From 9e13f1f9de1cb143fbae6f1170f26c8544b64cff Mon Sep 17 00:00:00 2001 From: Junxiao Bi Date: Fri, 25 Mar 2016 14:20:50 -0700 Subject: ocfs2: o2hb: fix double free bug This is a regression issue and caused the following kernel panic when do ocfs2 multiple test. BUG: unable to handle kernel paging request at 00000002000800c0 IP: [] kmem_cache_alloc+0x78/0x160 PGD 7bbe5067 PUD 0 Oops: 0000 [#1] SMP Modules linked in: ocfs2_dlmfs ocfs2_stack_o2cb ocfs2_dlm ocfs2_nodemanager ocfs2_stackglue iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi xen_kbdfront xen_netfront xen_fbfront xen_blkfront CPU: 2 PID: 4044 Comm: mpirun Not tainted 4.5.0-rc5-next-20160225 #1 Hardware name: Xen HVM domU, BIOS 4.3.1OVM 05/14/2014 task: ffff88007a521a80 ti: ffff88007aed0000 task.ti: ffff88007aed0000 RIP: 0010:[] [] kmem_cache_alloc+0x78/0x160 RSP: 0018:ffff88007aed3a48 EFLAGS: 00010282 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000001991 RDX: 0000000000001990 RSI: 00000000024000c0 RDI: 000000000001b330 RBP: ffff88007aed3a98 R08: ffff88007d29b330 R09: 00000002000800c0 R10: 0000000c51376d87 R11: ffff8800792cac38 R12: ffff88007cc30f00 R13: 00000000024000c0 R14: ffffffff811b053f R15: ffff88007aed3ce7 FS: 0000000000000000(0000) GS:ffff88007d280000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000002000800c0 CR3: 000000007aeb2000 CR4: 00000000000406e0 Call Trace: __d_alloc+0x2f/0x1a0 d_alloc+0x17/0x80 lookup_dcache+0x8a/0xc0 path_openat+0x3c3/0x1210 do_filp_open+0x80/0xe0 do_sys_open+0x110/0x200 SyS_open+0x19/0x20 do_syscall_64+0x72/0x230 entry_SYSCALL64_slow_path+0x25/0x25 Code: 05 e6 77 e7 7e 4d 8b 08 49 8b 40 10 4d 85 c9 0f 84 dd 00 00 00 48 85 c0 0f 84 d4 00 00 00 49 63 44 24 20 49 8b 3c 24 48 8d 4a 01 <49> 8b 1c 01 4c 89 c8 65 48 0f c7 0f 0f 94 c0 3c 01 75 b6 49 63 RIP kmem_cache_alloc+0x78/0x160 CR2: 00000002000800c0 ---[ end trace 823969e602e4aaac ]--- Fixes: a4a1dfa4bb8b("ocfs2/cluster: fix memory leak in o2hb_region_release") Signed-off-by: Junxiao Bi Reviewed-by: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/cluster/heartbeat.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index ef6a2ec494de..bd15929b5f92 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -1444,8 +1444,8 @@ static void o2hb_region_release(struct config_item *item) debugfs_remove(reg->hr_debug_dir); kfree(reg->hr_db_livenodes); kfree(reg->hr_db_regnum); - kfree(reg->hr_debug_elapsed_time); - kfree(reg->hr_debug_pinned); + kfree(reg->hr_db_elapsed_time); + kfree(reg->hr_db_pinned); spin_lock(&o2hb_live_lock); list_del(®->hr_all_item); -- cgit v1.2.3 From c1ad1e3ca3064b1f0ab5a5185b4732ab9ad8fa24 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:20:52 -0700 Subject: ocfs2: add ocfs2_write_type_t type to identify the caller of write Patchset: fix ocfs2 direct io code patch to support sparse file and data ordering semantics The idea is to use buffer io(more precisely use the interface ocfs2_write_begin_nolock & ocfs2_write_end_nolock) to do the zero work beyond block size. And clear UNWRITTEN flag until direct io data has been written to disk, which can prevent data corruption when system crashed during direct write. And we will also archive a better performance: eg. dd direct write new file with block size 4KB: before this patchset: 2.5 MB/s after this patchset: 66.4 MB/s This patch (of 8): To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. Remove unused args filp & flags. Add new arg type. The type is one of buffer/direct/mmap. Indicate 3 way to perform write. buffer/mmap type has implemented. direct type will be implemented later. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 20 ++++++++++++-------- fs/ocfs2/aops.h | 11 ++++++++--- fs/ocfs2/mmap.c | 4 ++-- 3 files changed, 22 insertions(+), 13 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 043110e5212d..69d7392f56d8 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1224,6 +1224,9 @@ struct ocfs2_write_ctxt { /* First cluster allocated in a nonsparse extend */ u32 w_first_new_cpos; + /* Type of caller. Must be one of buffer, mmap, direct. */ + ocfs2_write_type_t w_type; + struct ocfs2_write_cluster_desc w_desc[OCFS2_MAX_CLUSTERS_PER_PAGE]; /* @@ -1319,7 +1322,8 @@ static void ocfs2_free_write_ctxt(struct ocfs2_write_ctxt *wc) static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, struct ocfs2_super *osb, loff_t pos, - unsigned len, struct buffer_head *di_bh) + unsigned len, ocfs2_write_type_t type, + struct buffer_head *di_bh) { u32 cend; struct ocfs2_write_ctxt *wc; @@ -1334,6 +1338,7 @@ static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, wc->w_clen = cend - wc->w_cpos + 1; get_bh(di_bh); wc->w_di_bh = di_bh; + wc->w_type = type; if (unlikely(PAGE_CACHE_SHIFT > osb->s_clustersize_bits)) wc->w_large_pages = 1; @@ -2077,9 +2082,8 @@ out: return ret; } -int ocfs2_write_begin_nolock(struct file *filp, - struct address_space *mapping, - loff_t pos, unsigned len, unsigned flags, +int ocfs2_write_begin_nolock(struct address_space *mapping, + loff_t pos, unsigned len, ocfs2_write_type_t type, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page) { @@ -2096,7 +2100,7 @@ int ocfs2_write_begin_nolock(struct file *filp, int try_free = 1, ret1; try_again: - ret = ocfs2_alloc_write_ctxt(&wc, osb, pos, len, di_bh); + ret = ocfs2_alloc_write_ctxt(&wc, osb, pos, len, type, di_bh); if (ret) { mlog_errno(ret); return ret; @@ -2153,7 +2157,7 @@ try_again: (unsigned long long)OCFS2_I(inode)->ip_blkno, (long long)i_size_read(inode), le32_to_cpu(di->i_clusters), - pos, len, flags, mmap_page, + pos, len, type, mmap_page, clusters_to_alloc, extents_to_split); /* @@ -2323,8 +2327,8 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping, */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_write_begin_nolock(file, mapping, pos, len, flags, pagep, - fsdata, di_bh, NULL); + ret = ocfs2_write_begin_nolock(mapping, pos, len, OCFS2_WRITE_BUFFER, + pagep, fsdata, di_bh, NULL); if (ret) { mlog_errno(ret); goto out_fail; diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index 24e496d6bdcd..d06b80f58f83 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -47,9 +47,14 @@ int ocfs2_write_end_nolock(struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata); -int ocfs2_write_begin_nolock(struct file *filp, - struct address_space *mapping, - loff_t pos, unsigned len, unsigned flags, +typedef enum { + OCFS2_WRITE_BUFFER = 0, + OCFS2_WRITE_DIRECT, + OCFS2_WRITE_MMAP, +} ocfs2_write_type_t; + +int ocfs2_write_begin_nolock(struct address_space *mapping, + loff_t pos, unsigned len, ocfs2_write_type_t type, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page); diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index 77ebc2bc1cca..9ea081f4e6e4 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -104,8 +104,8 @@ static int __ocfs2_page_mkwrite(struct file *file, struct buffer_head *di_bh, if (page->index == last_index) len = ((size - 1) & ~PAGE_CACHE_MASK) + 1; - ret = ocfs2_write_begin_nolock(file, mapping, pos, len, 0, &locked_page, - &fsdata, di_bh, page); + ret = ocfs2_write_begin_nolock(mapping, pos, len, OCFS2_WRITE_MMAP, + &locked_page, &fsdata, di_bh, page); if (ret) { if (ret != -ENOSPC) mlog_errno(ret); -- cgit v1.2.3 From b46637d59f1160dee5d8e03498e667ab36c2be04 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:20:55 -0700 Subject: ocfs2: use c_new to indicate newly allocated extents To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. There is a problem in ocfs2's direct io implement: if system crashed after extents allocated, and before data return, we will get a extent with dirty data on disk. This problem violate the journal=order semantics, which means meta changes take effect after data written to disk. To resolve this issue, direct write can use the UNWRITTEN flag to describe a extent during direct data writeback. The direct write procedure should act in the following order: phase 1: alloc extent with UNWRITTEN flag phase 2: submit direct data to disk, add zero page to page cache phase 3: clear UNWRITTEN flag when data has been written to disk This patch is to change the 'c_unwritten' member of ocfs2_write_cluster_desc to 'c_clear_unwritten'. Means whether to clear the unwritten flag. It do not care if a extent is allocated or not. And use 'c_new' to specify a newly allocated extent. So the direct io procedure can use c_clear_unwritten to control the UNWRITTEN bit on extent. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 69d7392f56d8..516ea896f13c 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1212,7 +1212,7 @@ struct ocfs2_write_cluster_desc { * filled. */ unsigned c_new; - unsigned c_unwritten; + unsigned c_clear_unwritten; unsigned c_needs_zero; }; @@ -1588,19 +1588,19 @@ out: * Prepare a single cluster for write one cluster into the file. */ static int ocfs2_write_cluster(struct address_space *mapping, - u32 phys, unsigned int unwritten, + u32 phys, unsigned int new, + unsigned int clear_unwritten, unsigned int should_zero, struct ocfs2_alloc_context *data_ac, struct ocfs2_alloc_context *meta_ac, struct ocfs2_write_ctxt *wc, u32 cpos, loff_t user_pos, unsigned user_len) { - int ret, i, new; + int ret, i; u64 v_blkno, p_blkno; struct inode *inode = mapping->host; struct ocfs2_extent_tree et; - new = phys == 0 ? 1 : 0; if (new) { u32 tmp_pos; @@ -1610,9 +1610,9 @@ static int ocfs2_write_cluster(struct address_space *mapping, */ tmp_pos = cpos; ret = ocfs2_add_inode_data(OCFS2_SB(inode->i_sb), inode, - &tmp_pos, 1, 0, wc->w_di_bh, - wc->w_handle, data_ac, - meta_ac, NULL); + &tmp_pos, 1, !clear_unwritten, + wc->w_di_bh, wc->w_handle, + data_ac, meta_ac, NULL); /* * This shouldn't happen because we must have already * calculated the correct meta data allocation required. The @@ -1629,7 +1629,7 @@ static int ocfs2_write_cluster(struct address_space *mapping, mlog_errno(ret); goto out; } - } else if (unwritten) { + } else if (clear_unwritten) { ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), wc->w_di_bh); ret = ocfs2_mark_extent_written(inode, &et, @@ -1712,7 +1712,8 @@ static int ocfs2_write_cluster_by_desc(struct address_space *mapping, local_len = osb->s_clustersize - cluster_off; ret = ocfs2_write_cluster(mapping, desc->c_phys, - desc->c_unwritten, + desc->c_new, + desc->c_clear_unwritten, desc->c_needs_zero, data_ac, meta_ac, wc, desc->c_cpos, pos, local_len); @@ -1857,11 +1858,12 @@ static int ocfs2_populate_write_desc(struct inode *inode, if (phys == 0) { desc->c_new = 1; desc->c_needs_zero = 1; + desc->c_clear_unwritten = 1; *clusters_to_alloc = *clusters_to_alloc + 1; } if (ext_flags & OCFS2_EXT_UNWRITTEN) { - desc->c_unwritten = 1; + desc->c_clear_unwritten = 1; desc->c_needs_zero = 1; } -- cgit v1.2.3 From 65c4db8c8282cd239f0ba3226cfe59cc9768193c Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:20:58 -0700 Subject: ocfs2: test target page before change it To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. Direct io data will not appear in buffer. The w_target_page member will not be filled by direct io. So avoid to use it when it's NULL. Unlinke buffer io and mmap, direct io will call write_begin with more than 1 page a time. So the target_index is not sufficient to describe the actual data. change it to a range start at target_index, end in end_index. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 32 ++++++++++++++++++++++++++------ 1 file changed, 26 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 516ea896f13c..100f7a8a0de1 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1406,12 +1406,13 @@ static void ocfs2_write_failure(struct inode *inode, to = user_pos + user_len; struct page *tmppage; - ocfs2_zero_new_buffers(wc->w_target_page, from, to); + if (wc->w_target_page) + ocfs2_zero_new_buffers(wc->w_target_page, from, to); for(i = 0; i < wc->w_num_pages; i++) { tmppage = wc->w_pages[i]; - if (page_has_buffers(tmppage)) { + if (tmppage && page_has_buffers(tmppage)) { if (ocfs2_should_order_data(inode)) ocfs2_jbd2_file_inode(wc->w_handle, inode); @@ -1541,11 +1542,13 @@ static int ocfs2_grab_pages_for_write(struct address_space *mapping, wc->w_num_pages = 1; start = target_index; } + end_index = (user_pos + user_len - 1) >> PAGE_CACHE_SHIFT; for(i = 0; i < wc->w_num_pages; i++) { index = start + i; - if (index == target_index && mmap_page) { + if (index >= target_index && index <= end_index && + wc->w_type == OCFS2_WRITE_MMAP) { /* * ocfs2_pagemkwrite() is a little different * and wants us to directly use the page @@ -1564,6 +1567,11 @@ static int ocfs2_grab_pages_for_write(struct address_space *mapping, page_cache_get(mmap_page); wc->w_pages[i] = mmap_page; wc->w_target_locked = true; + } else if (index >= target_index && index <= end_index && + wc->w_type == OCFS2_WRITE_DIRECT) { + /* Direct write has no mapping page. */ + wc->w_pages[i] = NULL; + continue; } else { wc->w_pages[i] = find_or_create_page(mapping, index, GFP_NOFS); @@ -1665,6 +1673,12 @@ static int ocfs2_write_cluster(struct address_space *mapping, for(i = 0; i < wc->w_num_pages; i++) { int tmpret; + /* This is the direct io target page. */ + if (wc->w_pages[i] == NULL) { + p_blkno++; + continue; + } + tmpret = ocfs2_prepare_page_for_write(inode, &p_blkno, wc, wc->w_pages[i], cpos, user_pos, user_len, @@ -2266,7 +2280,8 @@ try_again: ocfs2_free_alloc_context(meta_ac); success: - *pagep = wc->w_target_page; + if (pagep) + *pagep = wc->w_target_page; *fsdata = wc; return 0; out_quota: @@ -2400,18 +2415,23 @@ int ocfs2_write_end_nolock(struct address_space *mapping, goto out_write_size; } - if (unlikely(copied < len)) { + if (unlikely(copied < len) && wc->w_target_page) { if (!PageUptodate(wc->w_target_page)) copied = 0; ocfs2_zero_new_buffers(wc->w_target_page, start+copied, start+len); } - flush_dcache_page(wc->w_target_page); + if (wc->w_target_page) + flush_dcache_page(wc->w_target_page); for(i = 0; i < wc->w_num_pages; i++) { tmppage = wc->w_pages[i]; + /* This is the direct io target page. */ + if (tmppage == NULL) + continue; + if (tmppage == wc->w_target_page) { from = wc->w_target_from; to = wc->w_target_to; -- cgit v1.2.3 From 46e625565950c5f31a193c4bbcfdef3fdc7235bf Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:01 -0700 Subject: ocfs2: do not change i_size in write_end for direct io To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. Append direct io do not change i_size in get block phase. It only move to orphan when starting write. After data is written to disk, it will delete itself from orphan and update i_size. So skip i_size change section in write_begin for direct io. And when there is no extents alloc, no meta data changes needed for direct io (since write_begin start trans for 2 reason: alloc extents & change i_size. Now none of them needed). So we can skip start trans procedure. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 79 +++++++++++++++++++++++++++++++++------------------------ 1 file changed, 46 insertions(+), 33 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 100f7a8a0de1..4ad77453438a 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2043,8 +2043,10 @@ static int ocfs2_expand_nonsparse_inode(struct inode *inode, if (ret) mlog_errno(ret); - wc->w_first_new_cpos = - ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)); + /* There is no wc if this is call from direct. */ + if (wc) + wc->w_first_new_cpos = + ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)); return ret; } @@ -2135,14 +2137,17 @@ try_again: } } - if (ocfs2_sparse_alloc(osb)) - ret = ocfs2_zero_tail(inode, di_bh, pos); - else - ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, len, - wc); - if (ret) { - mlog_errno(ret); - goto out; + /* Direct io change i_size late, should not zero tail here. */ + if (type != OCFS2_WRITE_DIRECT) { + if (ocfs2_sparse_alloc(osb)) + ret = ocfs2_zero_tail(inode, di_bh, pos); + else + ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, + len, wc); + if (ret) { + mlog_errno(ret); + goto out; + } } ret = ocfs2_check_range_for_refcount(inode, pos, len); @@ -2203,8 +2208,9 @@ try_again: credits = ocfs2_calc_extend_credits(inode->i_sb, &di->id2.i_list); - - } + } else if (type == OCFS2_WRITE_DIRECT) + /* direct write needs not to start trans if no extents alloc. */ + goto success; /* * We have to zero sparse allocated clusters, unwritten extent clusters, @@ -2402,12 +2408,14 @@ int ocfs2_write_end_nolock(struct address_space *mapping, handle_t *handle = wc->w_handle; struct page *tmppage; - ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (ret) { - copied = ret; - mlog_errno(ret); - goto out; + if (handle) { + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), + wc->w_di_bh, OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + copied = ret; + mlog_errno(ret); + goto