login
Search
Hi, Since it seems that another merge window will probably be opening shortly this is a posting of the current content of the GFS2/DLM -nmw git tree. There are no new features this time, its all fixes and cleanups. I have a few patches that I'm holding back which I'm intending to start off the new -nmw git tree with after this merge, those include: o Patch to use page_mkwrite() on page faults (and thus allow us to use the filemap_fault() function rather than having a special GFS2 version of this like we do at the moment) o Patch to clean up gfs2_internal_read() as requested by Christoph Hellwig o Patches to separate the GFS2 aops into three separate operations structures (for writeback, ordered & jdata) as per ext3 which is waiting for the merge of Nick Piggin's AOP changes so that I don't land up creating a merge conflict. Anyhow, back to what is in this set of patches... the major area of change is the journalling code where there has been a lot of clean up. Other than that, its mostly fixes for individual bugs and clean ups that other people have sent to me, Steve. -
From: Steven Whitehouse <swhiteho@redhat.com>
One of the races relates to referencing a variable while not holding
its protecting spinlock. The patch simply moves the test inside the
spin lock. The other races occurs when a demote to unlocked request
occurs during the time a demote to shared request is already running.
This of course only happens in the case that the lock was in the
exclusive mode to start with. The patch adds a check to see if another
demote request has occurred in the mean time and if it has, then it
performs a second demote.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 3f0974e..6a3eeba 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -545,12 +545,14 @@ static int rq_demote(struct gfs2_glock *gl)
return 0;
}
set_bit(GLF_LOCK, &gl->gl_flags);
- spin_unlock(&gl->gl_spin);
if (gl->gl_demote_state == LM_ST_UNLOCKED ||
- gl->gl_state != LM_ST_EXCLUSIVE)
+ gl->gl_state != LM_ST_EXCLUSIVE) {
+ spin_unlock(&gl->gl_spin);
gfs2_glock_drop_th(gl);
- else
+ } else {
+ spin_unlock(&gl->gl_spin);
gfs2_glock_xmote_th(gl, NULL);
+ }
spin_lock(&gl->gl_spin);
return 0;
@@ -760,10 +762,20 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
if (!gh) {
gl->gl_stamp = jiffies;
- if (ret & LM_OUT_CANCELED)
+ if (ret & LM_OUT_CANCELED) {
op_done = 0;
- else
+ } else {
+ spin_lock(&gl->gl_spin);
+ if (gl->gl_state != gl->gl_demote_state) {
+ gl->gl_req_bh = NULL;
+ spin_unlock(&gl->gl_spin);
+ gfs2_glock_drop_th(gl);
+ gfs2_glock_put(gl);
+ return;
+ }
gfs2_demote_wake(gl);
+ spin_unlock(&gl->gl_spin);
+ }
} else {
spin_lock(&gl->gl_spin);
list_del_init(&gh->gh_list);
@@ -817,7 +829,7 @@ out:
*
*/
-void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
+static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
{
struct gfs2_sbd *sdp = ...From: Josef Whiter <jwhiter@redhat.com>
If a glock is in the exclusive state and a request for demote to
deferred has been received, then further requests for demote to
shared are being ignored. This patch fixes that by ensuring that
we demote to unlocked in that case.
Signed-off-by: Josef Whiter <jwhiter@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 6a3eeba..6b6ae45 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -697,8 +697,9 @@ static void handle_callback(struct gfs2_glock *gl, unsigned int state, int remot
}
return;
}
- } else if (gl->gl_demote_state != LM_ST_UNLOCKED) {
- gl->gl_demote_state = state;
+ } else if (gl->gl_demote_state != LM_ST_UNLOCKED &&
+ gl->gl_demote_state != state) {
+ gl->gl_demote_state = LM_ST_UNLOCKED;
}
spin_unlock(&gl->gl_spin);
}
--
1.5.1.2
-
From: Jesper Juhl <jesper.juhl@gmail.com> This patch cleans up duplicate includes in fs/gfs2/ Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 6b6ae45..d403fd7 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -25,8 +25,6 @@ #include <asm/uaccess.h> #include <linux/seq_file.h> #include <linux/debugfs.h> -#include <linux/module.h> -#include <linux/kallsyms.h> #include "gfs2.h" #include "incore.h" diff --git a/fs/gfs2/locking/dlm/lock_dlm.h b/fs/gfs2/locking/dlm/lock_dlm.h index 24d70f7..9e8265d 100644 --- a/fs/gfs2/locking/dlm/lock_dlm.h +++ b/fs/gfs2/locking/dlm/lock_dlm.h @@ -13,7 +13,6 @@ #include <linux/module.h> #include <linux/slab.h> #include <linux/spinlock.h> -#include <linux/module.h> #include <linux/types.h> #include <linux/string.h> #include <linux/list.h> diff --git a/fs/gfs2/locking/nolock/main.c b/fs/gfs2/locking/nolock/main.c index 0d149c8..d3b8ce6 100644 --- a/fs/gfs2/locking/nolock/main.c +++ b/fs/gfs2/locking/nolock/main.c @@ -9,7 +9,6 @@ #include <linux/module.h> #include <linux/slab.h> -#include <linux/module.h> #include <linux/init.h> #include <linux/types.h> #include <linux/fs.h> -- 1.5.1.2 -
From: Steve French <smfrench@gmail.com>
When looking at an unrelated problem, I noticed that nfsd does not
set nameidata pointer on create (ie nd is NULL). This should
cause an oops in some cases in which when NFSd is mounted over GFS2.
Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index 911c115..5b8b994 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -69,7 +69,7 @@ static int gfs2_create(struct inode *dir, struct dentry *dentry,
mark_inode_dirty(inode);
break;
} else if (PTR_ERR(inode) != -EEXIST ||
- (nd->intent.open.flags & O_EXCL)) {
+ (nd && (nd->intent.open.flags & O_EXCL))) {
gfs2_holder_uninit(ghs);
return PTR_ERR(inode);
}
--
1.5.1.2
-
From: Steven Whitehouse <swhiteho@redhat.com>
This fixes an oops which was occurring during glock dumping due to the
seq file code not taking a reference to the glock. Also this fixes a
memory leak which occurred in certain cases, in turn preventing the
filesystem from unmounting.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index d403fd7..e4bc8ae 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -46,7 +46,6 @@ struct glock_iter {
int hash; /* hash bucket index */
struct gfs2_sbd *sdp; /* incore superblock */
struct gfs2_glock *gl; /* current glock struct */
- struct hlist_head *hb_list; /* current hash bucket ptr */
struct seq_file *seq; /* sequence file for debugfs */
char string[512]; /* scratch space */
};
@@ -1990,47 +1989,38 @@ int __init gfs2_glock_init(void)
static int gfs2_glock_iter_next(struct glock_iter *gi)
{
+ struct gfs2_glock *gl;
+
read_lock(gl_lock_addr(gi->hash));
- while (1) {
- if (!gi->hb_list) { /* If we don't have a hash bucket yet */
- gi->hb_list = &gl_hash_table[gi->hash].hb_list;
- if (hlist_empty(gi->hb_list)) {
- read_unlock(gl_lock_addr(gi->hash));
- gi->hash++;
- read_lock(gl_lock_addr(gi->hash));
- gi->hb_list = NULL;
- if (gi->hash >= GFS2_GL_HASH_SIZE) {
- read_unlock(gl_lock_addr(gi->hash));
- return 1;
- }
- else
- continue;
- }
- if (!hlist_empty(gi->hb_list)) {
- gi->gl = list_entry(gi->hb_list->first,
- struct gfs2_glock,
- gl_list);
- }
- } else {
- if (gi->gl->gl_list.next == NULL) {
- read_unlock(gl_lock_addr(gi->hash));
- gi->hash++;
- read_lock(gl_lock_addr(gi->hash));
- gi->hb_list = NULL;
- continue;
- }
- gi->gl = list_entry(gi->gl->gl_list.next,
- struct gfs2_glock, gl_list);
- }
+ gl = gi->gl;
+ if (gl) {
+ gi->gl = ...From: Bob Peterson <rpeterso@redhat.com>
This is the first of five patches for bug #248176:
There were still some critical variables being manipulated outside
the log_lock spinlock. That usually resulted in a hang.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 3b395c4..a0371f8 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -117,7 +117,7 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp)
struct buffer_head *bh;
struct gfs2_log_descriptor *ld;
struct gfs2_bufdata *bd1 = NULL, *bd2;
- unsigned int total = sdp->sd_log_num_buf;
+ unsigned int total;
unsigned int offset = BUF_OFFSET;
unsigned int limit;
unsigned int num;
@@ -127,12 +127,16 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp)
limit = buf_limit(sdp);
/* for 4k blocks, limit = 503 */
+ gfs2_log_lock(sdp);
+ total = sdp->sd_log_num_buf;
bd1 = bd2 = list_prepare_entry(bd1, &sdp->sd_log_le_buf, bd_le.le_list);
while(total) {
num = total;
if (total > limit)
num = limit;
+ gfs2_log_unlock(sdp);
bh = gfs2_log_get_buf(sdp);
+ gfs2_log_lock(sdp);
ld = (struct gfs2_log_descriptor *)bh->b_data;
ptr = (__be64 *)(bh->b_data + offset);
ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
@@ -152,21 +156,27 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp)
break;
}
+ gfs2_log_unlock(sdp);
set_buffer_dirty(bh);
ll_rw_block(WRITE, 1, &bh);
+ gfs2_log_lock(sdp);
n = 0;
list_for_each_entry_continue(bd2, &sdp->sd_log_le_buf,
bd_le.le_list) {
+ gfs2_log_unlock(sdp);
bh = gfs2_log_fake_buf(sdp, bd2->bd_bh);
set_buffer_dirty(bh);
ll_rw_block(WRITE, 1, &bh);
+ gfs2_log_lock(sdp);
if (++n >= num)
break;
}
+ BUG_ON(total < num);
total -= num;
}
+ gfs2_log_unlock(sdp);
}
static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
@@ -524,7 ...From: Bob Peterson <rpeterso@redhat.com> This is patch 2 of 5 for bug #248176. The list_move code previously concocted in log.c for bug #238162 (see https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=238162#c23) never runs as bh can now never be NULL at this point. Signed-off-by: Bob Peterson <rpeterso@redhat.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index f49a12e..f7c0608 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -83,11 +83,6 @@ static void gfs2_ail1_start_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai) gfs2_assert(sdp, bd->bd_ail == ai); - if (!bh){ - list_move(&bd->bd_ail_st_list, &ai->ai_ail2_list); - continue; - } - if (!buffer_busy(bh)) { if (!buffer_uptodate(bh)) { gfs2_log_unlock(sdp); @@ -130,11 +125,6 @@ static int gfs2_ail1_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai, int fl bd_ail_st_list) { bh = bd->bd_bh; - if (!bh){ - list_move(&bd->bd_ail_st_list, &ai->ai_ail2_list); - continue; - } - gfs2_assert(sdp, bd->bd_ail == ai); if (buffer_busy(bh)) { @@ -155,13 +145,14 @@ static int gfs2_ail1_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai, int fl static void gfs2_ail1_start(struct gfs2_sbd *sdp, int flags) { - struct list_head *head = &sdp->sd_ail1_list; + struct list_head *head; u64 sync_gen; struct list_head *first; struct gfs2_ail *first_ai, *ai, *tmp; int done = 0; gfs2_log_lock(sdp); + head = &sdp->sd_ail1_list; if (list_empty(head)) { gfs2_log_unlock(sdp); return; -- 1.5.1.2 -
From: Bob Peterson <rpeterso@redhat.com>
This is patch three of five for bug #248176.
