From 87124e581bfeaa5864662a435b6ee2a19e91b905 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 23 Jul 2007 09:54:36 +0100 Subject: [GFS2] Fix two races relating to glock callbacks 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 --- fs/gfs2/glock.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 3f0974e1afef..6a3eeba102f9 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 = gl->gl_sbd; int flags = gh ? gh->gh_flags : 0; -- cgit v1.2.3 From 26caee5bc643b318fa2e9bd4f66dace1755ec413 Mon Sep 17 00:00:00 2001 From: Josef Whiter Date: Mon, 23 Jul 2007 10:02:40 +0100 Subject: [GFS2] Fix calculation of demote state 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 6a3eeba102f9..6b6ae4537340 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); } -- cgit v1.2.3 From aa0481e58a9a97a97035725a712920b5fe32f348 Mon Sep 17 00:00:00 2001 From: Jesper Juhl Date: Sat, 21 Jul 2007 17:03:22 +0200 Subject: [GFS2] Clean up duplicate includes in fs/gfs2/ This patch cleans up duplicate includes in fs/gfs2/ Signed-off-by: Jesper Juhl Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 2 -- fs/gfs2/locking/dlm/lock_dlm.h | 1 - fs/gfs2/locking/nolock/main.c | 1 - 3 files changed, 4 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 6b6ae4537340..d403fd708075 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -25,8 +25,6 @@ #include #include #include -#include -#include #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 24d70f73b651..9e8265d28377 100644 --- a/fs/gfs2/locking/dlm/lock_dlm.h +++ b/fs/gfs2/locking/dlm/lock_dlm.h @@ -13,7 +13,6 @@ #include #include #include -#include #include #include #include diff --git a/fs/gfs2/locking/nolock/main.c b/fs/gfs2/locking/nolock/main.c index 0d149c8c493a..d3b8ce6fbbe3 100644 --- a/fs/gfs2/locking/nolock/main.c +++ b/fs/gfs2/locking/nolock/main.c @@ -9,7 +9,6 @@ #include #include -#include #include #include #include -- cgit v1.2.3 From afd0942d98f74296b74993739e41d2ca7cb9fd5a Mon Sep 17 00:00:00 2001 From: Steve French Date: Fri, 20 Jul 2007 13:07:26 -0500 Subject: [GFS2] GFS2 not checking pointer on create when running under nfsd 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c index 911c115b5c6c..5b8b994b9912 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); } -- cgit v1.2.3 From 7b08fc620109c2f66575e9ae884f45c37933ea18 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Tue, 24 Jul 2007 13:53:36 +0100 Subject: [GFS2] Fix an oops in glock dumping 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 --- fs/gfs2/glock.c | 73 +++++++++++++++++++++++++++------------------------------ 1 file changed, 35 insertions(+), 38 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index d403fd708075..e4bc8ae561aa 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 = hlist_entry(gl->gl_list.next, struct gfs2_glock, + gl_list); if (gi->gl) - break; + gfs2_glock_hold(gi->gl); } read_unlock(gl_lock_addr(gi->hash)); + if (gl) + gfs2_glock_put(gl); + + while(gi->gl == NULL) { + gi->hash++; + if (gi->hash >= GFS2_GL_HASH_SIZE) + return 1; + read_lock(gl_lock_addr(gi->hash)); + gi->gl = hlist_entry(gl_hash_table[gi->hash].hb_list.first, + struct gfs2_glock, gl_list); + if (gi->gl) + gfs2_glock_hold(gi->gl); + read_unlock(gl_lock_addr(gi->hash)); + } return 0; } static void gfs2_glock_iter_free(struct glock_iter *gi) { + if (gi->gl) + gfs2_glock_put(gi->gl); kfree(gi); } @@ -2044,12 +2034,17 @@ static struct glock_iter *gfs2_glock_iter_init(struct gfs2_sbd *sdp) gi->sdp = sdp; gi->hash = 0; - gi->gl = NULL; - gi->hb_list = NULL; gi->seq = NULL; memset(gi->string, 0, sizeof(gi->string)); - if (gfs2_glock_iter_next(gi)) { + read_lock(gl_lock_addr(gi->hash)); + gi->gl = hlist_entry(gl_hash_table[gi->hash].hb_list.first, + struct gfs2_glock, gl_list); + if (gi->gl) + gfs2_glock_hold(gi->gl); + read_unlock(gl_lock_addr(gi->hash)); + + if (!gi->gl && gfs2_glock_iter_next(gi)) { gfs2_glock_iter_free(gi); return NULL; } @@ -2066,7 +2061,7 @@ static void *gfs2_glock_seq_start(struct seq_file *file, loff_t *pos) if (!gi) return NULL; - while (n--) { + while(n--) { if (gfs2_glock_iter_next(gi)) { gfs2_glock_iter_free(gi); return NULL; @@ -2093,7 +2088,9 @@ static void *gfs2_glock_seq_next(struct seq_file *file, void *iter_ptr, static void gfs2_glock_seq_stop(struct seq_file *file, void *iter_ptr) { - /* nothing for now */ + struct glock_iter *gi = iter_ptr; + if (gi) + gfs2_glock_iter_free(gi); } static int gfs2_glock_seq_show(struct seq_file *file, void *iter_ptr) -- cgit v1.2.3 From 905d2aefa9e06ebb995df96920d273a516fcd3f9 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Tue, 24 Jul 2007 14:05:31 -0500 Subject: [GFS2] Move some code inside the log lock 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/lops.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 3b395c41b2f3..a0371f835cfb 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 +534,7 @@ static void databuf_lo_before_commit(struct gfs2_sbd *sdp) struct gfs2_log_descriptor *ld; unsigned int limit; unsigned int total_dbuf; - unsigned int total_jdata = sdp->sd_log_num_jdata; + unsigned int total_jdata; unsigned int num, n; __be64 *ptr = NULL; @@ -536,6 +546,7 @@ static void databuf_lo_before_commit(struct gfs2_sbd *sdp) */ gfs2_log_lock(sdp); total_dbuf = sdp->sd_log_num_databuf; + total_jdata = sdp->sd_log_num_jdata; bd2 = bd1 = list_prepare_entry(bd1, &sdp->sd_log_le_databuf, bd_le.le_list); while(total_dbuf) { @@ -621,10 +632,10 @@ static void databuf_lo_before_commit(struct gfs2_sbd *sdp) } gfs2_log_unlock(sdp); if (bh) { - set_buffer_mapped(bh); set_buffer_dirty(bh); ll_rw_block(WRITE, 1, &bh); bh = NULL; + ptr = NULL; } n = 0; gfs2_log_lock(sdp); -- cgit v1.2.3 From 693ddeabbb3e563f192a7ac74ec04168aa92e8d8 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Tue, 24 Jul 2007 14:07:33 -0500 Subject: [GFS2] Revert part of earlier log.c changes 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/log.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-) diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index f49a12e24086..f7c0608332fb 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; -- cgit v1.2.3 From 6760bdcd03a12d7d082794311ccbaf44bfc23b06 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Tue, 24 Jul 2007 14:09:32 -0500 Subject: [GFS2] Prevent infinite loop in try_rgrp_unlink() 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/rgrp.c | 30 +++++++++++++++++------------- 1 file changed, 17 insertions(+), 13 deletions(-) diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index ce48c4594ec8..b93ac45b88bb 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 +863,21 @@ static int try_rgrp_fit(struct gfs2_rgrpd *rgd, struct gfs2_alloc *al) static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked) { struct inode *inode; - u32 goal = 0; + u32 goal = 0, block; u64 no_addr; for(;;) { if (goal >= rgd->rd_data) break; - goal = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED, - GFS2_BLKST_UNLINKED); - if (goal == BFITNOENT) + block = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED, + GFS2_BLKST_UNLINKED); + if (block == BFITNOENT) break; - no_addr = goal + rgd->rd_data0; + /* rgblk_search can return a block < goal, so we need to + keep it marching forward. */ + no_addr = block + rgd->rd_data0; goal++; - if (no_addr < *last_unlinked) + if (*last_unlinked != NO_BLOCK && no_addr <= *last_unlinked) continue; *last_unlinked = no_addr; inode = gfs2_inode_lookup(rgd->rd_sbd->sd_vfs, DT_UNKNOWN, @@ -1152,7 +1158,7 @@ int gfs2_inplace_reserve_i(struct gfs2_inode *ip, char *file, unsigned int line) struct gfs2_alloc *al = &ip->i_alloc; struct inode *inode; int error = 0; - u64 last_unlinked = 0; + u64 last_unlinked = NO_BLOCK; if (gfs2_assert_warn(sdp, al->al_requested)) return -EINVAL; @@ -1305,9 +1311,7 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal, goal = 0; } - if (old_state != new_state) { - gfs2_assert_withdraw(rgd->rd_sbd, blk != BFITNOENT); - + if (blk != BFITNOENT && old_state != new_state) { gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1); gfs2_setbit(rgd, bi->bi_bh->b_data + bi->bi_offset, bi->bi_len, blk, new_state); -- cgit v1.2.3 From cee23c79d08c57bbbb9923703409af3b17518c58 Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Wed, 25 Jul 2007 17:53:58 +0800 Subject: [GFS2] use an temp variable to reduce a spin_unlock this is more clear. Signed-off-by: Denis Cheng Signed-off-by: David Teigland Signed-off-by: Steven Whitehouse --- fs/gfs2/locking/dlm/plock.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) diff --git a/fs/gfs2/locking/dlm/plock.c b/fs/gfs2/locking/dlm/plock.c index fba1f1d87e4f..1f7b038530b4 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 = { -- cgit v1.2.3 From 0f8468c8bef3d04637c924e7bef20ca53018b319 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 25 Jul 2007 10:06:22 -0500 Subject: [GFS2] Detach buf data during in-place writeback 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/log.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index f7c0608332fb..00ab6c070a15 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); } } -- cgit v1.2.3 From 4ef290025ccde7c52ba219cf733a4295acd5401f Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Tue, 31 Jul 2007 18:31:11 +0800 Subject: [GFS2] mark struct *_operations const these struct *_operations are all method tables, thus should be const. Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/eaops.c | 8 ++++---- fs/gfs2/eaops.h | 4 ++-- fs/gfs2/glock.c | 2 +- 3 files changed, 7 insertions(+), 7 deletions(-) diff --git a/fs/gfs2/eaops.c b/fs/gfs2/eaops.c index 1ab3e9d73886..aa8dbf303f6d 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 508b4f7a2449..da2f7fbbb40d 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 e4bc8ae561aa..0054b7df714a 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -2103,7 +2103,7 @@ static int gfs2_glock_seq_show(struct seq_file *file, void *iter_ptr) return 0; } -static struct seq_operations gfs2_glock_seq_ops = { +static const struct seq_operations gfs2_glock_seq_ops = { .start = gfs2_glock_seq_start, .next = gfs2_glock_seq_next, .stop = gfs2_glock_seq_stop, -- cgit v1.2.3 From ca5a939b33166a9f5a2556e6c4ec031524852ba2 Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Tue, 31 Jul 2007 18:31:12 +0800 Subject: [GFS2] use the declaration of gfs2_dops in the header file instead Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index cf5aa5050548..9a5e84038dd6 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; -- cgit v1.2.3 From 8fbbfd214c853102b614f4705c1904ed14f5a808 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Wed, 1 Aug 2007 13:57:10 +0100 Subject: [GFS2] Reduce number of gfs2_scand processes to one 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 --- fs/gfs2/daemon.c | 24 -------------------- fs/gfs2/daemon.h | 1 - fs/gfs2/glock.c | 63 +++++++++++++++++++++++++++++++++++++++------------- fs/gfs2/glock.h | 4 ++-- fs/gfs2/incore.h | 2 -- fs/gfs2/main.c | 3 +++ fs/gfs2/ops_fstype.c | 9 -------- fs/gfs2/ops_super.c | 1 - fs/gfs2/super.c | 1 - fs/gfs2/sys.c | 2 -- 10 files changed, 53 insertions(+), 57 deletions(-) diff --git a/fs/gfs2/daemon.c b/fs/gfs2/daemon.c index 3548d9f31e0d..3731ab0771d5 100644 --- a/fs/gfs2/daemon.c +++ b/fs/gfs2/daemon.c @@ -34,30 +34,6 @@ 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 801007120fb2..0de9b3557955 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 0054b7df714a..559937c710fc 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -25,6 +25,8 @@ #include #include #include +#include +#include #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 gfs2_glock *gl); static DECLARE_RWSEM(gfs2_umount_flush_sem); static struct dentry *gfs2_root; +static struct task_struct *scand_process; +static unsigned int scand_secs = 5; #define GFS2_GL_HASH_SHIFT 15 #define GFS2_GL_HASH_SIZE (1 << GFS2_GL_HASH_SHIFT) @@ -1627,7 +1631,7 @@ static int examine_bucket(glock_examiner examiner, struct gfs2_sbd *sdp, goto out; gl = list_entry(head->first, struct gfs2_glock, gl_list); while(1) { - if (gl->gl_sbd == sdp) { + if (!sdp || gl->gl_sbd == sdp) { gfs2_glock_hold(gl); read_unlock(gl_lock_addr(hash)); if (prev) @@ -1645,6 +1649,7 @@ out: read_unlock(gl_lock_addr(hash)); if (prev) gfs2_glock_put(prev); + cond_resched(); return has_entries; } @@ -1672,20 +1677,6 @@ out_schedule: gfs2_glock_schedule_for_reclaim(gl); } -/** - * gfs2_scand_internal - Look for glocks and inodes to toss from memory - * @sdp: the filesystem - * - */ - -void gfs2_scand_internal(struct gfs2_sbd *sdp) -{ - unsigned int x; - - for (x = 0; x < GFS2_GL_HASH_SIZE; x++) - examine_bucket(scan_glock, sdp, x); -} - /** * clear_glock - look at a glock and see if we can free it from glock cache * @gl: the glock to look at @@ -1973,6 +1964,35 @@ static int gfs2_dump_lockstate(struct gfs2_sbd *sdp) return error; } +/** + * 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() + */ + +static int gfs2_scand(void *data) +{ + unsigned x; + unsigned delay; + + while (!kthread_should_stop()) { + for (x = 0; x < GFS2_GL_HASH_SIZE; x++) + examine_bucket(scan_glock, NULL, x); + if (freezing(current)) + refrigerator(); + delay = scand_secs; + if (delay < 1) + delay = 1; + schedule_timeout_interruptible(delay * HZ); + } + + return 0; +} + + + int __init gfs2_glock_init(void) { unsigned i; @@ -1984,9 +2004,22 @@ int __init gfs2_glock_init(void) rwlock_init(&gl_hash_locks[i]); } #endif + + scand_process = kthread_run(gfs2_scand, NULL, "gfs2_scand"); + if (IS_ERR(scand_process)) + return PTR_ERR(scand_process); + return 0; } +void gfs2_glock_exit(void) +{ + kthread_stop(scand_process); +} + +module_param(scand_secs, uint, S_IRUGO|S_IWUSR); +MODULE_PARM_DESC(scand_secs, "The number of seconds between scand runs"); + static int gfs2_glock_iter_next(struct glock_iter *gi) { struct gfs2_glock *gl; diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h index 7721ca3fff9e..f7a8e626aa08 100644 --- a/fs/gfs2/glock.h +++ b/fs/gfs2/glock.h @@ -132,11 +132,11 @@ void gfs2_glock_cb(void *cb_data, unsigned int type, void *data); void gfs2_glock_schedule_for_reclaim(struct gfs2_glock *gl); void gfs2_reclaim_glock(struct gfs2_sbd *sdp); - -void gfs2_scand_internal(struct gfs2_sbd *sdp); void gfs2_gl_hash_clear(struct gfs2_sbd *sdp, int wait); int __init gfs2_glock_init(void); +void gfs2_glock_exit(void); + int gfs2_create_debugfs_file(struct gfs2_sbd *sdp); void gfs2_delete_debugfs_file(struct gfs2_sbd *sdp); int gfs2_register_debugfs(void); diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 170ba93829c0..1390b30daf19 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -429,7 +429,6 @@ struct gfs2_tune { unsigned int gt_log_flush_secs; unsigned int gt_jindex_refresh_secs; /* Check for new journal index */ - unsigned int gt_scand_secs; unsigned int gt_recoverd_secs; unsigned int gt_logd_secs; unsigned int gt_quotad_secs; @@ -574,7 +573,6 @@ struct gfs2_sbd { /* Daemon stuff */ - struct task_struct *sd_scand_process; struct task_struct *sd_recoverd_process; struct task_struct *sd_logd_process; struct task_struct *sd_quotad_process; diff --git a/fs/gfs2/main.c b/fs/gfs2/main.c index d5d4e68b8807..79c91fd8381b 100644 --- a/fs/gfs2/main.c +++ b/fs/gfs2/main.c @@ -107,6 +107,8 @@ static int __init init_gfs2_fs(void) fail_unregister: unregister_filesystem(&gfs2_fs_type); fail: + gfs2_glock_exit(); + if (gfs2_bufdata_cachep) kmem_cache_destroy(gfs2_bufdata_cachep); @@ -127,6 +129,7 @@ fail: static void __exit exit_gfs2_fs(void) { + gfs2_glock_exit(); gfs2_unregister_debugfs(); unregister_filesystem(&gfs2_fs_type); unregister_filesystem(&gfs2meta_fs_type); diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 9a5e84038dd6..58c730be2073 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -160,14 +160,6 @@ static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, if (undo) goto fail_trans; - p = kthread_run(gfs2_scand, sdp, "gfs2_scand"); - error = IS_ERR(p); - if (error) { - fs_err(sdp, "can't start scand thread: %d\n", error); - return error; - } - sdp->sd_scand_process = p; - for (sdp->sd_glockd_num = 0; sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd; sdp->sd_glockd_num++) { @@ -228,7 +220,6 @@ fail: while (sdp->sd_glockd_num--) kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]); - kthread_stop(sdp->sd_scand_process); return error; } diff --git a/fs/gfs2/ops_super.c b/fs/gfs2/ops_super.c index 603d940f1159..4316690d86f6 100644 --- a/fs/gfs2/ops_super.c +++ b/fs/gfs2/ops_super.c @@ -92,7 +92,6 @@ static void gfs2_put_super(struct super_block *sb) kthread_stop(sdp->sd_recoverd_process); while (sdp->sd_glockd_num--) kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]); - kthread_stop(sdp->sd_scand_process); if (!(sb->s_flags & MS_RDONLY)) { error = gfs2_make_fs_ro(sdp); diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c index f916b9740c75..55898023782b 100644 --- a/fs/gfs2/super.c +++ b/fs/gfs2/super.c @@ -58,7 +58,6 @@ void gfs2_tune_init(struct gfs2_tune *gt) gt->gt_incore_log_blocks = 1024; gt->gt_log_flush_secs = 60; gt->gt_jindex_refresh_secs = 60; - gt->gt_scand_secs = 15; gt->gt_recoverd_secs = 60; gt->gt_logd_secs = 1; gt->gt_quotad_secs = 5; diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c index c26c21b53c19..ba3a1729cc1a 100644 --- a/fs/gfs2/sys.c +++ b/fs/gfs2/sys.c @@ -442,7 +442,6 @@ TUNE_ATTR(quota_simul_sync, 1); TUNE_ATTR(quota_cache_secs, 1); TUNE_ATTR(stall_secs, 1); TUNE_ATTR(statfs_quantum, 1); -TUNE_ATTR_DAEMON(scand_secs, scand_process); TUNE_ATTR_DAEMON(recoverd_secs, recoverd_process); TUNE_ATTR_DAEMON(logd_secs, logd_process); TUNE_ATTR_DAEMON(quotad_secs, quotad_process); @@ -464,7 +463,6 @@ static struct attribute *tune_attrs[] = { &tune_attr_quota_cache_secs.attr, &tune_attr_stall_secs.attr, &tune_attr_statfs_quantum.attr, - &tune_attr_scand_secs.attr, &tune_attr_recoverd_secs.attr, &tune_attr_logd_secs.attr, &tune_attr_quotad_secs.attr, -- cgit v1.2.3 From 5f3eae7546093d845ca8ada1b95714202a136a1a Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 8 Aug 2007 16:52:09 -0500 Subject: [GFS2] invalid metadata block - REVISED 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/rgrp.