From 0e2bedaa394f74fa9f75ee937488c33d90039b5a Mon Sep 17 00:00:00 2001 From: Steve French Date: Fri, 30 Jan 2009 21:24:41 +0000 Subject: [CIFS] ipv6_addr_equal for address comparison Signed-off-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/connect.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c index 2209be9..005df85 100644 --- a/fs/cifs/connect.c +++ b/fs/cifs/connect.c @@ -23,7 +23,6 @@ #include #include #include -#include #include #include #include @@ -35,6 +34,7 @@ #include #include #include +#include #include "cifspdu.h" #include "cifsglob.h" #include "cifsproto.h" @@ -1379,8 +1379,8 @@ cifs_find_tcp_session(struct sockaddr_storage *addr) server->addr.sockAddr.sin_addr.s_addr)) continue; else if (addr->ss_family == AF_INET6 && - memcmp(&server->addr.sockAddr6.sin6_addr, - &addr6->sin6_addr, sizeof(addr6->sin6_addr))) + !ipv6_addr_equal(&server->addr.sockAddr6.sin6_addr, + &addr6->sin6_addr)) continue; ++server->srv_count; -- cgit v1.1 From c88ccea3143975294f5a52097546bcbb75975f52 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 10 Feb 2009 11:27:46 -0500 Subject: jbd2: Fix return value of jbd2_journal_start_commit() The function jbd2_journal_start_commit() returns 1 if either a transaction is committing or the function has queued a transaction commit. But it returns 0 if we raced with somebody queueing the transaction commit as well. This resulted in ext4_sync_fs() not functioning correctly (description from Arthur Jones): In the case of a data=ordered umount with pending long symlinks which are delayed due to a long list of other I/O on the backing block device, this causes the buffer associated with the long symlinks to not be moved to the inode dirty list in the second phase of fsync_super. Then, before they can be dirtied again, kjournald exits, seeing the UMOUNT flag and the dirty pages are never written to the backing block device, causing long symlink corruption and exposing new or previously freed block data to userspace. This can be reproduced with a script created by Eric Sandeen : #!/bin/bash umount /mnt/test2 mount /dev/sdb4 /mnt/test2 rm -f /mnt/test2/* dd if=/dev/zero of=/mnt/test2/bigfile bs=1M count=512 touch /mnt/test2/thisisveryveryveryveryveryveryveryveryveryveryveryveryveryveryveryverylongfilename ln -s /mnt/test2/thisisveryveryveryveryveryveryveryveryveryveryveryveryveryveryveryverylongfilename /mnt/test2/link umount /mnt/test2 mount /dev/sdb4 /mnt/test2 ls /mnt/test2/ This patch fixes jbd2_journal_start_commit() to always return 1 when there's a transaction committing or queued for commit. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" CC: Eric Sandeen CC: linux-ext4@vger.kernel.org --- fs/jbd2/journal.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c index eb34300..5814410 100644 --- a/fs/jbd2/journal.c +++ b/fs/jbd2/journal.c @@ -450,7 +450,7 @@ int __jbd2_log_space_left(journal_t *journal) } /* - * Called under j_state_lock. Returns true if a transaction was started. + * Called under j_state_lock. Returns true if a transaction commit was started. */ int __jbd2_log_start_commit(journal_t *journal, tid_t target) { @@ -518,7 +518,8 @@ int jbd2_journal_force_commit_nested(journal_t *journal) /* * Start a commit of the current running transaction (if any). Returns true - * if a transaction was started, and fills its tid in at *ptid + * if a transaction is going to be committed (or is currently already + * committing), and fills its tid in at *ptid */ int jbd2_journal_start_commit(journal_t *journal, tid_t *ptid) { @@ -528,15 +529,19 @@ int jbd2_journal_start_commit(journal_t *journal, tid_t *ptid) if (journal->j_running_transaction) { tid_t tid = journal->j_running_transaction->t_tid; - ret = __jbd2_log_start_commit(journal, tid); - if (ret && ptid) + __jbd2_log_start_commit(journal, tid); + /* There's a running transaction and we've just made sure + * it's commit has been scheduled. */ + if (ptid) *ptid = tid; - } else if (journal->j_committing_transaction && ptid) { + ret = 1; + } else if (journal->j_committing_transaction) { /* * If ext3_write_super() recently started a commit, then we * have to wait for completion of that transaction */ - *ptid = journal->j_committing_transaction->t_tid; + if (ptid) + *ptid = journal->j_committing_transaction->t_tid; ret = 1; } spin_unlock(&journal->j_state_lock); -- cgit v1.1 From 9eddacf9e9c03578ef2c07c9534423e823d677f8 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 10 Feb 2009 06:46:05 -0500 Subject: Revert "ext4: wait on all pending commits in ext4_sync_fs()" This undoes commit 14ce0cb411c88681ab8f3a4c9caa7f42e97a3184. Since jbd2_journal_start_commit() is now fixed to return 1 when we started a transaction commit, there's some transaction waiting to be committed or there's a transaction already committing, we don't need to call ext4_force_commit() in ext4_sync_fs(). Furthermore ext4_force_commit() can unnecessarily create sync transaction which is expensive so it's worthwhile to remove it when we can. http://bugzilla.kernel.org/show_bug.cgi?id=12224 Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: Eric Sandeen Cc: linux-ext4@vger.kernel.org --- fs/ext4/super.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ext4/super.c b/fs/ext4/super.c index e5f06a5..a5732c5 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -3046,14 +3046,17 @@ static void ext4_write_super(struct super_block *sb) static int ext4_sync_fs(struct super_block *sb, int wait) { int ret = 0; + tid_t target; trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait); sb->s_dirt = 0; if (EXT4_SB(sb)->s_journal) { - if (wait) - ret = ext4_force_commit(sb); - else - jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL); + if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, + &target)) { + if (wait) + jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, + target); + } } else { ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait); } -- cgit v1.1 From 7f5aa215088b817add9c71914b83650bdd49f8a9 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 10 Feb 2009 11:15:34 -0500 Subject: jbd2: Avoid possible NULL dereference in jbd2_journal_begin_ordered_truncate() If we race with commit code setting i_transaction to NULL, we could possibly dereference it. Proper locking requires the journal pointer (to access journal->j_list_lock), which we don't have. So we have to change the prototype of the function so that filesystem passes us the journal pointer. Also add a more detailed comment about why the function jbd2_journal_begin_ordered_truncate() does what it does and how it should be used. Thanks to Dan Carpenter for pointing to the suspitious code. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Acked-by: Joel Becker CC: linux-ext4@vger.kernel.org CC: ocfs2-devel@oss.oracle.com CC: mfasheh@suse.de CC: Dan Carpenter --- fs/ext4/inode.c | 6 ++++-- fs/jbd2/transaction.c | 42 +++++++++++++++++++++++++++++++----------- fs/ocfs2/journal.h | 6 ++++-- 3 files changed, 39 insertions(+), 15 deletions(-) (limited to 'fs') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 03ba20b..658c4a7 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -47,8 +47,10 @@ static inline int ext4_begin_ordered_truncate(struct inode *inode, loff_t new_size) { - return jbd2_journal_begin_ordered_truncate(&EXT4_I(inode)->jinode, - new_size); + return jbd2_journal_begin_ordered_truncate( + EXT4_SB(inode->i_sb)->s_journal, + &EXT4_I(inode)->jinode, + new_size); } static void ext4_invalidatepage(struct page *page, unsigned long offset); diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c index 46b4e34..28ce21d 100644 --- a/fs/jbd2/transaction.c +++ b/fs/jbd2/transaction.c @@ -2129,26 +2129,46 @@ done: } /* - * This function must be called when inode is journaled in ordered mode - * before truncation happens. It starts writeout of truncated part in - * case it is in the committing transaction so that we stand to ordered - * mode consistency guarantees. + * File truncate and transaction commit interact with each other in a + * non-trivial way. If a transaction writing data block A is + * committing, we cannot discard the data by truncate until we have + * written them. Otherwise if we crashed after the transaction with + * write has committed but before the transaction with truncate has + * committed, we could see stale data in block A. This function is a + * helper to solve this problem. It starts writeout of the truncated + * part in case it is in the committing transaction. + * + * Filesystem code must call this function when inode is journaled in + * ordered mode before truncation happens and after the inode has been + * placed on orphan list with the new inode size. The second condition + * avoids the race that someone writes new data and we start + * committing the transaction after this function has been called but + * before a transaction for truncate is started (and furthermore it + * allows us to optimize the case where the addition to orphan list + * happens in the same transaction as write --- we don't have to write + * any data in such case). */ -int jbd2_journal_begin_ordered_truncate(struct jbd2_inode *inode, +int jbd2_journal_begin_ordered_truncate(journal_t *journal, + struct jbd2_inode *jinode, loff_t new_size) { - journal_t *journal; - transaction_t *commit_trans; + transaction_t *inode_trans, *commit_trans; int ret = 0; - if (!inode->i_transaction && !inode->i_next_transaction) + /* This is a quick check to avoid locking if not necessary */ + if (!jinode->i_transaction) goto out; - journal = inode->i_transaction->t_journal; + /* Locks are here just to force reading of recent values, it is + * enough that the transaction was not committing before we started + * a transaction adding the inode to orphan list */ spin_lock(&journal->j_state_lock); commit_trans = journal->j_committing_transaction; spin_unlock(&journal->j_state_lock); - if (inode->i_transaction == commit_trans) { - ret = filemap_fdatawrite_range(inode->i_vfs_inode->i_mapping, + spin_lock(&journal->j_list_lock); + inode_trans = jinode->i_transaction; + spin_unlock(&journal->j_list_lock); + if (inode_trans == commit_trans) { + ret = filemap_fdatawrite_range(jinode->i_vfs_inode->i_mapping, new_size, LLONG_MAX); if (ret) jbd2_journal_abort(journal, ret); diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h index 3c3532e..172850a 100644 --- a/fs/ocfs2/journal.h +++ b/fs/ocfs2/journal.h @@ -513,8 +513,10 @@ static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode) static inline int ocfs2_begin_ordered_truncate(struct inode *inode, loff_t new_size) { - return jbd2_journal_begin_ordered_truncate(&OCFS2_I(inode)->ip_jinode, - new_size); + return jbd2_journal_begin_ordered_truncate( + OCFS2_SB(inode->i_sb)->journal->j_journal, + &OCFS2_I(inode)->ip_jinode, + new_size); } #endif /* OCFS2_JOURNAL_H */ -- cgit v1.1 From 7be2baaa0322c59ba888aa5260a8c130666acd41 Mon Sep 17 00:00:00 2001 From: Wei Yongjun Date: Tue, 10 Feb 2009 09:53:42 -0500 Subject: ext4: Fix to read empty directory blocks correctly in 64k The rec_len field in the directory entry is 16 bits, so there was a problem representing rec_len for filesystems with a 64k block size in the case where the directory entry takes the entire 64k block. Unfortunately, there were two schemes that were proposed; one where all zeros meant 65536 and one where all ones (65535) meant 65536. E2fsprogs used 0, whereas the kernel used 65535. Oops. Fortunately this case happens extremely rarely, with the most common case being the lost+found directory, created by mke2fs. So we will be liberal in what we accept, and accept both encodings, but we will continue to encode 65536 as 65535. This will require a change in e2fsprogs, but with fortunately ext4 filesystems normally have the dir_index feature enabled, which precludes having a completely empty directory block. Signed-off-by: Wei Yongjun Signed-off-by: "Theodore Ts'o" --- fs/ext4/ext4.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index aafc9eb..b0c87dc 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -868,7 +868,7 @@ static inline unsigned ext4_rec_len_from_disk(__le16 dlen) { unsigned len = le16_to_cpu(dlen); - if (len == EXT4_MAX_REC_LEN) + if (len == EXT4_MAX_REC_LEN || len == 0) return 1 << 16; return len; } -- cgit v1.1 From ba4439165f0f0d25b2fe065cf0c1ff8130b802eb Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Tue, 10 Feb 2009 11:14:34 -0500 Subject: ext4: Fix lockdep warning We should not call ext4_mb_add_n_trim while holding alloc_semp. ============================================= [ INFO: possible recursive locking detected ] 2.6.29-rc4-git1-dirty #124 --------------------------------------------- ffsb/3116 is trying to acquire lock: (&meta_group_info[i]->alloc_sem){----}, at: [] ext4_mb_load_buddy+0xd2/0x343 but task is already holding lock: (&meta_group_info[i]->alloc_sem){----}, at: [] ext4_mb_load_buddy+0xd2/0x343 http://bugzilla.kernel.org/show_bug.cgi?id=12672 Signed-off-by: Aneesh Kumar K.V Signed-off-by: "Theodore Ts'o" --- fs/ext4/mballoc.c | 29 ++++++++++++++++------------- 1 file changed, 16 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index deba54f..c962d06 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -4476,23 +4476,26 @@ static int ext4_mb_release_context(struct ext4_allocation_context *ac) pa->pa_free -= ac->ac_b_ex.fe_len; pa->pa_len -= ac->ac_b_ex.fe_len; spin_unlock(&pa->pa_lock); - /* - * We want to add the pa to the right bucket. - * Remove it from the list and while adding - * make sure the list to which we are adding - * doesn't grow big. - */ - if (likely(pa->pa_free)) { - spin_lock(pa->pa_obj_lock); - list_del_rcu(&pa->pa_inode_list); - spin_unlock(pa->pa_obj_lock); - ext4_mb_add_n_trim(ac); - } } - ext4_mb_put_pa(ac, ac->ac_sb, pa); } if (ac->alloc_semp) up_read(ac->alloc_semp); + if (pa) { + /* + * We want to add the pa to the right bucket. + * Remove it from the list and while adding + * make sure the list to which we are adding + * doesn't grow big. We need to release + * alloc_semp before calling ext4_mb_add_n_trim() + */ + if (pa->pa_linear && likely(pa->pa_free)) { + spin_lock(pa->pa_obj_lock); + list_del_rcu(&pa->pa_inode_list); + spin_unlock(pa->pa_obj_lock); + ext4_mb_add_n_trim(ac); + } + ext4_mb_put_pa(ac, ac->ac_sb, pa); + } if (ac->ac_bitmap_page) page_cache_release(ac->ac_bitmap_page); if (ac->ac_buddy_page) -- cgit v1.1 From eb099670895f22970cd143875467c2768d6d87e5 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Thu, 12 Feb 2009 09:27:38 -0500 Subject: Btrfs: make sure all pending extent operations are complete Theres a slight problem with finish_current_insert, if we set all to 1 and then go through and don't actually skip any of the extents on the pending list, we could exit right after we've added new extents. This is a problem because by inserting the new extents we could have gotten new COW's to happen and such, so we may have some pending updates to do or even more inserts to do after that. So this patch will only exit if we have never skipped any of the extents in the pending list, and we have no extents to insert, this will make sure that all of the pending work is truly done before we return. I've been running with this patch for a few days with all of my other testing and have not seen issues. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/extent-tree.c | 71 +++++++++++++++++++++++++++++--------------------- 1 file changed, 42 insertions(+), 29 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 7527523..376656f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -1323,8 +1323,25 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, int btrfs_extent_post_op(struct btrfs_trans_handle *trans, struct btrfs_root *root) { - finish_current_insert(trans, root->fs_info->extent_root, 1); - del_pending_extents(trans, root->fs_info->extent_root, 1); + u64 start; + u64 end; + int ret; + + while(1) { + finish_current_insert(trans, root->fs_info->extent_root, 1); + del_pending_extents(trans, root->fs_info->extent_root, 1); + + /* is there more work to do? */ + ret = find_first_extent_bit(&root->fs_info->pending_del, + 0, &start, &end, EXTENT_WRITEBACK); + if (!ret) + continue; + ret = find_first_extent_bit(&root->fs_info->extent_ins, + 0, &start, &end, EXTENT_WRITEBACK); + if (!ret) + continue; + break; + } return 0; } @@ -2211,13 +2228,12 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, u64 end; u64 priv; u64 search = 0; - u64 skipped = 0; struct btrfs_fs_info *info = extent_root->fs_info; struct btrfs_path *path; struct pending_extent_op *extent_op, *tmp; struct list_head insert_list, update_list; int ret; - int num_inserts = 0, max_inserts; + int num_inserts = 0, max_inserts, restart = 0; path = btrfs_alloc_path(); INIT_LIST_HEAD(&insert_list); @@ -2233,19 +2249,19 @@ again: ret = find_first_extent_bit(&info->extent_ins, search, &start, &end, EXTENT_WRITEBACK); if (ret) { - if (skipped && all && !num_inserts && + if (restart && !num_inserts && list_empty(&update_list)) { - skipped = 0; + restart = 0; search = 0; continue; } - mutex_unlock(&info->extent_ins_mutex); break; } ret = try_lock_extent(&info->extent_ins, start, end, GFP_NOFS); if (!ret) { - skipped = 1; + if (all) + restart = 1; search = end + 1; if (need_resched()) { mutex_unlock(&info->extent_ins_mutex); @@ -2264,7 +2280,7 @@ again: list_add_tail(&extent_op->list, &insert_list); search = end + 1; if (num_inserts == max_inserts) { - mutex_unlock(&info->extent_ins_mutex); + restart = 1; break; } } else if (extent_op->type == PENDING_BACKREF_UPDATE) { @@ -2280,7 +2296,6 @@ again: * somebody marked this thing for deletion then just unlock it and be * done, the free_extents will handle it */ - mutex_lock(&info->extent_ins_mutex); list_for_each_entry_safe(extent_op, tmp, &update_list, list) { clear_extent_bits(&info->extent_ins, extent_op->bytenr, extent_op->bytenr + extent_op->num_bytes - 1, @@ -2302,6 +2317,10 @@ again: if (!list_empty(&update_list)) { ret = update_backrefs(trans, extent_root, path, &update_list); BUG_ON(ret); + + /* we may have COW'ed new blocks, so lets start over */ + if (all) + restart = 1; } /* @@ -2309,9 +2328,9 @@ again: * need to make sure everything is cleaned then reset everything and * go back to the beginning */ - if (!num_inserts && all && skipped) { + if (!num_inserts && restart) { search = 0; - skipped = 0; + restart = 0; INIT_LIST_HEAD(&update_list); INIT_LIST_HEAD(&insert_list); goto again; @@ -2368,27 +2387,19 @@ again: BUG_ON(ret); /* - * if we broke out of the loop in order to insert stuff because we hit - * the maximum number of inserts at a time we can handle, then loop - * back and pick up where we left off + * if restart is set for whatever reason we need to go back and start + * searching through the pending list again. + * + * We just inserted some extents, which could have resulted in new + * blocks being allocated, which would result in new blocks needing + * updates, so if all is set we _must_ restart to get the updated + * blocks. */ - if (num_inserts == max_inserts) { - INIT_LIST_HEAD(&insert_list); - INIT_LIST_HEAD(&update_list); - num_inserts = 0; - goto again; - } - - /* - * again, if we need to make absolutely sure there are no more pending - * extent operations left and we know that we skipped some, go back to - * the beginning and do it all again - */ - if (all && skipped) { + if (restart || all) { INIT_LIST_HEAD(&insert_list); INIT_LIST_HEAD(&update_list); search = 0; - skipped = 0; + restart = 0; num_inserts = 0; goto again; } @@ -2709,6 +2720,8 @@ again: goto again; } + if (!err) + finish_current_insert(trans, extent_root, 0); return err; } -- cgit v1.1 From b288052e1779261ae80138074989ef50358c4e58 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 12 Feb 2009 09:37:35 -0500 Subject: Btrfs: process mount options on mount -o remount, Btrfs wasn't parsing any new mount options during remount, making it difficult to set mount options on a root drive. Signed-off-by: Chris Mason --- fs/btrfs/super.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index f3fd7e2..66b8341 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -511,6 +511,10 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data) struct btrfs_root *root = btrfs_sb(sb); int ret; + ret = btrfs_parse_options(root, data); + if (ret) + return -EINVAL; + if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) return 0; -- cgit v1.1 From 536ac8ae86e68bb5574d7cc81c7d229a86b82601 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 12 Feb 2009 09:41:38 -0500 Subject: Btrfs: use larger metadata clusters in ssd mode Larger metadata clusters can significantly improve writeback performance on ssd drives with large erasure blocks. The larger clusters make it more likely a given IO will completely overwrite the ssd block, so it doesn't have to do an internal rwm cycle. On spinning media, lager metadata clusters end up spreading out the metadata more over time, which makes fsck slower, so we don't want this to be the default. Signed-off-by: Chris Mason --- fs/btrfs/extent-tree.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 376656f..c59e120 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -2872,7 +2872,8 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans, if (data & BTRFS_BLOCK_GROUP_METADATA) { last_ptr = &root->fs_info->last_alloc; - empty_cluster = 64 * 1024; + if (!btrfs_test_opt(root, SSD)) + empty_cluster = 64 * 1024; } if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) -- cgit v1.1 From e1df36d2f18254d0690a0fbe036cece74ec311b8 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 12 Feb 2009 09:45:08 -0500 Subject: Btrfs: don't clean old snapshots on sync(1) Cleaning old snapshots can make sync(1) somewhat slow, and some users and applications still use it in a global fsync kind of workload. This patch changes btrfs not to clean old snapshots during sync, which is safe from a FS consistency point of view. The major downside is that it makes it difficult to tell when old snapshots have been reaped and the space they were using has been reclaimed. A new ioctl will be added for this purpose instead. Signed-off-by: Chris Mason --- fs/btrfs/super.c | 1 - 1 file changed, 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 66b8341..19a4daf 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -379,7 +379,6 @@ int btrfs_sync_fs(struct super_block *sb, int wait) btrfs_start_delalloc_inodes(root); btrfs_wait_ordered_extents(root, 0); - btrfs_clean_old_snapshots(root); trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); sb->s_dirt = 0; -- cgit v1.1 From b335b0034e252e79ec2e9c6697f5d663c4627bec Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 12 Feb 2009 10:06:04 -0500 Subject: Btrfs: Avoid using __GFP_HIGHMEM with slab allocator btrfs_releasepage may call kmem_cache_alloc indirectly, and provide same GFP flags it gets to kmem_cache_alloc. So it's possible to use __GFP_HIGHMEM with the slab allocator. Signed-off-by: Yan Zheng --- fs/btrfs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8f07062..638bcb5 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4263,7 +4263,7 @@ static int btrfs_releasepage(struct page *page, gfp_t gfp_flags) { if (PageWriteback(page) || PageDirty(page)) return 0; - return __btrfs_releasepage(page, gfp_flags); + return __btrfs_releasepage(page, gfp_flags & GFP_NOFS); } static void btrfs_invalidatepage(struct page *page, unsigned long offset) -- cgit v1.1 From 7951f3cefbd711f4429a0cd014aa83a844c399a0 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Thu, 12 Feb 2009 10:06:15 -0500 Subject: Btrfs: balance_level checks !child after access The BUG_ON() is in the wrong spot. Signed-off-by: Jeff Mahoney Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 35443cc..6674692 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -917,9 +917,9 @@ static noinline int balance_level(struct btrfs_trans_handle *trans, /* promote the child to a root */ child = read_node_slot(root, mid, 0); + BUG_ON(!child); btrfs_tree_lock(child); btrfs_set_lock_blocking(child); - BUG_ON(!child); ret = btrfs_cow_block(trans, root, child, mid, 0, &child, 0); BUG_ON(ret); -- cgit v1.1 From e00f7308658622fbd483cb0d9fe41165bf9050d0 Mon Sep 17 00:00:00 2001 From: Jeff Mahoney Date: Thu, 12 Feb 2009 14:11:25 -0500 Subject: Btrfs: remove btrfs_init_path btrfs_init_path was initially used when the path objects were on the stack. Now all the work is done by btrfs_alloc_path and btrfs_init_path isn't required. This patch removes it, and just uses kmem_cache_zalloc to zero out the object. Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 11 ++--------- fs/btrfs/ctree.h | 1 - fs/btrfs/inode-map.c | 1 - fs/btrfs/inode.c | 2 -- 4 files changed, 2 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 6674692..c8f4c54 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -38,19 +38,12 @@ static int balance_node_right(struct btrfs_trans_handle *trans, static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct btrfs_path *path, int level, int slot); -inline void btrfs_init_path(struct btrfs_path *p) -{ - memset(p, 0, sizeof(*p)); -} - struct btrfs_path *btrfs_alloc_path(void) { struct btrfs_path *path; - path = kmem_cache_alloc(btrfs_path_cachep, GFP_NOFS); - if (path) { - btrfs_init_path(path); + path = kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS); + if (path) path->reada = 1; - } return path; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 531db11..3f7a805 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1834,7 +1834,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans, void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); void btrfs_free_path(struct btrfs_path *p); -void btrfs_init_path(struct btrfs_path *p); void btrfs_set_path_blocking(struct btrfs_path *p); void btrfs_clear_path_blocking(struct btrfs_path *p); void btrfs_unlock_up_safe(struct btrfs_path *p, int level); diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 2aa7987..cc7334d 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -84,7 +84,6 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans, search_key.type = 0; search_key.offset = 0; - btrfs_init_path(path); start_found = 0; ret = btrfs_search_slot(trans, root, &search_key, path, 0, 0); if (ret < 0) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 638bcb5..3cee77a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2531,8 +2531,6 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, key.offset = (u64)-1; key.type = (u8)-1; - btrfs_init_path(path); - search_again: ret = btrfs_search_slot(trans, root, &key, path, -1, 1); if (ret < 0) -- cgit v1.1 From a48ddf08ba9bab91efd95e458737afa9d7699623 Mon Sep 17 00:00:00 2001 From: Qinghuang Feng Date: Thu, 12 Feb 2009 14:25:23 -0500 Subject: Btrfs: remove unused code in split_state() These two lines are not used, remove them. Signed-off-by: Qinghuang Feng Signed-off-by: Chris Mason --- fs/btrfs/extent_io.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 37d43b5..ebe6b29 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -415,8 +415,6 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig, node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node); if (node) { - struct extent_state *found; - found = rb_entry(node, struct extent_state, rb_node); free_extent_state(prealloc); return -EEXIST; } -- cgit v1.1 From 3f3420df505e47751ef76a652b5cb660e5360d6f Mon Sep 17 00:00:00 2001 From: Julia Lawall Date: Thu, 12 Feb 2009 10:16:03 -0500 Subject: Btrfs: fs/btrfs/volumes.c: remove useless kzalloc The call to kzalloc is followed by a kmalloc whose result is stored in the same variable. The semantic match that finds the problem is as follows: (http://www.emn.fr/x-info/coccinelle/) // @r exists@ local idexpression x; statement S; expression E; identifier f,l; position p1,p2; expression *ptr != NULL; @@ ( if ((x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...)) == NULL) S | x@p1 = \(kmalloc\|kzalloc\|kcalloc\)(...); ... if (x == NULL) S ) <... when != x when != if (...) { <+...x...+> } x->f = E ...> ( return \(0\|<+...x...