The try_rgrp_unlink code in rgrp.c had an infinite loop. This was
caused because the bitmap function rgblk_search can return a block
less than the "goal" block, in which case it was looping. The fix is
to make it always march forward as needed.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index ce48c45..b93ac45 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -31,6 +31,7 @@
#include "inode.h"
#define BFITNOENT ((u32)~0)
+#define NO_BLOCK ((u64)~0)
/*
* These routines are used by the resource group routines (rgrp.c)
@@ -116,8 +117,7 @@ static unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
* @buffer: the buffer that holds the bitmaps
* @buflen: the length (in bytes) of the buffer
* @goal: start search at this block's bit-pair (within @buffer)
- * @old_state: GFS2_BLKST_XXX the state of the block we're looking for;
- * bit 0 = alloc(1)/free(0), bit 1 = meta(1)/data(0)
+ * @old_state: GFS2_BLKST_XXX the state of the block we're looking for.
*
* Scope of @goal and returned block number is only within this bitmap buffer,
* not entire rgrp or filesystem. @buffer will be offset from the actual
@@ -137,9 +137,13 @@ static u32 gfs2_bitfit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
byte = buffer + (goal / GFS2_NBBY);
bit = (goal % GFS2_NBBY) * GFS2_BIT_SIZE;
end = buffer + buflen;
- alloc = (old_state & 1) ? 0 : 0x55;
+ alloc = (old_state == GFS2_BLKST_FREE) ? 0x55 : 0;
while (byte < end) {
+ /* If we're looking for a free block we can eliminate all
+ bitmap settings with 0x55, which represents four data
+ blocks in a row. If we're looking for a data block, we can
+ eliminate 0x00 which corresponds to four free blocks. */
if ((*byte & 0x55) == alloc) {
blk += (8 - bit) >> 1;
@@ -859,19 ...From: Denis Cheng <crquan@gmail.com>
this is more clear.
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/locking/dlm/plock.c b/fs/gfs2/locking/dlm/plock.c
index fba1f1d..1f7b038 100644
--- a/fs/gfs2/locking/dlm/plock.c
+++ b/fs/gfs2/locking/dlm/plock.c
@@ -346,15 +346,16 @@ static ssize_t dev_write(struct file *file, const char __user *u, size_t count,
static unsigned int dev_poll(struct file *file, poll_table *wait)
{
+ unsigned int mask = 0;
+
poll_wait(file, &send_wq, wait);
spin_lock(&ops_lock);
- if (!list_empty(&send_list)) {
- spin_unlock(&ops_lock);
- return POLLIN | POLLRDNORM;
- }
+ if (!list_empty(&send_list))
+ mask = POLLIN | POLLRDNORM;
spin_unlock(&ops_lock);
- return 0;
+
+ return mask;
}
static const struct file_operations dev_fops = {
--
1.5.1.2
-
From: Bob Peterson <rpeterso@redhat.com>
This is patch 5 of 5 for bug #248176
Metadata corruption was occurring because page references weren't
being removed in all cases. I previously added a function called
detach_bufdata, but I discovered there already WAS a function out
there to do the job. It's called gfs2_meta_cache_flush. So I added
a call to that to remove the page references.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index f7c0608..00ab6c0 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -219,6 +219,7 @@ static void gfs2_ail2_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
{
struct list_head *head = &ai->ai_ail2_list;
struct gfs2_bufdata *bd;
+ struct gfs2_inode *bh_ip;
while (!list_empty(head)) {
bd = list_entry(head->prev, struct gfs2_bufdata,
@@ -228,6 +229,8 @@ static void gfs2_ail2_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
list_del(&bd->bd_ail_st_list);
list_del(&bd->bd_ail_gl_list);
atomic_dec(&bd->bd_gl->gl_ail_count);
+ bh_ip = GFS2_I(bd->bd_bh->b_page->mapping->host);
+ gfs2_meta_cache_flush(bh_ip);
brelse(bd->bd_bh);
}
}
--
1.5.1.2
-
From: Denis Cheng <crquan@gmail.com>
these struct *_operations are all method tables, thus should be const.
