2006-01-16 23:50:04 +07:00
|
|
|
/*
|
|
|
|
* Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
|
2008-01-31 23:31:39 +07:00
|
|
|
* Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
|
2006-01-16 23:50:04 +07:00
|
|
|
*
|
|
|
|
* This copyrighted material is made available to anyone wishing to use,
|
|
|
|
* modify, copy, or redistribute it subject to the terms and conditions
|
2006-09-01 22:05:15 +07:00
|
|
|
* of the GNU General Public License version 2.
|
2006-01-16 23:50:04 +07:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
#include <linux/completion.h>
|
|
|
|
#include <linux/buffer_head.h>
|
2006-02-28 05:23:27 +07:00
|
|
|
#include <linux/gfs2_ondisk.h>
|
2006-09-19 12:56:29 +07:00
|
|
|
#include <linux/lm_interface.h>
|
2006-01-16 23:50:04 +07:00
|
|
|
|
|
|
|
#include "gfs2.h"
|
2006-02-28 05:23:27 +07:00
|
|
|
#include "incore.h"
|
2006-01-16 23:50:04 +07:00
|
|
|
#include "bmap.h"
|
|
|
|
#include "glock.h"
|
|
|
|
#include "glops.h"
|
|
|
|
#include "inode.h"
|
|
|
|
#include "log.h"
|
|
|
|
#include "meta_io.h"
|
|
|
|
#include "recovery.h"
|
|
|
|
#include "rgrp.h"
|
2006-02-28 05:23:27 +07:00
|
|
|
#include "util.h"
|
2006-10-03 22:10:41 +07:00
|
|
|
#include "trans.h"
|
2006-01-16 23:50:04 +07:00
|
|
|
|
2006-10-03 22:10:41 +07:00
|
|
|
/**
|
|
|
|
* ail_empty_gl - remove all buffers for a given lock from the AIL
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
* None of the buffers should be dirty, locked, or pinned.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void gfs2_ail_empty_gl(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp = gl->gl_sbd;
|
|
|
|
unsigned int blocks;
|
|
|
|
struct list_head *head = &gl->gl_ail_list;
|
|
|
|
struct gfs2_bufdata *bd;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
blocks = atomic_read(&gl->gl_ail_count);
|
|
|
|
if (!blocks)
|
|
|
|
return;
|
|
|
|
|
|
|
|
error = gfs2_trans_begin(sdp, 0, blocks);
|
|
|
|
if (gfs2_assert_withdraw(sdp, !error))
|
|
|
|
return;
|
|
|
|
|
|
|
|
gfs2_log_lock(sdp);
|
|
|
|
while (!list_empty(head)) {
|
|
|
|
bd = list_entry(head->next, struct gfs2_bufdata,
|
|
|
|
bd_ail_gl_list);
|
|
|
|
bh = bd->bd_bh;
|
2007-10-15 22:29:05 +07:00
|
|
|
gfs2_remove_from_ail(bd);
|
2007-09-03 17:01:33 +07:00
|
|
|
bd->bd_bh = NULL;
|
|
|
|
bh->b_private = NULL;
|
|
|
|
bd->bd_blkno = bh->b_blocknr;
|
|
|
|
gfs2_assert_withdraw(sdp, !buffer_busy(bh));
|
|
|
|
gfs2_trans_add_revoke(sdp, bd);
|
2006-10-03 22:10:41 +07:00
|
|
|
}
|
|
|
|
gfs2_assert_withdraw(sdp, !atomic_read(&gl->gl_ail_count));
|
|
|
|
gfs2_log_unlock(sdp);
|
|
|
|
|
|
|
|
gfs2_trans_end(sdp);
|
|
|
|
gfs2_log_flush(sdp, NULL);
|
|
|
|
}
|
2006-07-26 22:27:10 +07:00
|
|
|
|
|
|
|
/**
|
|
|
|
* gfs2_pte_inval - Sync and invalidate all PTEs associated with a glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void gfs2_pte_inval(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_inode *ip;
|
|
|
|
struct inode *inode;
|
|
|
|
|
|
|
|
ip = gl->gl_object;
|
|
|
|
inode = &ip->i_inode;
|
2006-11-02 00:22:46 +07:00
|
|
|
if (!ip || !S_ISREG(inode->i_mode))
|
2006-07-26 22:27:10 +07:00
|
|
|
return;
|
|
|
|
|
|
|
|
unmap_shared_mapping_range(inode->i_mapping, 0, 0);
|
|
|
|
if (test_bit(GIF_SW_PAGED, &ip->i_flags))
|
|
|
|
set_bit(GLF_DIRTY, &gl->gl_flags);
|
|
|
|
|
|
|
|
}
|
|
|
|
|
2006-01-16 23:50:04 +07:00
|
|
|
/**
|
|
|
|
* meta_go_sync - sync out the metadata for this glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
* Called when demoting or unlocking an EX glock. We must flush
|
|
|
|
* to disk all dirty buffers/pages relating to this glock, and must not
|
|
|
|
* not return to caller to demote/unlock the glock until I/O is complete.