out; + } } if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { @@ -2450,25 +2458,29 @@ int ocfs2_write_end_nolock(struct address_space *mapping, } if (page_has_buffers(tmppage)) { - if (ocfs2_should_order_data(inode)) - ocfs2_jbd2_file_inode(wc->w_handle, inode); + if (handle && ocfs2_should_order_data(inode)) + ocfs2_jbd2_file_inode(handle, inode); block_commit_write(tmppage, from, to); } } out_write_size: - pos += copied; - if (pos > i_size_read(inode)) { - i_size_write(inode, pos); - mark_inode_dirty(inode); - } - inode->i_blocks = ocfs2_inode_sector_count(inode); - di->i_size = cpu_to_le64((u64)i_size_read(inode)); - inode->i_mtime = inode->i_ctime = CURRENT_TIME; - di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec); - di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); - ocfs2_update_inode_fsync_trans(handle, inode, 1); - ocfs2_journal_dirty(handle, wc->w_di_bh); + /* Direct io do not update i_size here. */ + if (wc->w_type != OCFS2_WRITE_DIRECT) { + pos += copied; + if (pos > i_size_read(inode)) { + i_size_write(inode, pos); + mark_inode_dirty(inode); + } + inode->i_blocks = ocfs2_inode_sector_count(inode); + di->i_size = cpu_to_le64((u64)i_size_read(inode)); + inode->i_mtime = inode->i_ctime = CURRENT_TIME; + di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec); + di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); + ocfs2_update_inode_fsync_trans(handle, inode, 1); + } + if (handle) + ocfs2_journal_dirty(handle, wc->w_di_bh); out: /* unlock pages before dealloc since it needs acquiring j_trans_barrier @@ -2478,7 +2490,8 @@ out: */ ocfs2_unlock_pages(wc); - ocfs2_commit_trans(osb, handle); + if (handle) + ocfs2_commit_trans(osb, handle); ocfs2_run_deallocs(osb, &wc->w_dealloc); -- cgit v1.2.3 From 2de6a3c73180ef4071b45185756be51a6c767924 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:03 -0700 Subject: ocfs2: return the physical address in ocfs2_write_cluster To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. Direct io needs to get the physical address from write_begin, to map the user page. This patch is to change the arg 'phys' of ocfs2_write_cluster to a pointer, so it can be retrieved to write_begin. And we can retrieve it to the direct io procedure. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 28 +++++++++++++--------------- 1 file changed, 13 insertions(+), 15 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 4ad77453438a..7b268c357cf3 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1596,7 +1596,7 @@ out: * Prepare a single cluster for write one cluster into the file. */ static int ocfs2_write_cluster(struct address_space *mapping, - u32 phys, unsigned int new, + u32 *phys, unsigned int new, unsigned int clear_unwritten, unsigned int should_zero, struct ocfs2_alloc_context *data_ac, @@ -1605,9 +1605,10 @@ static int ocfs2_write_cluster(struct address_space *mapping, loff_t user_pos, unsigned user_len) { int ret, i; - u64 v_blkno, p_blkno; + u64 p_blkno; struct inode *inode = mapping->host; struct ocfs2_extent_tree et; + int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); if (new) { u32 tmp_pos; @@ -1641,7 +1642,7 @@ static int ocfs2_write_cluster(struct address_space *mapping, ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), wc->w_di_bh); ret = ocfs2_mark_extent_written(inode, &et, - wc->w_handle, cpos, 1, phys, + wc->w_handle, cpos, 1, *phys, meta_ac, &wc->w_dealloc); if (ret < 0) { mlog_errno(ret); @@ -1649,26 +1650,23 @@ static int ocfs2_write_cluster(struct address_space *mapping, } } - if (should_zero) - v_blkno = ocfs2_clusters_to_blocks(inode->i_sb, cpos); - else - v_blkno = user_pos >> inode->i_sb->s_blocksize_bits; - /* * The only reason this should fail is due to an inability to * find the extent added. */ - ret = ocfs2_extent_map_get_blocks(inode, v_blkno, &p_blkno, NULL, - NULL); + ret = ocfs2_get_clusters(inode, cpos, phys, NULL, NULL); if (ret < 0) { mlog(ML_ERROR, "Get physical blkno failed for inode %llu, " - "at logical block %llu", - (unsigned long long)OCFS2_I(inode)->ip_blkno, - (unsigned long long)v_blkno); + "at logical cluster %u", + (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos); goto out; } - BUG_ON(p_blkno == 0); + BUG_ON(*phys == 0); + + p_blkno = ocfs2_clusters_to_blocks(inode->i_sb, *phys); + if (!should_zero) + p_blkno += (user_pos >> inode->i_sb->s_blocksize_bits) & (u64)(bpc - 1); for(i = 0; i < wc->w_num_pages; i++) { int tmpret; @@ -1725,7 +1723,7 @@ static int ocfs2_write_cluster_by_desc(struct address_space *mapping, if ((cluster_off + local_len) > osb->s_clustersize) local_len = osb->s_clustersize - cluster_off; - ret = ocfs2_write_cluster(mapping, desc->c_phys, + ret = ocfs2_write_cluster(mapping, &desc->c_phys, desc->c_new, desc->c_clear_unwritten, desc->c_needs_zero, -- cgit v1.2.3 From 4506cfb6f8cad594ac73e0df2b2961ca10dbd25e Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:06 -0700 Subject: ocfs2: record UNWRITTEN extents when populate write desc To support direct io in ocfs2_write_begin_nolock & ocfs2_write_end_nolock. There is still one issue in the direct write procedure. phase 1: alloc extent with UNWRITTEN flag phase 2: submit direct data to disk, add zero page to page cache phase 3: clear UNWRITTEN flag when data has been written to disk When there are 2 direct write A(0~3KB),B(4~7KB) writing to the same cluster 0~7KB (cluster size 8KB). Write request A arrive phase 2 first, it will zero the region (4~7KB). Before request A enter to phase 3, request B arrive phase 2, it will zero region (0~3KB). This is just like request B steps request A. To resolve this issue, we should let request B knows this cluster is already under zero, to prevent it from steps the previous write request. This patch will add function ocfs2_unwritten_check() to do this job. It will record all clusters that are under direct write(it will be recorded in the 'ip_unwritten_list' member of inode info), and prevent the later direct write writing to the same cluster to do the zero work again. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 104 ++++++++++++++++++++++++++++++++++++++++++++++++++++--- fs/ocfs2/inode.c | 3 ++ fs/ocfs2/inode.h | 3 ++ fs/ocfs2/super.c | 1 + 4 files changed, 106 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 7b268c357cf3..c29d06634fd6 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1201,6 +1201,13 @@ next_bh: #define OCFS2_MAX_CLUSTERS_PER_PAGE (PAGE_CACHE_SIZE / OCFS2_MIN_CLUSTERSIZE) +struct ocfs2_unwritten_extent { + struct list_head ue_node; + struct list_head ue_ip_node; + u32 ue_cpos; + u32 ue_phys; +}; + /* * Describe the state of a single cluster to be written to. */ @@ -1275,6 +1282,8 @@ struct ocfs2_write_ctxt { struct buffer_head *w_di_bh; struct ocfs2_cached_dealloc_ctxt w_dealloc; + + struct list_head w_unwritten_list; }; void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages) @@ -1313,8 +1322,25 @@ static void ocfs2_unlock_pages(struct ocfs2_write_ctxt *wc) ocfs2_unlock_and_free_pages(wc->w_pages, wc->w_num_pages); } -static void ocfs2_free_write_ctxt(struct ocfs2_write_ctxt *wc) +static void ocfs2_free_unwritten_list(struct inode *inode, + struct list_head *head) +{ + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_unwritten_extent *dz = NULL, *tmp = NULL; + + list_for_each_entry_safe(dz, tmp, head, ue_node) { + list_del(&dz->ue_node); + spin_lock(&oi->ip_lock); + list_del(&dz->ue_ip_node); + spin_unlock(&oi->ip_lock); + kfree(dz); + } +} + +static void ocfs2_free_write_ctxt(struct inode *inode, + struct ocfs2_write_ctxt *wc) { + ocfs2_free_unwritten_list(inode, &wc->w_unwritten_list); ocfs2_unlock_pages(wc); brelse(wc->w_di_bh); kfree(wc); @@ -1346,6 +1372,7 @@ static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, wc->w_large_pages = 0; ocfs2_init_dealloc_ctxt(&wc->w_dealloc); + INIT_LIST_HEAD(&wc->w_unwritten_list); *wcp = wc; @@ -1795,6 +1822,66 @@ static void ocfs2_set_target_boundaries(struct ocfs2_super *osb, } } +/* + * Check if this extent is marked UNWRITTEN by direct io. If so, we need not to + * do the zero work. And should not to clear UNWRITTEN since it will be cleared + * by the direct io procedure. + * If this is a new extent that allocated by direct io, we should mark it in + * the ip_unwritten_list. + */ +static int ocfs2_unwritten_check(struct inode *inode, + struct ocfs2_write_ctxt *wc, + struct ocfs2_write_cluster_desc *desc) +{ + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_unwritten_extent *dz = NULL, *new = NULL; + int ret = 0; + + if (!desc->c_needs_zero) + return 0; + +retry: + spin_lock(&oi->ip_lock); + /* Needs not to zero no metter buffer or direct. The one who is zero + * the cluster is doing zero. And he will clear unwritten after all + * cluster io finished. */ + list_for_each_entry(dz, &oi->ip_unwritten_list, ue_ip_node) { + if (desc->c_cpos == dz->ue_cpos) { + BUG_ON(desc->c_new); + desc->c_needs_zero = 0; + desc->c_clear_unwritten = 0; + goto unlock; + } + } + + if (wc->w_type != OCFS2_WRITE_DIRECT) + goto unlock; + + if (new == NULL) { + spin_unlock(&oi->ip_lock); + new = kmalloc(sizeof(struct ocfs2_unwritten_extent), + GFP_NOFS); + if (new == NULL) { + ret = -ENOMEM; + goto out; + } + goto retry; + } + /* This direct write will doing zero. */ + new->ue_cpos = desc->c_cpos; + new->ue_phys = desc->c_phys; + desc->c_clear_unwritten = 0; + list_add_tail(&new->ue_ip_node, &oi->ip_unwritten_list); + list_add_tail(&new->ue_node, &wc->w_unwritten_list); + new = NULL; +unlock: + spin_unlock(&oi->ip_lock); +out: + if (new) + kfree(new); + return ret; +} + /* * Populate each single-cluster write descriptor in the write context * with information about the i/o to be done. @@ -1879,6 +1966,12 @@ static int ocfs2_populate_write_desc(struct inode *inode, desc->c_needs_zero = 1; } + ret = ocfs2_unwritten_check(inode, wc, desc); + if (ret) { + mlog_errno(ret); + goto out; + } + num_clusters--; } @@ -2215,9 +2308,8 @@ try_again: * and non-sparse clusters we just extended. For non-sparse writes, * we know zeros will only be needed in the first and/or last cluster. */ - if (clusters_to_alloc || extents_to_split || - (wc->w_clen && (wc->w_desc[0].c_needs_zero || - wc->w_desc[wc->w_clen - 1].c_needs_zero))) + if (wc->w_clen && (wc->w_desc[0].c_needs_zero || + wc->w_desc[wc->w_clen - 1].c_needs_zero)) cluster_of_pages = 1; else cluster_of_pages = 0; @@ -2296,7 +2388,7 @@ out_commit: ocfs2_commit_trans(osb, handle); out: - ocfs2_free_write_ctxt(wc); + ocfs2_free_write_ctxt(inode, wc); if (data_ac) { ocfs2_free_alloc_context(data_ac); @@ -2406,6 +2498,8 @@ int ocfs2_write_end_nolock(struct address_space *mapping, handle_t *handle = wc->w_handle; struct page *tmppage; + BUG_ON(!list_empty(&wc->w_unwritten_list)); + if (handle) { ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh, OCFS2_JOURNAL_ACCESS_WRITE); diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c index ba495beff1c2..12f4a9e9800f 100644 --- a/fs/ocfs2/inode.c +++ b/fs/ocfs2/inode.c @@ -1170,6 +1170,9 @@ static void ocfs2_clear_inode(struct inode *inode) mlog_bug_on_msg(!list_empty(&oi->ip_io_markers), "Clear inode of %llu, inode has io markers\n", (unsigned long long)oi->ip_blkno); + mlog_bug_on_msg(!list_empty(&oi->ip_unwritten_list), + "Clear