+>\|ptr\); | return@p2 ...; ) @script:python@ p1 << r.p1; p2 << r.p2; @@ print "* file: %s kmalloc %s return %s" % (p1[0].file,p1[0].line,p2[0].line) // Signed-off-by: Julia Lawall Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index bcd14eb..c793b6f 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2894,10 +2894,6 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key, free_extent_map(em); } - map = kzalloc(sizeof(*map), GFP_NOFS); - if (!map) - return -ENOMEM; - em = alloc_extent_map(GFP_NOFS); if (!em) return -ENOMEM; -- cgit v1.1 From 4008c04a07c73ec3cb1be4c1391d2159a8f75d6d Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 12 Feb 2009 14:09:45 -0500 Subject: Btrfs: make a lockdep class for the extent buffer locks Btrfs is currently using spin_lock_nested with a nested value based on the tree depth of the block. But, this doesn't quite work because the max tree depth is bigger than what spin_lock_nested can deal with, and because locks are sometimes taken before the level field is filled in. The solution here is to use lockdep_set_class_and_name instead, and to set the class before unlocking the pages when the block is read from the disk and just after init of a freshly allocated tree block. btrfs_clear_path_blocking is also changed to take the locks in the proper order, and it also makes sure all the locks currently held are properly set to blocking before it tries to retake the spinlocks. Otherwise, lockdep gets upset about bad lock orderin. The lockdep magic cam from Peter Zijlstra Signed-off-by: Chris Mason --- fs/btrfs/ctree.c | 45 ++++++++++++++++++++++++++++++++++----------- fs/btrfs/ctree.h | 10 +++------- fs/btrfs/disk-io.c | 46 +++++++++++++++++++++++++++++++++++++++++++++- fs/btrfs/disk-io.h | 10 ++++++++++ fs/btrfs/extent-tree.c | 7 +++++-- fs/btrfs/locking.c | 11 ----------- fs/btrfs/volumes.c | 2 ++ 7 files changed, 99 insertions(+), 32 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index c8f4c54..42491d7 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -62,14 +62,38 @@ noinline void btrfs_set_path_blocking(struct btrfs_path *p) /* * reset all the locked nodes in the patch to spinning locks. + * + * held is used to keep lockdep happy, when lockdep is enabled + * we set held to a blocking lock before we go around and + * retake all the spinlocks in the path. You can safely use NULL + * for held */ -noinline void btrfs_clear_path_blocking(struct btrfs_path *p) +noinline void btrfs_clear_path_blocking(struct btrfs_path *p, + struct extent_buffer *held) { int i; - for (i = 0; i < BTRFS_MAX_LEVEL; i++) { + +#ifdef CONFIG_DEBUG_LOCK_ALLOC + /* lockdep really cares that we take all of these spinlocks + * in the right order. If any of the locks in the path are not + * currently blocking, it is going to complain. So, make really + * really sure by forcing the path to blocking before we clear + * the path blocking. + */ + if (held) + btrfs_set_lock_blocking(held); + btrfs_set_path_blocking(p); +#endif + + for (i = BTRFS_MAX_LEVEL - 1; i >= 0; i--) { if (p->nodes[i] && p->locks[i]) btrfs_clear_lock_blocking(p->nodes[i]); } + +#ifdef CONFIG_DEBUG_LOCK_ALLOC + if (held) + btrfs_clear_lock_blocking(held); +#endif } /* this also releases the path */ @@ -279,7 +303,7 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans, trans->transid, level, &ins); BUG_ON(ret); cow = btrfs_init_new_buffer(trans, root, prealloc_dest, - buf->len); + buf->len, level); } else { cow = btrfs_alloc_free_block(trans, root, buf->len, parent_start, @@ -1559,7 +1583,7 @@ cow_done: if (!p->skip_locking) p->locks[level] = 1; - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, NULL); /* * we have a lock on b and as long as we aren't changing @@ -1598,7 +1622,7 @@ cow_done: btrfs_set_path_blocking(p); sret = split_node(trans, root, p, level); - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, NULL); BUG_ON(sret > 0); if (sret) { @@ -1618,7 +1642,7 @@ cow_done: btrfs_set_path_blocking(p); sret = balance_level(trans, root, p, level); - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, NULL); if (sret) { ret = sret; @@ -1681,13 +1705,13 @@ cow_done: if (!p->skip_locking) { int lret; - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, NULL); lret = btrfs_try_spin_lock(b); if (!lret) { btrfs_set_path_blocking(p); btrfs_tree_lock(b); - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, b); } } } else { @@ -1699,7 +1723,7 @@ cow_done: btrfs_set_path_blocking(p); sret = split_leaf(trans, root, key, p, ins_len, ret == 0); - btrfs_clear_path_blocking(p); + btrfs_clear_path_blocking(p, NULL); BUG_ON(sret > 0); if (sret) { @@ -3919,7 +3943,6 @@ find_next_key: btrfs_release_path(root, path); goto again; } else { - btrfs_clear_path_blocking(path); goto out; } } @@ -3939,7 +3962,7 @@ find_next_key: path->locks[level - 1] = 1; path->nodes[level - 1] = cur; unlock_up(path, level, 1); - btrfs_clear_path_blocking(path); + btrfs_clear_path_blocking(path, NULL); } out: if (ret == 0) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 3f7a805..766b31a 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -43,11 +43,7 @@ struct btrfs_ordered_sum; #define BTRFS_ACL_NOT_CACHED ((void *)-1) -#ifdef CONFIG_LOCKDEP -# define BTRFS_MAX_LEVEL 7 -#else -# define BTRFS_MAX_LEVEL 8 -#endif +#define BTRFS_MAX_LEVEL 8 /* holds pointers to all of the tree roots */ #define BTRFS_ROOT_TREE_OBJECTID 1ULL @@ -1715,7 +1711,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, u64 empty_size); struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 bytenr, u32 blocksize); + u64 bytenr, u32 blocksize, + int level); int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root, u64 num_bytes, u64 parent, u64 min_bytes, @@ -1835,7 +1832,6 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p); struct btrfs_path *btrfs_alloc_path(void); void btrfs_free_path(struct btrfs_path *p); void btrfs_set_path_blocking(struct btrfs_path *p); -void btrfs_clear_path_blocking(struct btrfs_path *p); void btrfs_unlock_up_safe(struct btrfs_path *p, int level); int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 5aebddd..adda739 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -75,6 +75,40 @@ struct async_submit_bio { struct btrfs_work work; }; +/* These are used to set the lockdep class on the extent buffer locks. + * The class is set by the readpage_end_io_hook after the buffer has + * passed csum validation but before the pages are unlocked. + * + * The lockdep class is also set by btrfs_init_new_buffer on freshly + * allocated blocks. + * + * The class is based on the level in the tree block, which allows lockdep + * to know that lower nodes nest inside the locks of higher nodes. + * + * We also add a check to make sure the highest level of the tree is + * the same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this + * code needs update as well. + */ +#ifdef CONFIG_DEBUG_LOCK_ALLOC +# if BTRFS_MAX_LEVEL != 8 +# error +# endif +static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1]; +static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = { + /* leaf */ + "btrfs-extent-00", + "btrfs-extent-01", + "btrfs-extent-02", + "btrfs-extent-03", + "btrfs-extent-04", + "btrfs-extent-05", + "btrfs-extent-06", + "btrfs-extent-07", + /* highest possible level */ + "btrfs-extent-08", +}; +#endif + /* * extents on the btree inode are pretty simple, there's one extent * that covers the entire device @@ -347,6 +381,15 @@ static int check_tree_block_fsid(struct btrfs_root *root, return ret; } +#ifdef CONFIG_DEBUG_LOCK_ALLOC +void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level) +{ + lockdep_set_class_and_name(&eb->lock, + &btrfs_eb_class[level], + btrfs_eb_name[level]); +} +#endif + static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, struct extent_state *state) { @@ -392,6 +435,8 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, } found_level = btrfs_header_level(eb); + btrfs_set_buffer_lockdep_class(eb, found_level); + ret = csum_tree_block(root, eb, 1); if (ret) ret = -EIO; @@ -1777,7 +1822,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, ret = find_and_setup_root(tree_root, fs_info, BTRFS_DEV_TREE_OBJECTID, dev_root); dev_root->track_dirty = 1; - if (ret) goto fail_extent_root; diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h index 494a56e..95029db 100644 --- a/fs/btrfs/disk-io.h +++ b/fs/btrfs/disk-io.h @@ -101,4 +101,14 @@ int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, int btrfs_add_log_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root); int btree_lock_page_hook(struct page *page); + + +#ifdef CONFIG_DEBUG_LOCK_ALLOC +void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level); +#else +static inline void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, + int level) +{ +} +#endif #endif diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c59e120..cd86bff 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3416,7 +3416,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans, struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root, - u64 bytenr, u32 blocksize) + u64 bytenr, u32 blocksize, + int level) { struct extent_buffer *buf; @@ -3424,6 +3425,7 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, if (!buf) return ERR_PTR(-ENOMEM); btrfs_set_header_generation(buf, trans->transid); + btrfs_set_buffer_lockdep_class(buf, level); btrfs_tree_lock(buf); clean_tree_block(trans, root, buf); @@ -3467,7 +3469,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, return ERR_PTR(ret); } - buf = btrfs_init_new_buffer(trans, root, ins.objectid, blocksize); + buf = btrfs_init_new_buffer(trans, root, ins.objectid, + blocksize, level); return buf; } diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c index 9ebe938..85506c4 100644 --- a/fs/btrfs/locking.c +++ b/fs/btrfs/locking.c @@ -25,21 +25,10 @@ #include "extent_io.h" #include "locking.h" -/* - * btrfs_header_level() isn't free, so don't call it when lockdep isn't - * on - */ -#ifdef CONFIG_DEBUG_LOCK_ALLOC -static inline void spin_nested(struct extent_buffer *eb) -{ - spin_lock_nested(&eb->lock, BTRFS_MAX_LEVEL - btrfs_header_level(eb)); -} -#else static inline void spin_nested(struct extent_buffer *eb) { spin_lock(&eb->lock); } -#endif /* * Setting a lock to blocking will drop the spinlock and set the diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c793b6f..1316139 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -3102,6 +3102,8 @@ int btrfs_read_sys_array(struct btrfs_root *root) if (!sb) return -ENOMEM; btrfs_set_buffer_uptodate(sb); + btrfs_set_buffer_lockdep_class(sb, 0); + write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE); array_size = btrfs_super_sys_array_size(super_copy); -- cgit v1.1 From 2456242530a21cfee82646ebeeda65d3f74faa4c Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Thu, 12 Feb 2009 14:14:53 -0500 Subject: Btrfs: hold trans_mutex when using btrfs_record_root_in_trans btrfs_record_root_in_trans needs the trans_mutex held to make sure two callers don't race to setup the root in a given transaction. This adds it to all the places that were missing it. Signed-off-by: Yan Zheng --- fs/btrfs/extent-tree.c | 2 ++ fs/btrfs/transaction.c | 2 ++ fs/btrfs/tree-log.c | 2 ++ 3 files changed, 6 insertions(+) (limited to 'fs') diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index cd86bff..0a5d796 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5658,7 +5658,9 @@ static noinline int relocate_one_extent(struct btrfs_root *extent_root, prev_block = block_start; } + mutex_lock(&extent_root->fs_info->trans_mutex); btrfs_record_root_in_trans(found_root); + mutex_unlock(&extent_root->fs_info->trans_mutex); if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) { /* * try to update data extent references while diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 919172de..4112d53 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -688,7 +688,9 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root, num_bytes -= btrfs_root_used(&dirty->root->root_item); bytes_used = btrfs_root_used(&root->root_item); if (num_bytes) { + mutex_lock(&root->fs_info->trans_mutex); btrfs_record_root_in_trans(root); + mutex_unlock(&root->fs_info->trans_mutex); btrfs_set_root_used(&root->root_item, bytes_used - num_bytes); } diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index 2079429..9c462fb 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -2832,7 +2832,9 @@ again: BUG_ON(!wc.replay_dest); wc.replay_dest->log_root = log; + mutex_lock(&fs_info->trans_mutex); btrfs_record_root_in_trans(wc.replay_dest); + mutex_unlock(&fs_info->trans_mutex); ret = walk_log_tree(trans, log, &wc); BUG_ON(ret); -- cgit v1.1 From d794bf8e0936dce45104565cd48c571061f4c1e3 Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Sat, 14 Feb 2009 10:31:16 -0500 Subject: ext4: Initialize preallocation list_head's properly When creating a new ext4_prealloc_space structure, we have to initialize its list_head pointers before we add them to any prealloc lists. Otherwise, with list debug enabled, we will get list corruption warnings. Signed-off-by: Aneesh Kumar K.V Signed-off-by: "Theodore Ts'o" --- fs/ext4/mballoc.