|
|
|
|
*/
|
|
|
|
|
2006-11-20 22:37:45 +07:00
|
|
|
static void meta_go_sync(struct gfs2_glock *gl)
|
2006-01-16 23:50:04 +07:00
|
|
|
{
|
2007-01-23 00:15:34 +07:00
|
|
|
if (gl->gl_state != LM_ST_EXCLUSIVE)
|
|
|
|
return;
|
|
|
|
|
2006-01-16 23:50:04 +07:00
|
|
|
if (test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) {
|
2006-04-07 22:17:32 +07:00
|
|
|
gfs2_log_flush(gl->gl_sbd, gl);
|
2006-09-22 04:05:23 +07:00
|
|
|
gfs2_meta_sync(gl);
|
2006-11-20 22:37:45 +07:00
|
|
|
gfs2_ail_empty_gl(gl);
|
2006-01-16 23:50:04 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* meta_go_inval - invalidate the metadata for this glock
|
|
|
|
* @gl: the glock
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void meta_go_inval(struct gfs2_glock *gl, int flags)
|
|
|
|
{
|
|
|
|
if (!(flags & DIO_METADATA))
|
|
|
|
return;
|
|
|
|
|
|
|
|
gfs2_meta_inval(gl);
|
2008-01-31 23:31:39 +07:00
|
|
|
if (gl->gl_object == GFS2_I(gl->gl_sbd->sd_rindex))
|
|
|
|
gl->gl_sbd->sd_rindex_uptodate = 0;
|
|
|
|
else if (gl->gl_ops == &gfs2_rgrp_glops && gl->gl_object) {
|
|
|
|
struct gfs2_rgrpd *rgd = (struct gfs2_rgrpd *)gl->gl_object;
|
|
|
|
|
|
|
|
rgd->rd_flags &= ~GFS2_RDF_UPTODATE;
|
|
|
|
}
|
2006-01-16 23:50:04 +07:00
|
|
|
}
|
|
|
|
|
2007-01-23 00:15:34 +07:00
|
|
|
/**
|
|
|
|
* inode_go_sync - Sync the dirty data and/or metadata for an inode glock
|
|
|
|
* @gl: the glock protecting the inode
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void inode_go_sync(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_inode *ip = gl->gl_object;
|
2007-11-02 15:39:34 +07:00
|
|
|
struct address_space *metamapping = gl->gl_aspace->i_mapping;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (gl->gl_state != LM_ST_UNLOCKED)
|
|
|
|
gfs2_pte_inval(gl);
|
|
|
|
if (gl->gl_state != LM_ST_EXCLUSIVE)
|
|
|
|
return;
|
2007-01-23 00:15:34 +07:00
|
|
|
|
|
|
|
if (ip && !S_ISREG(ip->i_inode.i_mode))
|
|
|
|
ip = NULL;
|
|
|
|
|
|
|
|
if (test_bit(GLF_DIRTY, &gl->gl_flags)) {
|
[GFS2] flush the glock completely in inode_go_sync
Fix for bz #231910
When filemap_fdatawrite() is called on the inode mapping in data=ordered mode,
it will add the glock to the log. In inode_go_sync(), if you do the
gfs2_log_flush() before this, after the filemap_fdatawrite() call, the glock
and its associated data buffers will be on the log again. This means you can
demote a lock from exclusive, without having it flushed from the log. The
attached patch simply moves the gfs2_log_flush up to after the
filemap_fdatawrite() call.
Originally, I tried moving the gfs2_log_flush to after gfs2_meta_sync(), but
that caused me to trip the following assert.
