aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-05-21 10:29:15 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-05-21 10:29:15 -0400
commit677abe49ad017679627af9d3bbd9a7ecb3d94c42 (patch)
tree7cdb7fbc95b51b288e923a3978e17ed2207a4118 /fs
parente90e4d9234c953b29267cc4fc9ad804128773313 (diff)
parent6a99be5d7b5973767b1ffa4fa68fed0738589c99 (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw
* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw: GFS2: Fix typo GFS2: stuck in inode wait, no glocks stuck GFS2: Eliminate useless err variable GFS2: Fix writing to non-page aligned gfs2_quota structures GFS2: Add some useful messages GFS2: fix quota state reporting GFS2: Various gfs2_logd improvements GFS2: glock livelock GFS2: Clean up stuffed file copying GFS2: docs update GFS2: Remove space from slab cache name
Diffstat (limited to 'fs')
-rw-r--r--fs/gfs2/aops.c8
-rw-r--r--fs/gfs2/bmap.c17
-rw-r--r--fs/gfs2/dir.c2
-rw-r--r--fs/gfs2/export.c2
-rw-r--r--fs/gfs2/glock.c3
-rw-r--r--fs/gfs2/incore.h11
-rw-r--r--fs/gfs2/inode.c101
-rw-r--r--fs/gfs2/inode.h5
-rw-r--r--fs/gfs2/log.c158
-rw-r--r--fs/gfs2/log.h1
-rw-r--r--fs/gfs2/lops.c2
-rw-r--r--fs/gfs2/main.c2
-rw-r--r--fs/gfs2/meta_io.c5
-rw-r--r--fs/gfs2/ops_fstype.c19
-rw-r--r--fs/gfs2/quota.c102
-rw-r--r--fs/gfs2/rgrp.c68
-rw-r--r--fs/gfs2/super.c11
-rw-r--r--fs/gfs2/sys.c6
-rw-r--r--fs/gfs2/trans.c18
19 files changed, 362 insertions, 179 deletions
diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index 0c1d0b82dcf1..a739a0a48067 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -418,6 +418,7 @@ static int gfs2_jdata_writepages(struct address_space *mapping,
418static int stuffed_readpage(struct gfs2_inode *ip, struct page *page) 418static int stuffed_readpage(struct gfs2_inode *ip, struct page *page)
419{ 419{
420 struct buffer_head *dibh; 420 struct buffer_head *dibh;
421 u64 dsize = i_size_read(&ip->i_inode);
421 void *kaddr; 422 void *kaddr;
422 int error; 423 int error;
423 424
@@ -437,9 +438,10 @@ static int stuffed_readpage(struct gfs2_inode *ip, struct page *page)
437 return error; 438 return error;
438 439
439 kaddr = kmap_atomic(page, KM_USER0); 440 kaddr = kmap_atomic(page, KM_USER0);
440 memcpy(kaddr, dibh->b_data + sizeof(struct gfs2_dinode), 441 if (dsize > (dibh->b_size - sizeof(struct gfs2_dinode)))
441 ip->i_disksize); 442 dsize = (dibh->b_size - sizeof(struct gfs2_dinode));
442 memset(kaddr + ip->i_disksize, 0, PAGE_CACHE_SIZE - ip->i_disksize); 443 memcpy(kaddr, dibh->b_data + sizeof(struct gfs2_dinode), dsize);
444 memset(kaddr + dsize, 0, PAGE_CACHE_SIZE - dsize);
443 kunmap_atomic(kaddr, KM_USER0); 445 kunmap_atomic(kaddr, KM_USER0);
444 flush_dcache_page(page); 446 flush_dcache_page(page);
445 brelse(dibh); 447 brelse(dibh);
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 5e411d5f4697..4a48c0f4b402 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -71,11 +71,13 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh,
71 71
72 if (!PageUptodate(page)) { 72 if (!PageUptodate(page)) {
73 void *kaddr = kmap(page); 73 void *kaddr = kmap(page);
74 u64 dsize = i_size_read(inode);
75
76 if (dsize > (dibh->b_size - sizeof(struct gfs2_dinode)))
77 dsize = dibh->b_size - sizeof(struct gfs2_dinode);
74 78
75 memcpy(kaddr, dibh->b_data + sizeof(struct gfs2_dinode), 79 memcpy(kaddr, dibh->b_data + sizeof(struct gfs2_dinode), dsize);
76 ip->i_disksize); 80 memset(kaddr + dsize, 0, PAGE_CACHE_SIZE - dsize);
77 memset(kaddr + ip->i_disksize, 0,
78 PAGE_CACHE_SIZE - ip->i_disksize);
79 kunmap(page); 81 kunmap(page);
80 82
81 SetPageUptodate(page); 83 SetPageUptodate(page);
@@ -1038,13 +1040,14 @@ static int trunc_start(struct gfs2_inode *ip, u64 size)
1038 goto out; 1040 goto out;
1039 1041
1040 if (gfs2_is_stuffed(ip)) { 1042 if (gfs2_is_stuffed(ip)) {
1041 ip->i_disksize = size; 1043 u64 dsize = size + sizeof(struct gfs2_inode);
1042 ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME; 1044 ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
1043 gfs2_trans_add_bh(ip->i_gl, dibh, 1); 1045 gfs2_trans_add_bh(ip->i_gl, dibh, 1);
1044 gfs2_dinode_out(ip, dibh->b_data); 1046 gfs2_dinode_out(ip, dibh->b_data);
1045 gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode) + size); 1047 if (dsize > dibh->b_size)
1048 dsize = dibh->b_size;
1049 gfs2_buffer_clear_tail(dibh, dsize);
1046 error = 1; 1050 error = 1;
1047
1048 } else { 1051 } else {
1049 if (size & (u64)(sdp->sd_sb.sb_bsize - 1)) 1052 if (size & (u64)(sdp->sd_sb.sb_bsize - 1))
1050 error = gfs2_block_truncate_page(ip->i_inode.i_mapping); 1053 error = gfs2_block_truncate_page(ip->i_inode.i_mapping);
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 25fddc100f18..8295c5b5d4a9 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1475,7 +1475,7 @@ struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name)
1475 inode = gfs2_inode_lookup(dir->i_sb, 1475 inode = gfs2_inode_lookup(dir->i_sb,
1476 be16_to_cpu(dent->de_type), 1476 be16_to_cpu(dent->de_type),
1477 be64_to_cpu(dent->de_inum.no_addr), 1477 be64_to_cpu(dent->de_inum.no_addr),
1478 be64_to_cpu(dent->de_inum.no_formal_ino), 0); 1478 be64_to_cpu(dent->de_inum.no_formal_ino));
1479 brelse(bh); 1479 brelse(bh);
1480 return inode; 1480 return inode;
1481 } 1481 }
diff --git a/fs/gfs2/export.c b/fs/gfs2/export.c
index c22c21174833..dfe237a3f8ad 100644
--- a/fs/gfs2/export.c
+++ b/fs/gfs2/export.c
@@ -168,7 +168,7 @@ static struct dentry *gfs2_get_dentry(struct super_block *sb,
168 if (error) 168 if (error)
169 goto fail; 169 goto fail;
170 170
171 inode = gfs2_inode_lookup(sb, DT_UNKNOWN, inum->no_addr, 0, 0); 171 inode = gfs2_inode_lookup(sb, DT_UNKNOWN, inum->no_addr, 0);
172 if (IS_ERR(inode)) { 172 if (IS_ERR(inode)) {
173 error = PTR_ERR(inode); 173 error = PTR_ERR(inode);
174 goto fail; 174 goto fail;
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 454d4b4eb36b..ddcdbf493536 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -855,6 +855,9 @@ void gfs2_holder_reinit(unsigned int state, unsigned flags, struct gfs2_holder *
855 gh->gh_flags = flags; 855 gh->gh_flags = flags;