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index b93ac45b88bb..2d7f7ea0c9a8 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, old_state); else -- cgit v1.2.3 From 75be73a8246ef96f7fa3f05a6a1450159fbb7a64 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 8 Aug 2007 17:08:14 -0500 Subject: [GFS2] Ensure journal file cache is flushed after recovery 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 Signed-off-by: Steven Whitehouse -- --- fs/gfs2/ops_fstype.c | 2 +- fs/gfs2/recovery.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 58c730be2073..f0bcaa25737c 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 5ada38c99a2c..beb6c7ac0086 100644 --- a/fs/gfs2/recovery.c +++ b/fs/gfs2/recovery.c @@ -469,7 +469,7 @@ int gfs2_recover_journal(struct gfs2_jdesc *jd) }; error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, - LM_FLAG_NOEXP, &ji_gh); + LM_FLAG_NOEXP | GL_NOCACHE, &ji_gh); if (error) goto fail_gunlock_j; } else { -- cgit v1.2.3 From adb4ec13cdddb6ab8280e4c7808ba30f46504e1d Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Sat, 11 Aug 2007 10:27:07 +0800 Subject: [GFS2] use list_for_each_entry instead Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index f0bcaa25737c..32b2859a89a5 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; -- cgit v1.2.3 From 2d3ba1ea97d839e60d742ccf9a6de5bf039c0b53 Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Sat, 11 Aug 2007 10:27:08 +0800 Subject: [GFS2] unneeded typecast sb->s_fs_info is a void pointer, thus the type cast is not needed. Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 32b2859a89a5..25cfab95eb9c 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; -- cgit v1.2.3 From 5d35e31f43c4910d0b6afc5160728a84bbaf86f0 Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Mon, 13 Aug 2007 11:01:58 +0800 Subject: [GFS2] better code for translating characters the original code could work, but I think this code could work better. Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 25cfab95eb9c..6c820cb23473 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: -- cgit v1.2.3 From 0fd5355470ea40355b8af76d01748ec7b9926d4d Mon Sep 17 00:00:00 2001 From: Abhijith Das Date: Tue, 14 Aug 2007 15:34:58 -0500 Subject: [GFS2] Force unstuff of hidden quota inode 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/quota.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c index 6e546ee8f3d4..5dfa4656122b 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; -- cgit v1.2.3 From 34eaae398e29dadeed95efd30f1eb694e5932b34 Mon Sep 17 00:00:00 2001 From: Denis Cheng Date: Wed, 15 Aug 2007 23:54:44 +0800 Subject: [GFS2] fixed a NULL pointer assignment BUG Signed-off-by: Denis Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 6c820cb23473..c1c6672ebb83 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; -- cgit v1.2.3 From 2d9a4bbf6d28673f4057682cc02d16bf288b4a35 Mon Sep 17 00:00:00 2001 From: Abhijith Das Date: Wed, 15 Aug 2007 11:25:05 -0500 Subject: [GFS2] Fix quota do_list operation hang 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/quota.c | 3 +++ include/linux/gfs2_ondisk.h | 30 +++++++++++++++++++++++++++++- 2 files changed, 32 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c index 5dfa4656122b..addb51e0f135 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 a44a6a078f0a..c3c19f926e6f 100644 --- a/include/linux/gfs2_ondisk.h +++ b/include/linux/gfs2_ondisk.h @@ -169,6 +169,33 @@ struct gfs2_rgrp { __u8 rg_reserved[80]; /* Several fields from gfs1 now reserved */ }; +/* + * 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 the quota file offsets for the corresponding +gfs2_quota structures on disk: + +for user quotas, given uid, +offset = uid * sizeof(struct gfs2_quota); + +for group quotas, given gid, +offset = (gid * sizeof(struct gfs2_quota)) + sizeof(struct gfs2_quota); + + + uid:0 gid:0 uid:12 gid:12 uid:17 gid:17 uid:5142 gid:5142 ++-------+-------+ +-------+-------+ +-------+- - - -+ +- - - -+-------+ +| valid | valid | :: | valid | valid | :: | valid | inval | :: | inval | valid | ++-------+-------+ +-------+-------+ +-------+- - - -+ +- - - -+-------+ +next:12 next:12 next:17 next:5142 next:NULL next:NULL + | | | | |<-- user quota list | + \______|___________/ \______|___________/ group quota list -->| + | | | + \__________________/ \_______________________________________/ + +*/ + /* * quota structure */ @@ -177,7 +204,8 @@ struct gfs2_quota { __be64 qu_limit; __be64 qu_warn; __be64 qu_value; - __u8 qu_reserved[64]; + __be32 qu_ll_next; /* location of next quota in list */ + __u8 qu_reserved[60]; }; /* -- cgit v1.2.3 From bb3b0e3df5420fdf2c6bbb4417525c6d2ef55bbb Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Thu, 16 Aug 2007 16:03:57 +0100 Subject: [GFS2] Clean up invalidatepage/releasepage 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 Cc: Bob Peterson --- fs/gfs2/glops.c | 4 +- fs/gfs2/log.c | 6 ++- fs/gfs2/ops_address.c | 132 ++++++-------------------------------------------- fs/gfs2/ops_fstype.c | 2 + 4 files changed, 24 insertions(+), 120 deletions(-) diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index 777ca46010e8..88342e0b4bc7 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 00ab6c070a15..d0e6b42c86e1 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 42a5f58f6fca..8407d1db4eac 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 = bh->b_private; - if (bd) { - bd->bd_bh = NULL; - bh->b_private = NULL; - if (!bd->bd_ail && list_empty(&bd->bd_le.le_list)) - kmem_cache_free(gfs2_bufdata_cachep, bd); - } - gfs2_log_unlock(sdp); - - lock_buffer(bh); - clear_buffer_dirty(bh); - bh->b_bdev = NULL; - clear_buffer_mapped(bh); - clear_buffer_req(bh); - clear_buffer_new(bh); - clear_buffer_delay(bh); - unlock_buffer(bh); -} - static void gfs2_invalidatepage(struct page *page, unsigned long offset) { - struct gfs2_sbd *sdp = GFS2_SB(page->mapping->host); - struct buffer_head *head, *bh, *next; - unsigned int curr_off = 0; - BUG_ON(!PageLocked(page)); if (offset == 0) ClearPageChecked(page); - if (!page_has_buffers(page)) - return; - bh = head = page_buffers(page); - do { - unsigned int next_off = curr_off + bh->b_size; - next = bh->b_this_page; - - if (offset <= curr_off) - discard_buffer(sdp, bh); - - curr_off = next_off; - bh = next; - } while (bh != head); - - if (!offset) - try_to_release_page(page, 0); - - return; + block_invalidatepage(page, offset); } /** @@ -735,59 +690,6 @@ out: return rv; } -/** - * stuck_releasepage - We're stuck in gfs2_releasepage(). Print stuff out. - * @bh: the buffer we're stuck on - * - */ - -static void stuck_releasepage(struct buffer_head *bh) -{ - struct inode *inode = bh->b_page->mapping->host; - struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; - struct gfs2_bufdata *bd = bh->b_private; - struct gfs2_glock *gl; -static unsigned limit = 0; - - if (limit > 3) - return; - limit++; - - fs_warn(sdp, "stuck in gfs2_releasepage() %p\n", inode); - fs_warn(sdp, "blkno = %llu, bh->b_count = %d\n", - (unsigned long long)bh->b_blocknr, atomic_read(&bh->b_count)); - fs_warn(sdp, "pinned = %u\n", buffer_pinned(bh)); - fs_warn(sdp, "bh->b_private = %s\n", (bd) ? "!NULL" : "NULL"); - - if (!bd) - return; - - gl = bd->bd_gl; - - fs_warn(sdp, "gl = (%u, %llu)\n", - gl->gl_name.ln_type, (unsigned long long)gl->gl_name.ln_number); - - fs_warn(sdp, "bd_list_tr = %s, bd_le.le_list = %s\n", - (list_empty(&bd->bd_list_tr)) ? "no" : "yes", - (list_empty(&bd->bd_le.le_list)) ? "no" : "yes"); - - if (gl->gl_ops == &gfs2_inode_glops) { - struct gfs2_inode *ip = gl->gl_object; - unsigned int x; - - if (!ip) - return; - - fs_warn(sdp, "ip = %llu %llu\n", - (unsigned long long)ip->i_no_formal_ino, - (unsigned long long)ip->i_no_addr); - - for (x = 0; x < GFS2_MAX_META_HEIGHT; x++) - fs_warn(sdp, "ip->i_cache[%u] = %s\n", - x, (ip->i_cache[x]) ? "!NULL" : "NULL"); - } -} - /** * gfs2_releasepage - free the metadata associated with a page * @page: the page that's being released @@ -805,38 +707,31 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask) struct gfs2_sbd *sdp = aspace->i_sb->s_fs_info; struct buffer_head *bh, *head; struct gfs2_bufdata *bd; - unsigned long t = jiffies + gfs2_tune_get(sdp, gt_stall_secs) * HZ; if (!page_has_buffers(page)) goto out; + gfs2_log_lock(sdp); head = bh = page_buffers(page); do { - while (atomic_read(&bh->b_count)) { - if (!atomic_read(&aspace->i_writecount)) - return 0; - - if (!(gfp_mask & __GFP_WAIT)) - return 0; - - if (time_after_eq(jiffies, t)) { - stuck_releasepage(bh); - /* should we withdraw here? */ - return 0; - } - - yield(); - } - + if (atomic_read(&bh->b_count)) + goto cannot_release; + bd = bh->b_private; + if (bd && bd->bd_ail) + goto cannot_release; gfs2_assert_warn(sdp, !buffer_pinned(bh)); gfs2_assert_warn(sdp, !buffer_dirty(bh)); + bh = bh->b_this_page; + } while(bh != head); + gfs2_log_unlock(sdp); + head = bh = page_buffers(page); + do { gfs2_log_lock(sdp); bd = bh->b_private; if (bd) { gfs2_assert_warn(sdp, bd->bd_bh == bh); gfs2_assert_warn(sdp, list_empty(&bd->bd_list_tr)); - gfs2_assert_warn(sdp, !bd->bd_ail); bd->bd_bh = NULL; if (!list_empty(&bd->bd_le.le_list)) bd = NULL; @@ -851,6 +746,9 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask) out: return try_to_free_buffers(page); +cannot_release: + gfs2_log_unlock(sdp); + return 0; } const struct address_space_operations gfs2_file_aops = { diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index c1c6672ebb83..9e0e9be1e41d 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -35,6 +35,7 @@ #include "super.h" #include "sys.h" #include "util.h" +#include "log.h" #define DO 0 #define UNDO 1 @@ -887,6 +888,7 @@ 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); kill_block_super(sb); } -- cgit v1.2.3 From 382e6e256b0cb1a84a45a520cef75d1b8ff44663 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Thu, 16 Aug 2007 17:08:20 +0100 Subject: [GFS2] Add a missing gfs2_trans_add_bh() 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 Cc: Wendy Cheng --- fs/gfs2/dir.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c index 2beb2f401aa2..08c6dd0c6713 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); -- cgit v1.2.3 From 9a5ad13856cbd10be429f09517c51277c02530f7 Mon Sep 17 00:00:00 2001 From: Benjamin Marzinski Date: Fri, 17 Aug 2007 20:22:07 -0500 Subject: [GFS2] Add NULL entry to token table 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/mount.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/mount.c b/fs/gfs2/mount.c index 4864659555d4..b941f9f9f958 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; -- cgit v1.2.3 From a13b8c5f2381495879e6facd3b3ada51c9e68194 Mon Sep 17 00:00:00 2001 From: Wendy Cheng Date: Mon, 20 Aug 2007 09:29:53 -0400 Subject: [GFS2] Reduce truncate IO traffic 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/bmap.c | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index cd805a66880d..9b8990444e6c 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; } -- cgit v1.2.3 From 61d96be0f474df354c2ff4a2b2bf410b23a5cd60 Mon Sep 17 00:00:00 2001 From: Patrick Caulfield Date: Mon, 20 Aug 2007 15:13:38 +0100 Subject: [DLM] Fix lowcomms socket closing 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 Signed-off-by: Steven Whitehouse --- fs/dlm/lowcomms.c | 16 ++++------------ 1 file changed, 4 insertions(+), 12 deletions(-) diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c index 9e9d2e82f40f..62a8a6ccd992 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; -- cgit v1.2.3 From a947e0335699a1d387c3826e5b8eff9e0afe505e Mon Sep 17 00:00:00 2001 From: Abhijith Das Date: Tue, 21 Aug 2007 09:57:29 -0500 Subject: [GFS2] Wendy's dump lockname in hex & fix glock dump 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 Signed-off-by: S. Wendy Cheng Signed-off-by: Abhijith Das --- fs/gfs2/glock.c | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 559937c710fc..3d949187fed0 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; memset(gi->string, 0, sizeof(gi->string)); - read_lock(gl_lock_addr(gi->hash)); - gi->gl = hlist_entry(gl_hash_table[gi->hash].hb_list.first, - struct gfs2_glock, gl_list); - if (gi->gl) - gfs2_glock_hold(gi->gl); - read_unlock(gl_lock_addr(gi->hash)); - - if (!gi->gl && gfs2_glock_iter_next(gi)) { + if (gfs2_glock_iter_next(gi)) { gfs2_glock_iter_free(gi); return NULL; } -- cgit v1.2.3 From ec217e0ece60f2240772e6f08e0529775846c627 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Wed, 22 Aug 2007 11:15:29 -0500 Subject: [GFS2] Patch to protect sd_log_num_jdata 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/lops.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index a0371f835cfb..7ef335623373 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); -- cgit v1.2.3 From d1e2777d4f419a865ddccdb9b3412021d0e4de51 Mon Sep 17 00:00:00 2001 From: Abhijith Das Date: Thu, 23 Aug 2007 13:33:01 -0500 Subject: [GFS2] panic after can't parse mount arguments 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 Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_fstype.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 9e0e9be1e41d..314c1134d12d 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); } -- cgit v1.2.3 From c4f68a130fc1795e4a75ec5bdaf9e85d86c22419 Mon Sep 17 00:00:00 2001 From: Benjamin Marzinski Date: Thu, 23 Aug 2007 13:19:05 -0500 Subject: [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 Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 75 ++++++++++++++++++++++++++++++++++++++++++++------------ fs/gfs2/glops.c | 2 ++ fs/gfs2/incore.h | 5 ++++ 3 files changed, 66 insertions(+), 16 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 3d949187fed0..931368a385c8 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -27,6 +27,8 @@ #include #include #include +#include +#include #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 *scand_process; static unsigned int scand_secs = 5; +static struct workqueue_struct *glock_workqueue; #define GFS2_GL_HASH_SHIFT 15 #define GFS2_GL_HASH_SIZE (1 << GFS2_GL_HASH_SHIFT) @@ -277,6 +282,18 @@ static struct gfs2_glock *gfs2_glock_find(const struct gfs2_sbd *sdp, return gl; } +static void glock_work_func(struct work_struct *work) +{ + struct gfs2_glock *gl = container_of(work, struct gfs2_glock, gl_work.work); + + spin_lock(&gl->gl_spin); + if (test_and_clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags)) + set_bit(GLF_DEMOTE, &gl->gl_flags); + run_queue(gl); + spin_unlock(&gl->gl_spin); + gfs2_glock_put(gl); +} + /** * gfs2_glock_get() - Get a glock, or create one if one doesn't exist * @sdp: The GFS2 superblock @@ -316,6 +333,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, gl->gl_name = name; atomic_set(&gl->gl_ref, 1); gl->gl_state = LM_ST_UNLOCKED; + gl->gl_demote_state = LM_ST_EXCLUSIVE; gl->gl_hash = hash; gl->gl_owner_pid = 0; gl->gl_ip = 0; @@ -324,10 +342,12 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, gl->gl_req_bh = NULL; gl->gl_vn = 0; gl->gl_stamp = jiffies; + gl->gl_tchange = jiffies; gl->gl_object = NULL; gl->gl_sbd = sdp; gl->gl_aspace = NULL; lops_init_le(&gl->gl_le, &gfs2_glock_lops); + INIT_DELAYED_WORK(&gl->gl_work, glock_work_func); /* If this glock protects actual on-disk data or metadata blocks, create a VFS inode to manage the pages/buffers holding them. */ @@ -441,6 +461,8 @@ static void wait_on_holder(struct gfs2_holder *gh) static void gfs2_demote_wake(struct gfs2_glock *gl) { + BUG_ON(!spin_is_locked(&gl->gl_spin)); + gl->gl_demote_state = LM_ST_EXCLUSIVE; clear_bit(GLF_DEMOTE, &gl->gl_flags); smp_mb__after_clear_bit(); wake_up_bit(&gl->gl_flags, GLF_DEMOTE); @@ -682,10 +704,14 @@ static void gfs2_glmutex_unlock(struct gfs2_glock *gl) * practise: LM_ST_SHARED and LM_ST_UNLOCKED */ -static void handle_callback(struct gfs2_glock *gl, unsigned int state, int remote) +static void handle_callback(struct gfs2_glock *gl, unsigned int state, + int remote, unsigned long delay) { + int bit = delay ? GLF_PENDING_DEMOTE : GLF_DEMOTE; + spin_lock(&gl->gl_spin); - if (test_and_set_bit(GLF_DEMOTE, &gl->gl_flags) == 0) { + set_bit(bit, &gl->gl_flags); + if (gl->gl_demote_state == LM_ST_EXCLUSIVE) { gl->gl_demote_state = state; gl->gl_demote_time = jiffies; if (remote && gl->gl_ops->go_type == LM_TYPE_IOPEN && @@ -727,6 +753,7 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state) } gl->gl_state = new_state; + gl->gl_tchange = jiffies; } /** @@ -813,7 +840,6 @@ out: gl->gl_req_gh = NULL; gl->gl_req_bh = NULL; clear_bit(GLF_LOCK, &gl->gl_flags); - run_queue(gl); spin_unlock(&gl->gl_spin); } @@ -885,7 +911,6 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret) gfs2_assert_warn(sdp, !ret); state_change(gl, LM_ST_UNLOCKED); - gfs2_demote_wake(gl); if (glops->go_inval) glops->go_inval(gl, DIO_METADATA); @@ -898,10 +923,10 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret) } spin_lock(&gl->gl_spin); + gfs2_demote_wake(gl); gl->gl_req_gh = NULL; gl->gl_req_bh = NULL; clear_bit(GLF_LOCK, &gl->gl_flags); - run_queue(gl); spin_unlock(&gl->gl_spin); gfs2_glock_put(gl); @@ -1209,9 +1234,10 @@ void gfs2_glock_dq(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; const struct gfs2_glock_operations *glops = gl->gl_ops; + unsigned delay = 0; if (gh->gh_flags & GL_NOCACHE) - handle_callback(gl, LM_ST_UNLOCKED, 0); + handle_callback(gl, LM_ST_UNLOCKED, 0, 0); gfs2_glmutex_lock(gl); @@ -1229,8 +1255,14 @@ void gfs2_glock_dq(struct gfs2_holder *gh) } clear_bit(GLF_LOCK, &gl->gl_flags); - run_queue(gl); spin_unlock(&gl->gl_spin); + + gfs2_glock_hold(gl); + if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) && + !test_bit(GLF_DEMOTE, &gl->gl_flags)) + delay = gl->gl_ops->go_min_hold_time; + if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0) + gfs2_glock_put(gl); } void gfs2_glock_dq_wait(struct gfs2_holder *gh) @@ -1457,18 +1489,21 @@ static void blocking_cb(struct gfs2_sbd *sdp, struct lm_lockname *name, unsigned int state) { struct gfs2_glock *gl; + unsigned long delay = 0; + unsigned long holdtime; + unsigned long now = jiffies; gl = gfs2_glock_find(sdp, name); if (!gl) return; - handle_callback(gl, state, 1); - - spin_lock(&gl->gl_spin); - run_queue(gl); - spin_unlock(&gl->gl_spin); + holdtime = gl->gl_tchange + gl->gl_ops->go_min_hold_time; + if (time_before(now, holdtime)) + delay = holdtime - now; - gfs2_glock_put(gl); + handle_callback(gl, state, 1, delay); + if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0) + gfs2_glock_put(gl); } /** @@ -1509,7 +1544,8 @@ void gfs2_glock_cb(void *cb_data, unsigned int type, void *data) return; if (!gfs2_assert_warn(sdp, gl->gl_req_bh)) gl->gl_req_bh(gl, async->lc_ret); - gfs2_glock_put(gl); + if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) + gfs2_glock_put(gl); up_read(&gfs2_umount_flush_sem); return; } @@ -1602,7 +1638,7 @@ void gfs2_reclaim_glock(struct gfs2_sbd *sdp) if (gfs2_glmutex_trylock(gl)) { if (list_empty(&gl->gl_holders) && gl->gl_state != LM_ST_UNLOCKED && demote_ok(gl)) - handle_callback(gl, LM_ST_UNLOCKED, 0); + handle_callback(gl, LM_ST_UNLOCKED, 0, 0); gfs2_glmutex_unlock(gl); } @@ -1702,7 +1738,7 @@ static void clear_glock(struct gfs2_glock *gl) if (gfs2_glmutex_trylock(gl)) { if (list_empty(&gl->gl_holders) && gl->gl_state != LM_ST_UNLOCKED) - handle_callback(gl, LM_ST_UNLOCKED, 0); + handle_callback(gl, LM_ST_UNLOCKED, 0, 0); gfs2_glmutex_unlock(gl); } } @@ -2009,11 +2045,18 @@ int __init gfs2_glock_init(void) if (IS_ERR(scand_process)) return PTR_ERR(scand_process); + glock_workqueue = create_workqueue("glock_workqueue"); + if (IS_ERR(glock_workqueue)) { + kthread_stop(scand_process); + return PTR_ERR(glock_workqueue); + } + return 0; } void gfs2_glock_exit(void) { + destroy_workqueue(glock_workqueue); kthread_stop(scand_process); } diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index 88342e0b4bc7..7ef6b23bb38a 100644 --- a/fs/gfs2/glops.c +++ b/fs/gfs2/glops.c @@ -454,6 +454,7 @@ const struct gfs2_glock_operations gfs2_inode_glops = { .go_lock = inode_go_lock, .go_unlock = inode_go_unlock, .go_type = LM_TYPE_INODE, + .go_min_hold_time = HZ / 10, }; const struct gfs2_glock_operations gfs2_rgrp_glops = { @@ -464,6 +465,7 @@ const struct gfs2_glock_operations gfs2_rgrp_glops = { .go_lock = rgrp_go_lock, .go_unlock = rgrp_go_unlock, .go_type = LM_TYPE_RGRP, + .go_min_hold_time = HZ / 10, }; const struct gfs2_glock_operations gfs2_trans_glops = { diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 1390b30daf19..23b611aa70d2 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -11,6 +11,7 @@ #define __INCORE_DOT_H__ #include +#include #define DIO_WAIT 0x00000010 #define DIO_METADATA 0x00000020 @@ -130,6 +131,7 @@ struct gfs2_glock_operations { int (*go_lock) (struct gfs2_holder *gh); void (*go_unlock) (struct gfs2_holder *gh); const int go_type; + const unsigned long go_min_hold_time; }; enum { @@ -161,6 +163,7 @@ enum { GLF_LOCK = 1, GLF_STICKY = 2, GLF_DEMOTE = 3, + GLF_PENDING_DEMOTE = 4, GLF_DIRTY = 5, }; @@ -193,6 +196,7 @@ struct gfs2_glock { u64 gl_vn; unsigned long gl_stamp; + unsigned long gl_tchange; void *gl_object; struct list_head gl_reclaim; @@ -203,6 +207,7 @@ struct gfs2_glock { struct gfs2_log_element gl_le; struct list_head gl_ail_list; atomic_t gl_ail_count; + struct delayed_work gl_work; }; struct gfs2_alloc { -- cgit v1.2.3 From e9bd2b3bafd29bf75522546207f0bba0ec4515c2 Mon Sep 17 00:00:00 2001 From: Wendy Cheng Date: Fri, 24 Aug 2007 09:15:01 -0400 Subject: [GFS2] fix inode meta data corruption Fix a nasty inode meta data corruption issue by keeping the buffer head in icache array. This buffer needs to stay in memory until journal flush occurs Otherwise, gfs2_meta_inode_buffer could do a disk read before the inode hits disk. It ends up with meta data corruptions. The buffer will be released as part of the existing journal flush logic. Signed-off-by: S. Wendy Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/inode.c | 20 +++++++++++++++----- 1 file changed, 15 insertions(+), 5 deletions(-) diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c index 34f7bcdea1e9..013f00b90cc2 100644 --- a/fs/gfs2/inode.c +++ b/fs/gfs2/inode.c @@ -244,6 +244,11 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) return 0; } +static void gfs2_inode_bh(struct gfs2_inode *ip, struct buffer_head *bh) +{ + ip->i_cache[0] = bh; +} + /** * gfs2_inode_refresh - Refresh the incore copy of the dinode * @ip: The GFS2 inode @@ -688,7 +693,7 @@ out: static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, const struct gfs2_inum_host *inum, unsigned int mode, unsigned int uid, unsigned int gid, - const u64 *generation, dev_t dev) + const u64 *generation, dev_t dev, struct buffer_head **bhp) { struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); struct gfs2_dinode *di; @@ -743,13 +748,15 @@ static void init_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, di->di_mtime_nsec = cpu_to_be32(tv.tv_nsec); di->di_ctime_nsec = cpu_to_be32(tv.tv_nsec); memset(&di->di_reserved, 0, sizeof(di->di_reserved)); + + set_buffer_uptodate(dibh); - brelse(dibh); + *bhp = dibh; } static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, unsigned int mode, const struct gfs2_inum_host *inum, - const u64 *generation, dev_t dev) + const u64 *generation, dev_t dev, struct buffer_head **bhp) { struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode); unsigned int uid, gid; @@ -770,7 +777,7 @@ static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl, if (error) goto out_quota; - init_dinode(dip, gl, inum, mode, uid, gid, generation, dev); + init_dinode(dip, gl, inum, mode, uid, gid, generation, dev, bhp); gfs2_quota_change(dip, +1, uid, gid); gfs2_trans_end(sdp); @@ -909,6 +916,7 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, struct gfs2_inum_host inum = { .no_addr = 0, .no_formal_ino = 0 }; int error; u64 generation; + struct buffer_head *bh=NULL; if (!name->len || name->len > GFS2_FNAMESIZE) return ERR_PTR(-ENAMETOOLONG); @@ -935,7 +943,7 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, if (error) goto fail_gunlock; - error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev); + error = make_dinode(dip, ghs[1].gh_gl, mode, &inum, &generation, dev, &bh); if (error) goto fail_gunlock2; @@ -945,6 +953,8 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, if (IS_ERR(inode)) goto fail_gunlock2; + gfs2_inode_bh(GFS2_I(inode), bh); + error = gfs2_inode_refresh(GFS2_I(inode)); if (error) goto fail_gunlock2; -- cgit v1.2.3 From 8497a46e178addb27ad1c981befaa17ca788b5c3 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Sun, 26 Aug 2007 14:23:56 +0100 Subject: [GFS2] Correct lock ordering in unlink This patch corrects the lock ordering in unlink to be the same as that in the rest of GFS2, i.e. parent -> child -> rgrp. Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_inode.c | 29 ++++++++++++++++++++--------- 1 file changed, 20 insertions(+), 9 deletions(-) diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c index 5b8b994b9912..2cbe5a321e89 100644 --- a/fs/gfs2/ops_inode.c +++ b/fs/gfs2/ops_inode.c @@ -278,17 +278,25 @@ static int gfs2_unlink(struct inode *dir, struct dentry *dentry) gfs2_holder_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, ghs + 2); - error = gfs2_glock_nq_m(3, ghs); + error = gfs2_glock_nq(ghs); /* parent */ if (error) - goto out; + goto out_parent; + + error = gfs2_glock_nq(ghs + 1); /* child */ + if (error) + goto out_child; + + error = gfs2_glock_nq(ghs + 2); /* rgrp */ + if (error) + goto out_rgrp; error = gfs2_unlink_ok(dip, &dentry->d_name, ip); if (error) - goto out_gunlock; + goto out_rgrp; error = gfs2_trans_begin(sdp, 2*RES_DINODE + RES_LEAF + RES_RG_BIT, 0); if (error) - goto out_gunlock; + goto out_rgrp; error = gfs2_dir_del(dip, &dentry->d_name); if (error) @@ -298,12 +306,15 @@ static int gfs2_unlink(struct inode *dir, struct dentry *dentry) out_end_trans: gfs2_trans_end(sdp); -out_gunlock: - gfs2_glock_dq_m(3, ghs); -out: - gfs2_holder_uninit(ghs); - gfs2_holder_uninit(ghs + 1); + gfs2_glock_dq(ghs + 2); +out_rgrp: gfs2_holder_uninit(ghs + 2); + gfs2_glock_dq(ghs + 1); +out_child: + gfs2_holder_uninit(ghs + 1); + gfs2_glock_dq(ghs); +out_parent: + gfs2_holder_uninit(ghs); gfs2_glock_dq_uninit(&ri_gh); return error; } -- cgit v1.2.3 From 1e1a3d03e927d39282208aed676e49d25129feea Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 27 Aug 2007 09:45:26 +0100 Subject: [GFS2] Introduce gfs2_remove_from_ail This collects together the operations required to remove a gfs2_bufdata from the ail lists. Its only called from two places to start with, but expect to see more of this function in future. Signed-off-by: Steven Whitehouse --- fs/gfs2/glops.c | 6 +----- fs/gfs2/log.c | 31 +++++++++++++++++++++---------- fs/gfs2/log.h | 1 + 3 files changed, 23 insertions(+), 15 deletions(-) diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index 7ef6b23bb38a..b17346a355bf 100644 --- a/fs/gfs2/glops.c +++ b/fs/gfs2/glops.c @@ -60,11 +60,7 @@ static void gfs2_ail_empty_gl(struct gfs2_glock *gl) blkno = bh->b_blocknr; gfs2_assert_withdraw(sdp, !buffer_busy(bh)); - bd->bd_ail = NULL; - list_del(&bd->bd_ail_st_list); - list_del(&bd->bd_ail_gl_list); - atomic_dec(&gl->gl_ail_count); - brelse(bh); + gfs2_remove_from_ail(NULL, bd); gfs2_log_unlock(sdp); gfs2_trans_add_revoke(sdp, blkno); diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index d0e6b42c86e1..d8232ec25397 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -59,6 +59,26 @@ unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct, return blks; } +/** + * gfs2_remove_from_ail - Remove an entry from the ail lists, updating counters + * @mapping: The associated mapping (maybe NULL) + * @bd: The gfs2_bufdata to remove + * + * The log lock _must_ be held when calling this function + * + */ + +void gfs2_remove_from_ail(struct address_space *mapping, struct gfs2_bufdata *bd) +{ + bd->bd_ail = NULL; + list_del(&bd->bd_ail_st_list); + list_del(&bd->bd_ail_gl_list); + atomic_dec(&bd->bd_gl->gl_ail_count); + if (mapping) + gfs2_meta_cache_flush(GFS2_I(mapping->host)); + brelse(bd->bd_bh); +} + /** * gfs2_ail1_start_one - Start I/O on a part of the AIL * @sdp: the filesystem @@ -219,21 +239,12 @@ 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, bd_ail_st_list); gfs2_assert(sdp, bd->bd_ail == ai); - bd->bd_ail = NULL; - list_del(&bd->bd_ail_st_list); - list_del(&bd->bd_ail_gl_list); - atomic_dec(&bd->bd_gl->gl_ail_count); - 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); + gfs2_remove_from_ail(bd->bd_bh->b_page->mapping, bd); } } diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h index 8e7aa0f29109..639423561b2d 100644 --- a/fs/gfs2/log.h +++ b/fs/gfs2/log.h @@ -58,6 +58,7 @@ struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, struct buffer_head *real); void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl); void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *trans); +void gfs2_remove_from_ail(struct address_space *mapping, struct gfs2_bufdata *bd); void gfs2_log_shutdown(struct gfs2_sbd *sdp); void gfs2_meta_syncfs(struct gfs2_sbd *sdp); -- cgit v1.2.3 From eaf965270ffff3086ef929e660ace45e862cfd2d Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 27 Aug 2007 09:49:37 +0100 Subject: [GFS2] Don't mark jdata dirty in gfs2_unstuffer_page() Journaled data is marked dirty by gfs2_unpin and should not be marked dirty here. Signed-off-by: Steven Whitehouse --- fs/gfs2/bmap.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index 9b8990444e6c..1e56f4de7358 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c @@ -95,7 +95,8 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh, set_buffer_uptodate(bh); if (sdp->sd_args.ar_data == GFS2_DATA_ORDERED || gfs2_is_jdata(ip)) gfs2_trans_add_bh(ip->i_gl, bh, 0); - mark_buffer_dirty(bh); + if (!gfs2_is_jdata(ip)) + mark_buffer_dirty(bh); if (release) { unlock_page(page); -- cgit v1.2.3 From 9b9107a5a8b190e6cf09bbdf893869c6a9c482cc Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 27 Aug 2007 13:54:05 +0100 Subject: [GFS2] Move pin/unpin into lops.c, clean up locking gfs2_pin and gfs2_unpin are only used in lops.c, despite being defined in meta_io.c, so this patch moves them into lops.c and makes them static. At the same time, its possible to clean up the locking in the buf and databuf _lo_add() functions so that we only need to grab the spinlock once. Also we have to move lock_buffer() around the _lo_add() functions since we can't do that in gfs2_pin() any more since we hold the spinlock for the duration of that function. As a result, the code shrinks by 12 lines and we do far fewer operations when adding buffers to the log. It also makes the code somewhat easier to read & understand. Signed-off-by: Steven Whitehouse --- fs/gfs2/lops.c | 117 +++++++++++++++++++++++++++++++++++++++++------------- fs/gfs2/meta_io.c | 70 -------------------------------- fs/gfs2/meta_io.h | 3 -- 3 files changed, 89 insertions(+), 101 deletions(-) diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 7ef335623373..3ec587135d46 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -27,7 +27,71 @@ #include "trans.h" #include "util.h" -static void glock_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) +/** + * gfs2_pin - Pin a buffer in memory + * @sdp: The superblock + * @bh: The buffer to be pinned + * + * The log lock must be held when calling this function + */ +static void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh) +{ + struct gfs2_bufdata *bd; + + gfs2_assert_withdraw(sdp, test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)); + + clear_buffer_dirty(bh); + if (test_set_buffer_pinned(bh)) + gfs2_assert_withdraw(sdp, 0); + if (!buffer_uptodate(bh)) + gfs2_io_error_bh(sdp, bh); + bd = bh->b_private; + /* If this buffer is in the AIL and it has already been written + * to in-place disk block, remove it from the AIL. + */ + if (bd->bd_ail) + list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list); + get_bh(bh); +} + +/** + * gfs2_unpin - Unpin a buffer + * @sdp: the filesystem the buffer belongs to + * @bh: The buffer to unpin + * @ai: + * + */ + +static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, + struct gfs2_ail *ai) +{ + struct gfs2_bufdata *bd = bh->b_private; + + gfs2_assert_withdraw(sdp, buffer_uptodate(bh)); + + if (!buffer_pinned(bh)) + gfs2_assert_withdraw(sdp, 0); + + lock_buffer(bh); + mark_buffer_dirty(bh); + clear_buffer_pinned(bh); + + gfs2_log_lock(sdp); + if (bd->bd_ail) { + list_del(&bd->bd_ail_st_list); + brelse(bh); + } else { + struct gfs2_glock *gl = bd->bd_gl; + list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list); + atomic_inc(&gl->gl_ail_count); + } + bd->bd_ail = ai; + list_add(&bd->bd_ail_st_list, &ai->ai_ail1_list); + gfs2_log_unlock(sdp); + unlock_buffer(bh); +} + +static void __glock_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) { struct gfs2_glock *gl; struct gfs2_trans *tr = current->journal_info; @@ -38,15 +102,19 @@ static void glock_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) if (gfs2_assert_withdraw(sdp, gfs2_glock_is_held_excl(gl))) return; - gfs2_log_lock(sdp); - if (!list_empty(&le->le_list)){ - gfs2_log_unlock(sdp); + if (!list_empty(&le->le_list)) return; - } + gfs2_glock_hold(gl); set_bit(GLF_DIRTY, &gl->gl_flags); sdp->sd_log_num_gl++; list_add(&le->le_list, &sdp->sd_log_le_gl); +} + +static void glock_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) +{ + gfs2_log_lock(sdp); + __glock_lo_add(sdp, le); gfs2_log_unlock(sdp); } @@ -71,30 +139,25 @@ static void buf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) struct gfs2_bufdata *bd = container_of(le, struct gfs2_bufdata, bd_le); struct gfs2_trans *tr; + lock_buffer(bd->bd_bh); gfs2_log_lock(sdp); - if (!list_empty(&bd->bd_list_tr)) { - gfs2_log_unlock(sdp); - return; - } + if (!list_empty(&bd->bd_list_tr)) + goto out; tr = current->journal_info; tr->tr_touched = 1; tr->tr_num_buf++; list_add(&bd->bd_list_tr, &tr->tr_list_buf); - gfs2_log_unlock(sdp); - if (!list_empty(&le->le_list)) - return; - - gfs2_trans_add_gl(bd->bd_gl); - + goto out; + __glock_lo_add(sdp, &bd->bd_gl->gl_le); gfs2_meta_check(sdp, bd->bd_bh); gfs2_pin(sdp, bd->bd_bh); - gfs2_log_lock(sdp); sdp->sd_log_num_buf++; list_add(&le->le_list, &sdp->sd_log_le_buf); - gfs2_log_unlock(sdp); - tr->tr_num_buf_new++; +out: + gfs2_log_unlock(sdp); + unlock_buffer(bd->bd_bh); } static void buf_lo_incore_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr) @@ -476,31 +539,29 @@ static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) struct address_space *mapping = bd->bd_bh->b_page->mapping; struct gfs2_inode *ip = GFS2_I(mapping->host); + lock_buffer(bd->bd_bh); gfs2_log_lock(sdp); - if (!list_empty(&bd->bd_list_tr)) { - gfs2_log_unlock(sdp); - return; - } + if (!list_empty(&bd->bd_list_tr)) + goto out; tr->tr_touched = 1; if (gfs2_is_jdata(ip)) { tr->tr_num_buf++; list_add(&bd->bd_list_tr, &tr->tr_list_buf); } - gfs2_log_unlock(sdp); if (!list_empty(&le->le_list)) - return; + goto out; - gfs2_trans_add_gl(bd->bd_gl); + __glock_lo_add(sdp, &bd->bd_gl->gl_le); if (gfs2_is_jdata(ip)) { 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); +out: gfs2_log_unlock(sdp); + unlock_buffer(bd->bd_bh); } static int gfs2_check_magic(struct buffer_head *bh) diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c index 8da343b34ae7..d762e4f7044e 100644 --- a/fs/gfs2/meta_io.c +++ b/fs/gfs2/meta_io.c @@ -297,76 +297,6 @@ void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh, unlock_page(bh->b_page); } -/** - * gfs2_pin - Pin a buffer in memory - * @sdp: the filesystem the buffer belongs to - * @bh: The buffer to be pinned - * - */ - -void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh) -{ - struct gfs2_bufdata *bd = bh->b_private; - - gfs2_assert_withdraw(sdp, test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)); - - if (test_set_buffer_pinned(bh)) - gfs2_assert_withdraw(sdp, 0); - - wait_on_buffer(bh); - - /* If this buffer is in the AIL and it has already been written - to in-place disk block, remove it from the AIL. */ - - gfs2_log_lock(sdp); - if (bd->bd_ail && !buffer_in_io(bh)) - list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list); - gfs2_log_unlock(sdp); - - clear_buffer_dirty(bh); - wait_on_buffer(bh); - - if (!buffer_uptodate(bh)) - gfs2_io_error_bh(sdp, bh); - - get_bh(bh); -} - -/** - * gfs2_unpin - Unpin a buffer - * @sdp: the filesystem the buffer belongs to - * @bh: The buffer to unpin - * @ai: - * - */ - -void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, - struct