GFS2: fsid=cypher-36:test.0: fatal: assertion "!buffer_busy(bh)" failed
GFS2: fsid=cypher-36:test.0: function = gfs2_ail_empty_gl, file = fs/gfs2/glops.c, line = 61
It appears that gfs2_log_flush() puts some of the glocks buffers in the busy
state and the filemap_fdatawrite() call is necessary to flush them. This makes
me worry slightly that a related problem could happen because of moving the
gfs2_log_flush() after the initial filemap_fdatawrite(), but I assume that
gfs2_ail_empty_gl() would catch that case as well.
Signed-off-by: Benjamin E. Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-05-02 21:44:03 +07:00
|
|
|
gfs2_log_flush(gl->gl_sbd, gl);
|
2007-11-02 15:39:34 +07:00
|
|
|
filemap_fdatawrite(metamapping);
|
2007-01-23 00:15:34 +07:00
|
|
|
if (ip) {
|
|
|
|
struct address_space *mapping = ip->i_inode.i_mapping;
|
2007-11-02 15:39:34 +07:00
|
|
|
filemap_fdatawrite(mapping);
|
|
|
|
error = filemap_fdatawait(mapping);
|
2007-05-08 14:23:25 +07:00
|
|
|
mapping_set_error(mapping, error);
|
2007-01-23 00:15:34 +07:00
|
|
|
}
|
2007-11-02 15:39:34 +07:00
|
|
|
error = filemap_fdatawait(metamapping);
|
|
|
|
mapping_set_error(metamapping, error);
|
2007-01-23 00:15:34 +07:00
|
|
|
clear_bit(GLF_DIRTY, &gl->gl_flags);
|
|
|
|
gfs2_ail_empty_gl(gl);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-01-16 23:50:04 +07:00
|
|
|
/**
|
|
|
|
* inode_go_xmote_bh - After promoting/demoting a glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void inode_go_xmote_bh(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_holder *gh = gl->gl_req_gh;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (gl->gl_state != LM_ST_UNLOCKED &&
|
|
|
|
(!gh || !(gh->gh_flags & GL_SKIP))) {
|
2006-09-22 04:05:23 +07:00
|
|
|
error = gfs2_meta_read(gl, gl->gl_name.ln_number, 0, &bh);
|
2006-01-16 23:50:04 +07:00
|
|
|
if (!error)
|
|
|
|
brelse(bh);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* inode_go_inval - prepare a inode glock to be released
|
|
|
|
* @gl: the glock
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void inode_go_inval(struct gfs2_glock *gl, int flags)
|
|
|
|
{
|
2006-11-23 22:51:34 +07:00
|
|
|
struct gfs2_inode *ip = gl->gl_object;
|
2006-01-16 23:50:04 +07:00
|
|
|
int meta = (flags & DIO_METADATA);
|
|
|
|
|
|
|
|
if (meta) {
|
|
|
|
gfs2_meta_inval(gl);
|
2006-11-23 22:51:34 +07:00
|
|
|
if (ip)
|
|
|
|
set_bit(GIF_INVALID, &ip->i_flags);
|
|
|
|
}
|
|
|
|
|
2007-10-15 21:40:33 +07:00
|
|
|
if (ip && S_ISREG(ip->i_inode.i_mode))
|
2006-11-23 22:51:34 +07:00
|
|
|
truncate_inode_pages(ip->i_inode.i_mapping, 0);
|
2006-01-16 23:50:04 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* inode_go_demote_ok - Check to see if it's ok to unlock an inode glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
* Returns: 1 if it's ok
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int inode_go_demote_ok(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp = gl->gl_sbd;
|
|
|
|
int demote = 0;
|
|
|
|
|
2006-02-28 05:23:27 +07:00
|
|
|
if (!gl->gl_object && !gl->gl_aspace->i_mapping->nrpages)
|
2006-01-16 23:50:04 +07:00
|
|
|
demote = 1;
|
|
|
|
else if (!sdp->sd_args.ar_localcaching &&
|
|
|
|
time_after_eq(jiffies, gl->gl_stamp +
|
|
|
|
gfs2_tune_get(sdp, gt_demote_secs) * HZ))
|
|
|
|
demote = 1;
|
|
|
|
|
|
|
|
return demote;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* inode_go_lock - operation done after an inode lock is locked by a process
|
|
|
|
* @gl: the glock
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
* Returns: errno
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int inode_go_lock(struct gfs2_holder *gh)
|
|
|
|
{
|
|
|
|
struct gfs2_glock *gl = gh->gh_gl;
|
2006-02-28 05:23:27 +07:00
|
|
|
struct gfs2_inode *ip = gl->gl_object;
|
2006-01-16 23:50:04 +07:00
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (!ip)
|
|
|
|
return 0;
|
|
|
|
|
2006-11-02 04:05:38 +07:00
|
|
|
if (test_bit(GIF_INVALID, &ip->i_flags)) {
|
2006-01-16 23:50:04 +07:00
|
|
|
error = gfs2_inode_refresh(ip);
|
|
|
|
if (error)
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((ip->i_di.di_flags & GFS2_DIF_TRUNC_IN_PROG) &&
|
|
|
|
(gl->gl_state == LM_ST_EXCLUSIVE) &&
|
2007-01-23 00:10:39 +07:00
|
|
|
(gh->gh_state == LM_ST_EXCLUSIVE))
|
2006-01-16 23:50:04 +07:00
|
|
|
error = gfs2_truncatei_resume(ip);
|
|
|
|
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rgrp_go_demote_ok - Check to see if it's ok to unlock a RG's glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
* Returns: 1 if it's ok
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int rgrp_go_demote_ok(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
return !gl->gl_aspace->i_mapping->nrpages;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rgrp_go_lock - operation done after an rgrp lock is locked by
|
|
|
|
* a first holder on this node.