856 gh->gh_iflags = 0; 856 gh->gh_iflags = 0;
857 gh->gh_ip = (unsigned long)__builtin_return_address(0); 857 gh->gh_ip = (unsigned long)__builtin_return_address(0);
858 if (gh->gh_owner_pid)
859 put_pid(gh->gh_owner_pid);
860 gh->gh_owner_pid = get_pid(task_pid(current));
858} 861}
859 862
860/** 863/**
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 3aac46f6853e..b5d7363b22da 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -439,9 +439,6 @@ struct gfs2_args {
439struct gfs2_tune { 439struct gfs2_tune {
440 spinlock_t gt_spin; 440 spinlock_t gt_spin;
441 441
442 unsigned int gt_incore_log_blocks;
443 unsigned int gt_log_flush_secs;
444
445 unsigned int gt_logd_secs; 442 unsigned int gt_logd_secs;
446 443
447 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */ 444 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
@@ -462,6 +459,7 @@ enum {
462 SDF_SHUTDOWN = 2, 459 SDF_SHUTDOWN = 2,
463 SDF_NOBARRIERS = 3, 460 SDF_NOBARRIERS = 3,
464 SDF_NORECOVERY = 4, 461 SDF_NORECOVERY = 4,
462 SDF_DEMOTE = 5,
465}; 463};
466 464
467#define GFS2_FSNAME_LEN 256 465#define GFS2_FSNAME_LEN 256
@@ -618,6 +616,7 @@ struct gfs2_sbd {
618 unsigned int sd_log_commited_databuf; 616 unsigned int sd_log_commited_databuf;
619 int sd_log_commited_revoke; 617 int sd_log_commited_revoke;
620 618
619 atomic_t sd_log_pinned;
621 unsigned int sd_log_num_buf; 620 unsigned int sd_log_num_buf;
622 unsigned int sd_log_num_revoke; 621 unsigned int sd_log_num_revoke;
623 unsigned int sd_log_num_rg; 622 unsigned int sd_log_num_rg;
@@ -629,15 +628,17 @@ struct gfs2_sbd {
629 struct list_head sd_log_le_databuf; 628 struct list_head sd_log_le_databuf;
630 struct list_head sd_log_le_ordered; 629 struct list_head sd_log_le_ordered;
631 630
631 atomic_t sd_log_thresh1;
632 atomic_t sd_log_thresh2;
632 atomic_t sd_log_blks_free; 633 atomic_t sd_log_blks_free;
633 struct mutex sd_log_reserve_mutex; 634 wait_queue_head_t sd_log_waitq;
635 wait_queue_head_t sd_logd_waitq;
634 636
635 u64 sd_log_sequence; 637 u64 sd_log_sequence;
636 unsigned int sd_log_head; 638 unsigned int sd_log_head;
637 unsigned int sd_log_tail; 639 unsigned int sd_log_tail;
638 int sd_log_idle; 640 int sd_log_idle;
639 641
640 unsigned long sd_log_flush_time;
641 struct rw_semaphore sd_log_flush_lock; 642 struct rw_semaphore sd_log_flush_lock;
642 atomic_t sd_log_in_flight; 643 atomic_t sd_log_in_flight;
643 wait_queue_head_t sd_log_flush_wait; 644 wait_queue_head_t sd_log_flush_wait;
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index b1bf2694fb2b..51d8061fa07a 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -158,7 +158,6 @@ void gfs2_set_iop(struct inode *inode)
158 * @sb: The super block 158 * @sb: The super block
159 * @no_addr: The inode number 159 * @no_addr: The inode number
160 * @type: The type of the inode 160 * @type: The type of the inode
161 * @skip_freeing: set this not return an inode if it is currently being freed.
162 * 161 *
163 * Returns: A VFS inode, or an error 162 * Returns: A VFS inode, or an error
164 */ 163 */
@@ -166,17 +165,14 @@ void gfs2_set_iop(struct inode *inode)
166struct inode *gfs2_inode_lookup(struct super_block *sb, 165struct inode *gfs2_inode_lookup(struct super_block *sb,
167 unsigned int type, 166 unsigned int type,
168 u64 no_addr, 167 u64 no_addr,
169 u64 no_formal_ino, int skip_freeing) 168 u64 no_formal_ino)
170{ 169{
171 struct inode *inode; 170 struct inode *inode;
172 struct gfs2_inode *ip; 171 struct gfs2_inode *ip;
173 struct gfs2_glock *io_gl; 172 struct gfs2_glock *io_gl;
174 int error; 173 int error;
175 174
176 if (skip_freeing) 175 inode = gfs2_iget(sb, no_addr);
177 inode = gfs2_iget_skip(sb, no_addr);
178 else
179 inode = gfs2_iget(sb, no_addr);
180 ip = GFS2_I(inode); 176 ip = GFS2_I(inode);
181 177
182 if (!inode) 178 if (!inode)
@@ -234,13 +230,100 @@ fail_glock:
234fail_iopen: 230fail_iopen:
235 gfs2_glock_put(io_gl); 231 gfs2_glock_put(io_gl);
236fail_put: 232fail_put:
237 ip->i_gl->gl_object = NULL; 233 if (inode->i_state & I_NEW)
234 ip->i_gl->gl_object = NULL;
238 gfs2_glock_put(ip->i_gl); 235 gfs2_glock_put(ip->i_gl);
239fail: 236fail:
240 iget_failed(inode); 237 if (inode->i_state & I_NEW)
238 iget_failed(inode);
239 else
240 iput(inode);
241 return ERR_PTR(error); 241 return ERR_PTR(error);
242} 242}
243 243
244/**
245 * gfs2_unlinked_inode_lookup - Lookup an unlinked inode for reclamation
246 * @sb: The super block
247 * no_addr: The inode number
248 * @@inode: A pointer to the inode found, if any
249 *
250 * Returns: 0 and *inode if no errors occurred. If an error occurs,
251 * the resulting *inode may or may not be NULL.
252 */
253
254int gfs2_unlinked_inode_lookup(struct super_block *sb, u64 no_addr,
255 struct inode **inode)
256{
257 struct gfs2_sbd *sdp;
258 struct gfs2_inode *ip;
259 struct gfs2_glock *io_gl;
260 int error;
261 struct gfs2_holder gh;
262
263 *inode = gfs2_iget_skip(sb, no_addr);
264
265 if (!(*inode))
266 return -ENOBUFS;
267
268 if (!((*inode)->i_state & I_NEW))
269 return -ENOBUFS;
270
271 ip = GFS2_I(*inode);
272 sdp = GFS2_SB(*inode);
273 ip->i_no_formal_ino = -1;
274
275 error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
276 if (unlikely(error))
277 goto fail;
278 ip->i_gl->gl_object = ip;
279
280 error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
281 if (unlikely(error))
282 goto fail_put;
283
284 set_bit(GIF_INVALID, &ip->i_flags);
285 error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, LM_FLAG_TRY | GL_EXACT,
286 &ip->i_iopen_gh);
287 if (unlikely(error)) {
288 if (error == GLR_TRYFAILED)
289 error = 0;
290 goto fail_iopen;
291 }
292 ip->i_iopen_gh.gh_gl->gl_object = ip;
293 gfs2_glock_put(io_gl);
294
295 (*inode)->i_mode = DT2IF(DT_UNKNOWN);
296
297 /*
298 * We must read the inode in order to work out its type in
299 * this case. Note that this doesn't happen often as we normally
300 * know the type beforehand. This code path only occurs during
301 * unlinked inode recovery (where it is safe to do this glock,
302 * which is not true in the general case).