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs') diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index c962d06..4415bee 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c @@ -3693,6 +3693,8 @@ ext4_mb_new_inode_pa(struct ext4_allocation_context *ac) pa->pa_free = pa->pa_len; atomic_set(&pa->pa_count, 1); spin_lock_init(&pa->pa_lock); + INIT_LIST_HEAD(&pa->pa_inode_list); + INIT_LIST_HEAD(&pa->pa_group_list); pa->pa_deleted = 0; pa->pa_linear = 0; @@ -3755,6 +3757,7 @@ ext4_mb_new_group_pa(struct ext4_allocation_context *ac) atomic_set(&pa->pa_count, 1); spin_lock_init(&pa->pa_lock); INIT_LIST_HEAD(&pa->pa_inode_list); + INIT_LIST_HEAD(&pa->pa_group_list); pa->pa_deleted = 0; pa->pa_linear = 1; -- cgit v1.1 From 2acf2c261b823d9d9ed954f348b97620297a36b5 Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Sat, 14 Feb 2009 10:42:58 -0500 Subject: ext4: Implement range_cyclic in ext4_da_writepages instead of write_cache_pages With delayed allocation we lock the page in write_cache_pages() and try to build an in memory extent of contiguous blocks. This is needed so that we can get large contiguous blocks request. If range_cyclic mode is enabled, write_cache_pages() will loop back to the 0 index if no I/O has been done yet, and try to start writing from the beginning of the range. That causes an attempt to take the page lock of lower index page while holding the page lock of higher index page, which can cause a dead lock with another writeback thread. The solution is to implement the range_cyclic behavior in ext4_da_writepages() instead. http://bugzilla.kernel.org/show_bug.cgi?id=12579 Signed-off-by: Aneesh Kumar K.V Signed-off-by: "Theodore Ts'o" --- fs/ext4/inode.c | 21 +++++++++++++++++++-- 1 file changed, 19 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 658c4a7..cbd2ca9 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -2439,6 +2439,7 @@ static int ext4_da_writepages(struct address_space *mapping, int no_nrwrite_index_update; int pages_written = 0; long pages_skipped; + int range_cyclic, cycled = 1, io_done = 0; int needed_blocks, ret = 0, nr_to_writebump = 0; struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb); @@ -2490,9 +2491,15 @@ static int ext4_da_writepages(struct address_space *mapping, if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX) range_whole = 1; - if (wbc->range_cyclic) + range_cyclic = wbc->range_cyclic; + if (wbc->range_cyclic) { index = mapping->writeback_index; - else + if (index) + cycled = 0; + wbc->range_start = index << PAGE_CACHE_SHIFT; + wbc->range_end = LLONG_MAX; + wbc->range_cyclic = 0; + } else index = wbc->range_start >> PAGE_CACHE_SHIFT; mpd.wbc = wbc; @@ -2506,6 +2513,7 @@ static int ext4_da_writepages(struct address_space *mapping, wbc->no_nrwrite_index_update = 1; pages_skipped = wbc->pages_skipped; +retry: while (!ret && wbc->nr_to_write > 0) { /* @@ -2548,6 +2556,7 @@ static int ext4_da_writepages(struct address_space *mapping, pages_written += mpd.pages_written; wbc->pages_skipped = pages_skipped; ret = 0; + io_done = 1; } else if (wbc->nr_to_write) /* * There is no more writeout needed @@ -2556,6 +2565,13 @@ static int ext4_da_writepages(struct address_space *mapping, */ break; } + if (!io_done && !cycled) { + cycled = 1; + index = 0; + wbc->range_start = index << PAGE_CACHE_SHIFT; + wbc->range_end = mapping->writeback_index - 1; + goto retry; + } if (pages_skipped != wbc->pages_skipped) printk(KERN_EMERG "This should not happen leaving %s " "with nr_to_write = %ld ret = %d\n", @@ -2563,6 +2579,7 @@ static int ext4_da_writepages(struct address_space *mapping, /* Update index */ index += pages_written; + wbc->range_cyclic = range_cyclic; if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) /* * set the writeback_index so that range_cyclic -- cgit v1.1 From 090542641de833c6f756895fc2f139f046e298f9 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Sun, 15 Feb 2009 20:02:19 -0500 Subject: ext4: Fix NULL dereference in ext4_ext_migrate()'s error handling This was found through a code checker (http://repo.or.cz/w/smatch.git/). It looks like you might be able to trigger the error by trying to migrate a readonly file system. Signed-off-by: Dan Carpenter Signed-off-by: "Theodore Ts'o" --- fs/ext4/migrate.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c index 734abca..fe64d9f 100644 --- a/fs/ext4/migrate.c +++ b/fs/ext4/migrate.c @@ -481,7 +481,7 @@ int ext4_ext_migrate(struct inode *inode) + 1); if (IS_ERR(handle)) { retval = PTR_ERR(handle); - goto err_out; + return retval; } tmp_inode = ext4_new_inode(handle, inode->i_sb->s_root->d_inode, @@ -489,8 +489,7 @@ int ext4_ext_migrate(struct inode *inode) if (IS_ERR(tmp_inode)) { retval = -ENOMEM; ext4_journal_stop(handle); - tmp_inode = NULL; - goto err_out; + return retval; } i_size_write(tmp_inode, i_size_read(inode)); /* @@ -618,8 +617,7 @@ err_out: ext4_journal_stop(handle); - if (tmp_inode) - iput(tmp_inode); + iput(tmp_inode); return retval; } -- cgit v1.1 From 1a88b5364b535edaa321d70a566e358390ff0872 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Mon, 16 Feb 2009 02:38:12 +0000 Subject: Fix incomplete __mntput locking Getting this wrong caused WARNING: at fs/namespace.c:636 mntput_no_expire+0xac/0xf2() due to optimistically checking cpu_writer->mnt outside the spinlock. Here's what we really want: * we know that nobody will set cpu_writer->mnt to mnt from now on * all changes to that sucker are done under cpu_writer->lock * we want the laziest equivalent of spin_lock(&cpu_writer->lock); if (likely(cpu_writer->mnt != mnt)) { spin_unlock(&cpu_writer->lock); continue; } /* do stuff */ that would make sure we won't miss earlier setting of ->mnt done by another CPU. Anyway, for now we just move the spin_lock() earlier and move the test into the properly locked region. Signed-off-by: Al Viro Reported-and-tested-by: Li Zefan Signed-off-by: Linus Torvalds --- fs/namespace.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/namespace.c b/fs/namespace.c index 228d8c4..06f8e63 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -614,9 +614,11 @@ static inline void __mntput(struct vfsmount *mnt) */ for_each_possible_cpu(cpu) { struct mnt_writer *cpu_writer = &per_cpu(mnt_writers, cpu); - if (cpu_writer->mnt != mnt) - continue; spin_lock(&cpu_writer->lock); + if (cpu_writer->mnt != mnt) { + spin_unlock(&cpu_writer->lock); + continue; + } atomic_add(cpu_writer->count, &mnt->__mnt_writers); cpu_writer->count = 0; /* -- cgit v1.1 From a60e78e57a17d55bbd5a96da16fe9649d364b987 Mon Sep 17 00:00:00 2001 From: Subhash Peddamallu Date: Mon, 16 Feb 2009 10:27:07 +0100 Subject: fs/bio: bio_alloc_bioset: pass right object ptr to mempool_free When freeing from bio pool use right ptr to account for bs->front_pad, instead of bio ptr, Signed-off-by: Subhash Peddamallu Signed-off-by: Jens Axboe --- fs/bio.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/bio.c b/fs/bio.c index 062299a..72ab251 100644 --- a/fs/bio.c +++ b/fs/bio.c @@ -302,9 +302,10 @@ void bio_init(struct bio *bio) struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) { struct bio *bio = NULL; + void *p; if (bs) { - void *p = mempool_alloc(bs->bio_pool, gfp_mask); + p = mempool_alloc(bs->bio_pool, gfp_mask); if (p) bio = p + bs->front_pad; @@ -329,7 +330,7 @@ struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) } if (unlikely(!bvl)) { if (bs) - mempool_free(bio, bs->bio_pool); + mempool_free(p, bs->bio_pool); else kfree(bio); bio = NULL; -- cgit v1.1 From 78f707bfc723552e8309b7c38a8d0cc51012e813 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 17 Feb 2009 13:59:08 +0100 Subject: block: revert part of 18ce3751ccd488c78d3827e9f6bf54e6322676fb The above commit added WRITE_SYNC and switched various places to using that for committing writes that will be waited upon immediately after submission. However, this causes a performance regression with AS and CFQ for ext3 at least, since sync_dirty_buffer() will submit some writes with WRITE_SYNC while ext3 has sumitted others dependent writes without the sync flag set. This causes excessive anticipation/idling in the IO scheduler because sync and async writes get interleaved, causing a big performance regression for the below test case (which is meant to simulate sqlite like behaviour). ---- test case ---- int main(int argc, char **argv) { int fdes, i; FILE *fp; struct timeval start; struct timeval end; struct timeval res; gettimeofday(&start, NULL); for (i=0; i --- fs/buffer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/buffer.c b/fs/buffer.c index 665d446..62b57e3 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -3108,7 +3108,7 @@ int sync_dirty_buffer(struct buffer_head *bh) if (test_clear_buffer_dirty(bh)) { get_bh(bh); bh->b_end_io = end_buffer_write_sync; - ret = submit_bh(WRITE_SYNC, bh); + ret = submit_bh(WRITE, bh); wait_on_buffer(bh); if (buffer_eopnotsupp(bh)) { clear_buffer_eopnotsupp(bh); -- cgit v1.1 From 8f19d472935c83d823fa4cf02bcc0a7b9952db30 Mon Sep 17 00:00:00 2001 From: Eric Biederman Date: Wed, 18 Feb 2009 14:48:16 -0800 Subject: seq_file: properly cope with pread Currently seq_read assumes that the offset passed to it is always the offset it passed to user space. In the case pread this assumption is broken and we do the wrong thing when presented with pread. To solve this I introduce an offset cache inside of struct seq_file so we know where our logical file position is. Then in seq_read if we try to read from another offset we reset our data structures and attempt to go to the offset user space wanted. [akpm@linux-foundation.org: restore FMODE_PWRITE] [pjt@google.com: seq_open needs its fmode opened up to take advantage of this] Signed-off-by: Eric Biederman Cc: Alexey Dobriyan Cc: Al Viro Cc: Paul Turner Cc: [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/seq_file.c | 36 ++++++++++++++++++++++++++++++++---- 1 file changed, 32 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/seq_file.c b/fs/seq_file.c index 5267098..a1a4cfe 100644 --- a/fs/seq_file.c +++ b/fs/seq_file.c @@ -48,8 +48,16 @@ int seq_open(struct file *file, const struct seq_operations *op) */ file->f_version = 0; - /* SEQ files support lseek, but not pread/pwrite */ - file->f_mode &= ~(FMODE_PREAD | FMODE_PWRITE); + /* + * seq_files support lseek() and pread(). They do not implement + * write() at all, but we clear FMODE_PWRITE here for historical + * reasons. + * + * If a client of seq_files a) implements file.write() and b) wishes to + * support pwrite() then that client will need to implement its own + * file.open() which calls seq_open() and then sets FMODE_PWRITE. + */ + file->f_mode &= ~FMODE_PWRITE; return 0; } EXPORT_SYMBOL(seq_open); @@ -131,6 +139,22 @@ ssize_t seq_read(struct file *file, char __user *buf, size_t size, loff_t *ppos) int err = 0; mutex_lock(&m->lock); + + /* Don't assume *ppos is where we left it */ + if (unlikely(*ppos != m->read_pos)) { + m->read_pos = *ppos; + while ((err = traverse(m, *ppos)) == -EAGAIN) + ; + if (err) { + /* With prejudice... */ + m->read_pos = 0; + m->version = 0; + m->index = 0; + m->count = 0; + goto Done; + } + } + /* * seq_file->op->..m_start/m_stop/m_next may do special actions * or optimisations based on the file->f_version, so we want to @@ -230,8 +254,10 @@ Fill: Done: if (!copied) copied = err; - else + else { *ppos += copied; + m->read_pos += copied; + } file->f_version = m->version; mutex_unlock(&m->lock); return copied; @@ -266,16 +292,18 @@ loff_t seq_lseek(struct file *file, loff_t offset, int origin) if (offset < 0) break; retval = offset; - if (offset != file->f_pos) { + if (offset != m->read_pos) { while ((retval=traverse(m, offset)) == -EAGAIN) ; if (retval) { /* with extreme prejudice... */ file->f_pos = 0; + m->read_pos = 0; m->version = 0; m->index = 0; m->count = 0; } else { + m->read_pos = offset; retval = file->f_pos = offset; } } -- cgit v1.1 From 610d18f4128ebbd88845d0fc60cce67b49af881e Mon Sep 17 00:00:00 2001 From: Davide Libenzi Date: Wed, 18 Feb 2009 14:48:18 -0800 Subject: timerfd: add flags check As requested by Michael, add a missing check for valid flags in timerfd_settime(), and make it return EINVAL in case some extra bits are set. Michael said: If this is to be any use to userland apps that want to check flag support (perhaps it is too late already), then the sooner we get it into the kernel the better: 2.6.29 would be good; earlier stables as well would be even better. [akpm@linux-foundation.org: remove unused TFD_FLAGS_SET] Acked-by: Michael Kerrisk Signed-off-by: Davide Libenzi Cc: [2.6.27.x, 2.6.28.x] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/timerfd.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/timerfd.c b/fs/timerfd.c index 6a123b8..b042bd7 100644 --- a/fs/timerfd.c +++ b/fs/timerfd.c @@ -186,10 +186,9 @@ SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags) BUILD_BUG_ON(TFD_CLOEXEC != O_CLOEXEC); BUILD_BUG_ON(TFD_NONBLOCK != O_NONBLOCK); - if (flags & ~(TFD_CLOEXEC | TFD_NONBLOCK)) - return -EINVAL; - if (clockid != CLOCK_MONOTONIC && - clockid != CLOCK_REALTIME) + if ((flags & ~TFD_CREATE_FLAGS) || + (clockid != CLOCK_MONOTONIC && + clockid != CLOCK_REALTIME)) return -EINVAL; ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); @@ -201,7 +200,7 @@ SYSCALL_DEFINE2(timerfd_create, int, clockid, int, flags) hrtimer_init(&ctx->tmr, clockid, HRTIMER_MODE_ABS); ufd = anon_inode_getfd("[timerfd]", &timerfd_fops, ctx, - flags & (O_CLOEXEC | O_NONBLOCK)); + flags & TFD_SHARED_FCNTL_FLAGS); if (ufd < 0) kfree(ctx); @@ -219,7 +218,8 @@ SYSCALL_DEFINE4(timerfd_settime, int, ufd, int, flags, if (copy_from_user(&ktmr, utmr, sizeof(ktmr))) return -EFAULT; - if (!timespec_valid(&ktmr.it_value) || + if ((flags & ~TFD_SETTIME_FLAGS) || + !timespec_valid(&ktmr.it_value) || !timespec_valid(&ktmr.it_interval)) return -EINVAL; -- cgit v1.1 From 1cf6e7d83bf334cc5916137862c920a97aabc018 Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Wed, 18 Feb 2009 14:48:18 -0800 Subject: mm: task dirty accounting fix YAMAMOTO-san noticed that task_dirty_inc doesn't seem to be called properly for cases where set_page_dirty is not used to dirty a page (eg. mark_buffer_dirty). Additionally, there is some inconsistency about when task_dirty_inc is called. It is used for dirty balancing, however it even gets called for __set_page_dirty_no_writeback. So rather than increment it in a set_page_dirty wrapper, move it down to exactly where the dirty page accounting stats are incremented. Cc: YAMAMOTO Takashi Signed-off-by: Nick Piggin Acked-by: Peter Zijlstra Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/buffer.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs') diff --git a/fs/buffer.c b/fs/buffer.c index 665d446..ff4d1cd 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -777,6 +777,7 @@ static int __set_page_dirty(struct page *page, __inc_zone_page_state(page, NR_FILE_DIRTY); __inc_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE); + task_dirty_inc(current); task_io_account_write(PAGE_CACHE_SIZE); } radix_tree_tag_set(&mapping->page_tree, -- cgit v1.1 From ada723dcd681e2dffd7d73345cc8fda0eb0df9bd Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 18 Feb 2009 14:48:30 -0800 Subject: fs/super.c: add lockdep annotation to s_umount Li Zefan said: Thread 1: for ((; ;)) { mount -t cpuset xxx /mnt > /dev/null 2>&1 cat /mnt/cpus > /dev/null 2>&1 umount /mnt > /dev/null 2>&1 } Thread 2: for ((; ;)) { mount -t cpuset xxx /mnt > /dev/null 2>&1 umount /mnt > /dev/null 2>&1 } (Note: It is irrelevant which cgroup subsys is used.) After a while a lockdep warning showed up: ============================================= [ INFO: possible recursive locking detected ] 2.6.28 #479 --------------------------------------------- mount/13554 is trying to acquire lock: (&type->s_umount_key#19){--..}, at: [] sget+0x5e/0x321 but task is already holding lock: (&type->s_umount_key#19){--..}, at: [] sget+0x1e2/0x321 other info that might help us debug this: 1 lock held by mount/13554: #0: (&type->s_umount_key#19){--..