gfs2_ail *ai) -{ - struct gfs2_bufdata *bd = bh->b_private; - - gfs2_assert_withdraw(sdp, buffer_uptodate(bh)); - - if (!buffer_pinned(bh)) - gfs2_assert_withdraw(sdp, 0); - - mark_buffer_dirty(bh); - clear_buffer_pinned(bh); - - gfs2_log_lock(sdp); - if (bd->bd_ail) { - list_del(&bd->bd_ail_st_list); - brelse(bh); - } else { - struct gfs2_glock *gl = bd->bd_gl; - list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list); - atomic_inc(&gl->gl_ail_count); - } - bd->bd_ail = ai; - list_add(&bd->bd_ail_st_list, &ai->ai_ail1_list); - gfs2_log_unlock(sdp); -} - /** * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore * @ip: the inode who owns the buffers diff --git a/fs/gfs2/meta_io.h b/fs/gfs2/meta_io.h index 527bf19d9690..fd67f07cd60e 100644 --- a/fs/gfs2/meta_io.h +++ b/fs/gfs2/meta_io.h @@ -50,9 +50,6 @@ int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh); void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh, int meta); -void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh); -void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, - struct gfs2_ail *ai); void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen); -- cgit v1.2.3 From d7b616e252b125f12b007c392f7644053bb6f140 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Sun, 2 Sep 2007 10:48:13 +0100 Subject: [GFS2] Clean up ordered write code The following patch removes the ordered write processing from databuf_lo_before_commit() and moves it to log.c. This has the effect of greatly simplyfying databuf_lo_before_commit() and well as potentially making the ordered write code more efficient. As a side effect of this, its now possible to remove ordered buffers from the ordered buffer list at any time, so we now make use of this in invalidatepage and releasepage to ensure timely release of these buffers. Signed-off-by: Steven Whitehouse --- fs/gfs2/incore.h | 2 +- fs/gfs2/log.c | 57 +++++++++++++++++- fs/gfs2/lops.c | 160 +++++++++++++------------------------------------- fs/gfs2/ops_address.c | 50 ++++++++++++++-- fs/gfs2/ops_fstype.c | 1 + 5 files changed, 143 insertions(+), 127 deletions(-) diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 23b611aa70d2..388dc1bd736f 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -612,13 +612,13 @@ struct gfs2_sbd { unsigned int sd_log_num_revoke; unsigned int sd_log_num_rg; unsigned int sd_log_num_databuf; - unsigned int sd_log_num_jdata; struct list_head sd_log_le_gl; struct list_head sd_log_le_buf; struct list_head sd_log_le_revoke; struct list_head sd_log_le_rg; struct list_head sd_log_le_databuf; + struct list_head sd_log_le_ordered; unsigned int sd_log_blks_free; struct mutex sd_log_reserve_mutex; diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index d8232ec25397..20fa528d457d 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -620,6 +620,57 @@ static void log_flush_commit(struct gfs2_sbd *sdp) } } +static void gfs2_ordered_write(struct gfs2_sbd *sdp) +{ + struct gfs2_bufdata *bd; + struct buffer_head *bh; + LIST_HEAD(written); + + gfs2_log_lock(sdp); + while (!list_empty(&sdp->sd_log_le_ordered)) { + bd = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_bufdata, bd_le.le_list); + list_move(&bd->bd_le.le_list, &written); + bh = bd->bd_bh; + if (!buffer_dirty(bh)) + continue; + get_bh(bh); + gfs2_log_unlock(sdp); + lock_buffer(bh); + if (test_clear_buffer_dirty(bh)) { + bh->b_end_io = end_buffer_write_sync; + submit_bh(WRITE, bh); + } else { + unlock_buffer(bh); + brelse(bh); + } + gfs2_log_lock(sdp); + } + list_splice(&written, &sdp->sd_log_le_ordered); + gfs2_log_unlock(sdp); +} + +static void gfs2_ordered_wait(struct gfs2_sbd *sdp) +{ + struct gfs2_bufdata *bd; + struct buffer_head *bh; + + gfs2_log_lock(sdp); + while (!list_empty(&sdp->sd_log_le_ordered)) { + bd = list_entry(sdp->sd_log_le_ordered.prev, struct gfs2_bufdata, bd_le.le_list); + bh = bd->bd_bh; + if (buffer_locked(bh)) { + get_bh(bh); + gfs2_log_unlock(sdp); + wait_on_buffer(bh); + brelse(bh); + gfs2_log_lock(sdp); + continue; + } + list_del_init(&bd->bd_le.le_list); + } + gfs2_log_unlock(sdp); +} + /** * gfs2_log_flush - flush incore transaction(s) * @sdp: the filesystem @@ -648,7 +699,7 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl) INIT_LIST_HEAD(&ai->ai_ail2_list); gfs2_assert_withdraw(sdp, - sdp->sd_log_num_buf + sdp->sd_log_num_jdata == + sdp->sd_log_num_buf + sdp->sd_log_num_databuf == sdp->sd_log_commited_buf + sdp->sd_log_commited_databuf); gfs2_assert_withdraw(sdp, @@ -658,7 +709,10 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl) sdp->sd_log_flush_wrapped = 0; ai->ai_first = sdp->sd_log_flush_head; + gfs2_ordered_write(sdp); lops_before_commit(sdp); + gfs2_ordered_wait(sdp); + if (!list_empty(&sdp->sd_log_flush_list)) log_flush_commit(sdp); else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){ @@ -751,7 +805,6 @@ void gfs2_log_shutdown(struct gfs2_sbd *sdp) gfs2_assert_withdraw(sdp, !sdp->sd_log_blks_reserved); gfs2_assert_withdraw(sdp, !sdp->sd_log_num_gl); gfs2_assert_withdraw(sdp, !sdp->sd_log_num_buf); - gfs2_assert_withdraw(sdp, !sdp->sd_log_num_jdata); gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke); gfs2_assert_withdraw(sdp, !sdp->sd_log_num_rg); gfs2_assert_withdraw(sdp, !sdp->sd_log_num_databuf); diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 3ec587135d46..7e2d4e692b50 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -555,10 +555,11 @@ static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) if (gfs2_is_jdata(ip)) { gfs2_pin(sdp, bd->bd_bh); tr->tr_num_databuf_new++; - sdp->sd_log_num_jdata++; + sdp->sd_log_num_databuf++; + list_add(&le->le_list, &sdp->sd_log_le_databuf); + } else { + list_add(&le->le_list, &sdp->sd_log_le_ordered); } - sdp->sd_log_num_databuf++; - list_add(&le->le_list, &sdp->sd_log_le_databuf); out: gfs2_log_unlock(sdp); unlock_buffer(bd->bd_bh); @@ -583,114 +584,59 @@ static int gfs2_check_magic(struct buffer_head *bh) /** * databuf_lo_before_commit - Scan the data buffers, writing as we go * - * Here we scan through the lists of buffers and make the assumption - * that any buffer thats been pinned is being journaled, and that - * any unpinned buffer is an ordered write data buffer and therefore - * will be written back rather than journaled. */ + static void databuf_lo_before_commit(struct gfs2_sbd *sdp) { - LIST_HEAD(started); - struct gfs2_bufdata *bd1 = NULL, *bd2, *bdt; + struct gfs2_bufdata *bd1 = NULL, *bd2; struct buffer_head *bh = NULL,*bh1 = NULL; struct gfs2_log_descriptor *ld; unsigned int limit; - unsigned int total_dbuf; - unsigned int total_jdata; + unsigned int total; unsigned int num, n; __be64 *ptr = NULL; + int magic; + limit = databuf_limit(sdp); - /* - * Start writing ordered buffers, write journaled buffers - * into the log along with a header - */ gfs2_log_lock(sdp); - total_dbuf = sdp->sd_log_num_databuf; - total_jdata = sdp->sd_log_num_jdata; + total = sdp->sd_log_num_databuf; bd2 = bd1 = list_prepare_entry(bd1, &sdp->sd_log_le_databuf, bd_le.le_list); - while(total_dbuf) { - num = total_jdata; + while(total) { + num = total; if (num > 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 + DATABUF_OFFSET); + ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); + ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); + ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); + ld->ld_type = cpu_to_be32(GFS2_LOG_DESC_JDATA); + ld->ld_length = cpu_to_be32(num + 1); + ld->ld_data1 = cpu_to_be32(num); + ld->ld_data2 = cpu_to_be32(0); + memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); + n = 0; - list_for_each_entry_safe_continue(bd1, bdt, - &sdp->sd_log_le_databuf, - bd_le.le_list) { - /* store off the buffer head in a local ptr since - * gfs2_bufdata might change when we drop the log lock - */ + list_for_each_entry_continue(bd1, &sdp->sd_log_le_databuf, + bd_le.le_list) { bh1 = bd1->bd_bh; - /* An ordered write buffer */ - if (bh1 && !buffer_pinned(bh1)) { - list_move(&bd1->bd_le.le_list, &started); - if (bd1 == bd2) { - bd2 = NULL; - bd2 = list_prepare_entry(bd2, - &sdp->sd_log_le_databuf, - bd_le.le_list); - } - total_dbuf--; - if (bh1) { - if (buffer_dirty(bh1)) { - get_bh(bh1); - - gfs2_log_unlock(sdp); - - ll_rw_block(SWRITE, 1, &bh1); - brelse(bh1); - - gfs2_log_lock(sdp); - } - continue; - } - continue; - } else if (bh1) { /* A journaled buffer */ - int magic; - gfs2_log_unlock(sdp); - if (!bh) { - bh = gfs2_log_get_buf(sdp); - ld = (struct gfs2_log_descriptor *) - bh->b_data; - ptr = (__be64 *)(bh->b_data + - DATABUF_OFFSET); - ld->ld_header.mh_magic = - cpu_to_be32(GFS2_MAGIC); - ld->ld_header.mh_type = - cpu_to_be32(GFS2_METATYPE_LD); - ld->ld_header.mh_format = - cpu_to_be32(GFS2_FORMAT_LD); - ld->ld_type = - cpu_to_be32(GFS2_LOG_DESC_JDATA); - ld->ld_length = cpu_to_be32(num + 1); - ld->ld_data1 = cpu_to_be32(num); - ld->ld_data2 = cpu_to_be32(0); - memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); - } - magic = gfs2_check_magic(bh1); - *ptr++ = cpu_to_be64(bh1->b_blocknr); - *ptr++ = cpu_to_be64((__u64)magic); - clear_buffer_escaped(bh1); - if (unlikely(magic != 0)) - set_buffer_escaped(bh1); - gfs2_log_lock(sdp); - if (++n >= num) - break; - } else if (!bh1) { - total_dbuf--; - sdp->sd_log_num_databuf--; - list_del_init(&bd1->bd_le.le_list); - if (bd1 == bd2) { - bd2 = NULL; - bd2 = list_prepare_entry(bd2, - &sdp->sd_log_le_databuf, - bd_le.le_list); - } - kmem_cache_free(gfs2_bufdata_cachep, bd1); - } + magic = gfs2_check_magic(bh1); + *ptr++ = cpu_to_be64(bh1->b_blocknr); + *ptr++ = cpu_to_be64((__u64)magic); + clear_buffer_escaped(bh1); + if (unlikely(magic != 0)) + set_buffer_escaped(bh1); + if (++n >= num) + break; } gfs2_log_unlock(sdp); if (bh) { @@ -727,34 +673,10 @@ static void databuf_lo_before_commit(struct gfs2_sbd *sdp) break; } bh = NULL; - BUG_ON(total_dbuf < num); - total_dbuf -= num; - total_jdata -= num; + BUG_ON(total < num); + total -= num; } gfs2_log_unlock(sdp); - - /* Wait on all ordered buffers */ - while (!list_empty(&started)) { - gfs2_log_lock(sdp); - bd1 = list_entry(started.next, struct gfs2_bufdata, - bd_le.le_list); - list_del_init(&bd1->bd_le.le_list); - sdp->sd_log_num_databuf--; - bh = bd1->bd_bh; - if (bh) { - bh->b_private = NULL; - get_bh(bh); - gfs2_log_unlock(sdp); - wait_on_buffer(bh); - brelse(bh); - } else - gfs2_log_unlock(sdp); - - kmem_cache_free(gfs2_bufdata_cachep, bd1); - } - - /* We've removed all the ordered write bufs here, so only jdata left */ - gfs2_assert_warn(sdp, sdp->sd_log_num_databuf == sdp->sd_log_num_jdata); } static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, @@ -838,11 +760,9 @@ static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai) bd = list_entry(head->next, struct gfs2_bufdata, bd_le.le_list); list_del_init(&bd->bd_le.le_list); sdp->sd_log_num_databuf--; - sdp->sd_log_num_jdata--; gfs2_unpin(sdp, bd->bd_bh, ai); } gfs2_assert_warn(sdp, !sdp->sd_log_num_databuf); - gfs2_assert_warn(sdp, !sdp->sd_log_num_jdata); } diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c index 8407d1db4eac..dd1ea491ddcb 100644 --- a/fs/gfs2/ops_address.c +++ b/fs/gfs2/ops_address.c @@ -616,13 +616,50 @@ static sector_t gfs2_bmap(struct address_space *mapping, sector_t lblock) return dblock; } +static void gfs2_discard(struct gfs2_sbd *sdp, struct buffer_head *bh) +{ + struct gfs2_bufdata *bd; + + lock_buffer(bh); + gfs2_log_lock(sdp); + clear_buffer_dirty(bh); + bd = bh->b_private; + if (bd) { + if (!list_empty(&bd->bd_le.le_list)) { + if (!buffer_pinned(bh)) + list_del_init(&bd->bd_le.le_list); + } + } + bh->b_bdev = NULL; + clear_buffer_mapped(bh); + clear_buffer_req(bh); + clear_buffer_new(bh); + gfs2_log_unlock(sdp); + unlock_buffer(bh); +} + static void gfs2_invalidatepage(struct page *page, unsigned long offset) { + struct gfs2_sbd *sdp = GFS2_SB(page->mapping->host); + struct buffer_head *bh, *head; + unsigned long pos = 0; + BUG_ON(!PageLocked(page)); if (offset == 0) ClearPageChecked(page); + if (!page_has_buffers(page)) + goto out; - block_invalidatepage(page, offset); + bh = head = page_buffers(page); + do { + if (offset <= pos) + gfs2_discard(sdp, bh); + pos += bh->b_size; + bh = bh->b_this_page; + } while (bh != head); +out: + if (offset == 0) + try_to_release_page(page, 0); } /** @@ -732,9 +769,14 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask) if (bd) { gfs2_assert_warn(sdp, bd->bd_bh == bh); gfs2_assert_warn(sdp, list_empty(&bd->bd_list_tr)); - bd->bd_bh = NULL; - if (!list_empty(&bd->bd_le.le_list)) - bd = NULL; + if (!list_empty(&bd->bd_le.le_list)) { + if (!buffer_pinned(bh)) + list_del_init(&bd->bd_le.le_list); + else + bd = NULL; + } + if (bd) + bd->bd_bh = NULL; bh->b_private = NULL; } gfs2_log_unlock(sdp); diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 314c1134d12d..35f3dfa9bfb1 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -82,6 +82,7 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb) INIT_LIST_HEAD(&sdp->sd_log_le_revoke); INIT_LIST_HEAD(&sdp->sd_log_le_rg); INIT_LIST_HEAD(&sdp->sd_log_le_databuf); + INIT_LIST_HEAD(&sdp->sd_log_le_ordered); mutex_init(&sdp->sd_log_reserve_mutex); INIT_LIST_HEAD(&sdp->sd_ail1_list); -- cgit v1.2.3 From 8475487befb29eeb038fef374a7433d276336a25 Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Sun, 2 Sep 2007 10:55:29 +0100 Subject: [GFS2] Fix ordering of dirty/journal for ordered buffer unstuffing Signed-off-by: Bob Peterson Signed-off-by: Steven Whitehouse --- fs/gfs2/bmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index 1e56f4de7358..93fa427bb5f5 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c @@ -93,10 +93,10 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh, map_bh(bh, inode->i_sb, block); set_buffer_uptodate(bh); - if (sdp->sd_args.ar_data == GFS2_DATA_ORDERED || gfs2_is_jdata(ip)) - gfs2_trans_add_bh(ip->i_gl, bh, 0); if (!gfs2_is_jdata(ip)) mark_buffer_dirty(bh); + if (sdp->sd_args.ar_data == GFS2_DATA_ORDERED || gfs2_is_jdata(ip)) + gfs2_trans_add_bh(ip->i_gl, bh, 0); if (release) { unlock_page(page); -- cgit v1.2.3 From 82e86087bb774cd54d47db4a7c771b5b29bea9ed Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Sun, 2 Sep 2007 15:39:43 +0100 Subject: [GFS2] Replace revoke structure with bufdata structure Both the revoke structure and the bufdata structure are quite similar. They are basically small tags which are put on lists. In addition to which the revoke structure is always allocated when there is a bufdata structure which is (or can be) freed. As such it should be possible to reduce the number of frees and allocations by using the same structure for both purposes. This patch is the first step along that path. It replaces existing uses of the revoke structure with the bufdata structure. Signed-off-by: Steven Whitehouse --- fs/gfs2/incore.h | 13 +++++++------ fs/gfs2/lops.c | 10 +++++----- fs/gfs2/trans.c | 20 ++++++++++---------- 3 files changed, 22 insertions(+), 21 deletions(-) diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 388dc1bd736f..8aa5780862be 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -114,7 +114,13 @@ struct gfs2_bufdata { struct buffer_head *bd_bh; struct gfs2_glock *bd_gl; - struct list_head bd_list_tr; + union { + struct list_head list_tr; + u64 blkno; + } u; +#define bd_list_tr u.list_tr +#define bd_blkno u.blkno + struct gfs2_log_element bd_le; struct gfs2_ail *bd_ail; @@ -298,11 +304,6 @@ struct gfs2_file { struct gfs2_holder f_fl_gh; }; -struct gfs2_revoke { - struct gfs2_log_element rv_le; - u64 rv_blkno; -}; - struct gfs2_revoke_replay { struct list_head rr_list; u64 rr_blkno; diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 7e2d4e692b50..cf6fe3631554 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -357,7 +357,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) struct buffer_head *bh; unsigned int offset; struct list_head *head = &sdp->sd_log_le_revoke; - struct gfs2_revoke *rv; + struct gfs2_bufdata *bd; if (!sdp->sd_log_num_revoke) return; @@ -376,8 +376,8 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) offset = sizeof(struct gfs2_log_descriptor); while (!list_empty(head)) { - rv = list_entry(head->next, struct gfs2_revoke, rv_le.le_list); - list_del_init(&rv->rv_le.le_list); + bd = list_entry(head->next, struct gfs2_bufdata, bd_le.le_list); + list_del_init(&bd->bd_le.le_list); sdp->sd_log_num_revoke--; if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) { @@ -392,8 +392,8 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) offset = sizeof(struct gfs2_meta_header); } - *(__be64 *)(bh->b_data + offset) = cpu_to_be64(rv->rv_blkno); - kfree(rv); + *(__be64 *)(bh->b_data + offset) = cpu_to_be64(bd->bd_blkno); + kfree(bd); offset += sizeof(u64); } diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c index f8dabf8446bb..eadf96e00510 100644 --- a/fs/gfs2/trans.c +++ b/fs/gfs2/trans.c @@ -144,23 +144,23 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta) void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, u64 blkno) { - struct gfs2_revoke *rv = kmalloc(sizeof(struct gfs2_revoke), - GFP_NOFS | __GFP_NOFAIL); - lops_init_le(&rv->rv_le, &gfs2_revoke_lops); - rv->rv_blkno = blkno; - lops_add(sdp, &rv->rv_le); + struct gfs2_bufdata *bd = kmalloc(sizeof(struct gfs2_bufdata), + GFP_NOFS | __GFP_NOFAIL); + lops_init_le(&bd->bd_le, &gfs2_revoke_lops); + bd->bd_blkno = blkno; + lops_add(sdp, &bd->bd_le); } void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno) { - struct gfs2_revoke *rv; + struct gfs2_bufdata *bd; int found = 0; gfs2_log_lock(sdp); - list_for_each_entry(rv, &sdp->sd_log_le_revoke, rv_le.le_list) { - if (rv->rv_blkno == blkno) { - list_del(&rv->rv_le.le_list); + list_for_each_entry(bd, &sdp->sd_log_le_revoke, bd_le.le_list) { + if (bd->bd_blkno == blkno) { + list_del(&bd->bd_le.le_list); gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke); sdp->sd_log_num_revoke--; found = 1; @@ -172,7 +172,7 @@ void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno) if (found) { struct gfs2_trans *tr = current->journal_info; - kfree(rv); + kfree(bd); tr->tr_num_revoke_rm++; } } -- cgit v1.2.3 From 0820ab517e1b100ee3f9584ec27f93309689ebe7 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Sun, 2 Sep 2007 16:47:38 +0100 Subject: [GFS2] Use slab operations for all gfs2_bufdata allocations The old revoke structure was allocated using kalloc/kfree but there is a slab cache for gfs2_bufdata, so we should use that now that the structures have been converted. This is part two of the patch series to merge the revoke and gfs2_bufdata structures. Signed-off-by: Steven Whitehouse --- fs/gfs2/lops.c | 2 +- fs/gfs2/trans.