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/eaops.c b/fs/gfs2/eaops.c
index 1ab3e9d..aa8dbf3 100644
--- a/fs/gfs2/eaops.c
+++ b/fs/gfs2/eaops.c
@@ -200,28 +200,28 @@ static int security_eo_remove(struct gfs2_inode *ip, struct gfs2_ea_request *er)
return gfs2_ea_remove_i(ip, er);
}
-static struct gfs2_eattr_operations gfs2_user_eaops = {
+static const struct gfs2_eattr_operations gfs2_user_eaops = {
.eo_get = user_eo_get,
.eo_set = user_eo_set,
.eo_remove = user_eo_remove,
.eo_name = "user",
};
-struct gfs2_eattr_operations gfs2_system_eaops = {
+const struct gfs2_eattr_operations gfs2_system_eaops = {
.eo_get = system_eo_get,
.eo_set = system_eo_set,
.eo_remove = system_eo_remove,
.eo_name = "system",
};
-static struct gfs2_eattr_operations gfs2_security_eaops = {
+static const struct gfs2_eattr_operations gfs2_security_eaops = {
.eo_get = security_eo_get,
.eo_set = security_eo_set,
.eo_remove = security_eo_remove,
.eo_name = "security",
};
-struct gfs2_eattr_operations *gfs2_ea_ops[] = {
+const struct gfs2_eattr_operations *gfs2_ea_ops[] = {
NULL,
&gfs2_user_eaops,
&gfs2_system_eaops,
diff --git a/fs/gfs2/eaops.h b/fs/gfs2/eaops.h
index 508b4f7..da2f7fb 100644
--- a/fs/gfs2/eaops.h
+++ b/fs/gfs2/eaops.h
@@ -22,9 +22,9 @@ struct gfs2_eattr_operations {
unsigned int gfs2_ea_name2type(const char *name, const char **truncated_name);
-extern struct gfs2_eattr_operations gfs2_system_eaops;
+extern const struct gfs2_eattr_operations gfs2_system_eaops;
-extern struct gfs2_eattr_operations *gfs2_ea_ops[];
+extern const struct gfs2_eattr_operations *gfs2_ea_ops[];
#endif /* __EAOPS_DOT_H__ */
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index e4bc8ae..0054b7d 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ ...From: Denis Cheng <crquan@gmail.com>
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index cf5aa50..9a5e840 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -28,6 +28,7 @@
#include "lm.h"
#include "mount.h"
#include "ops_fstype.h"
+#include "ops_dentry.h"
#include "ops_super.h"
#include "recovery.h"
#include "rgrp.h"
@@ -38,8 +39,6 @@
#define DO 0
#define UNDO 1
-extern struct dentry_operations gfs2_dops;
-
static struct gfs2_sbd *init_sbd(struct super_block *sb)
{
struct gfs2_sbd *sdp;
--
1.5.1.2
-
From: Steven Whitehouse <swhiteho@redhat.com>
We only need a single gfs2_scand process rather than the one
per filesystem which we had previously. As a result the parameter
determining the frequency of gfs2_scand runs becomes a module
parameter rather than a mount parameter as it was before.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/daemon.c b/fs/gfs2/daemon.c
index 3548d9f..3731ab0 100644
--- a/fs/gfs2/daemon.c
+++ b/fs/gfs2/daemon.c
@@ -35,30 +35,6 @@
The kthread functions used to start these daemons block and flush signals. */
/**
- * gfs2_scand - Look for cached glocks and inodes to toss from memory
- * @sdp: Pointer to GFS2 superblock
- *
- * One of these daemons runs, finding candidates to add to sd_reclaim_list.
- * See gfs2_glockd()
- */
-
-int gfs2_scand(void *data)
-{
- struct gfs2_sbd *sdp = data;
- unsigned long t;
-
- while (!kthread_should_stop()) {
- gfs2_scand_internal(sdp);
- t = gfs2_tune_get(sdp, gt_scand_secs) * HZ;
- if (freezing(current))
- refrigerator();
- schedule_timeout_interruptible(t);
- }
-
- return 0;
-}
-
-/**
* gfs2_glockd - Reclaim unused glock structures
* @sdp: Pointer to GFS2 superblock
*
diff --git a/fs/gfs2/daemon.h b/fs/gfs2/daemon.h
index 8010071..0de9b35 100644
--- a/fs/gfs2/daemon.h
+++ b/fs/gfs2/daemon.h
@@ -10,7 +10,6 @@
#ifndef __DAEMON_DOT_H__
#define __DAEMON_DOT_H__
-int gfs2_scand(void *data);
int gfs2_glockd(void *data);
int gfs2_recoverd(void *data);
int gfs2_logd(void *data);
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 0054b7d..559937c 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -25,6 +25,8 @@
#include <asm/uaccess.h>
#include <linux/seq_file.h>
#include <linux/debugfs.h>
+#include <linux/kthread.h>
+#include <linux/freezer.h>
#include "gfs2.h"
#include "incore.h"
@@ -58,6 +60,8 @@ static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh);
static void gfs2_glock_drop_th(struct ...From: Bob Peterson <rpeterso@redhat.com>
This is for bugzilla bug #248176: GFS2: invalid metadata block
Patches 1 thru 3 were accepted upstream, but there were problems
with 4 and 5. Those issues have been resolved and now the recovery
tests are passing without errors. This code has gone through
41 * 3 successful gfs2 recovery tests before it hit an
unrelated (openais) problem.
This is a complete rewrite of patch 4 for bug #248176.
Part of the problem was that inodes were being recycled
before their buffers were flushed to the journal logs.
Another problem was that the clone bitmaps were being
searched for deleted inodes to recycle, but only the
"real" bitmaps should be searched for that purpose.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index b93ac45..2d7f7ea 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -865,12 +865,15 @@ static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked)
struct inode *inode;
u32 goal = 0, block;
u64 no_addr;
+ struct gfs2_sbd *sdp = rgd->rd_sbd;
for(;;) {
if (goal >= rgd->rd_data)
break;
+ down_write(&sdp->sd_log_flush_lock);
block = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED,
GFS2_BLKST_UNLINKED);
+ up_write(&sdp->sd_log_flush_lock);
if (block == BFITNOENT)
break;
/* rgblk_search can return a block < goal, so we need to
@@ -1295,7 +1298,9 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
allocatable block anywhere else, we want to be able wrap around and
search in the first part of our first-searched bit block. */
for (x = 0; x <= length; x++) {
- if (bi->bi_clone)
+ /* The GFS2_BLKST_UNLINKED state doesn't apply to the clone
+ bitmaps, so we must search the originals for that. */
+ if (old_state != GFS2_BLKST_UNLINKED && bi->bi_clone)
blk = gfs2_bitfit(rgd, bi->bi_clone + bi->bi_offset,
bi->bi_len, goal, ...From: Bob Peterson <rpeterso@redhat.com>
This is for bugzilla bug #248176: GFS2: invalid metadata block
Patches 1 thru 3 were accepted upstream, but there were problems
with 4 and 5. Those issues have been resolved and now the recovery
tests are passing without errors. This code has gone through
41 * 3 successful gfs2 recovery tests before it hit an
unrelated (openais) problem. I'm continuing to test it.