303 */
304 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, LM_FLAG_TRY,
305 &gh);
306 if (unlikely(error)) {
307 if (error == GLR_TRYFAILED)
308 error = 0;
309 goto fail_glock;
310 }
311 /* Inode is now uptodate */
312 gfs2_glock_dq_uninit(&gh);
313 gfs2_set_iop(*inode);
314
315 return 0;
316fail_glock:
317 gfs2_glock_dq(&ip->i_iopen_gh);
318fail_iopen:
319 gfs2_glock_put(io_gl);
320fail_put:
321 ip->i_gl->gl_object = NULL;
322 gfs2_glock_put(ip->i_gl);
323fail:
324 return error;
325}
326
244static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf) 327static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
245{ 328{
246 const struct gfs2_dinode *str = buf; 329 const struct gfs2_dinode *str = buf;
@@ -862,7 +945,7 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
862 goto fail_gunlock2; 945 goto fail_gunlock2;
863 946
864 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr, 947 inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr,
865 inum.no_formal_ino, 0); 948 inum.no_formal_ino);
866 if (IS_ERR(inode)) 949 if (IS_ERR(inode))
867 goto fail_gunlock2; 950 goto fail_gunlock2;
868 951
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index c341aaf67adb..e161461d4c57 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -83,8 +83,9 @@ static inline void gfs2_inum_out(const struct gfs2_inode *ip,
83 83
84extern void gfs2_set_iop(struct inode *inode); 84extern void gfs2_set_iop(struct inode *inode);
85extern struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned type, 85extern struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned type,
86 u64 no_addr, u64 no_formal_ino, 86 u64 no_addr, u64 no_formal_ino);
87 int skip_freeing); 87extern int gfs2_unlinked_inode_lookup(struct super_block *sb, u64 no_addr,
88 struct inode **inode);
88extern struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr); 89extern struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr);
89 90
90extern int gfs2_inode_refresh(struct gfs2_inode *ip); 91extern int gfs2_inode_refresh(struct gfs2_inode *ip);
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index e5bf4b59d46e..b593f0e28f25 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -168,12 +168,11 @@ static int gfs2_ail1_empty_one(struct gfs2_sbd *sdp, struct gfs2_ail *ai, int fl
168 return list_empty(&ai->ai_ail1_list); 168 return list_empty(&ai->ai_ail1_list);
169} 169}
170 170
171static void gfs2_ail1_start(struct gfs2_sbd *sdp, int flags) 171static void gfs2_ail1_start(struct gfs2_sbd *sdp)
172{ 172{
173 struct list_head *head; 173 struct list_head *head;
174 u64 sync_gen; 174 u64 sync_gen;
175 struct list_head *first; 175 struct gfs2_ail *ai;
176 struct gfs2_ail *first_ai, *ai, *tmp;
177 int done = 0; 176 int done = 0;
178 177
179 gfs2_log_lock(sdp); 178 gfs2_log_lock(sdp);
@@ -184,21 +183,9 @@ static void gfs2_ail1_start(struct gfs2_sbd *sdp, int flags)
184 } 183 }
185 sync_gen = sdp->sd_ail_sync_gen++; 184 sync_gen = sdp->sd_ail_sync_gen++;
186 185
187 first = head->prev;
188 first_ai = list_entry(first, struct gfs2_ail, ai_list);
189 first_ai->ai_sync_gen = sync_gen;
190 gfs2_ail1_start_one(sdp, first_ai); /* This may drop log lock */
191
192 if (flags & DIO_ALL)
193 first = NULL;
194
195 while(!done) { 186 while(!done) {
196 if (first && (head->prev != first ||
197 gfs2_ail1_empty_one(sdp, first_ai, 0)))
198 break;
199
200 done = 1; 187 done = 1;
201 list_for_each_entry_safe_reverse(ai, tmp, head, ai_list) { 188 list_for_each_entry_reverse(ai, head, ai_list) {
202 if (ai->ai_sync_gen >= sync_gen) 189 if (ai->ai_sync_gen >= sync_gen)
203 continue; 190 continue;
204 ai->ai_sync_gen = sync_gen; 191 ai->ai_sync_gen = sync_gen;
@@ -290,58 +277,57 @@ static void ail2_empty(struct gfs2_sbd *sdp, unsigned int new_tail)
290 * flush time, so we ensure that we have just enough free blocks at all 277 * flush time, so we ensure that we have just enough free blocks at all
291 * times to avoid running out during a log flush. 278 * times to avoid running out during a log flush.
292 * 279 *
280 * We no longer flush the log here, instead we wake up logd to do that
281 * for us. To avoid the thundering herd and to ensure that we deal fairly
282 * with queued waiters, we use an exclusive wait. This means that when we
283 * get woken with enough journal space to get our reservation, we need to
284 * wake the next waiter on the list.
285 *
293 * Returns: errno 286 * Returns: errno
294 */ 287 */
295 288
296int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks) 289int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks)
297{ 290{
298 unsigned int try = 0;
299 unsigned reserved_blks = 6 * (4096 / sdp->sd_vfs->s_blocksize); 291 unsigned reserved_blks = 6 * (4096 / sdp->sd_vfs->s_blocksize);
292 unsigned wanted = blks + reserved_blks;
293 DEFINE_WAIT(wait);
294 int did_wait = 0;
295 unsigned int free_blocks;
300 296
301 if (gfs2_assert_warn(sdp, blks) || 297 if (gfs2_assert_warn(sdp, blks) ||
302 gfs2_assert_warn(sdp, blks <= sdp->sd_jdesc->jd_blocks)) 298 gfs2_assert_warn(sdp, blks <= sdp->sd_jdesc->jd_blocks))
303 return -EINVAL; 299 return -EINVAL;
304 300retry:
305 mutex_lock(&sdp->sd_log_reserve_mutex); 301 free_blocks = atomic_read(&sdp->sd_log_blks_free);
306 gfs2_log_lock(sdp); 302 if (unlikely(free_blocks <= wanted)) {
307 while(atomic_read(&sdp->sd_log_blks_free) <= (blks + reserved_blks)) { 303 do {
308 gfs2_log_unlock(sdp); 304 prepare_to_wait_exclusive(&sdp->sd_log_waitq, &wait,
309 gfs2_ail1_empty(sdp, 0); 305 TASK_UNINTERRUPTIBLE);
310 gfs2_log_flush(sdp, NULL); 306 wake_up(&sdp->sd_logd_waitq);
311 307 did_wait = 1;
312 if (try++) 308 if (atomic_read(&sdp->sd_log_blks_free) <= wanted)
313 gfs2_ail1_start(sdp, 0); 309 io_schedule();
314 gfs2_log_lock(sdp); 310 free_blocks = atomic_read(&sdp->sd_log_blks_free);
311 } while(free_blocks <= wanted);
312 finish_wait(&sdp->sd_log_waitq, &wait);
315 } 313 }
316 atomic_sub(blks, &sdp->sd_log_blks_free); 314 if (atomic_cmpxchg(&sdp->sd_log_blks_free, free_blocks,
315 free_blocks - blks) != free_blocks)
316 goto retry;
317 trace_gfs2_log_blocks(sdp, -blks); 317 trace_gfs2_log_blocks(sdp, -blks);
318 gfs2_log_unlock(sdp); 318
319 mutex_unlock(&sdp->sd_log_reserve_mutex); 319 /*
320 * If we waited, then so might others, wake them up _after_ we get
321 * our share of the log.