inode of %llu, inode has unwritten extents\n", + (unsigned long long)oi->ip_blkno); ocfs2_extent_map_trunc(inode, 0); diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index 01635e016b3e..68e8cf9bda97 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -57,6 +57,9 @@ struct ocfs2_inode_info u32 ip_flags; /* see below */ u32 ip_attr; /* inode attributes */ + /* Record unwritten extents during direct io. */ + struct list_head ip_unwritten_list; + /* protected by recovery_lock. */ struct inode *ip_next_orphan; diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index ccc9386c42c5..d142fb75740e 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -1745,6 +1745,7 @@ static void ocfs2_inode_init_once(void *data) spin_lock_init(&oi->ip_lock); ocfs2_extent_map_init(&oi->vfs_inode); INIT_LIST_HEAD(&oi->ip_io_markers); + INIT_LIST_HEAD(&oi->ip_unwritten_list); oi->ip_dir_start_lookup = 0; mutex_init(&oi->ip_unaligned_aio); init_rwsem(&oi->ip_alloc_sem); -- cgit v1.2.3 From c15471f79506830f80eca0e7fe09b8213953ab5f Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:09 -0700 Subject: ocfs2: fix sparse file & data ordering issue in direct io There are mainly three issues in the direct io code path after commit 24c40b329e03 ("ocfs2: implement ocfs2_direct_IO_write"): * Does not support sparse file. * Does not support data ordering. eg: when write to a file hole, it will alloc extent first. If system crashed before io finished, data will corrupt. * Potential risk when doing aio+dio. The -EIOCBQUEUED return value is likely to be ignored by ocfs2_direct_IO_write(). To resolve above problems, re-design direct io code with following ideas: * Use buffer io to fill in holes. And this will make better performance also. * Clear unwritten after direct write finished. So we can make sure meta data changes after data write to disk. (Unwritten extent is invisible to user, from user's view, meta data is not changed when allocate an unwritten extent.) * Clear ocfs2_direct_IO_write(). Do all ending work in end_io. This patch has passed fs,dio,ltp-aiodio.part1,ltp-aiodio.part2,ltp-aiodio.part4 test cases of ltp. For performance improvement, see following test result: ocfs2 cluster size 1MB, ocfs2 volume is mounted on /mnt/. The original way: + rm /mnt/test.img -f + dd if=/dev/zero of=/mnt/test.img bs=4K count=1048576 oflag=direct 1048576+0 records in 1048576+0 records out 4294967296 bytes (4.3 GB) copied, 1707.83 s, 2.5 MB/s + rm /mnt/test.img -f + dd if=/dev/zero of=/mnt/test.img bs=256K count=16384 oflag=direct 16384+0 records in 16384+0 records out 4294967296 bytes (4.3 GB) copied, 582.705 s, 7.4 MB/s After this patch: + rm /mnt/test.img -f + dd if=/dev/zero of=/mnt/test.img bs=4K count=1048576 oflag=direct 1048576+0 records in 1048576+0 records out 4294967296 bytes (4.3 GB) copied, 64.6412 s, 66.4 MB/s + rm /mnt/test.img -f + dd if=/dev/zero of=/mnt/test.img bs=256K count=16384 oflag=direct 16384+0 records in 16384+0 records out 4294967296 bytes (4.3 GB) copied, 34.7611 s, 124 MB/s Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 863 +++++++++++++++++++++++--------------------------------- 1 file changed, 346 insertions(+), 517 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index c29d06634fd6..578ce225cf83 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -499,158 +499,6 @@ bail: return status; } -/* - * TODO: Make this into a generic get_blocks function. - * - * From do_direct_io in direct-io.c: - * "So what we do is to permit the ->get_blocks function to populate - * bh.b_size with the size of IO which is permitted at this offset and - * this i_blkbits." - * - * This function is called directly from get_more_blocks in direct-io.c. - * - * called like this: dio->get_blocks(dio->inode, fs_startblk, - * fs_count, map_bh, dio->rw == WRITE); - */ -static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock, - struct buffer_head *bh_result, int create) -{ - int ret; - u32 cpos = 0; - int alloc_locked = 0; - u64 p_blkno, inode_blocks, contig_blocks; - unsigned int ext_flags; - unsigned char blocksize_bits = inode->i_sb->s_blocksize_bits; - unsigned long max_blocks = bh_result->b_size >> inode->i_blkbits; - unsigned long len = bh_result->b_size; - unsigned int clusters_to_alloc = 0, contig_clusters = 0; - - cpos = ocfs2_blocks_to_clusters(inode->i_sb, iblock); - - /* This function won't even be called if the request isn't all - * nicely aligned and of the right size, so there's no need - * for us to check any of that. */ - - inode_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); - - down_read(&OCFS2_I(inode)->ip_alloc_sem); - - /* This figures out the size of the next contiguous block, and - * our logical offset */ - ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno, - &contig_blocks, &ext_flags); - up_read(&OCFS2_I(inode)->ip_alloc_sem); - - if (ret) { - mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n", - (unsigned long long)iblock); - ret = -EIO; - goto bail; - } - - /* We should already CoW the refcounted extent in case of create. */ - BUG_ON(create && (ext_flags & OCFS2_EXT_REFCOUNTED)); - - /* allocate blocks if no p_blkno is found, and create == 1 */ - if (!p_blkno && create) { - ret = ocfs2_inode_lock(inode, NULL, 1); - if (ret < 0) { - mlog_errno(ret); - goto bail; - } - - alloc_locked = 1; - - down_write(&OCFS2_I(inode)->ip_alloc_sem); - - /* fill hole, allocate blocks can't be larger than the size - * of the hole */ - clusters_to_alloc = ocfs2_clusters_for_bytes(inode->i_sb, len); - contig_clusters = ocfs2_clusters_for_blocks(inode->i_sb, - contig_blocks); - if (clusters_to_alloc > contig_clusters) - clusters_to_alloc = contig_clusters; - - /* allocate extent and insert them into the extent tree */ - ret = ocfs2_extend_allocation(inode, cpos, - clusters_to_alloc, 0); - if (ret < 0) { - up_write(&OCFS2_I(inode)->ip_alloc_sem); - mlog_errno(ret); - goto bail; - } - - ret = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno, - &contig_blocks, &ext_flags); - if (ret < 0) { - up_write(&OCFS2_I(inode)->ip_alloc_sem); - mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n", - (unsigned long long)iblock); - ret = -EIO; - goto bail; - } - set_buffer_new(bh_result); - up_write(&OCFS2_I(inode)->ip_alloc_sem); - } - - /* - * get_more_blocks() expects us to describe a hole by clearing - * the mapped bit on bh_result(). - * - * Consider an unwritten extent as a hole. - */ - if (p_blkno && !(ext_flags & OCFS2_EXT_UNWRITTEN)) - map_bh(bh_result, inode->i_sb, p_blkno); - else - clear_buffer_mapped(bh_result); - - /* make sure we don't map more than max_blocks blocks here as - that's all the kernel will handle at this point. */ - if (max_blocks < contig_blocks) - contig_blocks = max_blocks; - bh_result->b_size = contig_blocks << blocksize_bits; -bail: - if (alloc_locked) - ocfs2_inode_unlock(inode, 1); - return ret; -} - -/* - * ocfs2_dio_end_io is called by the dio core when a dio is finished. We're - * particularly interested in the aio/dio case. We use the rw_lock DLM lock - * to protect io on one node from truncation on another. - */ -static int ocfs2_dio_end_io(struct kiocb *iocb, - loff_t offset, - ssize_t bytes, - void *private) -{ - struct inode *inode = file_inode(iocb->ki_filp); - int level; - - if (bytes <= 0) - return 0; - - /* this io's submitter should not have unlocked this before we could */ - BUG_ON(!ocfs2_iocb_is_rw_locked(iocb)); - - if (ocfs2_iocb_is_unaligned_aio(iocb)) { - ocfs2_iocb_clear_unaligned_aio(iocb); - - mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); - } - - /* Let rw unlock to be done later to protect append direct io write */ - if (offset + bytes <= i_size_read(inode)) { - ocfs2_iocb_clear_rw_locked(iocb); - - level = ocfs2_iocb_rw_locked_level(iocb); - ocfs2_rw_unlock(inode, level); - } - - return 0; -} - static int ocfs2_releasepage(struct page *page, gfp_t wait) { if (!page_has_buffers(page)) @@ -658,363 +506,6 @@ static int ocfs2_releasepage(struct page *page, gfp_t wait) return try_to_free_buffers(page); } -static int ocfs2_is_overwrite(struct ocfs2_super *osb, - struct inode *inode, loff_t offset) -{ - int ret = 0; - u32 v_cpos = 0; - u32 p_cpos = 0; - unsigned int num_clusters = 0; - unsigned int ext_flags = 0; - - v_cpos = ocfs2_bytes_to_clusters(osb->sb, offset); - ret = ocfs2_get_clusters(inode, v_cpos, &p_cpos, - &num_clusters, &ext_flags); - if (ret < 0) { - mlog_errno(ret); - return ret; - } - - if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) - return 1; - - return 0; -} - -static int ocfs2_direct_IO_zero_extend(struct ocfs2_super *osb, - struct inode *inode, loff_t offset, - u64 zero_len, int cluster_align) -{ - u32 p_cpos = 0; - u32 v_cpos = ocfs2_bytes_to_clusters(osb->sb, i_size_read(inode)); - unsigned int num_clusters = 0; - unsigned int ext_flags = 0; - int ret = 0; - - if (offset <= i_size_read(inode) || cluster_align) - return 0; - - ret = ocfs2_get_clusters(inode, v_cpos, &p_cpos, &num_clusters, - &ext_flags); - if (ret < 0) { - mlog_errno(ret); - return ret; - } - - if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) { - u64 s = i_size_read(inode); - sector_t sector = ((u64)p_cpos << (osb->s_clustersize_bits - 9)) + - (do_div(s, osb->s_clustersize) >> 9); - - ret = blkdev_issue_zeroout(osb->sb->s_bdev, sector, - zero_len >> 9, GFP_NOFS, false); - if (ret < 0) - mlog_errno(ret); - } - - return ret; -} - -static int ocfs2_direct_IO_extend_no_holes(struct ocfs2_super *osb, - struct inode *inode, loff_t offset) -{ - u64 zero_start, zero_len, total_zero_len; - u32 p_cpos = 0, clusters_to_add; - u32 v_cpos = ocfs2_bytes_to_clusters(osb->sb, i_size_read(inode)); - unsigned int num_clusters = 0; - unsigned int ext_flags = 0; - u32 size_div, offset_div; - int ret = 0; - - { - u64 o = offset; - u64 s = i_size_read(inode); - - offset_div = do_div(o, osb->s_clustersize); - size_div = do_div(s, osb->s_clustersize); - } - - if (offset <= i_size_read(inode)) - return 0; - - clusters_to_add = ocfs2_bytes_to_clusters(inode->i_sb, offset) - - ocfs2_bytes_to_clusters(inode->i_sb, i_size_read(inode)); - total_zero_len = offset - i_size_read(inode); - if (clusters_to_add) - total_zero_len -= offset_div; - - /* Allocate clusters to fill out holes, and this is only needed - * when we add more than one clusters. Otherwise the cluster will - * be allocated during direct IO */ - if (clusters_to_add > 1) { - ret = ocfs2_extend_allocation(inode, - OCFS2_I(inode)->ip_clusters, - clusters_to_add - 1, 0); - if (ret) { - mlog_errno(ret); - goto out; - } - } - - while (total_zero_len) { - ret = ocfs2_get_clusters(inode, v_cpos, &p_cpos, &num_clusters, - &ext_flags); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - - zero_start = ocfs2_clusters_to_bytes(osb->sb, p_cpos) + - size_div; - zero_len = ocfs2_clusters_to_bytes(osb->sb, num_clusters) - - size_div; - zero_len = min(total_zero_len, zero_len); - - if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) { - ret = blkdev_issue_zeroout(osb->sb->s_bdev, - zero_start >> 9, zero_len >> 9, - GFP_NOFS, false); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - } - - total_zero_len -= zero_len; - v_cpos += ocfs2_bytes_to_clusters(osb->sb, zero_len + size_div); - - /* Only at first iteration can be cluster not aligned. - * So set size_div to 0 for the rest */ - size_div = 0; - } - -out: - return ret; -} - -static ssize_t ocfs2_direct_IO_write(struct kiocb *iocb, - struct iov_iter *iter, - loff_t offset) -{ - ssize_t ret = 0; - ssize_t written = 0; - bool orphaned = false; - int is_overwrite = 0; - struct file *file = iocb->ki_filp; - struct inode *inode = file_inode(file)->i_mapping->host; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - struct buffer_head *di_bh = NULL; - size_t count = iter->count; - journal_t *journal = osb->journal->j_journal; - u64 zero_len_head, zero_len_tail; - int cluster_align_head, cluster_align_tail; - loff_t final_size = offset + count; - int append_write = offset >= i_size_read(inode) ? 