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index cf6fe3631554..4cbef4c23a6b 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -393,7 +393,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) } *(__be64 *)(bh->b_data + offset) = cpu_to_be64(bd->bd_blkno); - kfree(bd); + kmem_cache_free(gfs2_bufdata_cachep, bd); offset += sizeof(u64); } diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c index eadf96e00510..01cc27fefd84 100644 --- a/fs/gfs2/trans.c +++ b/fs/gfs2/trans.c @@ -144,7 +144,7 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta) void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, u64 blkno) { - struct gfs2_bufdata *bd = kmalloc(sizeof(struct gfs2_bufdata), + struct gfs2_bufdata *bd = kmem_cache_alloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL); lops_init_le(&bd->bd_le, &gfs2_revoke_lops); bd->bd_blkno = blkno; @@ -172,7 +172,7 @@ void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno) if (found) { struct gfs2_trans *tr = current->journal_info; - kfree(bd); + kmem_cache_free(gfs2_bufdata_cachep, bd); tr->tr_num_revoke_rm++; } } -- cgit v1.2.3 From 1ad38c437fa33f85ba4b6a85ea8c5478ee72d5bd Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 3 Sep 2007 11:01:33 +0100 Subject: [GFS2] Clean up gfs2_trans_add_revoke() The following alters gfs2_trans_add_revoke() to take a struct gfs2_bufdata as an argument. This eliminates the memory allocation which was previously required by making use of the already existing struct gfs2_bufdata. It makes some sanity checks to ensure that the gfs2_bufdata has been removed from all the lists before its recycled as a revoke structure. This saves one memory allocation and one free per revoke structure. Also as a result, and to simplify the locking, since there is no longer any blocking code in gfs2_trans_add_revoke() we must hold the log lock whenever this function is called. This reduces the amount of times we take and unlock the log lock. Signed-off-by: Steven Whitehouse --- fs/gfs2/glops.c | 14 +++++--------- fs/gfs2/log.c | 4 ++-- fs/gfs2/lops.c | 3 --- fs/gfs2/meta_io.c | 35 ++++++++++++----------------------- fs/gfs2/trans.c | 10 +++++----- fs/gfs2/trans.h | 2 +- 6 files changed, 25 insertions(+), 43 deletions(-) diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index b17346a355bf..4670dcb2a877 100644 --- a/fs/gfs2/glops.c +++ b/fs/gfs2/glops.c @@ -41,7 +41,6 @@ static void gfs2_ail_empty_gl(struct gfs2_glock *gl) struct list_head *head = &gl->gl_ail_list; struct gfs2_bufdata *bd; struct buffer_head *bh; - u64 blkno; int error; blocks = atomic_read(&gl->gl_ail_count); @@ -57,15 +56,12 @@ static void gfs2_ail_empty_gl(struct gfs2_glock *gl) bd = list_entry(head->next, struct gfs2_bufdata, bd_ail_gl_list); bh = bd->bd_bh; - blkno = bh->b_blocknr; - gfs2_assert_withdraw(sdp, !buffer_busy(bh)); - gfs2_remove_from_ail(NULL, bd); - gfs2_log_unlock(sdp); - - gfs2_trans_add_revoke(sdp, blkno); - - gfs2_log_lock(sdp); + 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); } gfs2_assert_withdraw(sdp, !atomic_read(&gl->gl_ail_count)); gfs2_log_unlock(sdp); diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index 20fa528d457d..4d04e6f19706 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -71,8 +71,8 @@ unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct, void gfs2_remove_from_ail(struct address_space *mapping, struct gfs2_bufdata *bd) { bd->bd_ail = NULL; - list_del(&bd->bd_ail_st_list); - list_del(&bd->bd_ail_gl_list); + list_del_init(&bd->bd_ail_st_list); + list_del_init(&bd->bd_ail_gl_list); atomic_dec(&bd->bd_gl->gl_ail_count); if (mapping) gfs2_meta_cache_flush(GFS2_I(mapping->host)); diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 4cbef4c23a6b..342c10e12af2 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -343,11 +343,8 @@ static void revoke_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) tr = current->journal_info; tr->tr_touched = 1; tr->tr_num_revoke++; - - gfs2_log_lock(sdp); sdp->sd_log_num_revoke++; list_add(&le->le_list, &sdp->sd_log_le_revoke); - gfs2_log_unlock(sdp); } static void revoke_lo_before_commit(struct gfs2_sbd *sdp) diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c index d762e4f7044e..19097bc7c81d 100644 --- a/fs/gfs2/meta_io.c +++ b/fs/gfs2/meta_io.c @@ -313,42 +313,31 @@ void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen) while (blen) { bh = getbuf(ip->i_gl, bstart, NO_CREATE); if (bh) { - struct gfs2_bufdata *bd = bh->b_private; + struct gfs2_bufdata *bd; + lock_buffer(bh); + gfs2_log_lock(sdp); + bd = bh->b_private; if (test_clear_buffer_pinned(bh)) { struct gfs2_trans *tr = current->journal_info; - struct gfs2_inode *bh_ip = - GFS2_I(bh->b_page->mapping->host); - - gfs2_log_lock(sdp); list_del_init(&bd->bd_le.le_list); gfs2_assert_warn(sdp, sdp->sd_log_num_buf); sdp->sd_log_num_buf--; - gfs2_log_unlock(sdp); - if (bh_ip->i_inode.i_private != NULL) - tr->tr_num_databuf_rm++; - else - tr->tr_num_buf_rm++; + tr->tr_num_buf_rm++; brelse(bh); } if (bd) { - gfs2_log_lock(sdp); if (bd->bd_ail) { - u64 blkno = bh->b_blocknr; - bd->bd_ail = NULL; - list_del(&bd->bd_ail_st_list); - list_del(&bd->bd_ail_gl_list); - atomic_dec(&bd->bd_gl->gl_ail_count); - brelse(bh); - gfs2_log_unlock(sdp); - gfs2_trans_add_revoke(sdp, blkno); - } else - gfs2_log_unlock(sdp); + gfs2_remove_from_ail(NULL, bd); + bh->b_private = NULL; + bd->bd_bh = NULL; + bd->bd_blkno = bh->b_blocknr; + gfs2_trans_add_revoke(sdp, bd); + } } - - lock_buffer(bh); clear_buffer_dirty(bh); clear_buffer_uptodate(bh); + gfs2_log_unlock(sdp); unlock_buffer(bh); brelse(bh); diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c index 01cc27fefd84..717983e2c2ae 100644 --- a/fs/gfs2/trans.c +++ b/fs/gfs2/trans.c @@ -142,12 +142,12 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta) lops_add(sdp, &bd->bd_le); } -void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, u64 blkno) +void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd) { - struct gfs2_bufdata *bd = kmem_cache_alloc(gfs2_bufdata_cachep, - GFP_NOFS | __GFP_NOFAIL); + BUG_ON(!list_empty(&bd->bd_le.le_list)); + BUG_ON(!list_empty(&bd->bd_ail_st_list)); + BUG_ON(!list_empty(&bd->bd_ail_gl_list)); lops_init_le(&bd->bd_le, &gfs2_revoke_lops); - bd->bd_blkno = blkno; lops_add(sdp, &bd->bd_le); } @@ -160,7 +160,7 @@ void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno) list_for_each_entry(bd, &sdp->sd_log_le_revoke, bd_le.le_list) { if (bd->bd_blkno == blkno) { - list_del(&bd->bd_le.le_list); + list_del_init(&bd->bd_le.le_list); gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke); sdp->sd_log_num_revoke--; found = 1; diff --git a/fs/gfs2/trans.h b/fs/gfs2/trans.h index 23d4cbe1de5b..043d5f4b9c4c 100644 --- a/fs/gfs2/trans.h +++ b/fs/gfs2/trans.h @@ -32,7 +32,7 @@ void gfs2_trans_end(struct gfs2_sbd *sdp); void gfs2_trans_add_gl(struct gfs2_glock *gl); void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta); -void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, u64 blkno); +void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd); void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno); void gfs2_trans_add_rg(struct gfs2_rgrpd *rgd); -- cgit v1.2.3 From b4c20166dcfca106f0f416bfce200099ed76ab18 Mon Sep 17 00:00:00 2001 From: Abhijith Das Date: Thu, 13 Sep 2007 23:35:27 -0500 Subject: [GFS2] flocks from same process trip kernel BUG at fs/gfs2/glock.c:1118! This patch adds a new flag to the gfs2_holder structure GL_FLOCK. It is set on holders of glocks representing flocks. This flag is checked in add_to_queue() and a process is permitted to queue more than one holder onto a glock if it is set. This solves the issue of a process not being able to do multiple flocks on the same file. Through a single descriptor, a process can now promote and demote flocks. Through multiple descriptors a process can now queue multiple flocks on the same file. There's still the problem of a process deadlocking itself (because gfs2 blocking locks are not interruptible) by queueing incompatible deadlock. Signed-off-by: Abhijith Das Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 43 +++++++++++++++++++++++++------------------ fs/gfs2/glock.h | 1 + fs/gfs2/ops_file.c | 13 ++++++------- 3 files changed, 32 insertions(+), 25 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 931368a385c8..d631cad0aeee 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -1106,24 +1106,31 @@ static void add_to_queue(struct gfs2_holder *gh) if (test_and_set_bit(HIF_WAIT, &gh->gh_iflags)) BUG(); - existing = find_holder_by_owner(&gl->gl_holders, gh->gh_owner_pid); - if (existing) { - print_symbol(KERN_WARNING "original: %s\n", existing->gh_ip); - printk(KERN_INFO "pid : %d\n", existing->gh_owner_pid); - printk(KERN_INFO "lock type : %d lock state : %d\n", - existing->gh_gl->gl_name.ln_type, existing->gh_gl->gl_state); - print_symbol(KERN_WARNING "new: %s\n", gh->gh_ip); - printk(KERN_INFO "pid : %d\n", gh->gh_owner_pid); - printk(KERN_INFO "lock type : %d lock state : %d\n", - gl->gl_name.ln_type, gl->gl_state); - BUG(); - } - - existing = find_holder_by_owner(&gl->gl_waiters3, gh->gh_owner_pid); - if (existing) { - print_symbol(KERN_WARNING "original: %s\n", existing->gh_ip); - print_symbol(KERN_WARNING "new: %s\n", gh->gh_ip); - BUG(); + if (!(gh->gh_flags & GL_FLOCK)) { + existing = find_holder_by_owner(&gl->gl_holders, + gh->gh_owner_pid); + if (existing) { + print_symbol(KERN_WARNING "original: %s\n", + existing->gh_ip); + printk(KERN_INFO "pid : %d\n", existing->gh_owner_pid); + printk(KERN_INFO "lock type : %d lock state : %d\n", + existing->gh_gl->gl_name.ln_type, + existing->gh_gl->gl_state); + print_symbol(KERN_WARNING "new: %s\n", gh->gh_ip); + printk(KERN_INFO "pid : %d\n", gh->gh_owner_pid); + printk(KERN_INFO "lock type : %d lock state : %d\n", + gl->gl_name.ln_type, gl->gl_state); + BUG(); + } + + existing = find_holder_by_owner(&gl->gl_waiters3, + gh->gh_owner_pid); + if (existing) { + print_symbol(KERN_WARNING "original: %s\n", + existing->gh_ip); + print_symbol(KERN_WARNING "new: %s\n", gh->gh_ip); + BUG(); + } } if (gh->gh_flags & LM_FLAG_PRIORITY) diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h index f7a8e626aa08..b16f604eea9f 100644 --- a/fs/gfs2/glock.h +++ b/fs/gfs2/glock.h @@ -26,6 +26,7 @@ #define GL_SKIP 0x00000100 #define GL_ATIME 0x00000200 #define GL_NOCACHE 0x00000400 +#define GL_FLOCK 0x00000800 #define GL_NOCANCEL 0x00001000 #define GLR_TRYFAILED 13 diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c index 94d76ace0b95..46a9e10ff17b 100644 --- a/fs/gfs2/ops_file.c +++ b/fs/gfs2/ops_file.c @@ -571,7 +571,8 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl) int error = 0; state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED; - flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE; + flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE + | GL_FLOCK; mutex_lock(&fp->f_fl_mutex); @@ -579,21 +580,19 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl) if (gl) { if (fl_gh->gh_state == state) goto out; - gfs2_glock_hold(gl); flock_lock_file_wait(file, &(struct file_lock){.fl_type = F_UNLCK}); - gfs2_glock_dq_uninit(fl_gh); + gfs2_glock_dq_wait(fl_gh); + gfs2_holder_reinit(state, flags, fl_gh); } else { error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr, &gfs2_flock_glops, CREATE, &gl); if (error) goto out; + gfs2_holder_init(gl, state, flags, fl_gh); + gfs2_glock_put(gl); } - - gfs2_holder_init(gl, state, flags, fl_gh); - gfs2_glock_put(gl); - error = gfs2_glock_nq(fl_gh); if (error) { gfs2_holder_uninit(fl_gh); -- cgit v1.2.3 From 49e61f2ef6f7d1d0296e3e30d366b28e0ca595c2 Mon Sep 17 00:00:00 2001 From: Wendy Cheng Date: Thu, 13 Sep 2007 17:52:42 -0400 Subject: [GFS2] Move inode deletion out of blocking_cb Move inode deletion code out of blocking_cb handle_callback route to avoid racy conditions that end up blocking lock_dlm1 thread. Fix bugzilla 286821. Signed-off-by: Wendy Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/glock.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index d631cad0aeee..a37efe4aae6f 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -716,12 +716,8 @@ static void handle_callback(struct gfs2_glock *gl, unsigned int state, gl->gl_demote_time = jiffies; if (remote && gl->gl_ops->go_type == LM_TYPE_IOPEN && gl->gl_object) { - struct inode *inode = igrab(gl->gl_object); + gfs2_glock_schedule_for_reclaim(gl); spin_unlock(&gl->gl_spin); - if (inode) { - d_prune_aliases(inode); - iput(inode); - } return; } } else if (gl->gl_demote_state != LM_ST_UNLOCKED && -- cgit v1.2.3 From d66f8277f53407754f50ae6bada68f1b68d04d48 Mon Sep 17 00:00:00 2001 From: Patrick Caulfield Date: Fri, 14 Sep 2007 08:49:21 +0100 Subject: [DLM] Make dlm_sendd cond_resched more Under high recovery loads dlm_sendd can monopolise the CPU and cause soft lockups. This one extra and one moved cond_resched() make it yield a little more during such times keeping work moving. Signed-Off-By: Patrick Caulfield Signed-off-by: Steven Whitehouse --- fs/dlm/lowcomms.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c index 62a8a6ccd992..58bf3f5cdbe2 100644 --- a/fs/dlm/lowcomms.c +++ b/fs/dlm/lowcomms.c @@ -1264,14 +1264,15 @@ static void send_to_sock(struct connection *con) if (len) { ret = sendpage(con->sock, e->page, offset, len, msg_flags); - if (ret == -EAGAIN || ret == 0) + if (ret == -EAGAIN || ret == 0) { + cond_resched(); goto out; + } if (ret <= 0) goto send_error; - } else { + } /* Don't starve people filling buffers */ cond_resched(); - } spin_lock(&con->writequeue_lock); e->offset += ret; -- cgit v1.2.3 From 55c0c4ac0be144014651b19e77c9b77f367955de Mon Sep 17 00:00:00 2001 From: Bob Peterson Date: Fri, 14 Sep 2007 09:27:59 -0500 Subject: [GFS2] GFS2: chmod hung - fix race in thread creation The problem boiled down to a race between the gdlm_init_threads() function initializing thread1 and its setting of blist = 1. Essentially, "if (current == ls->thread1)" was checked by the thread before the thread creator set ls->thread1. Since thread1 is the only thread who is allowed to work on the blocking queue, and since neither thread thought it was thread1, no one was working on the queue. So everything just sat. This patch reuses the ls->async_lock spin_lock to fix the race, and it fixes the problem. I've done more than 2000 iterations of the loop that was recreating the failure and it seems to work. Signed-off-by: Bob Peterson Signed-off-by: Steven Whitehouse -- --- fs/gfs2/locking/dlm/thread.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) diff --git a/fs/gfs2/locking/dlm/thread.c b/fs/gfs2/locking/dlm/thread.c index 1aca51e45092..bd938f06481d 100644 --- a/fs/gfs2/locking/dlm/thread.c +++ b/fs/gfs2/locking/dlm/thread.c @@ -268,20 +268,16 @@ static inline int check_drop(struct gdlm_ls *ls) return 0; } -static int gdlm_thread(void *data) +static int gdlm_thread(void *data, int blist) { struct gdlm_ls *ls = (struct gdlm_ls *) data; struct gdlm_lock *lp = NULL; - int blist = 0; uint8_t complete, blocking, submit, drop; DECLARE_WAITQUEUE(wait, current); /* Only thread1 is allowed to do blocking callbacks since gfs may wait for a completion callback within a blocking cb. */ - if (current == ls->thread1) - blist = 1; - while (!kthread_should_stop()) { set_current_state(TASK_INTERRUPTIBLE); add_wait_queue(&ls->thread_wait, &wait); @@ -333,12 +329,22 @@ static int gdlm_thread(void *data) return 0; } +static int gdlm_thread1(void *data) +{ + return gdlm_thread(data, 1); +} + +static int gdlm_thread2(void *data) +{ + return gdlm_thread(data, 0); +} + int gdlm_init_threads(struct gdlm_ls *ls) { struct task_struct *p; int error; - p = kthread_run(gdlm_thread, ls, "lock_dlm1"); + p = kthread_run(gdlm_thread1, ls, "lock_dlm1"); error = IS_ERR(p); if (error) { log_error("can't start lock_dlm1 thread %d", error); @@ -346,7 +352,7 @@ int gdlm_init_threads(struct gdlm_ls *ls) } ls->thread1 = p; - p = kthread_run(gdlm_thread, ls, "lock_dlm2"); + p = kthread_run(gdlm_thread2, ls, "lock_dlm2"); error = IS_ERR(p); if (error) { log_error("can't start lock_dlm2 thread %d", error); -- cgit v1.2.3 From 16615be18cadf53ee6f8a4f0bdd647f0753421b1 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Mon, 17 Sep 2007 10:59:52 +0100 Subject: [GFS2] Clean up journaled data writing This patch cleans up the code for writing journaled data into the log. It also removes the need to allocate a small "tag" structure for each block written into the log. Instead we just keep count of the outstanding I/O so that we can be sure that its all been written at the correct time. Another result of this patch is that a number of ll_rw_block() calls have become submit_bh() calls, closing some races at the same time. Signed-off-by: Steven Whitehouse --- fs/gfs2/incore.h | 9 +- fs/gfs2/log.c | 145 +++++++++++++++++------------- fs/gfs2/log.h | 1 + fs/gfs2/lops.c | 242 ++++++++++++++++++++++++++------------------------ fs/gfs2/meta_io.c | 55 +++++++----- fs/gfs2/meta_io.h | 3 + fs/gfs2/ops_address.c | 11 +-- fs/gfs2/ops_fstype.c | 3 +- fs/gfs2/ops_inode.c | 7 +- fs/gfs2/ops_super.c | 13 +-- 10 files changed, 269 insertions(+), 220 deletions(-) diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index 8aa5780862be..eaddfb5a8e6f 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -341,12 +341,6 @@ struct gfs2_quota_data { unsigned long qd_last_touched; }; -struct gfs2_log_buf { - struct list_head lb_list; - struct buffer_head *lb_bh; - struct buffer_head *lb_real; -}; - struct gfs2_trans { unsigned long tr_ip; @@ -631,7 +625,8 @@ struct gfs2_sbd { unsigned long sd_log_flush_time; struct rw_semaphore sd_log_flush_lock; - struct list_head sd_log_flush_list; + atomic_t sd_log_in_flight; + wait_queue_head_t sd_log_flush_wait; unsigned int sd_log_flush_head; u64 sd_log_flush_wrapped; diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index 