322 */
323 if (unlikely(did_wait))
324 wake_up(&sdp->sd_log_waitq);
320 325
321 down_read(&sdp->sd_log_flush_lock); 326 down_read(&sdp->sd_log_flush_lock);
322 327
323 return 0; 328 return 0;
324} 329}
325 330
326/**
327 * gfs2_log_release - Release a given number of log blocks
328 * @sdp: The GFS2 superblock
329 * @blks: The number of blocks
330 *
331 */
332
333void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks)
334{
335
336 gfs2_log_lock(sdp);
337 atomic_add(blks, &sdp->sd_log_blks_free);
338 trace_gfs2_log_blocks(sdp, blks);
339 gfs2_assert_withdraw(sdp,
340 atomic_read(&sdp->sd_log_blks_free) <= sdp->sd_jdesc->jd_blocks);
341 gfs2_log_unlock(sdp);
342 up_read(&sdp->sd_log_flush_lock);
343}
344
345static u64 log_bmap(struct gfs2_sbd *sdp, unsigned int lbn) 331static u64 log_bmap(struct gfs2_sbd *sdp, unsigned int lbn)
346{ 332{
347 struct gfs2_journal_extent *je; 333 struct gfs2_journal_extent *je;
@@ -559,11 +545,10 @@ static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail)
559 545
560 ail2_empty(sdp, new_tail); 546 ail2_empty(sdp, new_tail);
561 547
562 gfs2_log_lock(sdp);
563 atomic_add(dist, &sdp->sd_log_blks_free); 548 atomic_add(dist, &sdp->sd_log_blks_free);
564 trace_gfs2_log_blocks(sdp, dist); 549 trace_gfs2_log_blocks(sdp, dist);
565 gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <= sdp->sd_jdesc->jd_blocks); 550 gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <=
566 gfs2_log_unlock(sdp); 551 sdp->sd_jdesc->jd_blocks);
567 552
568 sdp->sd_log_tail = new_tail; 553 sdp->sd_log_tail = new_tail;
569} 554}
@@ -615,6 +600,7 @@ static void log_write_header(struct gfs2_sbd *sdp, u32 flags, int pull)
615 if (buffer_eopnotsupp(bh)) { 600 if (buffer_eopnotsupp(bh)) {
616 clear_buffer_eopnotsupp(bh); 601 clear_buffer_eopnotsupp(bh);
617 set_buffer_uptodate(bh); 602 set_buffer_uptodate(bh);
603 fs_info(sdp, "barrier sync failed - disabling barriers\n");
618 set_bit(SDF_NOBARRIERS, &sdp->sd_flags); 604 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
619 lock_buffer(bh); 605 lock_buffer(bh);
620skip_barrier: 606skip_barrier:
@@ -822,6 +808,13 @@ static void buf_lo_incore_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
822 * @sdp: the filesystem 808 * @sdp: the filesystem
823 * @tr: the transaction 809 * @tr: the transaction
824 * 810 *
811 * We wake up gfs2_logd if the number of pinned blocks exceed thresh1
812 * or the total number of used blocks (pinned blocks plus AIL blocks)
813 * is greater than thresh2.
814 *
815 * At mount time thresh1 is 1/3rd of journal size, thresh2 is 2/3rd of
816 * journal size.
817 *
825 * Returns: errno 818 * Returns: errno
826 */ 819 */
827 820
@@ -832,10 +825,10 @@ void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
832 825
833 up_read(&sdp->sd_log_flush_lock); 826 up_read(&sdp->sd_log_flush_lock);
834 827
835 gfs2_log_lock(sdp); 828 if (atomic_read(&sdp->sd_log_pinned) > atomic_read(&sdp->sd_log_thresh1) ||
836 if (sdp->sd_log_num_buf > gfs2_tune_get(sdp, gt_incore_log_blocks)) 829 ((sdp->sd_jdesc->jd_blocks - atomic_read(&sdp->sd_log_blks_free)) >
837 wake_up_process(sdp->sd_logd_process); 830 atomic_read(&sdp->sd_log_thresh2)))
838 gfs2_log_unlock(sdp); 831 wake_up(&sdp->sd_logd_waitq);
839} 832}
840 833
841/** 834/**
@@ -882,13 +875,23 @@ void gfs2_meta_syncfs(struct gfs2_sbd *sdp)
882{ 875{
883 gfs2_log_flush(sdp, NULL); 876 gfs2_log_flush(sdp, NULL);
884 for (;;) { 877 for (;;) {
885 gfs2_ail1_start(sdp, DIO_ALL); 878 gfs2_ail1_start(sdp);
886 if (gfs2_ail1_empty(sdp, DIO_ALL)) 879 if (gfs2_ail1_empty(sdp, DIO_ALL))
887 break; 880 break;
888 msleep(10); 881 msleep(10);
889 } 882 }
890} 883}
891 884
885static inline int gfs2_jrnl_flush_reqd(struct gfs2_sbd *sdp)
886{
887 return (atomic_read(&sdp->sd_log_pinned) >= atomic_read(&sdp->sd_log_thresh1));
888}
889
890static inline int gfs2_ail_flush_reqd(struct gfs2_sbd *sdp)
891{
892 unsigned int used_blocks = sdp->sd_jdesc->jd_blocks - atomic_read(&sdp->sd_log_blks_free);
893 return used_blocks >= atomic_read(&sdp->sd_log_thresh2);
894}
892 895
893/** 896/**
894 * gfs2_logd - Update log tail as Active Items get flushed to in-place blocks 897 * gfs2_logd - Update log tail as Active Items get flushed to in-place blocks
@@ -901,28 +904,43 @@ void gfs2_meta_syncfs(struct gfs2_sbd *sdp)
901int gfs2_logd(void *data) 904int gfs2_logd(void *data)
902{ 905{
903 struct gfs2_sbd *sdp = data; 906 struct gfs2_sbd *sdp = data;
904 unsigned long t; 907 unsigned long t = 1;
905 int need_flush; 908 DEFINE_WAIT(wait);
909 unsigned preflush;
906 910
907 while (!kthread_should_stop()) { 911 while (!kthread_should_stop()) {
908 /* Advance the log tail */
909 912
910 t = sdp->sd_log_flush_time + 913 preflush = atomic_read(&sdp->sd_log_pinned);
911 gfs2_tune_get(sdp, gt_log_flush_secs) * HZ; 914 if (gfs2_jrnl_flush_reqd(sdp) || t == 0) {
915 gfs2_ail1_empty(sdp, DIO_ALL);
916 gfs2_log_flush(sdp, NULL);
917 gfs2_ail1_empty(sdp, DIO_ALL);
918 }
912 919
913 gfs2_ail1_empty(sdp, DIO_ALL); 920 if (gfs2_ail_flush_reqd(sdp)) {
914 gfs2_log_lock(sdp); 921 gfs2_ail1_start(sdp);
915 need_flush = sdp->sd_log_num_buf > gfs2_tune_get(sdp, gt_incore_log_blocks); 922 io_schedule();
916 gfs2_log_unlock(sdp); 923 gfs2_ail1_empty(sdp, 0);
917 if (need_flush || time_after_eq(jiffies, t)) {
918 gfs2_log_flush(sdp, NULL); 924 gfs2_log_flush(sdp, NULL);
919 sdp->sd_log_flush_time = jiffies; 925 gfs2_ail1_empty(sdp, DIO_ALL);
920 } 926 }
921 927
928 wake_up(&sdp->sd_log_waitq);
922 t = gfs2_tune_get(sdp, gt_logd_secs) * HZ; 929 t = gfs2_tune_get(sdp, gt_logd_secs) * HZ;
923 if (freezing(current)) 930 if (freezing(current))
924 refrigerator(); 931 refrigerator();
925 schedule_timeout_interruptible(t); 932
933 do {
934 prepare_to_wait(&sdp->sd_logd_waitq, &wait,
935 TASK_UNINTERRUPTIBLE);
936 if (!gfs2_ail_flush_reqd(sdp) &&
937 !gfs2_jrnl_flush_reqd(sdp) &&
938 !kthread_should_stop())
939 t = schedule_timeout(t);
940 } while(t && !gfs2_ail_flush_reqd(sdp) &&
941 !gfs2_jrnl_flush_reqd(sdp) &&
942 !kthread_should_stop());
943 finish_wait(&sdp->sd_logd_waitq, &wait);
926 } 944 }
927 945
928 return 0; 946 return 0;
diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h
index 7c64510ccfd2..eb570b4ad443 100644
--- a/fs/gfs2/log.h
+++ b/fs/gfs2/log.h
@@ -51,7 +51,6 @@ unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct,
51 unsigned int ssize); 51 unsigned int ssize);
52 52
53int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks); 53int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks);
54void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks);
55void gfs2_log_incr_head(struct gfs2_sbd *sdp); 54void gfs2_log_incr_head(struct gfs2_sbd *sdp);
56 55
57struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp); 56struct buffer_head *gfs2_log_get_buf(struct gfs2_sbd *sdp);
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index adc260fbea90..bf33f822058d 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -54,6 +54,7 @@ static void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh)
54 if (bd->bd_ail) 54 if (bd->bd_ail)
55 list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list); 55 list_move(&bd->bd_ail_st_list, &bd->bd_ail->ai_ail2_list);
56 get_bh(bh); 56 get_bh(bh);
57 atomic_inc(&sdp->sd_log_pinned);
57 trace_gfs2_pin(bd, 1); 58 trace_gfs2_pin(bd, 1);
58} 59}
59 60
@@ -94,6 +95,7 @@ static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh,
94 trace_gfs2_pin(bd, 0); 95 trace_gfs2_pin(bd, 0);
95 gfs2_log_unlock(sdp); 96 gfs2_log_unlock(sdp);
96 unlock_buffer(bh); 97 unlock_buffer(bh);
98 atomic_dec(&sdp->sd_log_pinned);
97} 99}
98 100
99 101
diff --git a/fs/gfs2/main.c b/fs/gfs2/main.c
index a88fadc704bb..fb2a5f93b7c3 100644
--- a/fs/gfs2/main.c
+++ b/fs/gfs2/main.c
@@ -94,7 +94,7 @@ static int __init init_gfs2_fs(void)
94 if (!gfs2_glock_cachep) 94 if (!gfs2_glock_cachep)
95 goto fail; 95 goto fail;
96 96
97 gfs2_glock_aspace_cachep = kmem_cache_create("gfs2_glock (aspace)", 97 gfs2_glock_aspace_cachep = kmem_cache_create("gfs2_glock(aspace)",
98 sizeof(struct gfs2_glock) + 98 sizeof(struct gfs2_glock) +
99 sizeof(struct address_space), 99 sizeof(struct address_space),
100 0, 0, gfs2_init_gl_aspace_once); 100 0, 0, gfs2_init_gl_aspace_once);
diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c
index 0bb12c80937a..18176d0b75d7 100644
--- a/fs/gfs2/meta_io.c
+++ b/fs/gfs2/meta_io.c
@@ -34,7 +34,6 @@
34 34
35static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wbc) 35static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wbc)
36{ 36{
37 int err;
38 struct buffer_head *bh, *head; 37 struct buffer_head *bh, *head;
39 int nr_underway = 0; 38 int nr_underway = 0;
40 int write_op = (1 << BIO_RW_META) | ((wbc->sync_mode == WB_SYNC_ALL ? 39 int write_op = (1 << BIO_RW_META) | ((wbc->sync_mode == WB_SYNC_ALL ?
