
> return error;
> }
We will incorporate the above mentioned changes, and send the updated
patch at the earliest.
> Hmmm - I think something is still missing - what are the sagbno and
> eagbno parameters supposed to do?
Actually the parameters sagbno and eagbno are not needed in this
function and can be excluded.
> Cheers,
>
> Dave.
> --
> Dave Chinner
> david at fromorbit.com
Thank you for the detailed feedback. We will be sending the updated
patches at the earliest.
Regards,
A-DRS
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oss.sgi.com/pipermail/xfs/attachments/20150128/343bc521/attachment-0001.html>
More information about the xfs
mailing list