This is a complete rewrite of patch 5 for bug #248176, written by
Steve Whitehouse. This is referred to in the bugzilla record as
"new 6" and "a different solution".
The problem was that the journal inodes, although protected by
a glock, were not synched with the other nodes because they don't
use the inode glock synch operations (i.e. no "glops" were defined).
Therefore, journal recovery on a journal-recovering node were causing
the blocks to get out of sync with the node that was actually trying
to use that journal as it comes back up from a reboot.
There are two possible solutions: (1) To make the journals use the
normal inode glock sync operations, or (2) To make the journal
operations take effect immediately (i.e. no caching). Although
option 1 works, it turns out to be a lot more code. Steve opted
for option 2, which is much simpler and therefore less prone to
regression errors.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
--
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 58c730b..f0bcaa2 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -358,7 +358,7 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
ip = GFS2_I(sdp->sd_jdesc->jd_inode);
error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
- LM_FLAG_NOEXP | GL_EXACT,
+ LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
&sdp->sd_jinode_gh);
if (error) {
fs_err(sdp, "can't acquire journal inode glock: %d\n",
diff --git a/fs/gfs2/recovery.c b/fs/gfs2/recovery.c
index ...From: Denis Cheng <crquan@gmail.com>
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index f0bcaa2..32b2859 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -808,7 +808,6 @@ static struct super_block* get_gfs2_sb(const char *dev_name)
struct nameidata nd;
struct file_system_type *fstype;
struct super_block *sb = NULL, *s;
- struct list_head *l;
int error;
error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd);
@@ -820,8 +819,7 @@ static struct super_block* get_gfs2_sb(const char *dev_name)
error = vfs_getattr(nd.mnt, nd.dentry, &stat);
fstype = get_fs_type("gfs2");
- list_for_each(l, &fstype->fs_supers) {
- s = list_entry(l, struct super_block, s_instances);
+ list_for_each_entry(s, &fstype->fs_supers, s_instances) {
if ((S_ISBLK(stat.mode) && s->s_dev == stat.rdev) ||
(S_ISDIR(stat.mode) && s == nd.dentry->d_inode->i_sb)) {
sb = s;
--
1.5.1.2
-
From: Denis Cheng <crquan@gmail.com>
sb->s_fs_info is a void pointer, thus the type cast is not needed.
Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 32b2859..25cfab9 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -849,7 +849,7 @@ static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
error = -ENOENT;
goto error;
}
- sdp = (struct gfs2_sbd*) sb->s_fs_info;
+ sdp = sb->s_fs_info;
if (sdp->sd_vfs_meta) {
printk(KERN_WARNING "GFS2: gfs2meta mount already exists\n");
error = -EBUSY;
--
1.5.1.2
-
From: Denis Cheng <crquan@gmail.com> the original code could work, but I think this code could work better. Signed-off-by: Denis Cheng <crquan@gmail.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 25cfab9..6c820cb 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -144,7 +144,8 @@ static int init_names(struct gfs2_sbd *sdp, int silent) snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto); snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table); - while ((table = strchr(sdp->sd_table_name, '/'))) + table = sdp->sd_table_name; + while ((table = strchr(table, '/'))) *table = '_'; out: -- 1.5.1.2 -
From: Abhijith Das <adas@redhat.com>
This patch forcibly unstuffs (if stuffed) the hidden quota inode at the
first availble opportunity. In any practical scenario the quota inode
won't be stuffed, so this is ok to do. Unstuffing the quota inode allows
us to ignore the case of a stuffed quota inode in gfs2_adjust_quota().
Signed-off-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index 6e546ee..5dfa465 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -614,6 +614,16 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
s64 value;
int err = -EIO;
+ if (gfs2_is_stuffed(ip)) {
+ struct gfs2_alloc *al = NULL;
+ al = gfs2_alloc_get(ip);
+ /* just request 1 blk */
+ al->al_requested = 1;
+ gfs2_inplace_reserve(ip);
+ gfs2_unstuff_dinode(ip, NULL);
+ gfs2_inplace_release(ip);
+ gfs2_alloc_put(ip);
+ }
page = grab_cache_page(mapping, index);
if (!page)
return -ENOMEM;
--
1.5.1.2
-
From: Denis Cheng <crquan@gmail.com> Signed-off-by: Denis Cheng <crquan@gmail.com> Signed-off-by: Steven Whitehouse <swhiteho@redhat.com> diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 6c820cb..c1c6672 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -292,8 +292,9 @@ static int init_sb(struct gfs2_sbd *sdp, int silent, int undo) fs_err(sdp, "can't get root dentry\n"); error = -ENOMEM; iput(inode); - } - sb->s_root->d_op = &gfs2_dops; + } else + sb->s_root->d_op = &gfs2_dops; + out: gfs2_glock_dq_uninit(&sb_gh); return error; -- 1.5.1.2 -
From: Abhijith Das <adas@redhat.com>
This is the filesystem part of the patches to fix this bz. There are
additional userland patches (gfs2_quota, libgfs2) for the complete
solution. This patch adds a new field qu_ll_next to the gfs2_quota
structure. This field allows us to create linked lists of quotas in the
ondisk quota inode. Instead of scanning through the entire sparse quota
file for valid quotas, we can now simply walk through the user and group
quota linked lists to perform the do_list operation.