@@ -86,11 +85,10 @@ static int gfs2_aspace_writepage(struct page *page, struct writeback_control *wb
86 } while (bh != head); 85 } while (bh != head);
87 unlock_page(page); 86 unlock_page(page);
88 87
89 err = 0;
90 if (nr_underway == 0) 88 if (nr_underway == 0)
91 end_page_writeback(page); 89 end_page_writeback(page);
92 90
93 return err; 91 return 0;
94} 92}
95 93
96const struct address_space_operations gfs2_meta_aops = { 94const struct address_space_operations gfs2_meta_aops = {
@@ -313,6 +311,7 @@ void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int
313 struct gfs2_bufdata *bd = bh->b_private; 311 struct gfs2_bufdata *bd = bh->b_private;
314 312
315 if (test_clear_buffer_pinned(bh)) { 313 if (test_clear_buffer_pinned(bh)) {
314 atomic_dec(&sdp->sd_log_pinned);
316 list_del_init(&bd->bd_le.le_list); 315 list_del_init(&bd->bd_le.le_list);
317 if (meta) { 316 if (meta) {
318 gfs2_assert_warn(sdp, sdp->sd_log_num_buf); 317 gfs2_assert_warn(sdp, sdp->sd_log_num_buf);
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index c1309ed1c496..3593b3a7290e 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -57,8 +57,6 @@ static void gfs2_tune_init(struct gfs2_tune *gt)
57{ 57{
58 spin_lock_init(&gt->gt_spin); 58 spin_lock_init(&gt->gt_spin);
59 59
60 gt->gt_incore_log_blocks = 1024;
61 gt->gt_logd_secs = 1;
62 gt->gt_quota_simul_sync = 64; 60 gt->gt_quota_simul_sync = 64;
63 gt->gt_quota_warn_period = 10; 61 gt->gt_quota_warn_period = 10;
64 gt->gt_quota_scale_num = 1; 62 gt->gt_quota_scale_num = 1;
@@ -101,14 +99,15 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb)
101 spin_lock_init(&sdp->sd_trunc_lock); 99 spin_lock_init(&sdp->sd_trunc_lock);
102 100
103 spin_lock_init(&sdp->sd_log_lock); 101 spin_lock_init(&sdp->sd_log_lock);
104 102 atomic_set(&sdp->sd_log_pinned, 0);
105 INIT_LIST_HEAD(&sdp->sd_log_le_buf); 103 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
106 INIT_LIST_HEAD(&sdp->sd_log_le_revoke); 104 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
107 INIT_LIST_HEAD(&sdp->sd_log_le_rg); 105 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
108 INIT_LIST_HEAD(&sdp->sd_log_le_databuf); 106 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
109 INIT_LIST_HEAD(&sdp->sd_log_le_ordered); 107 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
110 108
111 mutex_init(&sdp->sd_log_reserve_mutex); 109 init_waitqueue_head(&sdp->sd_log_waitq);
110 init_waitqueue_head(&sdp->sd_logd_waitq);
112 INIT_LIST_HEAD(&sdp->sd_ail1_list); 111 INIT_LIST_HEAD(&sdp->sd_ail1_list);
113 INIT_LIST_HEAD(&sdp->sd_ail2_list); 112 INIT_LIST_HEAD(&sdp->sd_ail2_list);
114 113
@@ -487,7 +486,7 @@ static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
487 struct dentry *dentry; 486 struct dentry *dentry;
488 struct inode *inode; 487 struct inode *inode;
489 488
490 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); 489 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
491 if (IS_ERR(inode)) { 490 if (IS_ERR(inode)) {
492 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); 491 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
493 return PTR_ERR(inode); 492 return PTR_ERR(inode);
@@ -733,6 +732,8 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
733 if (sdp->sd_args.ar_spectator) { 732 if (sdp->sd_args.ar_spectator) {
734 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); 733 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
735 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); 734 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
735 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
736 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
736 } else { 737 } else {
737 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { 738 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
738 fs_err(sdp, "can't mount journal #%u\n", 739 fs_err(sdp, "can't mount journal #%u\n",
@@ -770,6 +771,8 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
770 goto fail_jinode_gh; 771 goto fail_jinode_gh;
771 } 772 }
772 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); 773 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
774 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
775 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
773 776
774 /* Map the extents for this journal's blocks */ 777 /* Map the extents for this journal's blocks */
775 map_journal_extents(sdp); 778 map_journal_extents(sdp);
@@ -951,8 +954,6 @@ static int init_threads(struct gfs2_sbd *sdp, int undo)
951 if (undo) 954 if (undo)
952 goto fail_quotad; 955 goto fail_quotad;
953 956
954 sdp->sd_log_flush_time = jiffies;
955
956 p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); 957 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
957 error = IS_ERR(p); 958 error = IS_ERR(p);
958 if (error) { 959 if (error) {
@@ -1160,7 +1161,7 @@ static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent
1160 GFS2_BASIC_BLOCK_SHIFT; 1161 GFS2_BASIC_BLOCK_SHIFT;
1161 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; 1162 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1162 1163
1163 sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit; 1164 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
1164 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum; 1165 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1165 if (sdp->sd_args.ar_statfs_quantum) { 1166 if (sdp->sd_args.ar_statfs_quantum) {
1166 sdp->sd_tune.gt_statfs_slow = 0; 1167 sdp->sd_tune.gt_statfs_slow = 0;
@@ -1323,7 +1324,7 @@ static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1323 memset(&args, 0, sizeof(args)); 1324 memset(&args, 0, sizeof(args));
1324 args.ar_quota = GFS2_QUOTA_DEFAULT; 1325 args.ar_quota = GFS2_QUOTA_DEFAULT;
1325 args.ar_data = GFS2_DATA_DEFAULT; 1326 args.ar_data = GFS2_DATA_DEFAULT;
1326 args.ar_commit = 60; 1327 args.ar_commit = 30;
1327 args.ar_statfs_quantum = 30; 1328 args.ar_statfs_quantum = 30;
1328 args.ar_quota_quantum = 60; 1329 args.ar_quota_quantum = 60;
1329 args.ar_errors = GFS2_ERRORS_DEFAULT; 1330 args.ar_errors = GFS2_ERRORS_DEFAULT;
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index 6dbcbad6ab17..d5f4661287f9 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -637,15 +637,40 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
637 unsigned blocksize, iblock, pos; 637 unsigned blocksize, iblock, pos;
638 struct buffer_head *bh, *dibh; 638 struct buffer_head *bh, *dibh;
639 struct page *page; 639 struct page *page;
640 void *kaddr; 640 void *kaddr, *ptr;
641 struct gfs2_quota *qp; 641 struct gfs2_quota q, *qp;