}, at: [] sget+0x1e2/0x321 stack backtrace: Pid: 13554, comm: mount Not tainted 2.6.28-mc #479 Call Trace: [] validate_chain+0x4c6/0xbbd [] __lock_acquire+0x676/0x700 [] lock_acquire+0x5d/0x7a [] ? sget+0x5e/0x321 [] down_write+0x34/0x50 [] ? sget+0x5e/0x321 [] sget+0x5e/0x321 [] ? cgroup_set_super+0x0/0x3e [] ? cgroup_test_super+0x0/0x2f [] cgroup_get_sb+0x98/0x2e7 [] cpuset_get_sb+0x4a/0x5f [] vfs_kern_mount+0x40/0x7b [] do_kern_mount+0x37/0xbf [] do_mount+0x5c3/0x61a [] ? copy_mount_options+0x2c/0x111 [] sys_mount+0x69/0xa0 [] sysenter_do_call+0x12/0x31 The cause is after alloc_super() and then retry, an old entry in list fs_supers is found, so grab_super(old) is called, but both functions hold s_umount lock: struct super_block *sget(...) { ... retry: spin_lock(&sb_lock); if (test) { list_for_each_entry(old, &type->fs_supers, s_instances) { if (!test(old, data)) continue; if (!grab_super(old)) <--- 2nd: down_write(&old->s_umount); goto retry; if (s) destroy_super(s); return old; } } if (!s) { spin_unlock(&sb_lock); s = alloc_super(type); <--- 1th: down_write(&s->s_umount) if (!s) return ERR_PTR(-ENOMEM); goto retry; } ... } It seems like a false positive, and seems like VFS but not cgroup needs to be fixed. Peter said: We can simply put the new s_umount instance in a but lockdep doesn't particularly cares about subclass order. If there's any issue with the callers of sget() assuming the s_umount lock being of sublcass 0, then there is another annotation we can use to fix that, but lets not bother with that if this is sufficient. Addresses http://bugzilla.kernel.org/show_bug.cgi?id=12673 Signed-off-by: Peter Zijlstra Tested-by: Li Zefan Reported-by: Li Zefan Cc: Al Viro Cc: Paul Menage Cc: Arjan van de Ven Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/super.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/super.c b/fs/super.c index 61dce00..8349ed6 100644 --- a/fs/super.c +++ b/fs/super.c @@ -82,7 +82,22 @@ static struct super_block *alloc_super(struct file_system_type *type) * lock ordering than usbfs: */ lockdep_set_class(&s->s_lock, &type->s_lock_key); - down_write(&s->s_umount); + /* + * sget() can have s_umount recursion. + * + * When it cannot find a suitable sb, it allocates a new + * one (this one), and tries again to find a suitable old + * one. + * + * In case that succeeds, it will acquire the s_umount + * lock of the old one. Since these are clearly distrinct + * locks, and this object isn't exposed yet, there's no + * risk of deadlocks. + * + * Annotate this by putting this lock in a different + * subclass. + */ + down_write_nested(&s->s_umount, SINGLE_DEPTH_NESTING); s->s_count = S_BIAS; atomic_set(&s->s_active, 1); mutex_init(&s->s_vfs_rename_mutex); -- cgit v1.1 From 2db69a9340da12a4db44edb7506dd68799aeff55 Mon Sep 17 00:00:00 2001 From: Bill Nottingham Date: Wed, 18 Feb 2009 14:48:39 -0800 Subject: vt: Declare PIO_CMAP/GIO_CMAP as compatbile ioctls. Otherwise, these don't work when called from 32-bit userspace on 64-bit kernels. Cc: Jiri Kosina Cc: Alan Cox Cc: [2.6.25.x, 2.6.26.x, 2.6.27.x, 2.6.28.x] Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/compat_ioctl.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/compat_ioctl.c b/fs/compat_ioctl.c index 9c6d815..39bd4d3 100644 --- a/fs/compat_ioctl.c +++ b/fs/compat_ioctl.c @@ -1938,6 +1938,8 @@ ULONG_IOCTL(SET_BITMAP_FILE) /* Big K */ COMPATIBLE_IOCTL(PIO_FONT) COMPATIBLE_IOCTL(GIO_FONT) +COMPATIBLE_IOCTL(PIO_CMAP) +COMPATIBLE_IOCTL(GIO_CMAP) ULONG_IOCTL(KDSIGACCEPT) COMPATIBLE_IOCTL(KDGETKEYCODE) COMPATIBLE_IOCTL(KDSETKEYCODE) -- cgit v1.1 From f04b30de3c82528f1ab4c58b3dd4c975f5341901 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 18 Feb 2009 14:48:43 -0800 Subject: inotify: fix GFP_KERNEL related deadlock Enhanced lockdep coverage of __GFP_NOFS turned up this new lockdep assert: [ 1093.677775] [ 1093.677781] ================================= [ 1093.680031] [ INFO: inconsistent lock state ] [ 1093.680031] 2.6.29-rc5-tip-01504-gb49eca1-dirty #1 [ 1093.680031] --------------------------------- [ 1093.680031] inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage. [ 1093.680031] kswapd0/308 [HC0[0]:SC0[0]:HE1:SE1] takes: [ 1093.680031] (&inode->inotify_mutex){+.+.?.}, at: [] inotify_inode_is_dead+0x20/0x80 [ 1093.680031] {RECLAIM_FS-ON-W} state was registered at: [ 1093.680031] [] mark_held_locks+0x43/0x5b [ 1093.680031] [] lockdep_trace_alloc+0x6c/0x6e [ 1093.680031] [] kmem_cache_alloc+0x20/0x150 [ 1093.680031] [] idr_pre_get+0x27/0x6c [ 1093.680031] [] inotify_handle_get_wd+0x25/0xad [ 1093.680031] [] inotify_add_watch+0x7a/0x129 [ 1093.680031] [] sys_inotify_add_watch+0x20f/0x250 [ 1093.680031] [] sysenter_do_call+0x12/0x35 [ 1093.680031] [] 0xffffffff [ 1093.680031] irq event stamp: 60417 [ 1093.680031] hardirqs last enabled at (60417): [] call_rcu+0x53/0x59 [ 1093.680031] hardirqs last disabled at (60416): [] call_rcu+0x17/0x59 [ 1093.680031] softirqs last enabled at (59656): [] __do_softirq+0x157/0x16b [ 1093.680031] softirqs last disabled at (59651): [] do_softirq+0x74/0x15d [ 1093.680031] [ 1093.680031] other info that might help us debug this: [ 1093.680031] 2 locks held by kswapd0/308: [ 1093.680031] #0: (shrinker_rwsem){++++..}, at: [] shrink_slab+0x36/0x189 [ 1093.680031] #1: (&type->s_umount_key#4){+++++.}, at: [] shrink_dcache_memory+0x110/0x1fb [ 1093.680031] [ 1093.680031] stack backtrace: [ 1093.680031] Pid: 308, comm: kswapd0 Not tainted 2.6.29-rc5-tip-01504-gb49eca1-dirty #1 [ 1093.680031] Call Trace: [ 1093.680031] [] valid_state+0x12a/0x13d [ 1093.680031] [] mark_lock+0xc1/0x1e9 [ 1093.680031] [] ? check_usage_forwards+0x0/0x3f [ 1093.680031] [] __lock_acquire+0x2c6/0xac8 [ 1093.680031] [] ? register_lock_class+0x17/0x228 [ 1093.680031] [] lock_acquire+0x5d/0x7a [ 1093.680031] [] ? inotify_inode_is_dead+0x20/0x80 [ 1093.680031] [] __mutex_lock_common+0x3a/0x4cb [ 1093.680031] [] ? inotify_inode_is_dead+0x20/0x80 [ 1093.680031] [] mutex_lock_nested+0x2e/0x36 [ 1093.680031] [] ? inotify_inode_is_dead+0x20/0x80 [ 1093.680031] [] inotify_inode_is_dead+0x20/0x80 [ 1093.680031] [] dentry_iput+0x90/0xc2 [ 1093.680031] [] d_kill+0x21/0x45 [ 1093.680031] [] __shrink_dcache_sb+0x27f/0x355 [ 1093.680031] [] shrink_dcache_memory+0x15e/0x1fb [ 1093.680031] [] shrink_slab+0x121/0x189 [ 1093.680031] [] kswapd+0x39f/0x561 [ 1093.680031] [] ? isolate_pages_global+0x0/0x233 [ 1093.680031] [] ? autoremove_wake_function+0x0/0x43 [ 1093.680031] [] ? kswapd+0x0/0x561 [ 1093.680031] [] kthread+0x41/0x82 [ 1093.680031] [] ? kthread+0x0/0x82 [ 1093.680031] [] kernel_thread_helper+0x7/0x10 inotify_handle_get_wd() does idr_pre_get() which does a kmem_cache_alloc() without __GFP_FS - and is hence deadlockable under extreme MM pressure. Signed-off-by: Ingo Molnar Acked-by: Peter Zijlstra Cc: MinChan Kim Cc: Nick Piggin Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/notify/inotify/inotify.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/notify/inotify/inotify.c b/fs/notify/inotify/inotify.c index dae3f28..331f2e8 100644 --- a/fs/notify/inotify/inotify.c +++ b/fs/notify/inotify/inotify.c @@ -156,7 +156,7 @@ static int inotify_handle_get_wd(struct inotify_handle *ih, int ret; do { - if (unlikely(!idr_pre_get(&ih->idr, GFP_KERNEL))) + if (unlikely(!idr_pre_get(&ih->idr, GFP_NOFS))) return -ENOSPC; ret = idr_get_new_above(&ih->idr, watch, ih->last_wd+1, &watch->wd); } while (ret == -EAGAIN); -- cgit v1.1 From 7fdf582447aa01658b624adc0a51a31e4278b68c Mon Sep 17 00:00:00 2001 From: Felix Blyakher Date: Wed, 18 Feb 2009 15:41:28 -0600 Subject: Revert "[XFS] use scalable vmap API" This reverts commit 95f8e302c04c0b0c6de35ab399a5551605eeb006. This commit caused regression. We'll try to fix use of new vmap API for next release. Signed-off-by: Christoph Hellwig Signed-off-by: Felix Blyakher --- fs/xfs/linux-2.6/xfs_buf.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index d71dc44..0b2177a 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -264,7 +264,7 @@ xfs_buf_free( uint i; if ((bp->b_flags & XBF_MAPPED) && (bp->b_page_count > 1)) - vm_unmap_ram(bp->b_addr - bp->b_offset, bp->b_page_count); + vunmap(bp->b_addr - bp->b_offset); for (i = 0; i < bp->b_page_count; i++) { struct page *page = bp->b_pages[i]; @@ -386,8 +386,8 @@ _xfs_buf_map_pages( bp->b_addr = page_address(bp->b_pages[0]) + bp->b_offset; bp->b_flags |= XBF_MAPPED; } else if (flags & XBF_MAPPED) { - bp->b_addr = vm_map_ram(bp->b_pages, bp->b_page_count, - -1, PAGE_KERNEL); + bp->b_addr = vmap(bp->b_pages, bp->b_page_count, + VM_MAP, PAGE_KERNEL); if (unlikely(bp->b_addr == NULL)) return -ENOMEM; bp->b_addr += bp->b_offset; -- cgit v1.1 From 27e88bf6af7d42adf790f7b2ed7d65475f191cf2 Mon Sep 17 00:00:00 2001 From: Felix Blyakher Date: Wed, 18 Feb 2009 15:56:51 -0600 Subject: Revert "[XFS] remove old vmap cache" This reverts commit d2859751cd0bf586941ffa7308635a293f943c17. This commit caused regression. We'll try to fix use of new vmap API for next release. Signed-off-by: Christoph Hellwig Signed-off-by: Felix Blyakher --- fs/xfs/linux-2.6/xfs_buf.c | 75 +++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 74 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c index 0b2177a..cb329ed 100644 --- a/fs/xfs/linux-2.6/xfs_buf.c +++ b/fs/xfs/linux-2.6/xfs_buf.c @@ -166,6 +166,75 @@ test_page_region( } /* + * Mapping of multi-page buffers into contiguous virtual space + */ + +typedef struct a_list { + void *vm_addr; + struct a_list *next; +} a_list_t; + +static a_list_t *as_free_head; +static int as_list_len; +static DEFINE_SPINLOCK(as_lock); + +/* + * Try to batch vunmaps because they are costly. + */ +STATIC void +free_address( + void *addr) +{ + a_list_t *aentry; + +#ifdef CONFIG_XEN + /* + * Xen needs to be able to make sure it can get an exclusive + * RO mapping of pages it wants to turn into a pagetable. If + * a newly allocated page is also still being vmap()ed by xfs, + * it will cause pagetable construction to fail. This is a + * quick workaround to always eagerly unmap pages so that Xen + * is happy. + */ + vunmap(addr); + return; +#endif + + aentry = kmalloc(sizeof(a_list_t), GFP_NOWAIT); + if (likely(aentry)) { + spin_lock(&as_lock); + aentry->next = as_free_head; + aentry->vm_addr = addr; + as_free_head = aentry; + as_list_len++; + spin_unlock(&as_lock); + } else { + vunmap(addr); + } +} + +STATIC void +purge_addresses(void) +{ + a_list_t *aentry, *old; + + if (as_free_head == NULL) + return; + + spin_lock(&as_lock); + aentry = as_free_head; + as_free_head = NULL; + as_list_len = 0; + spin_unlock(&as_lock); + + while ((old = aentry) != NULL) { + vunmap(aentry->vm_addr); + aentry = aentry->next; + kfree(old); + } +} + +/* * Internal xfs_buf_t object manipulation */ @@ -264,7 +333,7 @@ xfs_buf_free( uint i; if ((bp->b_flags & XBF_MAPPED) && (bp->b_page_count > 1)) - vunmap(bp->b_addr - bp->b_offset); + free_address(bp->b_addr - bp->b_offset); for (i = 0; i < bp->b_page_count; i++) { struct page *page = bp->b_pages[i]; @@ -386,6 +455,8 @@ _xfs_buf_map_pages( bp->b_addr = page_address(bp->b_pages[0]) + bp->b_offset; bp->b_flags |= XBF_MAPPED; } else if (flags & XBF_MAPPED) { + if (as_list_len > 64) + purge_addresses(); bp->b_addr = vmap(bp->b_pages, bp->b_page_count, VM_MAP, PAGE_KERNEL); if (unlikely(bp->b_addr == NULL)) @@ -1672,6 +1743,8 @@ xfsbufd( count++; } + if (as_list_len > 0) + purge_addresses(); if (count) blk_run_address_space(target->bt_mapping); -- cgit v1.1 From 2cfbd50b536c878e58ab3681c4e944fa3d99b415 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 20 Feb 2009 10:55:10 -0500 Subject: Btrfs: check file pointer in btrfs_sync_file fsync can be called by NFS with a null file pointer, and btrfs was oopsing in this case. Signed-off-by: Chris Mason --- fs/btrfs/file.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 3e8023e..872f104 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1222,7 +1222,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) /* * ok we haven't committed the transaction yet, lets do a commit */ - if (file->private_data) + if (file && file->private_data) btrfs_ioctl_trans_end(file); trans = btrfs_start_transaction(root, 1); @@ -1231,7 +1231,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) goto out; } - ret = btrfs_log_dentry_safe(trans, root, file->f_dentry); + ret = btrfs_log_dentry_safe(trans, root, dentry); if (ret < 0) goto out; @@ -1245,7 +1245,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) * file again, but that will end up using the synchronization * inside btrfs_sync_log to keep things safe. */ - mutex_unlock(&file->f_dentry->d_inode->i_mutex); + mutex_unlock(&dentry->d_inode->i_mutex); if (ret > 0) { ret = btrfs_commit_transaction(trans, root); @@ -1253,7 +1253,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync) btrfs_sync_log(trans, root); ret = btrfs_end_transaction(trans, root); } - mutex_lock(&file->f_dentry->d_inode->i_mutex); + mutex_lock(&dentry->d_inode->i_mutex); out: return ret > 0 ? EIO : ret; } -- cgit v1.1 From e4cce94c9c8797b08faf6a79396df4d175e377fa Mon Sep 17 00:00:00 2001 From: Igor Mammedov Date: Tue, 10 Feb 2009 14:10:26 +0300 Subject: [CIFS] Prevent OOPs when mounting with remote prefixpath. Fixes OOPs with message 'kernel BUG at fs/cifs/cifs_dfs_ref.c:274!'