|
|
|
|
* @gl: the glock
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
* Returns: errno
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int rgrp_go_lock(struct gfs2_holder *gh)
|
|
|
|
{
|
2006-02-28 05:23:27 +07:00
|
|
|
return gfs2_rgrp_bh_get(gh->gh_gl->gl_object);
|
2006-01-16 23:50:04 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* rgrp_go_unlock - operation done before an rgrp lock is unlocked by
|
|
|
|
* a last holder on this node.
|
|
|
|
* @gl: the glock
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void rgrp_go_unlock(struct gfs2_holder *gh)
|
|
|
|
{
|
2006-02-28 05:23:27 +07:00
|
|
|
gfs2_rgrp_bh_put(gh->gh_gl->gl_object);
|
2006-01-16 23:50:04 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2007-11-02 15:39:34 +07:00
|
|
|
* trans_go_sync - promote/demote the transaction glock
|
2006-01-16 23:50:04 +07:00
|
|
|
* @gl: the glock
|
|
|
|
* @state: the requested state
|
|
|
|
* @flags:
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2007-11-02 15:39:34 +07:00
|
|
|
static void trans_go_sync(struct gfs2_glock *gl)
|
2006-01-16 23:50:04 +07:00
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp = gl->gl_sbd;
|
|
|
|
|
|
|
|
if (gl->gl_state != LM_ST_UNLOCKED &&
|
|
|
|
test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) {
|
|
|
|
gfs2_meta_syncfs(sdp);
|
|
|
|
gfs2_log_shutdown(sdp);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* trans_go_xmote_bh - After promoting/demoting the transaction glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void trans_go_xmote_bh(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
struct gfs2_sbd *sdp = gl->gl_sbd;
|
2006-06-15 02:32:57 +07:00
|
|
|
struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode);
|
2006-02-28 05:23:27 +07:00
|
|
|
struct gfs2_glock *j_gl = ip->i_gl;
|
2006-10-14 08:47:13 +07:00
|
|
|
struct gfs2_log_header_host head;
|
2006-01-16 23:50:04 +07:00
|
|
|
int error;
|
|
|
|
|
|
|
|
if (gl->gl_state != LM_ST_UNLOCKED &&
|
|
|
|
test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) {
|
2006-11-20 22:37:45 +07:00
|
|
|
j_gl->gl_ops->go_inval(j_gl, DIO_METADATA);
|
2006-01-16 23:50:04 +07:00
|
|
|
|
|
|
|
error = gfs2_find_jhead(sdp->sd_jdesc, &head);
|
|
|
|
if (error)
|
|
|
|
gfs2_consist(sdp);
|
|
|
|
if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT))
|
|
|
|
gfs2_consist(sdp);
|
|
|
|
|
|
|
|
/* Initialize some head of the log stuff */
|
|
|
|
if (!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) {
|
|
|
|
sdp->sd_log_sequence = head.lh_sequence + 1;
|
|
|
|
gfs2_log_pointers_init(sdp, head.lh_blkno);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* quota_go_demote_ok - Check to see if it's ok to unlock a quota glock
|
|
|
|
* @gl: the glock
|
|
|
|
*
|
|
|
|
* Returns: 1 if it's ok
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int quota_go_demote_ok(struct gfs2_glock *gl)
|
|
|
|
{
|
|
|
|
return !atomic_read(&gl->gl_lvb_count);
|
|
|
|
}
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_meta_glops = {
|
2007-01-23 00:15:34 +07:00
|
|
|
.go_xmote_th = meta_go_sync,
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_META,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_inode_glops = {
|
2007-11-02 15:39:34 +07:00
|
|
|
.go_xmote_th = inode_go_sync,
|
2006-01-16 23:50:04 +07:00
|
|
|
.go_xmote_bh = inode_go_xmote_bh,
|
|
|
|
.go_inval = inode_go_inval,
|
|
|
|
.go_demote_ok = inode_go_demote_ok,
|
|
|
|
.go_lock = inode_go_lock,
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_INODE,
|
[GFS2] delay glock demote for a minimum hold time
When a lot of IO, with some distributed mmap IO, is run on a GFS2 filesystem in
a cluster, it will deadlock. The reason is that do_no_page() will repeatedly
call gfs2_sharewrite_nopage(), because each node keeps giving up the glock
too early, and is forced to call unmap_mapping_range(). This bumps the
mapping->truncate_count sequence count, forcing do_no_page() to retry. This
patch institutes a minimum glock hold time a tenth a second. This insures
that even in heavy contention cases, the node has enough time to get some
useful work done before it gives up the glock.