642 s64 value; 642 int err, nbytes;
643 int err = -EIO;
644 u64 size; 643 u64 size;
645 644
646 if (gfs2_is_stuffed(ip)) 645 if (gfs2_is_stuffed(ip))
647 gfs2_unstuff_dinode(ip, NULL); 646 gfs2_unstuff_dinode(ip, NULL);
648 647
648 memset(&q, 0, sizeof(struct gfs2_quota));
649 err = gfs2_internal_read(ip, NULL, (char *)&q, &loc, sizeof(q));
650 if (err < 0)
651 return err;
652
653 err = -EIO;
654 qp = &q;
655 qp->qu_value = be64_to_cpu(qp->qu_value);
656 qp->qu_value += change;
657 qp->qu_value = cpu_to_be64(qp->qu_value);
658 qd->qd_qb.qb_value = qp->qu_value;
659 if (fdq) {
660 if (fdq->d_fieldmask & FS_DQ_BSOFT) {
661 qp->qu_warn = cpu_to_be64(fdq->d_blk_softlimit);
662 qd->qd_qb.qb_warn = qp->qu_warn;
663 }
664 if (fdq->d_fieldmask & FS_DQ_BHARD) {
665 qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit);
666 qd->qd_qb.qb_limit = qp->qu_limit;
667 }
668 }
669
670 /* Write the quota into the quota file on disk */
671 ptr = qp;
672 nbytes = sizeof(struct gfs2_quota);
673get_a_page:
649 page = grab_cache_page(mapping, index); 674 page = grab_cache_page(mapping, index);
650 if (!page) 675 if (!page)
651 return -ENOMEM; 676 return -ENOMEM;
@@ -667,7 +692,12 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
667 if (!buffer_mapped(bh)) { 692 if (!buffer_mapped(bh)) {
668 gfs2_block_map(inode, iblock, bh, 1); 693 gfs2_block_map(inode, iblock, bh, 1);
669 if (!buffer_mapped(bh)) 694 if (!buffer_mapped(bh))
670 goto unlock; 695 goto unlock_out;
696 /* If it's a newly allocated disk block for quota, zero it */
697 if (buffer_new(bh)) {
698 memset(bh->b_data, 0, bh->b_size);
699 set_buffer_uptodate(bh);
700 }
671 } 701 }
672 702
673 if (PageUptodate(page)) 703 if (PageUptodate(page))
@@ -677,32 +707,34 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
677 ll_rw_block(READ_META, 1, &bh); 707 ll_rw_block(READ_META, 1, &bh);
678 wait_on_buffer(bh); 708 wait_on_buffer(bh);
679 if (!buffer_uptodate(bh)) 709 if (!buffer_uptodate(bh))
680 goto unlock; 710 goto unlock_out;
681 } 711 }
682 712
683 gfs2_trans_add_bh(ip->i_gl, bh, 0); 713 gfs2_trans_add_bh(ip->i_gl, bh, 0);
684 714
685 kaddr = kmap_atomic(page, KM_USER0); 715 kaddr = kmap_atomic(page, KM_USER0);
686 qp = kaddr + offset; 716 if (offset + sizeof(struct gfs2_quota) > PAGE_CACHE_SIZE)
687 value = (s64)be64_to_cpu(qp->qu_value) + change; 717 nbytes = PAGE_CACHE_SIZE - offset;
688 qp->qu_value = cpu_to_be64(value); 718 memcpy(kaddr + offset, ptr, nbytes);
689 qd->qd_qb.qb_value = qp->qu_value;
690 if (fdq) {
691 if (fdq->d_fieldmask & FS_DQ_BSOFT) {
692 qp->qu_warn = cpu_to_be64(fdq->d_blk_softlimit);
693 qd->qd_qb.qb_warn = qp->qu_warn;
694 }
695 if (fdq->d_fieldmask & FS_DQ_BHARD) {
696 qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit);
697 qd->qd_qb.qb_limit = qp->qu_limit;
698 }
699 }
700 flush_dcache_page(page); 719 flush_dcache_page(page);
701 kunmap_atomic(kaddr, KM_USER0); 720 kunmap_atomic(kaddr, KM_USER0);
721 unlock_page(page);
722 page_cache_release(page);
702 723
724 /* If quota straddles page boundary, we need to update the rest of the
725 * quota at the beginning of the next page */
726 if (offset != 0) { /* first page, offset is closer to PAGE_CACHE_SIZE */
727 ptr = ptr + nbytes;
728 nbytes = sizeof(struct gfs2_quota) - nbytes;
729 offset = 0;
730 index++;
731 goto get_a_page;
732 }
733
734 /* Update the disk inode timestamp and size (if extended) */
703 err = gfs2_meta_inode_buffer(ip, &dibh); 735 err = gfs2_meta_inode_buffer(ip, &dibh);
704 if (err) 736 if (err)
705 goto unlock; 737 goto out;
706 738
707 size = loc + sizeof(struct gfs2_quota); 739 size = loc + sizeof(struct gfs2_quota);
708 if (size > inode->i_size) { 740 if (size > inode->i_size) {
@@ -715,7 +747,9 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
715 brelse(dibh); 747 brelse(dibh);
716 mark_inode_dirty(inode); 748 mark_inode_dirty(inode);
717 749
718unlock: 750out:
751 return err;
752unlock_out:
719 unlock_page(page); 753 unlock_page(page);
720 page_cache_release(page); 754 page_cache_release(page);
721 return err; 755 return err;
@@ -779,8 +813,10 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
779 * rgrp since it won't be allocated during the transaction 813 * rgrp since it won't be allocated during the transaction
780 */ 814 */
781 al->al_requested = 1; 815 al->al_requested = 1;
782 /* +1 in the end for block requested above for unstuffing */ 816 /* +3 in the end for unstuffing block, inode size update block
783 blocks = num_qd * data_blocks + RES_DINODE + num_qd + 1; 817 * and another block in case quota straddles page boundary and
818 * two blocks need to be updated instead of 1 */
819 blocks = num_qd * data_blocks + RES_DINODE + num_qd + 3;
784 820
785 if (nalloc) 821 if (nalloc)
786 al->al_requested += nalloc * (data_blocks + ind_blocks); 822 al->al_requested += nalloc * (data_blocks + ind_blocks);
@@ -1418,10 +1454,18 @@ static int gfs2_quota_get_xstate(struct super_block *sb,
1418 1454
1419 memset(fqs, 0, sizeof(struct fs_quota_stat)); 1455 memset(fqs, 0, sizeof(struct fs_quota_stat));
1420 fqs->qs_version = FS_QSTAT_VERSION; 1456 fqs->qs_version = FS_QSTAT_VERSION;
1421 if (sdp->sd_args.ar_quota == GFS2_QUOTA_ON) 1457
1422 fqs->qs_flags = (XFS_QUOTA_UDQ_ENFD | XFS_QUOTA_GDQ_ENFD); 1458 switch (sdp->sd_args.ar_quota) {
1423 else if (sdp->sd_args.ar_quota == GFS2_QUOTA_ACCOUNT) 1459 case GFS2_QUOTA_ON:
1424 fqs->qs_flags = (XFS_QUOTA_UDQ_ACCT | XFS_QUOTA_GDQ_ACCT); 1460 fqs->qs_flags |= (XFS_QUOTA_UDQ_ENFD | XFS_QUOTA_GDQ_ENFD);
1461 /*FALLTHRU*/
1462 case GFS2_QUOTA_ACCOUNT:
1463 fqs->qs_flags |= (XFS_QUOTA_UDQ_ACCT | XFS_QUOTA_GDQ_ACCT);
1464 break;
1465 case GFS2_QUOTA_OFF:
1466 break;
1467 }
1468
1425 if (sdp->sd_quota_inode) { 1469 if (sdp->sd_quota_inode) {
1426 fqs->qs_uquota.qfs_ino = GFS2_I(sdp->sd_quota_inode)->i_no_addr; 1470 fqs->qs_uquota.qfs_ino = GFS2_I(sdp->sd_quota_inode)->i_no_addr;
1427 fqs->qs_uquota.qfs_nblks = sdp->sd_quota_inode->i_blocks; 1471 fqs->qs_uquota.qfs_nblks = sdp->sd_quota_inode->i_blocks;
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 503b842f3ba2..8bce73ed4d8e 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -948,13 +948,13 @@ static int try_rgrp_fit(struct gfs2_rgrpd *rgd, struct gfs2_alloc *al)
948 * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes 948 * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes
949 * @rgd: The rgrp 949 * @rgd: The rgrp
950 * 950 *
951 * Returns: The inode, if one has been found 951 * Returns: 0 if no error
952 * The inode, if one has been found, in inode.