. Checks if the prefixpath in an accesible while we are still in cifs_mount and fails with reporting a error if we can't access the prefixpath Should fix Samba bugs 6086 and 5861 and kernel bug 12192 Signed-off-by: Igor Mammedov Acked-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/cifsproto.h | 1 + fs/cifs/connect.c | 45 +++++++++++++++++++++++++++++++++++++++++++++ fs/cifs/inode.c | 4 ++-- 3 files changed, 48 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h index 382ba62..ec9f9c1 100644 --- a/fs/cifs/cifsproto.h +++ b/fs/cifs/cifsproto.h @@ -42,6 +42,7 @@ extern void _FreeXid(unsigned int); #define GetXid() (int)_GetXid(); cFYI(1,("CIFS VFS: in %s as Xid: %d with uid: %d",__func__, xid,current_fsuid())); #define FreeXid(curr_xid) {_FreeXid(curr_xid); cFYI(1,("CIFS VFS: leaving %s (xid = %d) rc = %d",__func__,curr_xid,(int)rc));} extern char *build_path_from_dentry(struct dentry *); +extern char *cifs_build_path_to_root(struct cifs_sb_info *cifs_sb); extern char *build_wildcard_path_from_dentry(struct dentry *direntry); /* extern void renew_parental_timestamps(struct dentry *direntry);*/ extern int SendReceive(const unsigned int /* xid */ , struct cifsSesInfo *, diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c index 005df85..da0f4ff 100644 --- a/fs/cifs/connect.c +++ b/fs/cifs/connect.c @@ -2180,6 +2180,33 @@ static void setup_cifs_sb(struct smb_vol *pvolume_info, "mount option supported")); } +static int +is_path_accessible(int xid, struct cifsTconInfo *tcon, + struct cifs_sb_info *cifs_sb, const char *full_path) +{ + int rc; + __u64 inode_num; + FILE_ALL_INFO *pfile_info; + + rc = CIFSGetSrvInodeNumber(xid, tcon, full_path, &inode_num, + cifs_sb->local_nls, + cifs_sb->mnt_cifs_flags & + CIFS_MOUNT_MAP_SPECIAL_CHR); + if (rc != -EOPNOTSUPP) + return rc; + + pfile_info = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL); + if (pfile_info == NULL) + return -ENOMEM; + + rc = CIFSSMBQPathInfo(xid, tcon, full_path, pfile_info, + 0 /* not legacy */, cifs_sb->local_nls, + cifs_sb->mnt_cifs_flags & + CIFS_MOUNT_MAP_SPECIAL_CHR); + kfree(pfile_info); + return rc; +} + int cifs_mount(struct super_block *sb, struct cifs_sb_info *cifs_sb, char *mount_data, const char *devname) @@ -2190,6 +2217,7 @@ cifs_mount(struct super_block *sb, struct cifs_sb_info *cifs_sb, struct cifsSesInfo *pSesInfo = NULL; struct cifsTconInfo *tcon = NULL; struct TCP_Server_Info *srvTcp = NULL; + char *full_path; xid = GetXid(); @@ -2426,6 +2454,23 @@ mount_fail_check: cifs_sb->rsize = min(cifs_sb->rsize, (tcon->ses->server->maxBuf - MAX_CIFS_HDR_SIZE)); + if (!rc && cifs_sb->prepathlen) { + /* build_path_to_root works only when we have a valid tcon */ + full_path = cifs_build_path_to_root(cifs_sb); + if (full_path == NULL) { + rc = -ENOMEM; + goto mount_fail_check; + } + rc = is_path_accessible(xid, tcon, cifs_sb, full_path); + if (rc) { + cERROR(1, ("Path %s in not accessible: %d", + full_path, rc)); + kfree(full_path); + goto mount_fail_check; + } + kfree(full_path); + } + /* volume_info->password is freed above when existing session found (in which case it is not needed anymore) but when new sesion is created the password ptr is put in the new session structure (in which case the diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c index bcf7b51..7342bfb 100644 --- a/fs/cifs/inode.c +++ b/fs/cifs/inode.c @@ -621,7 +621,7 @@ static const struct inode_operations cifs_ipc_inode_ops = { .lookup = cifs_lookup, }; -static char *build_path_to_root(struct cifs_sb_info *cifs_sb) +char *cifs_build_path_to_root(struct cifs_sb_info *cifs_sb) { int pplen = cifs_sb->prepathlen; int dfsplen; @@ -678,7 +678,7 @@ struct inode *cifs_iget(struct super_block *sb, unsigned long ino) return inode; cifs_sb = CIFS_SB(inode->i_sb); - full_path = build_path_to_root(cifs_sb); + full_path = cifs_build_path_to_root(cifs_sb); if (full_path == NULL) return ERR_PTR(-ENOMEM); -- cgit v1.1 From 132ac7b77cc95a22d6118d327c96586759fbf006 Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Tue, 10 Feb 2009 07:33:57 -0500 Subject: cifs: refactor new_inode() calls and inode initialization Move new inode creation into a separate routine and refactor the callers to take advantage of it. Signed-off-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/cifsproto.h | 2 ++ fs/cifs/inode.c | 96 +++++++++++++++++++++++++++++++++-------------------- fs/cifs/readdir.c | 54 ++++++++++++++---------------- 3 files changed, 86 insertions(+), 66 deletions(-) (limited to 'fs') diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h index ec9f9c1..62fd5bd 100644 --- a/fs/cifs/cifsproto.h +++ b/fs/cifs/cifsproto.h @@ -92,6 +92,8 @@ extern u64 cifs_UnixTimeToNT(struct timespec); extern __le64 cnvrtDosCifsTm(__u16 date, __u16 time); extern struct timespec cnvrtDosUnixTm(__u16 date, __u16 time); +extern struct inode *cifs_new_inode(struct super_block *sb, + unsigned long *inum); extern int cifs_get_inode_info(struct inode **pinode, const unsigned char *search_path, FILE_ALL_INFO *pfile_info, diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c index 7342bfb..c7674f5 100644 --- a/fs/cifs/inode.c +++ b/fs/cifs/inode.c @@ -199,6 +199,49 @@ static void fill_fake_finddataunix(FILE_UNIX_BASIC_INFO *pfnd_dat, pfnd_dat->Gid = cpu_to_le64(pinode->i_gid); } +/** + * cifs_new inode - create new inode, initialize, and hash it + * @sb - pointer to superblock + * @inum - if valid pointer and serverino is enabled, replace i_ino with val + * + * Create a new inode, initialize it for CIFS and hash it. Returns the new + * inode or NULL if one couldn't be allocated. + * + * If the share isn't mounted with "serverino" or inum is a NULL pointer then + * we'll just use the inode number assigned by new_inode(). Note that this can + * mean i_ino collisions since the i_ino assigned by new_inode is not + * guaranteed to be unique. + */ +struct inode * +cifs_new_inode(struct super_block *sb, unsigned long *inum) +{ + struct inode *inode; + + inode = new_inode(sb); + if (inode == NULL) + return NULL; + + /* + * BB: Is i_ino == 0 legal? Here, we assume that it is. If it isn't we + * stop passing inum as ptr. Are there sanity checks we can use to + * ensure that the server is really filling in that field? Also, + * if serverino is disabled, perhaps we should be using iunique()? + */ + if (inum && (CIFS_SB(sb)->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM)) + inode->i_ino = *inum; + + /* + * must set this here instead of cifs_alloc_inode since VFS will + * clobber i_flags + */ + if (sb->s_flags & MS_NOATIME) + inode->i_flags |= S_NOATIME | S_NOCMTIME; + + insert_inode_hash(inode); + + return inode; +} + int cifs_get_inode_info_unix(struct inode **pinode, const unsigned char *full_path, struct super_block *sb, int xid) { @@ -233,22 +276,12 @@ int cifs_get_inode_info_unix(struct inode **pinode, /* get new inode */ if (*pinode == NULL) { - *pinode = new_inode(sb); + *pinode = cifs_new_inode(sb, (unsigned long *) + &find_data.UniqueId); if (*pinode == NULL) { rc = -ENOMEM; goto cgiiu_exit; } - /* Is an i_ino of zero legal? */ - /* note ino incremented to unique num in new_inode */ - /* Are there sanity checks we can use to ensure that - the server is really filling in that field? */ - if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) - (*pinode)->i_ino = (unsigned long)find_data.UniqueId; - - if (sb->s_flags & MS_NOATIME) - (*pinode)->i_flags |= S_NOATIME | S_NOCMTIME; - - insert_inode_hash(*pinode); } inode = *pinode; @@ -465,11 +498,8 @@ int cifs_get_inode_info(struct inode **pinode, /* get new inode */ if (*pinode == NULL) { - *pinode = new_inode(sb); - if (*pinode == NULL) { - rc = -ENOMEM; - goto cgii_exit; - } + __u64 inode_num; + /* Is an i_ino of zero legal? Can we use that to check if the server supports returning inode numbers? Are there other sanity checks we can use to ensure that @@ -486,7 +516,6 @@ int cifs_get_inode_info(struct inode **pinode, if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) { int rc1 = 0; - __u64 inode_num; rc1 = CIFSGetSrvInodeNumber(xid, pTcon, full_path, &inode_num, @@ -496,12 +525,17 @@ int cifs_get_inode_info(struct inode **pinode, if (rc1) { cFYI(1, ("GetSrvInodeNum rc %d", rc1)); /* BB EOPNOSUPP disable SERVER_INUM? */ - } else /* do we need cast or hash to ino? */ - (*pinode)->i_ino = inode_num; - } /* else ino incremented to unique num in new_inode*/ - if (sb->s_flags & MS_NOATIME) - (*pinode)->i_flags |= S_NOATIME | S_NOCMTIME; - insert_inode_hash(*pinode); + } + *pinode = cifs_new_inode(sb, (unsigned long *) + &inode_num); + } else { + *pinode = cifs_new_inode(sb, NULL); + } + + if (*pinode == NULL) { + rc = -ENOMEM; + goto cgii_exit; + } } inode = *pinode; cifsInfo = CIFS_I(inode); @@ -1114,24 +1148,14 @@ int cifs_mkdir(struct inode *inode, struct dentry *direntry, int mode) else direntry->d_op = &cifs_dentry_ops; - newinode = new_inode(inode->i_sb); + newinode = cifs_new_inode(inode->i_sb, (unsigned long *) + &pInfo->UniqueId); if (newinode == NULL) { kfree(pInfo); goto mkdir_get_info; } - /* Is an i_ino of zero legal? */ - /* Are there sanity checks we can use to ensure that - the server is really filling in that field? */ - if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) { - newinode->i_ino = - (unsigned long)pInfo->UniqueId; - } /* note ino incremented to unique num in new_inode */ - if (inode->i_sb->s_flags & MS_NOATIME) - newinode->i_flags |= S_NOATIME | S_NOCMTIME; newinode->i_nlink = 2; - - insert_inode_hash(newinode); d_instantiate(direntry, newinode); /* we already checked in POSIXCreate whether diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c index 9f51f9b..02a2022 100644 --- a/fs/cifs/readdir.c +++ b/fs/cifs/readdir.c @@ -56,35 +56,34 @@ static inline void dump_cifs_file_struct(struct file *file, char *label) } #endif /* DEBUG2 */ -/* Returns one if new inode created (which therefore needs to be hashed) */ +/* Returns 1 if new inode created, 2 if both dentry and inode were */ /* Might check in the future if inode number changed so we can rehash inode */ -static int construct_dentry(struct qstr *qstring, struct file *file, - struct inode **ptmp_inode, struct dentry **pnew_dentry) +static int +construct_dentry(struct qstr *qstring, struct file *file, + struct inode **ptmp_inode, struct dentry **pnew_dentry, + unsigned long *inum) { - struct dentry *tmp_dentry; - struct cifs_sb_info *cifs_sb; - struct cifsTconInfo *pTcon; + struct dentry *tmp_dentry = NULL; + struct super_block *sb = file->f_path.dentry->d_sb; int rc = 0; cFYI(1, ("For %s", qstring->name)); - cifs_sb = CIFS_SB(file->f_path.dentry->d_sb); - pTcon = cifs_sb->tcon; qstring->hash = full_name_hash(qstring->name, qstring->len); tmp_dentry = d_lookup(file->f_path.dentry, qstring); if (tmp_dentry) { + /* BB: overwrite old name? i.e. tmp_dentry->d_name and + * tmp_dentry->d_name.len?? + */ cFYI(0, ("existing dentry with inode 0x%p", tmp_dentry->d_inode)); *ptmp_inode = tmp_dentry->d_inode; -/* BB overwrite old name? i.e. tmp_dentry->d_name and tmp_dentry->d_name.len??*/ if (*ptmp_inode == NULL) { - *ptmp_inode = new_inode(file->f_path.dentry->d_sb); + *ptmp_inode = cifs_new_inode(sb, inum); if (*ptmp_inode == NULL) return rc; rc = 1; } - if (file->f_path.dentry->d_sb->s_flags & MS_NOATIME) - (*ptmp_inode)->i_flags |= S_NOATIME | S_NOCMTIME; } else { tmp_dentry = d_alloc(file->f_path.dentry, qstring); if (tmp_dentry == NULL) { @@ -93,15 +92,14 @@ static int construct_dentry(struct qstr *qstring, struct file *file, return rc; } - *ptmp_inode = new_inode(file->f_path.dentry->d_sb); - if (pTcon->nocase) + if (CIFS_SB(sb)->tcon->nocase) tmp_dentry->d_op = &cifs_ci_dentry_ops; else tmp_dentry->d_op = &cifs_dentry_ops; + + *ptmp_inode = cifs_new_inode(sb, inum); if (*ptmp_inode == NULL) return rc; - if (file->f_path.dentry->d_sb->s_flags & MS_NOATIME) - (*ptmp_inode)->i_flags |= S_NOATIME | S_NOCMTIME; rc = 2; } @@ -842,9 +840,7 @@ static int cifs_get_name_from_search_buf(struct qstr *pqst, len = strnlen(filename, PATH_MAX); } - /* BB fixme - hash low and high 32 bits if not 64 bit arch BB */ - if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM) - *pinum = pFindData->UniqueId; + *pinum = pFindData->UniqueId; } else if (level == SMB_FIND_FILE_DIRECTORY_INFO) { FILE_DIRECTORY_INFO *pFindData = (FILE_DIRECTORY_INFO *)current_entry; @@ -940,20 +936,18 @@ static int cifs_filldir(char *pfindEntry, struct file *file, if (rc) return rc; - rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry); + /* only these two infolevels return valid inode numbers */ + if (pCifsF->srch_inf.info_level == SMB_FIND_FILE_UNIX || + pCifsF->srch_inf.info_level == SMB_FIND_FILE_ID_FULL_DIR_INFO) + rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry, + &inum); + else + rc = construct_dentry(&qstring, file, &tmp_inode, &tmp_dentry, + NULL); + if ((tmp_inode == NULL) || (tmp_dentry == NULL)) return -ENOMEM; - if (rc) { - /* inode created, we need to hash it with right inode number */ - if (inum != 0) { - /* BB fixme - hash the 2 32 quantities bits together if - * necessary BB */ - tmp_inode->i_ino = inum; - } - insert_inode_hash(tmp_inode); - } - /* we pass in rc below, indicating whether it is a new inode, so we can figure out whether to invalidate the inode cached data if the file has changed */ -- cgit v1.1 From 950ec52880fab89b957c7dc45e8b8476dd63741f Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Wed, 11 Feb 2009 08:08:26 -0500 Subject: cifs: properly handle case where CIFSGetSrvInodeNumber fails ...if it does then we pass a pointer to an unintialized variable for the inode number to cifs_new_inode. Have it pass a NULL pointer instead. Also tweak the function prototypes to reduce the amount of casting. Signed-off-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/cifsproto.h | 3 +-- fs/cifs/inode.c | 20 ++++++++++---------- fs/cifs/readdir.c | 6 +++--- 3 files changed, 14 insertions(+), 15 deletions(-) (limited to 'fs') diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h index 62fd5bd..446e62c 100644 --- a/fs/cifs/cifsproto.h +++ b/fs/cifs/cifsproto.h @@ -92,8 +92,7 @@ extern u64 cifs_UnixTimeToNT(struct timespec); extern __le64 cnvrtDosCifsTm(__u16 date, __u16 time); extern struct timespec cnvrtDosUnixTm(__u16 date, __u16 time); -extern struct inode *cifs_new_inode(struct super_block *sb, - unsigned long *inum); +extern struct inode *cifs_new_inode(struct super_block *sb, __u64 *inum); extern int cifs_get_inode_info(struct inode **pinode, const unsigned char *search_path, FILE_ALL_INFO *pfile_info, diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c index c7674f5..475115c 100644 --- a/fs/cifs/inode.c +++ b/fs/cifs/inode.c @@ -213,7 +213,7 @@ static void fill_fake_finddataunix(FILE_UNIX_BASIC_INFO *pfnd_dat, * guaranteed to be unique. */ struct inode * -cifs_new_inode(struct super_block *sb, unsigned long *inum) +cifs_new_inode(struct super_block *sb, __u64 *inum) { struct inode *inode; @@ -228,7 +228,7 @@ cifs_new_inode(struct super_block *sb, unsigned long *inum) * if serverino is disabled, perhaps we should be using iunique()? */ if (inum && (CIFS_SB(sb)->mnt_cifs_flags & CIFS_MOUNT_SERVER_INUM)) - inode->i_ino = *inum; + inode->i_ino = (unsigned long) *inum; /* * must set this here instead of cifs_alloc_inode since VFS will @@ -276,8 +276,7 @@ int cifs_get_inode_info_unix(struct inode **pinode, /* get new inode */ if (*pinode == NULL) { - *pinode = cifs_new_inode(sb, (unsigned long *) - &find_data.UniqueId); + *pinode = cifs_new_inode(sb, &find_data.UniqueId); if (*pinode == NULL) { rc = -ENOMEM; goto cgiiu_exit; @@ -499,6 +498,7 @@ int cifs_get_inode_info(struct inode **pinode, /* get new inode */ if (*pinode == NULL) { __u64 inode_num; + __u64 *pinum = &inode_num; /* Is an i_ino of zero legal? Can we use that to check if the server supports returning inode numbers? Are @@ -518,20 +518,20 @@ int cifs_get_inode_info(struct inode **pinode, int rc1 = 0; rc1 = CIFSGetSrvInodeNumber(xid, pTcon, - full_path, &inode_num, + full_path, pinum, cifs_sb->local_nls, cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR); if (rc1) { cFYI(1, ("GetSrvInodeNum rc %d", rc1)); + pinum = NULL; /* BB EOPNOSUPP disable SERVER_INUM? */ } - *pinode = cifs_new_inode(sb, (unsigned long *) - &inode_num); } else { - *pinode = cifs_new_inode(sb, NULL); + pinum = NULL; } + *pinode = cifs_new_inode(sb, pinum); if (*pinode == NULL) { rc = -ENOMEM; goto cgii_exit; @@ -1148,8 +1148,8 @@ int cifs_mkdir(struct inode *inode, struct dentry *direntry, int mode) else direntry->d_op = &cifs_dentry_ops; - newinode = cifs_new_inode(inode->i_sb, (unsigned long *) - &pInfo->UniqueId); + newinode = cifs_new_inode(inode->i_sb, + &pInfo->UniqueId); if (newinode == NULL) { kfree(pInfo); goto mkdir_get_info; diff --git a/fs/cifs/readdir.c b/fs/cifs/readdir.c index 02a2022..c2c01ff 100644 --- a/fs/cifs/readdir.c +++ b/fs/cifs/readdir.c @@ -61,7 +61,7 @@ static inline void dump_cifs_file_struct(struct file *file, char *label) static int construct_dentry(struct qstr *qstring, struct file *file, struct inode **ptmp_inode, struct dentry **pnew_dentry, - unsigned long *inum) + __u64 *inum) { struct dentry *tmp_dentry = NULL; struct super_block *sb = file->f_path.dentry->d_sb; @@ -820,7 +820,7 @@ static int find_cifs_entry(const int xid, struct cifsTconInfo *pTcon, /* inode num, inode type and filename returned */ static int cifs_get_name_from_search_buf(struct qstr *pqst, char *current_entry, __u16 level, unsigned int unicode, - struct cifs_sb_info *cifs_sb, int max_len, ino_t *pinum) + struct cifs_sb_info *cifs_sb, int max_len, __u64 *pinum) { int rc = 0; unsigned int len = 0; @@ -903,7 +903,7 @@ static int cifs_filldir(char *pfindEntry, struct file *file, struct qstr qstring; struct cifsFileInfo *pCifsF; unsigned int obj_type; - ino_t inum; + __u64 inum; struct cifs_sb_info *cifs_sb; struct inode *tmp_inode; struct dentry *tmp_dentry; -- cgit v1.1 From 44f68fadd865bb288ebdcea2b602f0b1cab27a0c Mon Sep 17 00:00:00 2001 From: Jeff Layton Date: Wed, 11 Feb 2009 08:08:28 -0500 Subject: cifs: posix fill in inode needed by posix open function needed to prepare for posix open Signed-off-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/cifsproto.h | 2 ++ fs/cifs/inode.c | 2 +- 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/cifs/cifsproto.h b/fs/cifs/cifsproto.h index 446e62c..083dfc5 100644 --- a/fs/cifs/cifsproto.h +++ b/fs/cifs/cifsproto.h @@ -92,6 +92,8 @@ extern u64 cifs_UnixTimeToNT(struct timespec); extern __le64 cnvrtDosCifsTm(__u16 date, __u16 time); extern struct timespec cnvrtDosUnixTm(__u16 date, __u16 time); +extern void posix_fill_in_inode(struct inode *tmp_inode, + FILE_UNIX_BASIC_INFO *pData, int isNewInode); extern struct inode *cifs_new_inode(struct super_block *sb, __u64 *inum); extern int cifs_get_inode_info(struct inode **pinode, const unsigned char *search_path, diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c index 475115c..4690a36 100644 --- a/fs/cifs/inode.c +++ b/fs/cifs/inode.c @@ -1051,7 +1051,7 @@ out_reval: return rc; } -static void posix_fill_in_inode(struct inode *tmp_inode, +void posix_fill_in_inode(struct inode *tmp_inode, FILE_UNIX_BASIC_INFO *pData, int isNewInode) { struct cifsInodeInfo *cifsInfo = CIFS_I(tmp_inode); -- cgit v1.1 From 69765529d701c838df19ea1f5ad2f33a528261ae Mon Sep 17 00:00:00 2001 From: Steve French Date: Tue, 17 Feb 2009 01:29:40 +0000 Subject: [CIFS] Fix oops in cifs_strfromUCS_le mounting to servers which do not specify their OS Fixes kernel bug #10451 http://bugzilla.kernel.org/show_bug.cgi?id=10451 Certain NAS appliances do not set the operating system or network operating system fields in the session setup response on the wire. cifs was oopsing on the unexpected zero length response fields (when trying to null terminate a zero length field). This fixes the oops. Acked-by: Jeff Layton CC: stable Signed-off-by: Steve French --- fs/cifs/CHANGES | 3 ++- fs/cifs/sess.c | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES index 73ac7eb..1cfa72e 100644 --- a/fs/cifs/CHANGES +++ b/fs/cifs/CHANGES @@ -7,7 +7,8 @@ specified and user does not have access to query information about the top of the share. Fix problem in 2.6.28 resolving DFS paths to Samba servers (worked to Windows). Fix rmdir so that pending search (readdir) requests do not get invalid results which include the now -removed directory. +removed directory. Fix oops in cifs_dfs_ref.c when prefixpath is not reachable +when using DFS. Version 1.55 ------------ diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c index 5f22de7..b234407 100644 --- a/fs/cifs/sess.c +++ b/fs/cifs/sess.c @@ -228,7 +228,7 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft, kfree(ses->serverOS); /* UTF-8 string will not grow more than four times as big as UCS-16 */ - ses->serverOS = kzalloc(4 * len, GFP_KERNEL); + ses->serverOS = kzalloc((4 * len) + 2 /* trailing null */, GFP_KERNEL); if (ses->serverOS != NULL) cifs_strfromUCS_le(ses->serverOS, (__le16 *)data, len, nls_cp); data += 2 * (len + 1); @@ -241,7 +241,7 @@ static int decode_unicode_ssetup(char **pbcc_area, int bleft, return rc; kfree(ses->serverNOS); - ses->serverNOS = kzalloc(4 * len, GFP_KERNEL); /* BB this is wrong length FIXME BB */ + ses->serverNOS = kzalloc((4 * len) + 2 /* trailing null */, GFP_KERNEL); if (ses->serverNOS != NULL) { cifs_strfromUCS_le(ses->serverNOS, (__le16 *)data, len, nls_cp); -- cgit v1.1 From c3b2a0c640bff7df85d79fb4f89674949a267ec2 Mon Sep 17 00:00:00 2001 From: Steve French Date: Fri, 20 Feb 2009 04:32:45 +0000 Subject: [CIFS] improve posix semantics of file create Samba server added support for a new posix open/create/mkdir operation a year or so ago, and we added support to cifs for mkdir to use it, but had not added the corresponding code to file create. The following patch helps improve the performance of the cifs create path (to Samba and servers which support the cifs posix protocol extensions). Using Connectathon basic test1, with 2000 files, the performance improved about 15%, and also helped reduce network traffic (17% fewer SMBs sent over the wire) due to saving a network round trip for the SetPathInfo on every file create. It should also help the semantics (and probably the performance) of write (e.g. when posix byte range locks are on the file) on file handles opened with posix create, and adds support for a few flags which would have to be ignored otherwise. Signed-off-by: Steve French --- fs/cifs/CHANGES | 4 +- fs/cifs/dir.c | 307 +++++++++++++++++++++++++++++++++++++------------------- 2 files changed, 208 insertions(+), 103 deletions(-) (limited to 'fs') diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES index 1cfa72e..72063f5 100644 --- a/fs/cifs/CHANGES +++ b/fs/cifs/CHANGES @@ -8,7 +8,9 @@ top of the share. Fix problem in 2.6.28 resolving DFS paths to Samba servers (worked to Windows). Fix rmdir so that pending search (readdir) requests do not get invalid results which include the now removed directory. Fix oops in cifs_dfs_ref.c when prefixpath is not reachable -when using DFS. +when using DFS. Add better file create support to servers which support +the CIFS POSIX protocol extensions (this adds support for new flags +on create, and improves semantics for write of locked ranges). Version 1.55 ------------ diff --git a/fs/cifs/dir.c b/fs/cifs/dir.c index 964aad0..89fb728 100644 --- a/fs/cifs/dir.c +++ b/fs/cifs/dir.c @@ -3,7 +3,7 @@ * * vfs operations that deal with dentries * - * Copyright (C) International Business Machines Corp., 2002,2008 + * Copyright (C) International Business Machines Corp., 2002,2009 * Author(s): Steve French (sfrench@us.ibm.com) * * This library is free software; you can redistribute it and/or modify @@ -129,6 +129,78 @@ cifs_bp_rename_retry: return full_path; } +static int cifs_posix_open(char *full_path, struct inode **pinode, + struct super_block *sb, int mode, int oflags, + int *poplock, __u16 *pnetfid, int xid) +{ + int rc; + __u32 oplock; + FILE_UNIX_BASIC_INFO *presp_data; + __u32 posix_flags = 0; + struct cifs_sb_info *cifs_sb = CIFS_SB(sb); + + cFYI(1, ("posix open %s", full_path)); + + presp_data = kzalloc(sizeof(FILE_UNIX_BASIC_INFO), GFP_KERNEL); + if (presp_data == NULL) + return -ENOMEM; + +/* So far cifs posix extensions can only map the following flags. + There are other valid fmode oflags such as FMODE_LSEEK, FMODE_PREAD, but + so far we do not seem to need them, and we can treat them as local only */ + if ((oflags & (FMODE_READ | FMODE_WRITE)) == + (FMODE_READ | FMODE_WRITE)) + posix_flags = SMB_O_RDWR; + else if (oflags & FMODE_READ) + posix_flags = SMB_O_RDONLY; + else if (oflags & FMODE_WRITE) + posix_flags = SMB_O_WRONLY; + if (oflags & O_CREAT) + posix_flags |= SMB_O_CREAT; + if (oflags & O_EXCL) + posix_flags |= SMB_O_EXCL; + if (oflags & O_TRUNC) + posix_flags |= SMB_O_TRUNC; + if (oflags & O_APPEND) + posix_flags |= SMB_O_APPEND; + if (oflags & O_SYNC) + posix_flags |= SMB_O_SYNC; + if (oflags & O_DIRECTORY) + posix_flags |= SMB_O_DIRECTORY; + if (oflags & O_NOFOLLOW) + posix_flags |= SMB_O_NOFOLLOW; + if (oflags & O_DIRECT) + posix_flags |= SMB_O_DIRECT; + + + rc = CIFSPOSIXCreate(xid, cifs_sb->tcon, posix_flags, mode, + pnetfid, presp_data, &oplock, full_path, + cifs_sb->local_nls, cifs_sb->mnt_cifs_flags & + CIFS_MOUNT_MAP_SPECIAL_CHR); + if (rc) + goto posix_open_ret; + + if (presp_data->Type == cpu_to_le32(-1)) + goto posix_open_ret; /* open ok, caller does qpathinfo */ + + /* get new inode and set it up */ + if (!pinode) + goto posix_open_ret; /* caller does not need info */ + + *pinode = cifs_new_inode(sb, &presp_data->UniqueId); + + /* We do not need to close the file if new_inode fails since + the caller will retry qpathinfo as long as inode is null */ + if (*pinode == NULL) + goto posix_open_ret; + + posix_fill_in_inode(*pinode, presp_data, 1); + +posix_open_ret: + kfree(presp_data); + return rc; +} + static void setup_cifs_dentry(struct cifsTconInfo *tcon, struct dentry *direntry, struct inode *newinode) @@ -150,7 +222,14 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode, int xid; int create_options = CREATE_NOT_DIR; int oplock = 0; - /* BB below access is too much for the mknod to request */ + int oflags; + /* + * BB below access is probably too much for mknod to request + * but we have to do query and setpathinfo so requesting + * less could fail (unless we want to request getatr and setatr + * permissions (only). At least for POSIX we do not have to + * request so much. + */ int desiredAccess = GENERIC_READ | GENERIC_WRITE; __u16 fileHandle; struct cifs_sb_info *cifs_sb; @@ -174,13 +253,43 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode, } mode &= ~current->fs->umask; + if (oplockEnabled) + oplock = REQ_OPLOCK; - if (nd && (nd->flags & LOOKUP_OPEN)) { - int oflags = nd->intent.open.flags; + if (nd && (nd->flags & LOOKUP_OPEN)) + oflags = nd->intent.open.flags; + else + oflags = FMODE_READ; + + if (tcon->unix_ext && (tcon->ses->capabilities & CAP_UNIX) && + (CIFS_UNIX_POSIX_PATH_OPS_CAP & + le64_to_cpu(tcon->fsUnixInfo.Capability))) { + rc = cifs_posix_open(full_path, &newinode, inode->i_sb, + mode, oflags, &oplock, &fileHandle, xid); + /* EIO could indicate that (posix open) operation is not + supported, despite what server claimed in capability + negotation. EREMOTE indicates DFS junction, which is not + handled in posix open */ + + if ((rc == 0) && (newinode == NULL)) + goto cifs_create_get_file_info; /* query inode info */ + else if (rc == 0) /* success, no need to query */ + goto cifs_create_set_dentry; + else if ((rc != -EIO) && (rc != -EREMOTE) && + (rc != -EOPNOTSUPP)) /* path not found or net err */ + goto cifs_create_out; + /* else fallthrough to retry, using older open call, this is + case where server does not support this SMB level, and + falsely claims capability (also get here for DFS case + which should be rare for path not covered on files) */ + } + if (nd && (nd->flags & LOOKUP_OPEN)) { + /* if the file is going to stay open, then we + need to set the desired access properly */ desiredAccess = 0; if (oflags & FMODE_READ) - desiredAccess |= GENERIC_READ; + desiredAccess |= GENERIC_READ; /* is this too little? */ if (oflags & FMODE_WRITE) { desiredAccess |= GENERIC_WRITE; if (!(oflags & FMODE_READ)) @@ -199,8 +308,6 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode, /* BB add processing to set equivalent of mode - e.g. via CreateX with ACLs */ - if (oplockEnabled) - oplock = REQ_OPLOCK; buf = kmalloc(sizeof(FILE_ALL_INFO), GFP_KERNEL); if (buf == NULL) { @@ -233,116 +340,112 @@ cifs_create(struct inode *inode, struct dentry *direntry, int mode, } if (rc) { cFYI(1, ("cifs_create returned 0x%x", rc)); - } else { - /* If Open reported that we actually created a file - then we now have to set the mode if possible */ - if ((tcon->unix_ext) && (oplock & CIFS_CREATE_ACTION)) { - struct cifs_unix_set_info_args args = { + goto cifs_create_out; + } + + /* If Open reported that we actually created a file + then we now have to set the mode if possible */ + if ((tcon->unix_ext) && (oplock & CIFS_CREATE_ACTION)) { + struct cifs_unix_set_info_args args = { .mode = mode, .ctime = NO_CHANGE_64, .atime = NO_CHANGE_64, .mtime = NO_CHANGE_64, .device = 0, - }; + }; - if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) { - args.uid = (__u64) current_fsuid(); - if (inode->i_mode & S_ISGID) - args.gid = (__u64) inode->i_gid; - else - args.gid = (__u64) current_fsgid(); - } else { - args.uid = NO_CHANGE_64; - args.gid = NO_CHANGE_64; - } - CIFSSMBUnixSetInfo(xid, tcon, full_path, &args, - cifs_sb->local_nls, - cifs_sb->mnt_cifs_flags & - CIFS_MOUNT_MAP_SPECIAL_CHR); + if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID) { + args.uid = (__u64) current_fsuid(); + if (inode->i_mode & S_ISGID) + args.gid = (__u64) inode->i_gid; + else + args.gid = (__u64) current_fsgid(); } else { - /* BB implement mode setting via Windows security - descriptors e.g. */ - /* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/ - - /* Could set r/o dos attribute if mode & 0222 == 0 */ + args.uid = NO_CHANGE_64; + args.gid = NO_CHANGE_64; } + CIFSSMBUnixSetInfo(xid, tcon, full_path, &args, + cifs_sb->local_nls, + cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MAP_SPECIAL_CHR); + } else { + /* BB implement mode setting via Windows security + descriptors e.g. */ + /* CIFSSMBWinSetPerms(xid,tcon,path,mode,-1,-1,nls);*/ - /* server might mask mode so we have to query for it */ - if (tcon->unix_ext) - rc = cifs_get_inode_info_unix(&newinode, full_path, - inode->i_sb, xid); - else { - rc = cifs_get_inode_info(&newinode, full_path, - buf, inode->i_sb, xid, - &fileHandle); - if (newinode) { - if (cifs_sb->mnt_cifs_flags & - CIFS_MOUNT_DYNPERM) - newinode->i_mode = mode; - if ((oplock & CIFS_CREATE_ACTION) && - (cifs_sb->mnt_cifs_flags & - CIFS_MOUNT_SET_UID)) { - newinode->i_uid = current_fsuid(); - if (inode->i_mode & S_ISGID) - newinode->i_gid = - inode->i_gid; - else - newinode->i_gid = - current_fsgid(); - } + /* Could set r/o dos attribute if mode & 0222 == 0 */ + } + +cifs_create_get_file_info: + /* server might mask mode so we have to query for it */ + if (tcon->unix_ext) + rc = cifs_get_inode_info_unix(&newinode, full_path, + inode->i_sb, xid); + else { + rc = cifs_get_inode_info(&newinode, full_path, buf, + inode->i_sb, xid, &fileHandle); + if (newinode) { + if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_DYNPERM) + newinode->i_mode = mode; + if ((oplock & CIFS_CREATE_ACTION) && + (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_SET_UID)) { + newinode->i_uid = current_fsuid(); + if (inode->i_mode & S_ISGID) + newinode->i_gid = inode->i_gid; + else + newinode->i_gid = current_fsgid(); } } + } - if (rc != 0) { - cFYI(1, ("Create worked, get_inode_info failed rc = %d", - rc)); - } else - setup_cifs_dentry(tcon, direntry, newinode); - - if ((nd == NULL /* nfsd case - nfs srv does not set nd */) || - (!