Signed-off-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index 5dfa465..addb51e 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -70,6 +70,7 @@ struct gfs2_quota_host {
u64 qu_limit;
u64 qu_warn;
s64 qu_value;
+ u32 qu_ll_next;
};
struct gfs2_quota_change_host {
@@ -580,6 +581,7 @@ static void gfs2_quota_in(struct gfs2_quota_host *qu, const void *buf)
qu->qu_limit = be64_to_cpu(str->qu_limit);
qu->qu_warn = be64_to_cpu(str->qu_warn);
qu->qu_value = be64_to_cpu(str->qu_value);
+ qu->qu_ll_next = be32_to_cpu(str->qu_ll_next);
}
static void gfs2_quota_out(const struct gfs2_quota_host *qu, void *buf)
@@ -589,6 +591,7 @@ static void gfs2_quota_out(const struct gfs2_quota_host *qu, void *buf)
str->qu_limit = cpu_to_be64(qu->qu_limit);
str->qu_warn = cpu_to_be64(qu->qu_warn);
str->qu_value = cpu_to_be64(qu->qu_value);
+ str->qu_ll_next = cpu_to_be32(qu->qu_ll_next);
memset(&str->qu_reserved, 0, sizeof(str->qu_reserved));
}
diff --git a/include/linux/gfs2_ondisk.h b/include/linux/gfs2_ondisk.h
index a44a6a0..c3c19f9 100644
--- a/include/linux/gfs2_ondisk.h
+++ b/include/linux/gfs2_ondisk.h
@@ -170,6 +170,33 @@ struct gfs2_rgrp {
};
/*
+ * quota linked list: user quotas and group quotas form two separate
+ * singly linked lists. ll_next stores uids or gids of next quotas in the
+ * linked list.
+
+Given the uid/gid, how to calculate ...From: Steven Whitehouse <swhiteho@redhat.com>
This patch fixes some bugs relating to journaled data files by cleaning
up the gfs2_invalidatepage() and gfs2_releasepage() functions. We now
never block during gfs2_releasepage(), instead we always either release
or refuse to release depending on the status of the buffers.
This fixes Red Hat bugzillas #248969 and #252392.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Bob Peterson <rpeterso@redhat.com>
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index 777ca46..88342e0 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -156,9 +156,11 @@ static void inode_go_sync(struct gfs2_glock *gl)
ip = NULL;
if (test_bit(GLF_DIRTY, &gl->gl_flags)) {
- if (ip)
+ if (ip && !gfs2_is_jdata(ip))
filemap_fdatawrite(ip->i_inode.i_mapping);
gfs2_log_flush(gl->gl_sbd, gl);
+ if (ip && gfs2_is_jdata(ip))
+ filemap_fdatawrite(ip->i_inode.i_mapping);
gfs2_meta_sync(gl);
if (ip) {
struct address_space *mapping = ip->i_inode.i_mapping;
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index 00ab6c0..d0e6b42 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -229,8 +229,10 @@ static void gfs2_ail2_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
list_del(&bd->bd_ail_st_list);
list_del(&bd->bd_ail_gl_list);
atomic_dec(&bd->bd_gl->gl_ail_count);
- bh_ip = GFS2_I(bd->bd_bh->b_page->mapping->host);
- gfs2_meta_cache_flush(bh_ip);
+ if (bd->bd_bh->b_page->mapping) {
+ bh_ip = GFS2_I(bd->bd_bh->b_page->mapping->host);
+ gfs2_meta_cache_flush(bh_ip);
+ }
brelse(bd->bd_bh);
}
}
diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index 42a5f58..8407d1d 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -616,58 +616,13 @@ static sector_t gfs2_bmap(struct address_space *mapping, sector_t lblock)
return dblock;
}
-static void discard_buffer(struct gfs2_sbd *sdp, struct buffer_head *bh)
-{
- struct gfs2_bufdata *bd;
-
- gfs2_log_lock(sdp);
- bd = ...From: Steven Whitehouse <swhiteho@redhat.com>
This was missing from the dir_split_leaf() function although in
most cases its not a problem due to other functions having
already previously called gfs2_trans_add_bh. This makes certain
that it is correct.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Wendy Cheng <wcheng@redhat.com>
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 2beb2f4..08c6dd0 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1043,6 +1043,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
error = gfs2_meta_inode_buffer(dip, &dibh);
if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
+ gfs2_trans_add_bh(dip->i_gl, dibh, 1);
dip->i_di.di_blocks++;
gfs2_set_inode_blocks(&dip->i_inode);
gfs2_dinode_out(dip, dibh->b_data);
--
1.5.1.2
-
From: Benjamin Marzinski <bmarzins@redhat.com>
match_token() was returning garbage data instead of a fail value. This data
happened to match a valid option id for an option that required an argument (in
this case, lockproto=%s) For match_token() to correctly fail if the option
doesn't match any of the tokens, the token table must end with a NULL entry.
This patch adds the NULL entry.