952 */ 953 */
953 954
954static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, 955static u64 try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked,
955 u64 skip) 956 u64 skip)
956{ 957{
957 struct inode *inode;
958 u32 goal = 0, block; 958 u32 goal = 0, block;
959 u64 no_addr; 959 u64 no_addr;
960 struct gfs2_sbd *sdp = rgd->rd_sbd; 960 struct gfs2_sbd *sdp = rgd->rd_sbd;
@@ -979,14 +979,11 @@ static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked,
979 if (no_addr == skip) 979 if (no_addr == skip)
980 continue; 980 continue;
981 *last_unlinked = no_addr; 981 *last_unlinked = no_addr;
982 inode = gfs2_inode_lookup(rgd->rd_sbd->sd_vfs, DT_UNKNOWN, 982 return no_addr;
983 no_addr, -1, 1);
984 if (!IS_ERR(inode))
985 return inode;
986 } 983 }
987 984
988 rgd->rd_flags &= ~GFS2_RDF_CHECK; 985 rgd->rd_flags &= ~GFS2_RDF_CHECK;
989 return NULL; 986 return 0;
990} 987}
991 988
992/** 989/**
@@ -1067,11 +1064,12 @@ static void forward_rgrp_set(struct gfs2_sbd *sdp, struct gfs2_rgrpd *rgd)
1067 * Try to acquire rgrp in way which avoids contending with others. 1064 * Try to acquire rgrp in way which avoids contending with others.
1068 * 1065 *
1069 * Returns: errno 1066 * Returns: errno
1067 * unlinked: the block address of an unlinked block to be reclaimed
1070 */ 1068 */
1071 1069
1072static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked) 1070static int get_local_rgrp(struct gfs2_inode *ip, u64 *unlinked,
1071 u64 *last_unlinked)
1073{ 1072{
1074 struct inode *inode = NULL;
1075 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); 1073 struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
1076 struct gfs2_rgrpd *rgd, *begin = NULL; 1074 struct gfs2_rgrpd *rgd, *begin = NULL;
1077 struct gfs2_alloc *al = ip->i_alloc; 1075 struct gfs2_alloc *al = ip->i_alloc;
@@ -1080,6 +1078,7 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
1080 int loops = 0; 1078 int loops = 0;
1081 int error, rg_locked; 1079 int error, rg_locked;
1082 1080
1081 *unlinked = 0;
1083 rgd = gfs2_blk2rgrpd(sdp, ip->i_goal); 1082 rgd = gfs2_blk2rgrpd(sdp, ip->i_goal);
1084 1083
1085 while (rgd) { 1084 while (rgd) {
@@ -1096,19 +1095,24 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
1096 case 0: 1095 case 0:
1097 if (try_rgrp_fit(rgd, al)) 1096 if (try_rgrp_fit(rgd, al))
1098 goto out; 1097 goto out;
1099 if (rgd->rd_flags & GFS2_RDF_CHECK) 1098 /* If the rg came in already locked, there's no
1100 inode = try_rgrp_unlink(rgd, last_unlinked, ip->i_no_addr); 1099 way we can recover from a failed try_rgrp_unlink
1100 because that would require an iput which can only
1101 happen after the rgrp is unlocked. */
1102 if (!rg_locked && rgd->rd_flags & GFS2_RDF_CHECK)
1103 *unlinked = try_rgrp_unlink(rgd, last_unlinked,
1104 ip->i_no_addr);
1101 if (!rg_locked) 1105 if (!rg_locked)
1102 gfs2_glock_dq_uninit(&al->al_rgd_gh); 1106 gfs2_glock_dq_uninit(&al->al_rgd_gh);
1103 if (inode) 1107 if (*unlinked)
1104 return inode; 1108 return -EAGAIN;
1105 /* fall through */ 1109 /* fall through */
1106 case GLR_TRYFAILED: 1110 case GLR_TRYFAILED:
1107 rgd = recent_rgrp_next(rgd); 1111 rgd = recent_rgrp_next(rgd);
1108 break; 1112 break;
1109 1113
1110 default: 1114 default:
1111 return ERR_PTR(error); 1115 return error;
1112 } 1116 }
1113 } 1117 }
1114 1118
@@ -1130,12 +1134,13 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
1130 case 0: 1134 case 0:
1131 if (try_rgrp_fit(rgd, al)) 1135 if (try_rgrp_fit(rgd, al))
1132 goto out; 1136 goto out;
1133 if (rgd->rd_flags & GFS2_RDF_CHECK) 1137 if (!rg_locked && rgd->rd_flags & GFS2_RDF_CHECK)
1134 inode = try_rgrp_unlink(rgd, last_unlinked, ip->i_no_addr); 1138 *unlinked = try_rgrp_unlink(rgd, last_unlinked,
1139 ip->i_no_addr);
1135 if (!rg_locked) 1140 if (!rg_locked)
1136 gfs2_glock_dq_uninit(&al->al_rgd_gh); 1141 gfs2_glock_dq_uninit(&al->al_rgd_gh);
1137 if (inode) 1142 if (*unlinked)
1138 return inode; 1143 return -EAGAIN;
1139 break; 1144 break;
1140 1145
1141 case GLR_TRYFAILED: 1146 case GLR_TRYFAILED:
@@ -1143,7 +1148,7 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
1143 break; 1148 break;
1144 1149
1145 default: 1150 default:
1146 return ERR_PTR(error); 1151 return error;
1147 } 1152 }
1148 1153
1149 rgd = gfs2_rgrpd_get_next(rgd); 1154 rgd = gfs2_rgrpd_get_next(rgd);
@@ -1152,7 +1157,7 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
1152 1157
1153 if (rgd == begin) { 1158 if (rgd == begin) {
1154 if (++loops >= 3) 1159 if (++loops >= 3)
1155 return ERR_PTR(-ENOSPC); 1160 return -ENOSPC;
1156 if (!skipped) 1161 if (!skipped)
1157 loops++; 1162 loops++;
1158 flags = 0; 1163 flags = 0;
@@ -1172,7 +1177,7 @@ out:
1172 forward_rgrp_set(sdp, rgd); 1177 forward_rgrp_set(sdp, rgd);
1173 } 1178 }
1174 1179
1175 return NULL; 1180 return 0;
1176} 1181}
1177 1182
1178/** 1183/**
@@ -1188,7 +1193,7 @@ int gfs2_inplace_reserve_i(struct gfs2_inode *ip, char *file, unsigned int line)
1188 struct gfs2_alloc *al = ip->i_alloc; 1193 struct gfs2_alloc *al = ip->i_alloc;
1189 struct inode *inode; 1194 struct inode *inode;
1190 int error = 0; 1195 int error = 0;
1191 u64 last_unlinked = NO_BLOCK; 1196 u64 last_unlinked = NO_BLOCK, unlinked;
1192 1197
1193 if (gfs2_assert_warn(sdp, al->al_requested)) 1198 if (gfs2_assert_warn(sdp, al->al_requested))
1194 return -EINVAL; 1199 return -EINVAL;
@@ -1204,14 +1209,19 @@ try_again:
1204 if (error) 1209 if (error)
1205 return error; 1210 return error;
1206 1211
1207 inode = get_local_rgrp(ip, &last_unlinked); 1212 error = get_local_rgrp(ip, &unlinked, &last_unlinked);
1208 if (inode) { 1213 if (error) {
