From: Steven Whitehouse <swhiteho@redhat.com>
To: cluster-devel.redhat.com
Subject: [Cluster-devel] [GFS2 PATCH] [TRY #2] GFS2: Implement a "bitmap has no extents longer than X" scheme
Date: Thu, 07 Nov 2013 11:32:25 +0000 [thread overview]
Message-ID: <1383823945.2712.4.camel@menhir> (raw)
In-Reply-To: <103368646.17689932.1383753576352.JavaMail.root@redhat.com>
Hi,
On Wed, 2013-11-06 at 10:59 -0500, Bob Peterson wrote:
> Hi,
>
> This is a slightly revised version of a patch I recently sent out:
>
> Before this patch, we used a bit, GBF_FULL, to determine when a bitmap
> was full. With the preceding patch, we started accepting block
> reservations smaller than the ideal size, which requires a lot more
> parsing of the bitmaps. To reduce the amount of bitmap searching, this
> patch implements a scheme whereby each rgrp keeps track of the point
> at this multi-block reservations will fail.
>
The first two patches look good. However this one I'm still not so keen
on. I'd much prefer if we can avoid dropping the GBF_FULL support at
bitmap level if the new test is at rgrp level. It doesn't look like that
is too tricky to achieve either - all we need to do is to not remove the
GBF_FULL support since there doesn't appear to be any conflict between
the two mechanisms.
The net result is a diffstat that looks like this:
[root at chywoon linux-2.6]# diffstat -p1 ./patch3.diff
fs/gfs2/incore.h | 1 +
fs/gfs2/lops.c | 1 +
fs/gfs2/rgrp.c | 32 ++++++++++++++++++++++++++------
3 files changed, 28 insertions(+), 6 deletions(-)
as opposed to your patch which looks like:
[root at chywoon linux-2.6]# diffstat -p1 ../bob-a-3.mbox
fs/gfs2/incore.h | 2 +-
fs/gfs2/lops.c | 2 +-
fs/gfs2/rgrp.c | 53 ++++++++++++++++++++++++++---------------------------
3 files changed, 28 insertions(+), 29 deletions(-)
So there would be fewer changes overall. I've given this a quick spin,
and it hasn't broken yet, so let me know what you think,
Steve.
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index ba1ea67..0132816 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -93,6 +93,7 @@ struct gfs2_rgrpd {
struct gfs2_rgrp_lvb *rd_rgl;
u32 rd_last_alloc;
u32 rd_flags;
+ u32 rd_extfail_pt; /* extent failure point */
#define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
#define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
#define GFS2_RDF_ERROR 0x40000000 /* error in rg */
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 010b9fb..cdadb92 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -83,6 +83,7 @@ static void maybe_release_space(struct gfs2_bufdata *bd)
bd->bd_bh->b_data + bi->bi_offset, bi->bi_len);
clear_bit(GBF_FULL, &bi->bi_flags);
rgd->rd_free_clone = rgd->rd_free;
+ rgd->rd_extfail_pt = rgd->rd_free;
}
/**
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index ba5eb41..af4c8cf 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -641,9 +641,13 @@ static void __rs_deltree(struct gfs2_blkreserv *rs)
/* return reserved blocks to the rgrp */
BUG_ON(rs->rs_rbm.rgd->rd_reserved < rs->rs_free);
rs->rs_rbm.rgd->rd_reserved -= rs->rs_free;
+ /* The rgrp extent failure point is likely not to increase;
+ it will only do so if the freed blocks are somehow
+ contiguous with a span of free blocks that follows. Still,
+ it will force the number to be recalculated later. */
+ rgd->rd_extfail_pt += rs->rs_free;
rs->rs_free = 0;
clear_bit(GBF_FULL, &bi->bi_flags);
- smp_mb__after_clear_bit();
}
}
@@ -1132,6 +1136,8 @@ int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data);
rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK);
rgd->rd_free_clone = rgd->rd_free;
+ /* max out the rgrp allocation failure point */
+ rgd->rd_extfail_pt = rgd->rd_free;
}
if (be32_to_cpu(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) {
rgd->rd_rgl->rl_unlinked = cpu_to_be32(count_unlinked(rgd));
@@ -1593,6 +1599,8 @@ fail:
* Side effects:
* - If looking for free blocks, we set GBF_FULL on each bitmap which
* has no free blocks in it.