Signed-off-by: Benjamin E. Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/mount.c b/fs/gfs2/mount.c
index 4864659..b941f9f 100644
--- a/fs/gfs2/mount.c
+++ b/fs/gfs2/mount.c
@@ -42,6 +42,7 @@ enum {
Opt_nosuiddir,
Opt_data_writeback,
Opt_data_ordered,
+ Opt_err,
};
static match_table_t tokens = {
@@ -64,7 +65,8 @@ static match_table_t tokens = {
{Opt_suiddir, "suiddir"},
{Opt_nosuiddir, "nosuiddir"},
{Opt_data_writeback, "data=writeback"},
- {Opt_data_ordered, "data=ordered"}
+ {Opt_data_ordered, "data=ordered"},
+ {Opt_err, NULL}
};
/**
@@ -237,6 +239,7 @@ int gfs2_mount_args(struct gfs2_sbd *sdp, char *data_arg, int remount)
case Opt_data_ordered:
args->ar_data = GFS2_DATA_ORDERED;
break;
+ case Opt_err:
default:
fs_info(sdp, "unknown option: %s\n", o);
error = -EINVAL;
--
1.5.1.2
-
From: Wendy Cheng <wcheng@redhat.com>
Current GFS2 setattr call unconditionally invokes do_shrink even the
requested size and actual file size are equal. This has generated large
amount of extra IOs found during NFS benchmark runs. This patch moves
the relevant logic out of shrink code path. Since setattr is a system
call, the time stamps update is still required.
Signed-off-by: S. Wendy Cheng <wcheng@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index cd805a6..9b89904 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -1085,6 +1085,33 @@ static int do_shrink(struct gfs2_inode *ip, u64 size)
return error;
}
+static int do_touch(struct gfs2_inode *ip, u64 size)
+{
+ struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+ struct buffer_head *dibh;
+ int error;
+
+ error = gfs2_trans_begin(sdp, RES_DINODE, 0);
+ if (error)
+ return error;
+
+ down_write(&ip->i_rw_mutex);
+
+ error = gfs2_meta_inode_buffer(ip, &dibh);
+ if (error)
+ goto do_touch_out;
+
+ ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
+ gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+ gfs2_dinode_out(ip, dibh->b_data);
+ brelse(dibh);
+
+do_touch_out:
+ up_write(&ip->i_rw_mutex);
+ gfs2_trans_end(sdp);
+ return error;
+}
+
/**
* gfs2_truncatei - make a file a given size
* @ip: the inode
@@ -1105,8 +1132,11 @@ int gfs2_truncatei(struct gfs2_inode *ip, u64 size)
if (size > ip->i_di.di_size)
error = do_grow(ip, size);
- else
+ else if (size < ip->i_di.di_size)
error = do_shrink(ip, size);
+ else
+ /* update time stamps */
+ error = do_touch(ip, size);
return error;
}
--
1.5.1.2
-
From: Patrick Caulfield <pcaulfie@redhat.com>
This patch fixes the slight mess made in lowcomms closing by previous patches
and fixes all sorts of DLM hangs.
Signed-Off-By: Patrick Caulfield <pcaulfie@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 9e9d2e8..62a8a6c 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -334,18 +334,8 @@ static void close_connection(struct connection *con, bool and_other)
con->rx_page = NULL;
}
- /* If we are an 'othercon' then NULL the pointer to us
- from the parent and tidy ourself up */
- if (test_bit(CF_IS_OTHERCON, &con->flags)) {
- struct connection *parent = __nodeid2con(con->nodeid, 0);
- parent->othercon = NULL;
- kmem_cache_free(con_cache, con);
- }
- else {
- /* Parent connections get reused */
- con->retries = 0;
- mutex_unlock(&con->sock_mutex);
- }
+ con->retries = 0;
+ mutex_unlock(&con->sock_mutex);
}
/* We only send shutdown messages to nodes that are not part of the cluster */
@@ -731,6 +721,8 @@ static int tcp_accept_from_sock(struct connection *con)
INIT_WORK(&othercon->swork, process_send_sockets);
INIT_WORK(&othercon->rwork, process_recv_sockets);
set_bit(CF_IS_OTHERCON, &othercon->flags);
+ }
+ if (!othercon->sock) {
newcon->othercon = othercon;
othercon->sock = newsock;
newsock->sk->sk_user_data = othercon;
--
1.5.1.2
-
From: Abhijith Das <adas@redhat.com>
With this patch, gfs2 glockdump through the debugfs filesystem will only
dump glocks for the specified filesystem instead of all glocks. Also, to
aid debugging, the glock number is dumped in hex instead of decimal.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: S. Wendy Cheng <wcheng@redhat.com>
Signed-off-by: Abhijith Das <adas@redhat.com>
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 559937c..3d94918 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1844,7 +1844,7 @@ static int dump_glock(struct glock_iter *gi, struct gfs2_glock *gl)
spin_lock(&gl->gl_spin);
- print_dbg(gi, "Glock 0x%p (%u, %llu)\n", gl, gl->gl_name.ln_type,
+ print_dbg(gi, "Glock 0x%p (%u, 0x%llx)\n", gl, gl->gl_name.ln_type,
(unsigned long long)gl->gl_name.ln_number);
print_dbg(gi, " gl_flags =");
for (x = 0; x < 32; x++) {
@@ -2024,20 +2024,21 @@ static int gfs2_glock_iter_next(struct glock_iter *gi)
{
struct gfs2_glock *gl;
+restart:
read_lock(gl_lock_addr(gi->hash));
gl = gi->gl;
if (gl) {
- gi->gl = hlist_entry(gl->gl_list.next, struct gfs2_glock,
- gl_list);
+ gi->gl = hlist_entry(gl->gl_list.next,
+ struct gfs2_glock, gl_list);
if (gi->gl)
gfs2_glock_hold(gi->gl);
}
read_unlock(gl_lock_addr(gi->hash));
if (gl)
gfs2_glock_put(gl);
-
- while(gi->gl == NULL) {