1209 if (ip != GFS2_I(sdp->sd_rindex)) 1214 if (ip != GFS2_I(sdp->sd_rindex))
1210 gfs2_glock_dq_uninit(&al->al_ri_gh); 1215 gfs2_glock_dq_uninit(&al->al_ri_gh);
1211 if (IS_ERR(inode)) 1216 if (error != -EAGAIN)
1212 return PTR_ERR(inode); 1217 return error;
1213 iput(inode); 1218 error = gfs2_unlinked_inode_lookup(ip->i_inode.i_sb,
1219 unlinked, &inode);
1220 if (inode)
1221 iput(inode);
1214 gfs2_log_flush(sdp, NULL); 1222 gfs2_log_flush(sdp, NULL);
1223 if (error == GLR_TRYFAILED)
1224 error = 0;
1215 goto try_again; 1225 goto try_again;
1216 } 1226 }
1217 1227
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 50aac606b990..4d1aad38f1b1 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1113,7 +1113,7 @@ static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
1113 int error; 1113 int error;
1114 1114
1115 spin_lock(&gt->gt_spin); 1115 spin_lock(&gt->gt_spin);
1116 args.ar_commit = gt->gt_log_flush_secs; 1116 args.ar_commit = gt->gt_logd_secs;
1117 args.ar_quota_quantum = gt->gt_quota_quantum; 1117 args.ar_quota_quantum = gt->gt_quota_quantum;
1118 if (gt->gt_statfs_slow) 1118 if (gt->gt_statfs_slow)
1119 args.ar_statfs_quantum = 0; 1119 args.ar_statfs_quantum = 0;
@@ -1160,7 +1160,7 @@ static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
1160 else 1160 else
1161 clear_bit(SDF_NOBARRIERS, &sdp->sd_flags); 1161 clear_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1162 spin_lock(&gt->gt_spin); 1162 spin_lock(&gt->gt_spin);
1163 gt->gt_log_flush_secs = args.ar_commit; 1163 gt->gt_logd_secs = args.ar_commit;
1164 gt->gt_quota_quantum = args.ar_quota_quantum; 1164 gt->gt_quota_quantum = args.ar_quota_quantum;
1165 if (args.ar_statfs_quantum) { 1165 if (args.ar_statfs_quantum) {
1166 gt->gt_statfs_slow = 0; 1166 gt->gt_statfs_slow = 0;
@@ -1305,8 +1305,8 @@ static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1305 } 1305 }
1306 if (args->ar_discard) 1306 if (args->ar_discard)
1307 seq_printf(s, ",discard"); 1307 seq_printf(s, ",discard");
1308 val = sdp->sd_tune.gt_log_flush_secs; 1308 val = sdp->sd_tune.gt_logd_secs;
1309 if (val != 60) 1309 if (val != 30)
1310 seq_printf(s, ",commit=%d", val); 1310 seq_printf(s, ",commit=%d", val);
1311 val = sdp->sd_tune.gt_statfs_quantum; 1311 val = sdp->sd_tune.gt_statfs_quantum;
1312 if (val != 30) 1312 if (val != 30)
@@ -1334,7 +1334,8 @@ static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1334 } 1334 }
1335 if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags)) 1335 if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags))
1336 seq_printf(s, ",nobarrier"); 1336 seq_printf(s, ",nobarrier");
1337 1337 if (test_bit(SDF_DEMOTE, &sdp->sd_flags))
1338 seq_printf(s, ",demote_interface_used");
1338 return 0; 1339 return 0;
1339} 1340}
1340 1341
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index 54fd98425991..37f5393e68e6 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -232,6 +232,8 @@ static ssize_t demote_rq_store(struct gfs2_sbd *sdp, const char *buf, size_t len
232 glops = gfs2_glops_list[gltype]; 232 glops = gfs2_glops_list[gltype];
233 if (glops == NULL) 233 if (glops == NULL)
234 return -EINVAL; 234 return -EINVAL;
235 if (!test_and_set_bit(SDF_DEMOTE, &sdp->sd_flags))
236 fs_info(sdp, "demote interface used\n");
235 rv = gfs2_glock_get(sdp, glnum, glops, 0, &gl); 237 rv = gfs2_glock_get(sdp, glnum, glops, 0, &gl);
236 if (rv) 238 if (rv)
237 return rv; 239 return rv;
@@ -468,8 +470,6 @@ static ssize_t name##_store(struct gfs2_sbd *sdp, const char *buf, size_t len)\
468} \ 470} \
469TUNE_ATTR_2(name, name##_store) 471TUNE_ATTR_2(name, name##_store)
470 472
471TUNE_ATTR(incore_log_blocks, 0);
472TUNE_ATTR(log_flush_secs, 0);
473TUNE_ATTR(quota_warn_period, 0); 473TUNE_ATTR(quota_warn_period, 0);
474TUNE_ATTR(quota_quantum, 0); 474TUNE_ATTR(quota_quantum, 0);
475TUNE_ATTR(max_readahead, 0); 475TUNE_ATTR(max_readahead, 0);
@@ -481,8 +481,6 @@ TUNE_ATTR(statfs_quantum, 1);
481TUNE_ATTR_3(quota_scale, quota_scale_show, quota_scale_store); 481TUNE_ATTR_3(quota_scale, quota_scale_show, quota_scale_store);
482 482
483static struct attribute *tune_attrs[] = { 483static struct attribute *tune_attrs[] = {
484 &tune_attr_incore_log_blocks.attr,
485 &tune_attr_log_flush_secs.attr,
486 &tune_attr_quota_warn_period.attr, 484 &tune_attr_quota_warn_period.attr,
487 &tune_attr_quota_quantum.attr, 485 &tune_attr_quota_quantum.attr,
488 &tune_attr_max_readahead.attr, 486 &tune_attr_max_readahead.attr,
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 4ef0e9fa3549..9ec73a854111 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -23,6 +23,7 @@
23#include "meta_io.h" 23#include "meta_io.h"
24#include "trans.h" 24#include "trans.h"
25#include "util.h" 25#include "util.h"
26#include "trace_gfs2.h"
26 27
27int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks, 28int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
28 unsigned int revokes) 29 unsigned int revokes)
@@ -75,6 +76,23 @@ fail_holder_uninit:
75 return error; 76 return error;
76} 77}
77 78
79/**
80 * gfs2_log_release - Release a given number of log blocks
81 * @sdp: The GFS2 superblock
82 * @blks: The number of blocks
83 *
84 */
85
86static void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks)
87{
88
89 atomic_add(blks, &sdp->sd_log_blks_free);
90 trace_gfs2_log_blocks(sdp, blks);
91 gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <=
92 sdp->sd_jdesc->jd_blocks);
93 up_read(&sdp->sd_log_flush_lock);
94}
95
78void gfs2_trans_end(struct gfs2_sbd *sdp) 96void gfs2_trans_end(struct gfs2_sbd *sdp)
79{ 97{
80 struct gfs2_trans *tr = current->journal_info; 98 struct gfs2_trans *tr = current->journal_info;