(nd->flags & LOOKUP_OPEN))) { - /* mknod case - do not leave file open */ - CIFSSMBClose(xid, tcon, fileHandle); - } else if (newinode) { - struct cifsFileInfo *pCifsFile = - kzalloc(sizeof(struct cifsFileInfo), GFP_KERNEL); - - if (pCifsFile == NULL) - goto cifs_create_out; - pCifsFile->netfid = fileHandle; - pCifsFile->pid = current->tgid; - pCifsFile->pInode = newinode; - pCifsFile->invalidHandle = false; - pCifsFile->closePend = false; - init_MUTEX(&pCifsFile->fh_sem); - mutex_init(&pCifsFile->lock_mutex); - INIT_LIST_HEAD(&pCifsFile->llist); - atomic_set(&pCifsFile->wrtPending, 0); - - /* set the following in open now +cifs_create_set_dentry: + if (rc == 0) + setup_cifs_dentry(tcon, direntry, newinode); + else + cFYI(1, ("Create worked, get_inode_info failed rc = %d", rc)); + + /* nfsd case - nfs srv does not set nd */ + if ((nd == NULL) || (!(nd->flags & LOOKUP_OPEN))) { + /* mknod case - do not leave file open */ + CIFSSMBClose(xid, tcon, fileHandle); + } else if (newinode) { + struct cifsFileInfo *pCifsFile = + kzalloc(sizeof(struct cifsFileInfo), GFP_KERNEL); + + if (pCifsFile == NULL) + goto cifs_create_out; + pCifsFile->netfid = fileHandle; + pCifsFile->pid = current->tgid; + pCifsFile->pInode = newinode; + pCifsFile->invalidHandle = false; + pCifsFile->closePend = false; + init_MUTEX(&pCifsFile->fh_sem); + mutex_init(&pCifsFile->lock_mutex); + INIT_LIST_HEAD(&pCifsFile->llist); + atomic_set(&pCifsFile->wrtPending, 0); + + /* set the following in open now pCifsFile->pfile = file; */ - write_lock(&GlobalSMBSeslock); - list_add(&pCifsFile->tlist, &tcon->openFileList); - pCifsInode = CIFS_I(newinode); - if (pCifsInode) { - /* if readable file instance put first in list*/ - if (write_only) { - list_add_tail(&pCifsFile->flist, - &pCifsInode->openFileList); - } else { - list_add(&pCifsFile->flist, - &pCifsInode->openFileList); - } - if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) { - pCifsInode->clientCanCacheAll = true; - pCifsInode->clientCanCacheRead = true; - cFYI(1, ("Exclusive Oplock inode %p", - newinode)); - } else if ((oplock & 0xF) == OPLOCK_READ) - pCifsInode->clientCanCacheRead = true; + write_lock(&GlobalSMBSeslock); + list_add(&pCifsFile->tlist, &tcon->openFileList); + pCifsInode = CIFS_I(newinode); + if (pCifsInode) { + /* if readable file instance put first in list*/ + if (write_only) { + list_add_tail(&pCifsFile->flist, + &pCifsInode->openFileList); + } else { + list_add(&pCifsFile->flist, + &pCifsInode->openFileList); } - write_unlock(&GlobalSMBSeslock); + if ((oplock & 0xF) == OPLOCK_EXCLUSIVE) { + pCifsInode->clientCanCacheAll = true; + pCifsInode->clientCanCacheRead = true; + cFYI(1, ("Exclusive Oplock inode %p", + newinode)); + } else if ((oplock & 0xF) == OPLOCK_READ) + pCifsInode->clientCanCacheRead = true; } + write_unlock(&GlobalSMBSeslock); } cifs_create_out: kfree(buf); -- cgit v1.1 From eca6acf91552a9b2e997cc76339115c95eac0217 Mon Sep 17 00:00:00 2001 From: Steve French Date: Fri, 20 Feb 2009 05:43:09 +0000 Subject: [CIFS] Fix multiuser mounts so server does not invalidate earlier security contexts When two different users mount the same Windows 2003 Server share using CIFS, the first session mounted can be invalidated. Some servers invalidate the first smb session when a second similar user (e.g. two users who get mapped by server to "guest") authenticates an smb session from the same client. By making sure that we set the 2nd and subsequent vc numbers to nonzero values, this ensures that we will not have this problem. Fixes Samba bug 6004, problem description follows: How to reproduce: - configure an "open share" (full permissions to Guest user) on Windows 2003 Server (I couldn't reproduce the problem with Samba server or Windows older than 2003) - mount the share twice with different users who will be authenticated as guest. noacl,noperm,user=john,dir_mode=0700,domain=DOMAIN,rw noacl,noperm,user=jeff,dir_mode=0700,domain=DOMAIN,rw Result: - just the mount point mounted last is accessible: Signed-off-by: Steve French --- fs/cifs/CHANGES | 10 +++++++ fs/cifs/cifsfs.h | 2 +- fs/cifs/cifsglob.h | 6 +++- fs/cifs/cifssmb.c | 7 +++-- fs/cifs/sess.c | 87 ++++++++++++++++++++++++++++++++++++++++++++++++++++-- 5 files changed, 105 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES index 72063f5..851388f 100644 --- a/fs/cifs/CHANGES +++ b/fs/cifs/CHANGES @@ -1,3 +1,13 @@ +Version 1.57 +------------ +Improve support for multiple security contexts to the same server. We +used to use the same "vcnumber" for all connections which could cause +the server to treat subsequent connections, especially those that +are authenticated as guest, as reconnections, invalidating the earlier +user's smb session. This fix allows cifs to mount multiple times to the +same server with different userids without risking invalidating earlier +established security contexts. + Version 1.56 ------------ Add "forcemandatorylock" mount option to allow user to use mandatory diff --git a/fs/cifs/cifsfs.h b/fs/cifs/cifsfs.h index 7ac4818..2b1d28a 100644 --- a/fs/cifs/cifsfs.h +++ b/fs/cifs/cifsfs.h @@ -100,5 +100,5 @@ extern long cifs_ioctl(struct file *filep, unsigned int cmd, unsigned long arg); extern const struct export_operations cifs_export_ops; #endif /* EXPERIMENTAL */ -#define CIFS_VERSION "1.56" +#define CIFS_VERSION "1.57" #endif /* _CIFSFS_H */ diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h index 94c1ca0..e004f6d 100644 --- a/fs/cifs/cifsglob.h +++ b/fs/cifs/cifsglob.h @@ -164,9 +164,12 @@ struct TCP_Server_Info { /* multiplexed reads or writes */ unsigned int maxBuf; /* maxBuf specifies the maximum */ /* message size the server can send or receive for non-raw SMBs */ - unsigned int maxRw; /* maxRw specifies the maximum */ + unsigned int max_rw; /* maxRw specifies the maximum */ /* message size the server can send or receive for */ /* SMB_COM_WRITE_RAW or SMB_COM_READ_RAW. */ + unsigned int max_vcs; /* maximum number of smb sessions, at least + those that can be specified uniquely with + vcnumbers */ char sessid[4]; /* unique token id for this session */ /* (returned on Negotiate */ int capabilities; /* allow selective disabling of caps by smb sess */ @@ -210,6 +213,7 @@ struct cifsSesInfo { unsigned overrideSecFlg; /* if non-zero override global sec flags */ __u16 ipc_tid; /* special tid for connection to IPC share */ __u16 flags; + __u16 vcnum; char *serverOS; /* name of operating system underlying server */ char *serverNOS; /* name of network operating system of server */ char *serverDomain; /* security realm of server */ diff --git a/fs/cifs/cifssmb.c b/fs/cifs/cifssmb.c index 552642a..939e2f7 100644 --- a/fs/cifs/cifssmb.c +++ b/fs/cifs/cifssmb.c @@ -528,14 +528,15 @@ CIFSSMBNegotiate(unsigned int xid, struct cifsSesInfo *ses) server->maxReq = le16_to_cpu(rsp->MaxMpxCount); server->maxBuf = min((__u32)le16_to_cpu(rsp->MaxBufSize), (__u32)CIFSMaxBufSize + MAX_CIFS_HDR_SIZE); + server->max_vcs = le16_to_cpu(rsp->MaxNumberVcs); GETU32(server->sessid) = le32_to_cpu(rsp->SessionKey); /* even though we do not use raw we might as well set this accurately, in case we ever find a need for it */ if ((le16_to_cpu(rsp->RawMode) & RAW_ENABLE) == RAW_ENABLE) { - server->maxRw = 0xFF00; + server->max_rw = 0xFF00; server->capabilities = CAP_MPX_MODE | CAP_RAW_MODE; } else { - server->maxRw = 0;/* we do not need to use raw anyway */ + server->max_rw = 0;/* do not need to use raw anyway */ server->capabilities = CAP_MPX_MODE; } tmp = (__s16)le16_to_cpu(rsp->ServerTimeZone); @@ -638,7 +639,7 @@ CIFSSMBNegotiate(unsigned int xid, struct cifsSesInfo *ses) /* probably no need to store and check maxvcs */ server->maxBuf = min(le32_to_cpu(pSMBr->MaxBufferSize), (__u32) CIFSMaxBufSize + MAX_CIFS_HDR_SIZE); - server->maxRw = le32_to_cpu(pSMBr->MaxRawSize); + server->max_rw = le32_to_cpu(pSMBr->MaxRawSize); cFYI(DBG2, ("Max buf = %d", ses->server->maxBuf)); GETU32(ses->server->sessid) = le32_to_cpu(pSMBr->SessionKey); server->capabilities = le32_to_cpu(pSMBr->Capabilities); diff --git a/fs/cifs/sess.c b/fs/cifs/sess.c index b234407..5c68b42 100644 --- a/fs/cifs/sess.c +++ b/fs/cifs/sess.c @@ -34,15 +34,99 @@ extern void SMBNTencrypt(unsigned char *passwd, unsigned char *c8, unsigned char *p24); +/* Checks if this is the first smb session to be reconnected after + the socket has been reestablished (so we know whether to use vc 0). + Called while holding the cifs_tcp_ses_lock, so do not block */ +static bool is_first_ses_reconnect(struct cifsSesInfo *ses) +{ + struct list_head *tmp; + struct cifsSesInfo *tmp_ses; + + list_for_each(tmp, &ses->server->smb_ses_list) { + tmp_ses = list_entry(tmp, struct cifsSesInfo, + smb_ses_list); + if (tmp_ses->need_reconnect == false) + return false; + } + /* could not find a session that was already connected, + this must be the first one we are reconnecting */ + return true; +} + +/* + * vc number 0 is treated specially by some servers, and should be the + * first one we request. After that we can use vcnumbers up to maxvcs, + * one for each smb session (some Windows versions set maxvcs incorrectly + * so maxvc=1 can be ignored). If we have too many vcs, we can reuse + * any vc but zero (some servers reset the connection on vcnum zero) + * + */ +static __le16 get_next_vcnum(struct cifsSesInfo *ses) +{ + __u16 vcnum = 0; + struct list_head *tmp; + struct cifsSesInfo *tmp_ses; + __u16 max_vcs = ses->server->max_vcs; + __u16 i; + int free_vc_found = 0; + + /* Quoting the MS-SMB specification: "Windows-based SMB servers set this + field to one but do not enforce this limit, which allows an SMB client + to establish more virtual circuits than allowed by this value ... but + other server implementations can enforce this limit." */ + if (max_vcs < 2) + max_vcs = 0xFFFF; + + write_lock(&cifs_tcp_ses_lock); + if ((ses->need_reconnect) && is_first_ses_reconnect(ses)) + goto get_vc_num_exit; /* vcnum will be zero */ + for (i = ses->server->srv_count - 1; i < max_vcs; i++) { + if (i == 0) /* this is the only connection, use vc 0 */ + break; + + free_vc_found = 1; + + list_for_each(tmp, &ses->server->smb_ses_list) { + tmp_ses = list_entry(tmp, struct cifsSesInfo, + smb_ses_list); + if (tmp_ses->vcnum == i) { + free_vc_found = 0; + break; /* found duplicate, try next vcnum */ + } + } + if (free_vc_found) + break; /* we found a vcnumber that will work - use it */ + } + + if (i == 0) + vcnum = 0; /* for most common case, ie if one smb session, use + vc zero. Also for case when no free vcnum, zero + is safest to send (some clients only send zero) */ + else if (free_vc_found == 0) + vcnum = 1; /* we can not reuse vc=0 safely, since some servers + reset all uids on that, but 1 is ok. */ + else + vcnum = i; + ses->vcnum = vcnum; +get_vc_num_exit: + write_unlock(&cifs_tcp_ses_lock); + + return le16_to_cpu(vcnum); +} + static __u32 cifs_ssetup_hdr(struct cifsSesInfo *ses, SESSION_SETUP_ANDX *pSMB) { __u32 capabilities = 0; /* init fields common to all four types of SessSetup */ - /* note that header is initialized to zero in header_assemble */ + /* Note that offsets for first seven fields in req struct are same */ + /* in CIFS Specs so does not matter which of 3 forms of struct */ + /* that we use in next few lines */ + /* Note that header is initialized to zero in header_assemble */ pSMB->req.AndXCommand = 0xFF; pSMB->req.MaxBufferSize = cpu_to_le16(ses->server->maxBuf); pSMB->req.MaxMpxCount = cpu_to_le16(ses->server->maxReq); + pSMB->req.VcNumber = get_next_vcnum(ses); /* Now no need to set SMBFLG_CASELESS or obsolete CANONICAL PATH */ @@ -71,7 +155,6 @@ static __u32 cifs_ssetup_hdr(struct cifsSesInfo *ses, SESSION_SETUP_ANDX *pSMB) if (ses->capabilities & CAP_UNIX) capabilities |= CAP_UNIX; - /* BB check whether to init vcnum BB */ return capabilities; } -- cgit v1.1 From cac711211a039ae2e2dc6322ffb3c2279d093bf1 Mon Sep 17 00:00:00 2001 From: Krzysztof Sachanowicz Date: Mon, 23 Feb 2009 22:21:55 +0100 Subject: proc: proc_get_inode should de_put when inode already initialized de_get is called before every proc_get_inode, but corresponding de_put is called only when dropping last reference to an inode. This might cause something like remove_proc_entry: /proc/stats busy, count=14496 to be printed to the syslog. The fix is to call de_put in case of an already initialized inode in proc_get_inode. Signed-off-by: Krzysztof Sachanowicz Tested-by: Marcin Pilipczuk Acked-by: Al Viro Signed-off-by: Linus Torvalds --- fs/proc/inode.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/proc/inode.c b/fs/proc/inode.c index 3e76bb9..d8bb5c6 100644 --- a/fs/proc/inode.c +++ b/fs/proc/inode.c @@ -485,8 +485,10 @@ struct inode *proc_get_inode(struct super_block *sb, unsigned int ino, } } unlock_new_inode(inode); - } else + } else { module_put(de->owner); + de_put(de); + } return inode; out_ino: -- cgit v1.1