4d04e6f19706..ee704676b2f1 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -104,11 +104,8 @@ static void gfs2_ail1_start_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai) gfs2_assert(sdp, bd->bd_ail == ai); if (!buffer_busy(bh)) { - if (!buffer_uptodate(bh)) { - gfs2_log_unlock(sdp); + if (!buffer_uptodate(bh)) gfs2_io_error_bh(sdp, bh); - gfs2_log_lock(sdp); - } list_move(&bd->bd_ail_st_list, &ai->ai_ail2_list); continue; } @@ -118,9 +115,16 @@ static void gfs2_ail1_start_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai) list_move(&bd->bd_ail_st_list, &ai->ai_ail1_list); + get_bh(bh); gfs2_log_unlock(sdp); - wait_on_buffer(bh); - ll_rw_block(WRITE, 1, &bh); + lock_buffer(bh); + if (test_clear_buffer_dirty(bh)) { + bh->b_end_io = end_buffer_write_sync; + submit_bh(WRITE, bh); + } else { + unlock_buffer(bh); + brelse(bh); + } gfs2_log_lock(sdp); retry = 1; @@ -446,10 +450,10 @@ static unsigned int current_tail(struct gfs2_sbd *sdp) return tail; } -static inline void log_incr_head(struct gfs2_sbd *sdp) +void gfs2_log_incr_head(struct gfs2_sbd *sdp) { if (sdp->sd_log_flush_head == sdp->sd_log_tail) - gfs2_assert_withdraw(sdp, sdp->sd_log_flush_head == sdp->sd_log_head); + BUG_ON(sdp->sd_log_flush_head != sdp->sd_log_head); if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) { sdp->sd_log_flush_head = 0; @@ -457,6 +461,23 @@ static inline void log_incr_head(struct gfs2_sbd *sdp) } } +/** + * gfs2_log_write_endio - End of I/O for a log buffer + * @bh: The buffer head + * @uptodate: I/O Status + * + */ + +static void gfs2_log_write_endio(struct buffer_head *bh, int uptodate) +{ + struct gfs2_sbd *sdp = bh->b_private; + bh->b_private = NULL; + + end_buffer_write_sync(bh, uptodate); + if (atomic_dec_and_test(&sdp->sd_log_in_flight)) + wake_up(&sdp->sd_log_flush_wait); +} + /** * gfs2_log_get_buf - Get and initialize a buffer to use for log control data * @sdp: The GFS2 superblock @@ -467,24 +488,41 @@ static inline void log_incr_head(struct gfs2_sbd *sdp) struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp) { u64 blkno = log_bmap(sdp, sdp->sd_log_flush_head); - struct gfs2_log_buf *lb; struct buffer_head *bh; - lb = kzalloc(sizeof(struct gfs2_log_buf), GFP_NOFS | __GFP_NOFAIL); - list_add(&lb->lb_list, &sdp->sd_log_flush_list); - - bh = lb->lb_bh = sb_getblk(sdp->sd_vfs, blkno); + bh = sb_getblk(sdp->sd_vfs, blkno); lock_buffer(bh); memset(bh->b_data, 0, bh->b_size); set_buffer_uptodate(bh); clear_buffer_dirty(bh); - unlock_buffer(bh); - - log_incr_head(sdp); + gfs2_log_incr_head(sdp); + atomic_inc(&sdp->sd_log_in_flight); + bh->b_private = sdp; + bh->b_end_io = gfs2_log_write_endio; return bh; } +/** + * gfs2_fake_write_endio - + * @bh: The buffer head + * @uptodate: The I/O Status + * + */ + +static void gfs2_fake_write_endio(struct buffer_head *bh, int uptodate) +{ + struct buffer_head *real_bh = bh->b_private; + struct gfs2_sbd *sdp = GFS2_SB(real_bh->b_page->mapping->host); + + end_buffer_write_sync(bh, uptodate); + free_buffer_head(bh); + unlock_buffer(real_bh); + brelse(real_bh); + if (atomic_dec_and_test(&sdp->sd_log_in_flight)) + wake_up(&sdp->sd_log_flush_wait); +} + /** * gfs2_log_fake_buf - Build a fake buffer head to write metadata buffer to log * @sdp: the filesystem @@ -497,22 +535,20 @@ struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, struct buffer_head *real) { u64 blkno = log_bmap(sdp, sdp->sd_log_flush_head); - struct gfs2_log_buf *lb; struct buffer_head *bh; - lb = kzalloc(sizeof(struct gfs2_log_buf), GFP_NOFS | __GFP_NOFAIL); - list_add(&lb->lb_list, &sdp->sd_log_flush_list); - lb->lb_real = real; - - bh = lb->lb_bh = alloc_buffer_head(GFP_NOFS | __GFP_NOFAIL); + bh = alloc_buffer_head(GFP_NOFS | __GFP_NOFAIL); atomic_set(&bh->b_count, 1); - bh->b_state = (1 << BH_Mapped) | (1 << BH_Uptodate); + bh->b_state = (1 << BH_Mapped) | (1 << BH_Uptodate) | (1 << BH_Lock); set_bh_page(bh, real->b_page, bh_offset(real)); bh->b_blocknr = blkno; bh->b_size = sdp->sd_sb.sb_bsize; bh->b_bdev = sdp->sd_vfs->s_bdev; + bh->b_private = real; + bh->b_end_io = gfs2_fake_write_endio; - log_incr_head(sdp); + gfs2_log_incr_head(sdp); + atomic_inc(&sdp->sd_log_in_flight); return bh; } @@ -579,45 +615,24 @@ static void log_write_header(struct gfs2_sbd *sdp, u32 flags, int pull) gfs2_assert_withdraw(sdp, !pull); sdp->sd_log_idle = (tail == sdp->sd_log_flush_head); - log_incr_head(sdp); + gfs2_log_incr_head(sdp); } static void log_flush_commit(struct gfs2_sbd *sdp) { - struct list_head *head = &sdp->sd_log_flush_list; - struct gfs2_log_buf *lb; - struct buffer_head *bh; - int flushcount = 0; - - while (!list_empty(head)) { - lb = list_entry(head->next, struct gfs2_log_buf, lb_list); - list_del(&lb->lb_list); - bh = lb->lb_bh; - - wait_on_buffer(bh); - if (!buffer_uptodate(bh)) - gfs2_io_error_bh(sdp, bh); - if (lb->lb_real) { - while (atomic_read(&bh->b_count) != 1) /* Grrrr... */ - schedule(); - free_buffer_head(bh); - } else - brelse(bh); - kfree(lb); - flushcount++; + DEFINE_WAIT(wait); + + if (atomic_read(&sdp->sd_log_in_flight)) { + do { + prepare_to_wait(&sdp->sd_log_flush_wait, &wait, + TASK_UNINTERRUPTIBLE); + if (atomic_read(&sdp->sd_log_in_flight)) + io_schedule(); + } while(atomic_read(&sdp->sd_log_in_flight)); + finish_wait(&sdp->sd_log_flush_wait, &wait); } - /* If nothing was journaled, the header is unplanned and unwanted. */ - if (flushcount) { - log_write_header(sdp, 0, 0); - } else { - unsigned int tail; - tail = current_tail(sdp); - - gfs2_ail1_empty(sdp, 0); - if (sdp->sd_log_tail != tail) - log_pull_tail(sdp, tail); - } + log_write_header(sdp, 0, 0); } static void gfs2_ordered_write(struct gfs2_sbd *sdp) @@ -698,10 +713,16 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl) INIT_LIST_HEAD(&ai->ai_ail1_list); INIT_LIST_HEAD(&ai->ai_ail2_list); - gfs2_assert_withdraw(sdp, - sdp->sd_log_num_buf + sdp->sd_log_num_databuf == - sdp->sd_log_commited_buf + - sdp->sd_log_commited_databuf); + if (sdp->sd_log_num_buf != sdp->sd_log_commited_buf) { + printk(KERN_INFO "GFS2: log buf %u %u\n", sdp->sd_log_num_buf, + sdp->sd_log_commited_buf); + gfs2_assert_withdraw(sdp, 0); + } + if (sdp->sd_log_num_databuf != sdp->sd_log_commited_databuf) { + printk(KERN_INFO "GFS2: log databuf %u %u\n", + sdp->sd_log_num_databuf, sdp->sd_log_commited_databuf); + gfs2_assert_withdraw(sdp, 0); + } gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke == sdp->sd_log_commited_revoke); @@ -713,7 +734,7 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl) lops_before_commit(sdp); gfs2_ordered_wait(sdp); - if (!list_empty(&sdp->sd_log_flush_list)) + if (sdp->sd_log_head != sdp->sd_log_flush_head) log_flush_commit(sdp); else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){ gfs2_log_lock(sdp); diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h index 639423561b2d..dae282400627 100644 --- a/fs/gfs2/log.h +++ b/fs/gfs2/log.h @@ -52,6 +52,7 @@ int gfs2_ail1_empty(struct gfs2_sbd *sdp, int flags); int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks); void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks); +void gfs2_log_incr_head(struct gfs2_sbd *sdp); struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp); struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp, diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c index 342c10e12af2..6c27cea761c6 100644 --- a/fs/gfs2/lops.c +++ b/fs/gfs2/lops.c @@ -91,6 +91,39 @@ static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh, unlock_buffer(bh); } + +static inline struct gfs2_log_descriptor *bh_log_desc(struct buffer_head *bh) +{ + return (struct gfs2_log_descriptor *)bh->b_data; +} + +static inline __be64 *bh_log_ptr(struct buffer_head *bh) +{ + struct gfs2_log_descriptor *ld = bh_log_desc(bh); + return (__force __be64 *)(ld + 1); +} + +static inline __be64 *bh_ptr_end(struct buffer_head *bh) +{ + return (__force __be64 *)(bh->b_data + bh->b_size); +} + + +static struct buffer_head *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type) +{ + struct buffer_head *bh = gfs2_log_get_buf(sdp); + struct gfs2_log_descriptor *ld = bh_log_desc(bh); + ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); + ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); + ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); + ld->ld_type = cpu_to_be32(ld_type); + ld->ld_length = 0; + ld->ld_data1 = 0; + ld->ld_data2 = 0; + memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); + return bh; +} + static void __glock_lo_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le) { struct gfs2_glock *gl; @@ -181,7 +214,6 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp) struct gfs2_log_descriptor *ld; struct gfs2_bufdata *bd1 = NULL, *bd2; unsigned int total; - unsigned int offset = BUF_OFFSET; unsigned int limit; unsigned int num; unsigned n; @@ -198,18 +230,12 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp) if (total > limit) num = limit; gfs2_log_unlock(sdp); - bh = gfs2_log_get_buf(sdp); + bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_METADATA); 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); - ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); - ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); - ld->ld_type = cpu_to_be32(GFS2_LOG_DESC_METADATA); + ld = bh_log_desc(bh); + ptr = bh_log_ptr(bh); ld->ld_length = cpu_to_be32(num + 1); ld->ld_data1 = cpu_to_be32(num); - ld->ld_data2 = cpu_to_be32(0); - memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); n = 0; list_for_each_entry_continue(bd1, &sdp->sd_log_le_buf, @@ -220,17 +246,17 @@ static void buf_lo_before_commit(struct gfs2_sbd *sdp) } gfs2_log_unlock(sdp); - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); + submit_bh(WRITE, bh); gfs2_log_lock(sdp); n = 0; list_for_each_entry_continue(bd2, &sdp->sd_log_le_buf, bd_le.le_list) { + get_bh(bd2->bd_bh); gfs2_log_unlock(sdp); + lock_buffer(bd2->bd_bh); bh = gfs2_log_fake_buf(sdp, bd2->bd_bh); - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); + submit_bh(WRITE, bh); gfs2_log_lock(sdp); if (++n >= num) break; @@ -359,17 +385,11 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) if (!sdp->sd_log_num_revoke) return; - bh = gfs2_log_get_buf(sdp); - ld = (struct gfs2_log_descriptor *)bh->b_data; - ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); - ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); - ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); - ld->ld_type = cpu_to_be32(GFS2_LOG_DESC_REVOKE); + bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_REVOKE); + ld = bh_log_desc(bh); ld->ld_length = cpu_to_be32(gfs2_struct2blk(sdp, sdp->sd_log_num_revoke, sizeof(u64))); ld->ld_data1 = cpu_to_be32(sdp->sd_log_num_revoke); - ld->ld_data2 = cpu_to_be32(0); - memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); offset = sizeof(struct gfs2_log_descriptor); while (!list_empty(head)) { @@ -378,8 +398,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) sdp->sd_log_num_revoke--; if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) { - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); + submit_bh(WRITE, bh); bh = gfs2_log_get_buf(sdp); mh = (struct gfs2_meta_header *)bh->b_data; @@ -396,8 +415,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp) } gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke); - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); + submit_bh(WRITE, bh); } static void revoke_lo_before_scan(struct gfs2_jdesc *jd, @@ -562,118 +580,110 @@ out: unlock_buffer(bd->bd_bh); } -static int gfs2_check_magic(struct buffer_head *bh) +static void gfs2_check_magic(struct buffer_head *bh) { - struct page *page = bh->b_page; void *kaddr; __be32 *ptr; - int rv = 0; - kaddr = kmap_atomic(page, KM_USER0); + clear_buffer_escaped(bh); + kaddr = kmap_atomic(bh->b_page, KM_USER0); ptr = kaddr + bh_offset(bh); if (*ptr == cpu_to_be32(GFS2_MAGIC)) - rv = 1; + set_buffer_escaped(bh); kunmap_atomic(kaddr, KM_USER0); - - return rv; } -/** - * databuf_lo_before_commit - Scan the data buffers, writing as we go - * - */ - -static void databuf_lo_before_commit(struct gfs2_sbd *sdp) +static void gfs2_write_blocks(struct gfs2_sbd *sdp, struct buffer_head *bh, + struct list_head *list, struct list_head *done, + unsigned int n) { - struct gfs2_bufdata *bd1 = NULL, *bd2; - struct buffer_head *bh = NULL,*bh1 = NULL; + struct buffer_head *bh1; struct gfs2_log_descriptor *ld; - unsigned int limit; - unsigned int total; - unsigned int num, n; - __be64 *ptr = NULL; - int magic; + struct gfs2_bufdata *bd; + __be64 *ptr; + if (!bh) + return; - limit = databuf_limit(sdp); + ld = bh_log_desc(bh); + ld->ld_length = cpu_to_be32(n + 1); + ld->ld_data1 = cpu_to_be32(n); + ptr = bh_log_ptr(bh); + + get_bh(bh); + submit_bh(WRITE, bh); gfs2_log_lock(sdp); - total = sdp->sd_log_num_databuf; - bd2 = bd1 = list_prepare_entry(bd1, &sdp->sd_log_le_databuf, - bd_le.le_list); - while(total) { - num = total; - if (num > 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 + DATABUF_OFFSET); - ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC); - ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD); - ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD); - ld->ld_type = cpu_to_be32(GFS2_LOG_DESC_JDATA); - ld->ld_length = cpu_to_be32(num + 1); - ld->ld_data1 = cpu_to_be32(num); - ld->ld_data2 = cpu_to_be32(0); - memset(ld->ld_reserved, 0, sizeof(ld->ld_reserved)); - - n = 0; - list_for_each_entry_continue(bd1, &sdp->sd_log_le_databuf, - bd_le.le_list) { - bh1 = bd1->bd_bh; - - magic = gfs2_check_magic(bh1); - *ptr++ = cpu_to_be64(bh1->b_blocknr); - *ptr++ = cpu_to_be64((__u64)magic); - clear_buffer_escaped(bh1); - if (unlikely(magic != 0)) - set_buffer_escaped(bh1); - if (++n >= num) - break; + while(!list_empty(list)) { + bd = list_entry(list->next, struct gfs2_bufdata, bd_le.le_list); + list_move_tail(&bd->bd_le.le_list, done); + get_bh(bd->bd_bh); + while (be64_to_cpu(*ptr) != bd->bd_bh->b_blocknr) { + gfs2_log_incr_head(sdp); + ptr += 2; } gfs2_log_unlock(sdp); - if (bh) { - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); - bh = NULL; - ptr = NULL; + lock_buffer(bd->bd_bh); + if (buffer_escaped(bd->bd_bh)) { + void *kaddr; + bh1 = gfs2_log_get_buf(sdp); + kaddr = kmap_atomic(bd->bd_bh->b_page, KM_USER0); + memcpy(bh1->b_data, kaddr + bh_offset(bd->bd_bh), + bh1->b_size); + kunmap_atomic(kaddr, KM_USER0); + *(__be32 *)bh1->b_data = 0; + clear_buffer_escaped(bd->bd_bh); + unlock_buffer(bd->bd_bh); + brelse(bd->bd_bh); + } else { + bh1 = gfs2_log_fake_buf(sdp, bd->bd_bh); } - n = 0; + submit_bh(WRITE, bh1); gfs2_log_lock(sdp); - list_for_each_entry_continue(bd2, &sdp->sd_log_le_databuf, - bd_le.le_list) { - if (!bd2->bd_bh) - continue; - /* copy buffer if it needs escaping */ + ptr += 2; + } + gfs2_log_unlock(sdp); + brelse(bh); +} + +/** + * databuf_lo_before_commit - Scan the data buffers, writing as we go + * + */ + +static void databuf_lo_before_commit(struct gfs2_sbd *sdp) +{ + struct gfs2_bufdata *bd = NULL; + struct buffer_head *bh = NULL; + unsigned int n = 0; + __be64 *ptr = NULL, *end = NULL; + LIST_HEAD(processed); + LIST_HEAD(in_progress); + + gfs2_log_lock(sdp); + while (!list_empty(&sdp->sd_log_le_databuf)) { + if (ptr == end) { gfs2_log_unlock(sdp); - if (unlikely(buffer_escaped(bd2->bd_bh))) { - void *kaddr; - struct page *page = bd2->bd_bh->b_page; - bh = gfs2_log_get_buf(sdp); - kaddr = kmap_atomic(page, KM_USER0); - memcpy(bh->b_data, - kaddr + bh_offset(bd2->bd_bh), - sdp->sd_sb.sb_bsize); - kunmap_atomic(kaddr, KM_USER0); - *(__be32 *)bh->b_data = 0; - } else { - bh = gfs2_log_fake_buf(sdp, bd2->bd_bh); - } - set_buffer_dirty(bh); - ll_rw_block(WRITE, 1, &bh); + gfs2_write_blocks(sdp, bh, &in_progress, &processed, n); + n = 0; + bh = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_JDATA); + ptr = bh_log_ptr(bh); + end = bh_ptr_end(bh) - 1; gfs2_log_lock(sdp); - if (++n >= num) - break; + continue; } - bh = NULL; - BUG_ON(total < num); - total -= num; + bd = list_entry(sdp->sd_log_le_databuf.next, struct gfs2_bufdata, bd_le.le_list); + list_move_tail(&bd->bd_le.le_list, &in_progress); + gfs2_check_magic(bd->bd_bh); + *ptr++ = cpu_to_be64(bd->bd_bh->b_blocknr); + *ptr++ = cpu_to_be64(buffer_escaped(bh) ? 