+ * - If looking for free blocks, we set rd_extfail_pt on each rgrp which
+ * has come up short on a free block search.
*
* Returns: 0 on success, -ENOSPC if there is no block of the requested state
*/
@@ -1604,6 +1612,8 @@ static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 *minext,
struct buffer_head *bh;
int initial_bii;
u32 initial_offset;
+ int first_bii = rbm->bii;
+ u32 first_offset = rbm->offset;
u32 offset;
u8 *buffer;
int n = 0;
@@ -1679,6 +1689,13 @@ next_iter:
if (minext == NULL || state != GFS2_BLKST_FREE)
return -ENOSPC;
+ /* If the extent was too small, and it's smaller than the smallest
+ to have failed before, remember for future reference that it's
+ useless to search this rgrp again for this amount or more. */
+ if ((first_offset == 0) && (first_bii == 0) &&
+ (*minext < rbm->rgd->rd_extfail_pt))
+ rbm->rgd->rd_extfail_pt = *minext;
+
/* If the maximum extent we found is big enough to fulfill the
minimum requirements, use it anyway. */
if (maxext.len) {
@@ -1924,7 +1941,9 @@ int gfs2_inplace_reserve(struct gfs2_inode *ip, const struct gfs2_alloc_parms *a
}
/* Skip unuseable resource groups */
- if (rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC | GFS2_RDF_ERROR))
+ if ((rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC |
+ GFS2_RDF_ERROR)) ||
+ (ap && (ap->target > rs->rs_rbm.rgd->rd_extfail_pt)))
goto skip_rgrp;
if (sdp->sd_args.ar_rgrplvb)
@@ -2106,10 +2125,10 @@ int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl)
if (rgd == NULL)
return 0;
- gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u r:%u\n",
+ gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u r:%u e:%u\n",
(unsigned long long)rgd->rd_addr, rgd->rd_flags,
rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes,
- rgd->rd_reserved);
+ rgd->rd_reserved, rgd->rd_extfail_pt);
spin_lock(&rgd->rd_rsspin);
for (n = rb_first(&rgd->rd_rstree); n; n = rb_next(&trs->rs_node)) {
trs = rb_entry(n, struct gfs2_blkreserv, rs_node);
@@ -2228,9 +2247,10 @@ int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks,
/* Since all blocks are reserved in advance, this shouldn't happen */
if (error) {
- fs_warn(sdp, "inum=%llu error=%d, nblocks=%u, full=%d\n",
+ fs_warn(sdp, "inum=%llu error=%d, nblocks=%u, full=%d fail_pt=%d\n",
(unsigned long long)ip->i_no_addr, error, *nblocks,
- test_bit(GBF_FULL, &rbm.rgd->rd_bits->bi_flags));
+ test_bit(GBF_FULL, &rbm.rgd->rd_bits->bi_flags),
+ rbm.rgd->rd_extfail_pt);
goto rgrp_error;
}
next prev parent reply other threads:[~2013-11-07 11:32 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <1450086691.17689091.1383753511869.JavaMail.root@redhat.com>
2013-11-06 15:59 ` [Cluster-devel] [GFS2 PATCH] [TRY #2] GFS2: Implement a "bitmap has no extents longer than X" scheme Bob Peterson
2013-11-07 11:32 ` Steven Whitehouse [this message]
2013-11-11 14:01 ` Bob Peterson
2013-11-11 14:52 ` Steven Whitehouse
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1383823945.2712.4.camel@menhir \
--to=swhiteho@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).