+ if (gl && gi->gl == NULL)
gi->hash++;
+ while(gi->gl == NULL) {
if (gi->hash >= GFS2_GL_HASH_SIZE)
return 1;
read_lock(gl_lock_addr(gi->hash));
@@ -2046,7 +2047,12 @@ static int gfs2_glock_iter_next(struct glock_iter *gi)
if (gi->gl)
gfs2_glock_hold(gi->gl);
read_unlock(gl_lock_addr(gi->hash));
+ gi->hash++;
}
+
+ if (gi->sdp != gi->gl->gl_sbd)
+ goto restart;
+
return 0;
}
@@ -2068,16 +2074,10 @@ static struct glock_iter *gfs2_glock_iter_init(struct gfs2_sbd *sdp)
gi->sdp = sdp;
gi->hash = 0;
gi->seq = NULL;
+ gi->gl = NULL;
...From: Bob Peterson <rpeterso@redhat.com>
This is a patch to GFS2 to protect sd_log_num_jdata with the
gfs2_log_lock. Without this patch, there is a timing window
where you can get hit the following assert from function
gfs2_log_flush():
gfs2_assert_withdraw(sdp,
sdp->sd_log_num_buf + sdp->sd_log_num_jdata ==
sdp->sd_log_commited_buf +
sdp->sd_log_commited_databuf);
I've tested it on my roth cluster and it fixes the problem.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index a0371f8..7ef3356 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -492,11 +492,12 @@ static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
gfs2_trans_add_gl(bd->bd_gl);
if (gfs2_is_jdata(ip)) {
- sdp->sd_log_num_jdata++;
gfs2_pin(sdp, bd->bd_bh);
tr->tr_num_databuf_new++;
}
gfs2_log_lock(sdp);
+ if (gfs2_is_jdata(ip))
+ sdp->sd_log_num_jdata++;
sdp->sd_log_num_databuf++;
list_add(&le->le_list, &sdp->sd_log_le_databuf);
gfs2_log_unlock(sdp);
--
1.5.1.2
-
From: Abhijith Das <adas@redhat.com>
When you try to mount gfs2 with -o garbage, the mount fails and the gfs2
superblock is deallocated and becomes NULL. The vfs comes around later
on and calls gfs2_kill_sb. At this point the hidden gfs2 superblock
pointer (sb->s_fs_info) is NULL and dereferencing it through
gfs2_meta_syncfs causes the panic. (the other function call to
gfs2_delete_debugfs_file() succeeds because this function already checks
for a NULL pointer)
Signed-off-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 9e0e9be..314c113 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -887,8 +887,10 @@ error:
static void gfs2_kill_sb(struct super_block *sb)
{
- gfs2_delete_debugfs_file(sb->s_fs_info);
- gfs2_meta_syncfs(sb->s_fs_info);
+ if (sb->s_fs_info) {
+ gfs2_delete_debugfs_file(sb->s_fs_info);
+ gfs2_meta_syncfs(sb->s_fs_info);
+ }
kill_block_super(sb);
}
--
1.5.1.2
-
From: Benjamin Marzinski <bmarzins@redhat.com> 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> diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 3d94918..931368a 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -27,6 +27,8 @@ #include <linux/debugfs.h> #include <linux/kthread.h> #include <linux/freezer.h> +#include <linux/workqueue.h> +#include <linux/jiffies.h> #include "gfs2.h" #include "incore.h" @@ -58,10 +60,13 @@ static int gfs2_dump_lockstate(struct gfs2_sbd *sdp); static int dump_glock(struct glock_iter *gi, struct gfs2_glock *gl); static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh); static void gfs2_glock_drop_th(struct gfs2_glock *gl); +static void run_queue(struct gfs2_glock *gl); + static DECLARE_RWSEM(gfs2_umount_flush_sem); static struct dentry *gfs2_root; static struct task_struct ...
Hi,
There have been no changes since I posted the patches for review a few
days ago. Please consider pulling the following changes from the
GFS2/DLM -nmw git tree:
The following changes since commit bbf25010f1a6b761914430f5fca081ec8c7accd1:
Linus Torvalds (1):
Linux 2.6.23
are found in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw.git
Abhijith Das (5):
[GFS2] Force unstuff of hidden quota inode
[GFS2] Fix quota do_list operation hang
[GFS2] Wendy's dump lockname in hex & fix glock dump
[GFS2] panic after can't parse mount arguments
[GFS2] flocks from same process trip kernel BUG at fs/gfs2/glock.c:1118!
Benjamin Marzinski (3):
[GFS2] Add NULL entry to token table
[GFS2] delay glock demote for a minimum hold time
[GFS2] Alternate gfs2_iget to avoid looking up inodes being freed
Bob Peterson (9):
[GFS2] Move some code inside the log lock
[GFS2] Revert part of earlier log.c changes
[GFS2] Prevent infinite loop in try_rgrp_unlink()
[GFS2] Detach buf data during in-place writeback
[GFS2] invalid metadata block - REVISED
[GFS2] Ensure journal file cache is flushed after recovery
[GFS2] Patch to protect sd_log_num_jdata
[GFS2] Fix ordering of dirty/journal for ordered buffer unstuffing
[GFS2] GFS2: chmod hung - fix race in thread creation
David Teigland (1):
[DLM] block dlm_recv in recovery transition
Denis Cheng (7):
[GFS2] use an temp variable to reduce a spin_unlock
[GFS2] mark struct *_operations const
[GFS2] use the declaration of gfs2_dops in the header file instead
[GFS2] use list_for_each_entry instead
[GFS2] unneeded typecast
[GFS2] better code for translating characters
[GFS2] fixed a NULL pointer assignment BUG
Jesper Juhl (1):
[GFS2] Clean up duplicate includes in fs/gfs2/
Josef Whiter (1):
[GFS2] Fix calculation of demote ...