1 : 0); + n++; } gfs2_log_unlock(sdp); + gfs2_write_blocks(sdp, bh, &in_progress, &processed, n); + gfs2_log_lock(sdp); + list_splice(&processed, &sdp->sd_log_le_databuf); + gfs2_log_unlock(sdp); } static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start, @@ -807,10 +817,10 @@ const struct gfs2_log_operations gfs2_databuf_lops = { const struct gfs2_log_operations *gfs2_log_ops[] = { &gfs2_glock_lops, + &gfs2_databuf_lops, &gfs2_buf_lops, - &gfs2_revoke_lops, &gfs2_rg_lops, - &gfs2_databuf_lops, + &gfs2_revoke_lops, NULL, }; diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c index 19097bc7c81d..1d80f2d42122 100644 --- a/fs/gfs2/meta_io.c +++ b/fs/gfs2/meta_io.c @@ -297,6 +297,37 @@ void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh, unlock_page(bh->b_page); } +void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta) +{ + struct gfs2_sbd *sdp = GFS2_SB(bh->b_page->mapping->host); + struct gfs2_bufdata *bd = bh->b_private; + if (test_clear_buffer_pinned(bh)) { + list_del_init(&bd->bd_le.le_list); + if (meta) { + gfs2_assert_warn(sdp, sdp->sd_log_num_buf); + sdp->sd_log_num_buf--; + tr->tr_num_buf_rm++; + } else { + gfs2_assert_warn(sdp, sdp->sd_log_num_databuf); + sdp->sd_log_num_databuf--; + tr->tr_num_databuf_rm++; + } + tr->tr_touched = 1; + brelse(bh); + } + if (bd) { + if (bd->bd_ail) { + gfs2_remove_from_ail(NULL, bd); + bh->b_private = NULL; + bd->bd_bh = NULL; + bd->bd_blkno = bh->b_blocknr; + gfs2_trans_add_revoke(sdp, bd); + } + } + clear_buffer_dirty(bh); + clear_buffer_uptodate(bh); +} + /** * gfs2_meta_wipe - make inode's buffers so they aren't dirty/pinned anymore * @ip: the inode who owns the buffers @@ -313,33 +344,11 @@ void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen) while (blen) { bh = getbuf(ip->i_gl, bstart, NO_CREATE); if (bh) { - struct gfs2_bufdata *bd; - lock_buffer(bh); gfs2_log_lock(sdp); - bd = bh->b_private; - if (test_clear_buffer_pinned(bh)) { - struct gfs2_trans *tr = current->journal_info; - list_del_init(&bd->bd_le.le_list); - gfs2_assert_warn(sdp, sdp->sd_log_num_buf); - sdp->sd_log_num_buf--; - tr->tr_num_buf_rm++; - brelse(bh); - } - if (bd) { - if (bd->bd_ail) { - gfs2_remove_from_ail(NULL, bd); - bh->b_private = NULL; - bd->bd_bh = NULL; - bd->bd_blkno = bh->b_blocknr; - gfs2_trans_add_revoke(sdp, bd); - } - } - clear_buffer_dirty(bh); - clear_buffer_uptodate(bh); + gfs2_remove_from_journal(bh, current->journal_info, 1); gfs2_log_unlock(sdp); unlock_buffer(bh); - brelse(bh); } diff --git a/fs/gfs2/meta_io.h b/fs/gfs2/meta_io.h index fd67f07cd60e..b7048222ebb4 100644 --- a/fs/gfs2/meta_io.h +++ b/fs/gfs2/meta_io.h @@ -51,6 +51,9 @@ int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh); void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh, int meta); +void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, + int meta); + void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen); void gfs2_meta_cache_flush(struct gfs2_inode *ip); diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c index dd1ea491ddcb..b7baf1831912 100644 --- a/fs/gfs2/ops_address.c +++ b/fs/gfs2/ops_address.c @@ -414,7 +414,8 @@ static int gfs2_prepare_write(struct file *file, struct page *page, if (ind_blocks || data_blocks) rblocks += RES_STATFS + RES_QUOTA; - error = gfs2_trans_begin(sdp, rblocks, 0); + error = gfs2_trans_begin(sdp, rblocks, + PAGE_CACHE_SIZE/sdp->sd_sb.sb_bsize); if (error) goto out_trans_fail; @@ -625,10 +626,10 @@ static void gfs2_discard(struct gfs2_sbd *sdp, struct buffer_head *bh) clear_buffer_dirty(bh); bd = bh->b_private; if (bd) { - if (!list_empty(&bd->bd_le.le_list)) { - if (!buffer_pinned(bh)) - list_del_init(&bd->bd_le.le_list); - } + if (!list_empty(&bd->bd_le.le_list) && !buffer_pinned(bh)) + list_del_init(&bd->bd_le.le_list); + else + gfs2_remove_from_journal(bh, current->journal_info, 0); } bh->b_bdev = NULL; clear_buffer_mapped(bh); diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 35f3dfa9bfb1..1ac9afa58c65 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -89,7 +89,8 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb) INIT_LIST_HEAD(&sdp->sd_ail2_list); init_rwsem(&sdp->sd_log_flush_lock); - INIT_LIST_HEAD(&sdp->sd_log_flush_list); + atomic_set(&sdp->sd_log_in_flight, 0); + init_waitqueue_head(&sdp->sd_log_flush_wait); INIT_LIST_HEAD(&sdp->sd_revoke_list); diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c index 2cbe5a321e89..291f0c7eaa3b 100644 --- a/fs/gfs2/ops_inode.c +++ b/fs/gfs2/ops_inode.c @@ -905,12 +905,17 @@ static int gfs2_permission(struct inode *inode, int mask, struct nameidata *nd) static int setattr_size(struct inode *inode, struct iattr *attr) { struct gfs2_inode *ip = GFS2_I(inode); + struct gfs2_sbd *sdp = GFS2_SB(inode); int error; if (attr->ia_size != ip->i_di.di_size) { - error = vmtruncate(inode, attr->ia_size); + error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks); if (error) return error; + error = vmtruncate(inode, attr->ia_size); + gfs2_trans_end(sdp); + if (error) + return error; } error = gfs2_truncatei(ip, attr->ia_size); diff --git a/fs/gfs2/ops_super.c b/fs/gfs2/ops_super.c index 4316690d86f6..950f31460e8b 100644 --- a/fs/gfs2/ops_super.c +++ b/fs/gfs2/ops_super.c @@ -455,12 +455,15 @@ static void gfs2_delete_inode(struct inode *inode) } error = gfs2_dinode_dealloc(ip); - /* - * Must do this before unlock to avoid trying to write back - * potentially dirty data now that inode no longer exists - * on disk. - */ + if (error) + goto out_unlock; + + error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks); + if (error) + goto out_unlock; + /* Needs to be done before glock release & also in a transaction */ truncate_inode_pages(&inode->i_data, 0); + gfs2_trans_end(sdp); out_unlock: gfs2_glock_dq(&ip->i_iopen_gh); -- cgit v1.2.3 From de986e859a29097fb9211b052d86a9a2c868f6cd Mon Sep 17 00:00:00 2001 From: Wendy Cheng Date: Tue, 18 Sep 2007 09:19:13 -0400 Subject: [GFS2] Data corruption fix * GFS2 has been using i_cache array to store its indirect meta blocks. Its flush routine doesn't correctly clean up all the entries. The problem would show while multiple nodes do simultaneous writes to the same file. Upon glock exclusive lock transfer, if the file is a sparse file with large file size where the indirect meta blocks span multiple array entries with "zero" entries in between. The flush routine prematurely stops the flushing that leaves old (stale) entries around. This leads to several nasty issues, including data corruption. * Fix gfs2_get_block_noalloc checking to correctly return EIO upon unmapped buffer. Signed-off-by: Wendy Cheng Signed-off-by: Steven Whitehouse --- fs/gfs2/meta_io.c | 8 ++++---- fs/gfs2/ops_address.c | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c index 1d80f2d42122..4da423985e4f 100644 --- a/fs/gfs2/meta_io.c +++ b/fs/gfs2/meta_io.c @@ -374,10 +374,10 @@ void gfs2_meta_cache_flush(struct gfs2_inode *ip) for (x = 0; x < GFS2_MAX_META_HEIGHT; x++) { bh_slot = &ip->i_cache[x]; - if (!*bh_slot) - break; - brelse(*bh_slot); - *bh_slot = NULL; + if (*bh_slot) { + brelse(*bh_slot); + *bh_slot = NULL; + } } spin_unlock(&ip->i_spin); diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c index b7baf1831912..4002f417dc19 100644 --- a/fs/gfs2/ops_address.c +++ b/fs/gfs2/ops_address.c @@ -90,7 +90,7 @@ static int gfs2_get_block_noalloc(struct inode *inode, sector_t lblock, error = gfs2_block_map(inode, lblock, 0, bh_result); if (error) return error; - if (bh_result->b_blocknr == 0) + if (!buffer_mapped(bh_result)) return -EIO; return 0; } -- cgit v1.2.3 From 7a9f53b3c1875bef22ad4588e818bc046ef183da Mon Sep 17 00:00:00 2001 From: Benjamin Marzinski Date: Tue, 18 Sep 2007 13:33:18 -0500 Subject: [GFS2] Alternate gfs2_iget to avoid looking up inodes being freed There is a possible deadlock between two processes on the same node, where one process is deleting an inode, and another process is looking for allocated but unused inodes to delete in order to create more space. process A does an iput() on inode X, and it's i_count drops to 0. This causes iput_final() to be called, which puts an inode into state I_FREEING at generic_delete_inode(). There no point between when iput_final() is called, and when I_FREEING is set where GFS2 could acquire any glocks. Once I_FREEING is set, no other process on that node can successfully look up that inode until the delete finishes. process B locks the the resource group for the same inode in get_local_rgrp(), which is called by gfs2_inplace_reserve_i() process A tries to lock the resource group for the inode in gfs2_dinode_dealloc(), but it's already locked by process B process B waits in find_inode for the inode to have the I_FREEING state cleared. Deadlock. This patch solves the problem by adding an alternative to gfs2_iget(), gfs2_iget_skip(), that simply skips any inodes that are in the I_FREEING state.o The alternate test function is just like the original one, except that it fails if the inode is being freed, and sets a skipped flag. The alternate set function is just like the original, except that it fails if the skipped flag is set. Only try_rgrp_unlink() calls gfs2_iget_skip() instead of gfs2_iget(). Signed-off-by: Benjamin E. Marzinski Signed-off-by: Steven Whitehouse --- fs/gfs2/dir.c | 2 +- fs/gfs2/inode.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++---- fs/gfs2/inode.h | 3 ++- fs/gfs2/ops_export.c | 2 +- fs/gfs2/ops_fstype.c | 2 +- fs/gfs2/rgrp.c | 2 +- 6 files changed, 60 insertions(+), 9 deletions(-) diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c index 08c6dd0c6713..9949bb746a52 100644 --- a/fs/gfs2/dir.c +++ b/fs/gfs2/dir.c @@ -1502,7 +1502,7 @@ struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name) inode = gfs2_inode_lookup(dir->i_sb, be16_to_cpu(dent->de_type), be64_to_cpu(dent->de_inum.no_addr), - be64_to_cpu(dent->de_inum.no_formal_ino)); + be64_to_cpu(dent->de_inum.no_formal_ino), 0); brelse(bh); return inode; } diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c index 013f00b90cc2..5f6dc32946cd 100644 --- a/fs/gfs2/inode.c +++ b/fs/gfs2/inode.c @@ -77,6 +77,49 @@ static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr) return iget5_locked(sb, hash, iget_test, iget_set, &no_addr); } +struct gfs2_skip_data { + u64 no_addr; + int skipped; +}; + +static int iget_skip_test(struct inode *inode, void *opaque) +{ + struct gfs2_inode *ip = GFS2_I(inode); + struct gfs2_skip_data *data = opaque; + + if (ip->i_no_addr == data->no_addr && inode->i_private != NULL){ + if (inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)){ + data->skipped = 1; + return 0; + } + return 1; + } + return 0; +} + +static int iget_skip_set(struct inode *inode, void *opaque) +{ + struct gfs2_inode *ip = GFS2_I(inode); + struct gfs2_skip_data *data = opaque; + + if (data->skipped) + return 1; + inode->i_ino = (unsigned long)(data->no_addr); + ip->i_no_addr = data->no_addr; + return 0; +} + +static struct inode *gfs2_iget_skip(struct super_block *sb, + u64 no_addr) +{ + struct gfs2_skip_data data; + unsigned long hash = (unsigned long)no_addr; + + data.no_addr = no_addr; + data.skipped = 0; + return iget5_locked(sb, hash, iget_skip_test, iget_skip_set, &data); +} + /** * GFS2 lookup code fills in vfs inode contents based on info obtained * from directory entry inside gfs2_inode_lookup(). This has caused issues @@ -112,6 +155,7 @@ void gfs2_set_iop(struct inode *inode) * @sb: The super block * @no_addr: The inode number * @type: The type of the inode + * @skip_freeing: set this not return an inode if it is currently being freed. * * Returns: A VFS inode, or an error */ @@ -119,13 +163,19 @@ void gfs2_set_iop(struct inode *inode) struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type, u64 no_addr, - u64 no_formal_ino) + u64 no_formal_ino, int skip_freeing) { - struct inode *inode = gfs2_iget(sb, no_addr); - struct gfs2_inode *ip = GFS2_I(inode); + struct inode *inode; + struct gfs2_inode *ip; struct gfs2_glock *io_gl; int error; + if (skip_freeing) + inode = gfs2_iget_skip(sb, no_addr); + else + inode = gfs2_iget(sb, no_addr); + ip = GFS2_I(inode); + if (!inode) return ERR_PTR(-ENOBUFS); @@ -949,7 +999,7 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name, inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr, - inum.no_formal_ino); + inum.no_formal_ino, 0); if (IS_ERR(inode)) goto fail_gunlock2; diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h index 4517ac82c01c..351ac87ab384 100644 --- a/fs/gfs2/inode.h +++ b/fs/gfs2/inode.h @@ -49,7 +49,8 @@ static inline void gfs2_inum_out(const struct gfs2_inode *ip, void gfs2_inode_attr_in(struct gfs2_inode *ip); void gfs2_set_iop(struct inode *inode); struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned type, - u64 no_addr, u64 no_formal_ino); + u64 no_addr, u64 no_formal_ino, + int skip_freeing); struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr); int gfs2_inode_refresh(struct gfs2_inode *ip); diff --git a/fs/gfs2/ops_export.c b/fs/gfs2/ops_export.c index b8312edee0e4..e2d1347796a9 100644 --- a/fs/gfs2/ops_export.c +++ b/fs/gfs2/ops_export.c @@ -237,7 +237,7 @@ static struct dentry *gfs2_get_dentry(struct super_block *sb, void *inum_obj) inode = gfs2_inode_lookup(sb, DT_UNKNOWN, inum->no_addr, - 0); + 0, 0); if (!inode) goto fail; if (IS_ERR(inode)) { diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 1ac9afa58c65..17de58e83d92 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -230,7 +230,7 @@ fail: static inline struct inode *gfs2_lookup_root(struct super_block *sb, u64 no_addr) { - return gfs2_inode_lookup(sb, DT_DIR, no_addr, 0); + return gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); } static int init_sb(struct gfs2_sbd *sdp, int silent, int undo) diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index 2d7f7ea0c9a8..708c287e1d0e 100644 --- a/fs/gfs2/rgrp.c +++ b/fs/gfs2/rgrp.c @@ -884,7 +884,7 @@ static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked) continue; *last_unlinked = no_addr; inode = gfs2_inode_lookup(rgd->rd_sbd->sd_vfs, DT_UNKNOWN, - no_addr, -1); + no_addr, -1, 1); if (!IS_ERR(inode)) return inode; } -- cgit v1.2.3 From 891ba6d4a5f9e6302bb6542592d73feb4d0d3687 Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Thu, 20 Sep 2007 15:26:33 +0100 Subject: [GFS2] Don't try to remove buffers that don't exist Signed-off-by: Steven Whitehouse --- fs/gfs2/ops_address.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c index 4002f417dc19..873a511ef2be 100644 --- a/fs/gfs2/ops_address.c +++ b/fs/gfs2/ops_address.c @@ -747,7 +747,7 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask) struct gfs2_bufdata *bd; if (!page_has_buffers(page)) - goto out; + return 0; gfs2_log_lock(sdp); head = bh = page_buffers(page); @@ -787,7 +787,6 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask) bh = bh->b_this_page; } while (bh != head); -out: return try_to_free_buffers(page); cannot_release: gfs2_log_unlock(sdp); -- cgit v1.2.3 From 5a60c532c9224babc172fafccc9e2fec6937af6f Mon Sep 17 00:00:00 2001 From: Steven Whitehouse Date: Wed, 26 Sep 2007 09:39:31 +0100 Subject: [GFS2] Get superblock a different way The mapping may be NULL by the time the I/O has completed, so we now get the superblock by a different route (via the bd and glock) to avoid this problem. Signed-off-by: Steven Whitehouse Cc: Wendy Cheng --- fs/gfs2/log.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c index ee704676b2f1..7df702473252 100644 --- a/fs/gfs2/log.c +++ b/fs/gfs2/log.c @@ -513,7 +513,8 @@ struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp) static void gfs2_fake_write_endio(struct buffer_head *bh, int uptodate) { struct buffer_head *real_bh = bh->b_private; - struct gfs2_sbd *sdp = GFS2_SB(real_bh->b_page->mapping->host); + struct gfs2_bufdata *bd = real_bh->b_private; + struct gfs2_sbd *sdp = bd->bd_gl->gl_sbd; end_buffer_write_sync(bh, uptodate); free_buffer_head(bh); -- cgit v1.2.3 From b434eda6fda5bcdcc2dd918e5ffbf7184f2d4e17 Mon Sep 17 00:00:00 2001 From: Patrick Caulfield Date: Mon, 1 Oct 2007 15:28:42 +0100 Subject: [DLM] don't overwrite castparam if it's NULL If the castaddr passed to the userland API is NULL then don't overwrite the existing castparam. This allows a different thread to cancel a lock request and the CANCEL AST gets delivered to the original thread. bz#306391 (for RHEL4) refers. Signed-Off-By: Patrick Caulfield Signed-off-by: Steven Whitehouse --- fs/dlm/lock.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/dlm/lock.c b/fs/dlm/lock.c index 2082daf083d8..031229f144fa 100644 --- a/fs/dlm/lock.c +++ b/fs/dlm/lock.c @@ -4429,7 +4429,8 @@ int dlm_user_unlock(struct dlm_ls *ls, struct dlm_user_args *ua_tmp, if (lvb_in && ua->lksb.sb_lvbptr) memcpy(ua->lksb.sb_lvbptr, lvb_in, DLM_USER_LVB_LEN); - ua->castparam = ua_tmp->castparam; + if (ua_tmp->castparam) + ua->castparam = ua_tmp->castparam; ua->user_lksb = ua_tmp->user_lksb; error = set_unlock_args(flags, ua, &args); @@ -4474,7 +4475,8 @@ int dlm_user_cancel(struct dlm_ls *ls, struct dlm_user_args *ua_tmp, goto out; ua = (struct dlm_user_args *)lkb->lkb_astparam; - ua->castparam = ua_tmp->castparam; + if (ua_tmp->castparam) + ua->castparam = ua_tmp->castparam; ua->user_lksb = ua_tmp->user_lksb; error = set_unlock_args(flags, ua, &args); -- cgit v1.2.3 From c36258b5925e6cf6bf72904635100593573bfcff Mon Sep 17 00:00:00 2001 From: David Teigland Date: Thu, 27 Sep 2007 15:53:38 -0500 Subject: [DLM] block dlm_recv in recovery transition Introduce a per-lockspace rwsem that's held in read mode by dlm_recv threads while working in the dlm. This allows dlm_recv activity to be suspended when the lockspace transitions to, from and between recovery cycles. The specific bug prompting this change is one where an in-progress recovery cycle is aborted by a new recovery cycle. While dlm_recv was processing a recovery message, the recovery cycle was aborted and dlm_recoverd began cleaning up. dlm_recv decremented recover_locks_count on an rsb after dlm_recoverd had reset it to zero. This is fixed by suspending dlm_recv (taking write lock on the rwsem) before aborting the current recovery. The transitions to/from normal and recovery modes are simplified by using this new ability to block dlm_recv. The switch from normal to recovery mode means dlm_recv goes from processing locking messages, to saving them for later, and vice versa. Races are avoided by blocking dlm_recv when setting the flag that switches between modes. Signed-off-by: David Teigland Signed-off-by: Steven Whitehouse --- fs/dlm/dlm_internal.h | 1 + fs/dlm/lock.c | 136 ++++++++++++++++++++++++++++++-------------------- fs/dlm/lock.h | 3 +- fs/dlm/lockspace.c | 1 + fs/dlm/member.c | 41 +++++++++------ fs/dlm/midcomms.c | 17 +------ fs/dlm/rcom.c | 36 +++---------- fs/dlm/rcom.h | 5 +- fs/dlm/recoverd.c | 11 +++- fs/dlm/requestqueue.c | 58 +++++++++------------ fs/dlm/requestqueue.h | 4 +- 11 files changed, 161 insertions(+), 152 deletions(-) diff --git a/fs/dlm/dlm_internal.h b/fs/dlm/dlm_internal.h index 74901e981e10..d2fc2384c3be 100644 --- a/fs/dlm/dlm_internal.h +++ b/fs/dlm/dlm_internal.h @@ -491,6 +491,7 @@ struct dlm_ls { uint64_t ls_recover_seq; struct dlm_recover *ls_recover_args; struct rw_semaphore ls_in_recovery; /* block local requests */ + struct rw_semaphore ls_recv_active; /* block dlm_recv */ struct list_head ls_requestqueue;/* queue remote requests */ struct mutex ls_requestqueue_mutex; char *ls_recover_buf; diff --git a/fs/dlm/lock.c b/fs/dlm/lock.c index 031229f144fa..3915b8e14146 100644 --- a/fs/dlm/lock.c +++ b/fs/dlm/lock.c @@ -3638,55 +3638,8 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms) dlm_put_lkb(lkb); } -int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery) +static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms) { - struct dlm_message *ms = (struct dlm_message *) hd; - struct dlm_ls *ls; - int error = 0; - - if (!recovery) - dlm_message_in(ms); - - ls = dlm_find_lockspace_global(hd->h_lockspace); - if (!ls) { - log_print("drop message %d from %d for unknown lockspace %d", - ms->m_type, nodeid, hd->h_lockspace); - return -EINVAL; - } - - /* recovery may have just ended leaving a bunch of backed-up requests - in the requestqueue; wait