A second issue is that when gfs2_glock_dq() is called from within a page fault
to demote a lock, and the associated page needs to be written out, it will
try to acqire a lock on it, but it has already been locked at a higher level.
This patch puts makes gfs2_glock_dq() use the work queue as well, to avoid this
issue. This is the same patch as Steve Whitehouse originally proposed to fix
this issue, execpt that gfs2_glock_dq() now grabs a reference to the glock
before it queues up the work on it.
Signed-off-by: Benjamin E. Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-08-24 01:19:05 +07:00
|
|
|
.go_min_hold_time = HZ / 10,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_rgrp_glops = {
|
2007-02-28 21:03:00 +07:00
|
|
|
.go_xmote_th = meta_go_sync,
|
2006-01-16 23:50:04 +07:00
|
|
|
.go_inval = meta_go_inval,
|
|
|
|
.go_demote_ok = rgrp_go_demote_ok,
|
|
|
|
.go_lock = rgrp_go_lock,
|
|
|
|
.go_unlock = rgrp_go_unlock,
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_RGRP,
|
[GFS2] delay glock demote for a minimum hold time
When a lot of IO, with some distributed mmap IO, is run on a GFS2 filesystem in
a cluster, it will deadlock. The reason is that do_no_page() will repeatedly
call gfs2_sharewrite_nopage(), because each node keeps giving up the glock
too early, and is forced to call unmap_mapping_range(). This bumps the
mapping->truncate_count sequence count, forcing do_no_page() to retry. This
patch institutes a minimum glock hold time a tenth a second. This insures
that even in heavy contention cases, the node has enough time to get some
useful work done before it gives up the glock.
A second issue is that when gfs2_glock_dq() is called from within a page fault
to demote a lock, and the associated page needs to be written out, it will
try to acqire a lock on it, but it has already been locked at a higher level.
This patch puts makes gfs2_glock_dq() use the work queue as well, to avoid this
issue. This is the same patch as Steve Whitehouse originally proposed to fix
this issue, execpt that gfs2_glock_dq() now grabs a reference to the glock
before it queues up the work on it.
Signed-off-by: Benjamin E. Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
2007-08-24 01:19:05 +07:00
|
|
|
.go_min_hold_time = HZ / 10,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_trans_glops = {
|
2007-11-02 15:39:34 +07:00
|
|
|
.go_xmote_th = trans_go_sync,
|
2006-01-16 23:50:04 +07:00
|
|
|
.go_xmote_bh = trans_go_xmote_bh,
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_NONDISK,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_iopen_glops = {
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_IOPEN,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_flock_glops = {
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_FLOCK,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_nondisk_glops = {
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_NONDISK,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_quota_glops = {
|
2006-01-16 23:50:04 +07:00
|
|
|
.go_demote_ok = quota_go_demote_ok,
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_QUOTA,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|
2006-08-30 20:30:00 +07:00
|
|
|
const struct gfs2_glock_operations gfs2_journal_glops = {
|
2006-09-05 21:53:09 +07:00
|
|
|
.go_type = LM_TYPE_JOURNAL,
|
2006-01-16 23:50:04 +07:00
|
|
|
};
|
|
|
|
|