while dlm_recoverd clears them */ - - if (!recovery) - dlm_wait_requestqueue(ls); - - /* recovery may have just started while there were a bunch of - in-flight requests -- save them in requestqueue to be processed - after recovery. we can't let dlm_recvd block on the recovery - lock. if dlm_recoverd is calling this function to clear the - requestqueue, it needs to be interrupted (-EINTR) if another - recovery operation is starting. */ - - while (1) { - if (dlm_locking_stopped(ls)) { - if (recovery) { - error = -EINTR; - goto out; - } - error = dlm_add_requestqueue(ls, nodeid, hd); - if (error == -EAGAIN) - continue; - else { - error = -EINTR; - goto out; - } - } - - if (dlm_lock_recovery_try(ls)) - break; - schedule(); - } - switch (ms->m_type) { /* messages sent to a master node */ @@ -3761,17 +3714,90 @@ int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery) log_error(ls, "unknown message type %d", ms->m_type); } - dlm_unlock_recovery(ls); - out: - dlm_put_lockspace(ls); dlm_astd_wake(); - return error; } +/* If the lockspace is in recovery mode (locking stopped), then normal + messages are saved on the requestqueue for processing after recovery is + done. When not in recovery mode, we wait for dlm_recoverd to drain saved + messages off the requestqueue before we process new ones. This occurs right + after recovery completes when we transition from saving all messages on + requestqueue, to processing all the saved messages, to processing new + messages as they arrive. */ -/* - * Recovery related - */ +static void dlm_receive_message(struct dlm_ls *ls, struct dlm_message *ms, + int nodeid) +{ + if (dlm_locking_stopped(ls)) { + dlm_add_requestqueue(ls, nodeid, (struct dlm_header *) ms); + } else { + dlm_wait_requestqueue(ls); + _receive_message(ls, ms); + } +} + +/* This is called by dlm_recoverd to process messages that were saved on + the requestqueue. */ + +void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms) +{ + _receive_message(ls, ms); +} + +/* This is called by the midcomms layer when something is received for + the lockspace. It could be either a MSG (normal message sent as part of + standard locking activity) or an RCOM (recovery message sent as part of + lockspace recovery). */ + +void dlm_receive_buffer(struct dlm_header *hd, int nodeid) +{ + struct dlm_message *ms = (struct dlm_message *) hd; + struct dlm_rcom *rc = (struct dlm_rcom *) hd; + struct dlm_ls *ls; + int type = 0; + + switch (hd->h_cmd) { + case DLM_MSG: + dlm_message_in(ms); + type = ms->m_type; + break; + case DLM_RCOM: + dlm_rcom_in(rc); + type = rc->rc_type; + break; + default: + log_print("invalid h_cmd %d from %u", hd->h_cmd, nodeid); + return; + } + + if (hd->h_nodeid != nodeid) { + log_print("invalid h_nodeid %d from %d lockspace %x", + hd->h_nodeid, nodeid, hd->h_lockspace); + return; + } + + ls = dlm_find_lockspace_global(hd->h_lockspace); + if (!ls) { + log_print("invalid h_lockspace %x from %d cmd %d type %d", + hd->h_lockspace, nodeid, hd->h_cmd, type); + + if (hd->h_cmd == DLM_RCOM && type == DLM_RCOM_STATUS) + dlm_send_ls_not_ready(nodeid, rc); + return; + } + + /* this rwsem allows dlm_ls_stop() to wait for all dlm_recv threads to + be inactive (in this ls) before transitioning to recovery mode */ + + down_read(&ls->ls_recv_active); + if (hd->h_cmd == DLM_MSG) + dlm_receive_message(ls, ms, nodeid); + else + dlm_receive_rcom(ls, rc, nodeid); + up_read(&ls->ls_recv_active); + + dlm_put_lockspace(ls); +} static void recover_convert_waiter(struct dlm_ls *ls, struct dlm_lkb *lkb) { diff --git a/fs/dlm/lock.h b/fs/dlm/lock.h index 1720313c22df..ada04680a1e5 100644 --- a/fs/dlm/lock.h +++ b/fs/dlm/lock.h @@ -16,7 +16,8 @@ void dlm_print_rsb(struct dlm_rsb *r); void dlm_dump_rsb(struct dlm_rsb *r); void dlm_print_lkb(struct dlm_lkb *lkb); -int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery); +void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms); +void dlm_receive_buffer(struct dlm_header *hd, int nodeid); int dlm_modes_compat(int mode1, int mode2); int dlm_find_rsb(struct dlm_ls *ls, char *name, int namelen, unsigned int flags, struct dlm_rsb **r_ret); diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c index 1dc72105ab12..628eaa669e68 100644 --- a/fs/dlm/lockspace.c +++ b/fs/dlm/lockspace.c @@ -519,6 +519,7 @@ static int new_lockspace(char *name, int namelen, void **lockspace, ls->ls_recover_seq = 0; ls->ls_recover_args = NULL; init_rwsem(&ls->ls_in_recovery); + init_rwsem(&ls->ls_recv_active); INIT_LIST_HEAD(&ls->ls_requestqueue); mutex_init(&ls->ls_requestqueue_mutex); mutex_init(&ls->ls_clear_proc_locks); diff --git a/fs/dlm/member.c b/fs/dlm/member.c index d09977528f69..e9cdcab306e2 100644 --- a/fs/dlm/member.c +++ b/fs/dlm/member.c @@ -18,10 +18,6 @@ #include "rcom.h" #include "config.h" -/* - * Following called by dlm_recoverd thread - */ - static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new) { struct dlm_member *memb = NULL; @@ -250,18 +246,30 @@ int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out) return error; } -/* - * Following called from lockspace.c - */ +/* Userspace guarantees that dlm_ls_stop() has completed on all nodes before + dlm_ls_start() is called on any of them to start the new recovery. */ int dlm_ls_stop(struct dlm_ls *ls) { int new; /* - * A stop cancels any recovery that's in progress (see RECOVERY_STOP, - * dlm_recovery_stopped()) and prevents any new locks from being - * processed (see RUNNING, dlm_locking_stopped()). + * Prevent dlm_recv from being in the middle of something when we do + * the stop. This includes ensuring dlm_recv isn't processing a + * recovery message (rcom), while dlm_recoverd is aborting and + * resetting things from an in-progress recovery. i.e. we want + * dlm_recoverd to abort its recovery without worrying about dlm_recv + * processing an rcom at the same time. Stopping dlm_recv also makes + * it easy for dlm_receive_message() to check locking stopped and add a + * message to the requestqueue without races. + */ + + down_write(&ls->ls_recv_active); + + /* + * Abort any recovery that's in progress (see RECOVERY_STOP, + * dlm_recovery_stopped()) and tell any other threads running in the + * dlm to quit any processing (see RUNNING, dlm_locking_stopped()). */ spin_lock(&ls->ls_recover_lock); @@ -270,9 +278,15 @@ int dlm_ls_stop(struct dlm_ls *ls) ls->ls_recover_seq++; spin_unlock(&ls->ls_recover_lock); + /* + * Let dlm_recv run again, now any normal messages will be saved on the + * requestqueue for later. + */ + + up_write(&ls->ls_recv_active); + /* * This in_recovery lock does two things: - * * 1) Keeps this function from returning until all threads are out * of locking routines and locking is truely stopped. * 2) Keeps any new requests from being processed until it's unlocked @@ -284,9 +298,8 @@ int dlm_ls_stop(struct dlm_ls *ls) /* * The recoverd suspend/resume makes sure that dlm_recoverd (if - * running) has noticed the clearing of RUNNING above and quit - * processing the previous recovery. This will be true for all nodes - * before any nodes start the new recovery. + * running) has noticed RECOVERY_STOP above and quit processing the + * previous recovery. */ dlm_recoverd_suspend(ls); diff --git a/fs/dlm/midcomms.c b/fs/dlm/midcomms.c index a5126e0c68a6..f8c69dda16a0 100644 --- a/fs/dlm/midcomms.c +++ b/fs/dlm/midcomms.c @@ -2,7 +2,7 @@ ******************************************************************************* ** ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. -** Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved. +** Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved. ** ** This copyrighted material is made available to anyone wishing to use, ** modify, copy, or redistribute it subject to the terms and conditions @@ -27,7 +27,6 @@ #include "dlm_internal.h" #include "lowcomms.h" #include "config.h" -#include "rcom.h" #include "lock.h" #include "midcomms.h" @@ -117,19 +116,7 @@ int dlm_process_incoming_buffer(int nodeid, const void *base, offset &= (limit - 1); len -= msglen; - switch (msg->h_cmd) { - case DLM_MSG: - dlm_receive_message(msg, nodeid, 0); - break; - - case DLM_RCOM: - dlm_receive_rcom(msg, nodeid); - break; - - default: - log_print("unknown msg type %x from %u: %u %u %u %u", - msg->h_cmd, nodeid, msglen, len, offset, ret); - } + dlm_receive_buffer(msg, nodeid); } if (msg != (struct dlm_header *) __tmp) diff --git a/fs/dlm/rcom.c b/fs/dlm/rcom.c index 188b91c027e4..ae2fd97fa4ad 100644 --- a/fs/dlm/rcom.c +++ b/fs/dlm/rcom.c @@ -2,7 +2,7 @@ ******************************************************************************* ** ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. -** Copyright (C) 2005 Red Hat, Inc. All rights reserved. +** Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved. ** ** This copyrighted material is made available to anyone wishing to use, ** modify, copy, or redistribute it subject to the terms and conditions @@ -386,7 +386,10 @@ static void receive_rcom_lock_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in) dlm_recover_process_copy(ls, rc_in); } -static int send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in) +/* If the lockspace doesn't exist then still send a status message + back; it's possible that it just doesn't have its global_id yet. */ + +int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in) { struct dlm_rcom *rc; struct rcom_config *rf; @@ -446,28 +449,11 @@ static int is_old_reply(struct dlm_ls *ls, struct dlm_rcom *rc) return rv; } -/* Called by dlm_recvd; corresponds to dlm_receive_message() but special +/* Called by dlm_recv; corresponds to dlm_receive_message() but special recovery-only comms are sent through here. */ -void dlm_receive_rcom(struct dlm_header *hd, int nodeid) +void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid) { - struct dlm_rcom *rc = (struct dlm_rcom *) hd; - struct dlm_ls *ls; - - dlm_rcom_in(rc); - - /* If the lockspace doesn't exist then still send a status message - back; it's possible that it just doesn't have its global_id yet. */ - - ls = dlm_find_lockspace_global(hd->h_lockspace); - if (!ls) { - log_print("lockspace %x from %d type %x not found", - hd->h_lockspace, nodeid, rc->rc_type); - if (rc->rc_type == DLM_RCOM_STATUS) - send_ls_not_ready(nodeid, rc); - return; - } - if (dlm_recovery_stopped(ls) && (rc->rc_type != DLM_RCOM_STATUS)) { log_debug(ls, "ignoring recovery message %x from %d", rc->rc_type, nodeid); @@ -477,12 +463,6 @@ void dlm_receive_rcom(struct dlm_header *hd, int nodeid) if (is_old_reply(ls, rc)) goto out; - if (nodeid != rc->rc_header.h_nodeid) { - log_error(ls, "bad rcom nodeid %d from %d", - rc->rc_header.h_nodeid, nodeid); - goto out; - } - switch (rc->rc_type) { case DLM_RCOM_STATUS: receive_rcom_status(ls, rc); @@ -520,6 +500,6 @@ void dlm_receive_rcom(struct dlm_header *hd, int nodeid) DLM_ASSERT(0, printk("rc_type=%x\n", rc->rc_type);); } out: - dlm_put_lockspace(ls); + return; } diff --git a/fs/dlm/rcom.h b/fs/dlm/rcom.h index d7984321ff41..b09abd29ba38 100644 --- a/fs/dlm/rcom.h +++ b/fs/dlm/rcom.h @@ -2,7 +2,7 @@ ******************************************************************************* ** ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. -** Copyright (C) 2005 Red Hat, Inc. All rights reserved. +** Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved. ** ** This copyrighted material is made available to anyone wishing to use, ** modify, copy, or redistribute it subject to the terms and conditions @@ -18,7 +18,8 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid); int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name,int last_len); int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid); int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb); -void dlm_receive_rcom(struct dlm_header *hd, int nodeid); +void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid); +int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in); #endif diff --git a/fs/dlm/recoverd.c b/fs/dlm/recoverd.c index 66575997861c..4b89e20eebe7 100644 --- a/fs/dlm/recoverd.c +++ b/fs/dlm/recoverd.c @@ -24,19 +24,28 @@ /* If the start for which we're re-enabling locking (seq) has been superseded - by a newer stop (ls_recover_seq), we need to leave locking disabled. */ + by a newer stop (ls_recover_seq), we need to leave locking disabled. + + We suspend dlm_recv threads here to avoid the race where dlm_recv a) sees + locking stopped and b) adds a message to the requestqueue, but dlm_recoverd + enables locking and clears the requestqueue between a and b. */ static int enable_locking(struct dlm_ls *ls, uint64_t seq) { int error = -EINTR; + down_write(&ls->ls_recv_active); + spin_lock(&ls->ls_recover_lock); if (ls->ls_recover_seq == seq) { set_bit(LSFL_RUNNING, &ls->ls_flags); + /* unblocks processes waiting to enter the dlm */ up_write(&ls->ls_in_recovery); error = 0; } spin_unlock(&ls->ls_recover_lock); + + up_write(&ls->ls_recv_active); return error; } diff --git a/fs/dlm/requestqueue.c b/fs/dlm/requestqueue.c index 65008d79c96d..0de04f17ccea 100644 --- a/fs/dlm/requestqueue.c +++ b/fs/dlm/requestqueue.c @@ -1,7 +1,7 @@ /****************************************************************************** ******************************************************************************* ** -** Copyright (C) 2005 Red Hat, Inc. All rights reserved. +** Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved. ** ** This copyrighted material is made available to anyone wishing to use, ** modify, copy, or redistribute it subject to the terms and conditions @@ -20,7 +20,7 @@ struct rq_entry { struct list_head list; int nodeid; - char request[1]; + char request[0]; }; /* @@ -30,42 +30,39 @@ struct rq_entry { * lockspace is enabled on some while still suspended on others. */ -int dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd) +void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd) { struct rq_entry *e; int length = hd->h_length; - int rv = 0; e = kmalloc(sizeof(struct rq_entry) + length, GFP_KERNEL); if (!e) { - log_print("dlm_add_requestqueue: out of memory\n"); - return 0; + log_print("dlm_add_requestqueue: out of memory len %d", length); + return; } e->nodeid = nodeid; memcpy(e->request, hd, length); - /* We need to check dlm_locking_stopped() after taking the mutex to - avoid a race where dlm_recoverd enables locking and runs - process_requestqueue between our earlier dlm_locking_stopped check - and this addition to the requestqueue. */ - mutex_lock(&ls->ls_requestqueue_mutex); - if (dlm_locking_stopped(ls)) - list_add_tail(&e->list, &ls->ls_requestqueue); - else { - log_debug(ls, "dlm_add_requestqueue skip from %d", nodeid); - kfree(e); - rv = -EAGAIN; - } + list_add_tail(&e->list, &ls->ls_requestqueue); mutex_unlock(&ls->ls_requestqueue_mutex); - return rv; } +/* + * Called by dlm_recoverd to process normal messages saved while recovery was + * happening. Normal locking has been enabled before this is called. dlm_recv + * upon receiving a message, will wait for all saved messages to be drained + * here before processing the message it got. If a new dlm_ls_stop() arrives + * while we're processing these saved messages, it may block trying to suspend + * dlm_recv if dlm_recv is waiting for us in dlm_wait_requestqueue. In that + * case, we don't abort since locking_stopped is still 0. If dlm_recv is not + * waiting for us, then this processing may be aborted due to locking_stopped. + */ + int dlm_process_requestqueue(struct dlm_ls *ls) { struct rq_entry *e; - struct dlm_header *hd; int error = 0; mutex_lock(&ls->ls_requestqueue_mutex); @@ -79,14 +76,7 @@ int dlm_process_requestqueue(struct dlm_ls *ls) e = list_entry(ls->ls_requestqueue.next, struct rq_entry, list); mutex_unlock(&ls->ls_requestqueue_mutex); - hd = (struct dlm_header *) e->request; - error = dlm_receive_message(hd, e->nodeid, 1); - - if (error == -EINTR) { - /* entry is left on requestqueue */ - log_debug(ls, "process_requestqueue abort eintr"); - break; - } + dlm_receive_message_saved(ls, (struct dlm_message *)e->request); mutex_lock(&ls->ls_requestqueue_mutex); list_del(&e->list); @@ -106,10 +96,12 @@ int dlm_process_requestqueue(struct dlm_ls *ls) /* * After recovery is done, locking is resumed and dlm_recoverd takes all the - * saved requests and processes them as they would have been by dlm_recvd. At - * the same time, dlm_recvd will start receiving new requests from remote - * nodes. We want to delay dlm_recvd processing new requests until - * dlm_recoverd has finished processing the old saved requests. + * saved requests and processes them as they would have been by dlm_recv. At + * the same time, dlm_recv will start receiving new requests from remote nodes. + * We want to delay dlm_recv processing new requests until dlm_recoverd has + * finished processing the old saved requests. We don't check for locking + * stopped here because dlm_ls_stop won't stop locking until it's suspended us + * (dlm_recv). */ void dlm_wait_requestqueue(struct dlm_ls *ls) @@ -118,8 +110,6 @@ void dlm_wait_requestqueue(struct dlm_ls *ls) mutex_lock(&ls->ls_requestqueue_mutex); if (list_empty(&ls->ls_requestqueue)) break; - if (dlm_locking_stopped(ls)) - break; mutex_unlock(&ls->ls_requestqueue_mutex); schedule(); } diff --git a/fs/dlm/requestqueue.h b/fs/dlm/requestqueue.h index 6a53ea03335d..aba34fc05ee4 100644 --- a/fs/dlm/requestqueue.h +++ b/fs/dlm/requestqueue.h @@ -1,7 +1,7 @@ /****************************************************************************** ******************************************************************************* ** -** Copyright (C) 2005 Red Hat, Inc. All rights reserved. +** Copyright (C) 2005-2007 Red Hat, Inc. All rights reserved. ** ** This copyrighted material is made available to anyone wishing to use, ** modify, copy, or redistribute it subject to the terms and conditions @@ -13,7 +13,7 @@ #ifndef __REQUESTQUEUE_DOT_H__ #define __REQUESTQUEUE_DOT_H__ -int dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd); +void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd); int dlm_process_requestqueue(struct dlm_ls *ls); void dlm_wait_requestqueue(struct dlm_ls *ls); void dlm_purge_requestqueue(struct dlm_ls *ls); -- cgit v1.2.3