1 : 0; - unsigned int num_clusters = 0; - unsigned int ext_flags = 0; - - { - u64 o = offset; - u64 s = i_size_read(inode); - - zero_len_head = do_div(o, 1 << osb->s_clustersize_bits); - cluster_align_head = !zero_len_head; - - zero_len_tail = osb->s_clustersize - - do_div(s, osb->s_clustersize); - if ((offset - i_size_read(inode)) < zero_len_tail) - zero_len_tail = offset - i_size_read(inode); - cluster_align_tail = !zero_len_tail; - } - - /* - * when final_size > inode->i_size, inode->i_size will be - * updated after direct write, so add the inode to orphan - * dir first. - */ - if (final_size > i_size_read(inode)) { - ret = ocfs2_add_inode_to_orphan(osb, inode); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - orphaned = true; - } - - if (append_write) { - ret = ocfs2_inode_lock(inode, NULL, 1); - if (ret < 0) { - mlog_errno(ret); - goto clean_orphan; - } - - /* zeroing out the previously allocated cluster tail - * that but not zeroed */ - if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) { - down_read(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_direct_IO_zero_extend(osb, inode, offset, - zero_len_tail, cluster_align_tail); - up_read(&OCFS2_I(inode)->ip_alloc_sem); - } else { - down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_direct_IO_extend_no_holes(osb, inode, - offset); - up_write(&OCFS2_I(inode)->ip_alloc_sem); - } - if (ret < 0) { - mlog_errno(ret); - ocfs2_inode_unlock(inode, 1); - goto clean_orphan; - } - - is_overwrite = ocfs2_is_overwrite(osb, inode, offset); - if (is_overwrite < 0) { - mlog_errno(is_overwrite); - ret = is_overwrite; - ocfs2_inode_unlock(inode, 1); - goto clean_orphan; - } - - ocfs2_inode_unlock(inode, 1); - } - - written = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter, - offset, ocfs2_direct_IO_get_blocks, - ocfs2_dio_end_io, NULL, 0); - /* overwrite aio may return -EIOCBQUEUED, and it is not an error */ - if ((written < 0) && (written != -EIOCBQUEUED)) { - loff_t i_size = i_size_read(inode); - - if (offset + count > i_size) { - ret = ocfs2_inode_lock(inode, &di_bh, 1); - if (ret < 0) { - mlog_errno(ret); - goto clean_orphan; - } - - if (i_size == i_size_read(inode)) { - ret = ocfs2_truncate_file(inode, di_bh, - i_size); - if (ret < 0) { - if (ret != -ENOSPC) - mlog_errno(ret); - - ocfs2_inode_unlock(inode, 1); - brelse(di_bh); - di_bh = NULL; - goto clean_orphan; - } - } - - ocfs2_inode_unlock(inode, 1); - brelse(di_bh); - di_bh = NULL; - - ret = jbd2_journal_force_commit(journal); - if (ret < 0) - mlog_errno(ret); - } - } else if (written > 0 && append_write && !is_overwrite && - !cluster_align_head) { - /* zeroing out the allocated cluster head */ - u32 p_cpos = 0; - u32 v_cpos = ocfs2_bytes_to_clusters(osb->sb, offset); - - ret = ocfs2_inode_lock(inode, NULL, 0); - if (ret < 0) { - mlog_errno(ret); - goto clean_orphan; - } - - ret = ocfs2_get_clusters(inode, v_cpos, &p_cpos, - &num_clusters, &ext_flags); - if (ret < 0) { - mlog_errno(ret); - ocfs2_inode_unlock(inode, 0); - goto clean_orphan; - } - - BUG_ON(!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN)); - - ret = blkdev_issue_zeroout(osb->sb->s_bdev, - (u64)p_cpos << (osb->s_clustersize_bits - 9), - zero_len_head >> 9, GFP_NOFS, false); - if (ret < 0) - mlog_errno(ret); - - ocfs2_inode_unlock(inode, 0); - } - -clean_orphan: - if (orphaned) { - int tmp_ret; - int update_isize = written > 0 ? 1 : 0; - loff_t end = update_isize ? offset + written : 0; - - tmp_ret = ocfs2_inode_lock(inode, &di_bh, 1); - if (tmp_ret < 0) { - ret = tmp_ret; - mlog_errno(ret); - goto out; - } - - tmp_ret = ocfs2_del_inode_from_orphan(osb, inode, di_bh, - update_isize, end); - if (tmp_ret < 0) { - ocfs2_inode_unlock(inode, 1); - ret = tmp_ret; - mlog_errno(ret); - brelse(di_bh); - goto out; - } - - ocfs2_inode_unlock(inode, 1); - brelse(di_bh); - - tmp_ret = jbd2_journal_force_commit(journal); - if (tmp_ret < 0) { - ret = tmp_ret; - mlog_errno(tmp_ret); - } - } - -out: - if (ret >= 0) - ret = written; - return ret; -} - -static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter, - loff_t offset) -{ - struct file *file = iocb->ki_filp; - struct inode *inode = file_inode(file)->i_mapping->host; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - int full_coherency = !(osb->s_mount_opt & - OCFS2_MOUNT_COHERENCY_BUFFERED); - - /* - * Fallback to buffered I/O if we see an inode without - * extents. - */ - if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) - return 0; - - /* Fallback to buffered I/O if we are appending and - * concurrent O_DIRECT writes are allowed. - */ - if (i_size_read(inode) <= offset && !full_coherency) - return 0; - - if (iov_iter_rw(iter) == READ) - return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, - iter, offset, - ocfs2_direct_IO_get_blocks, - ocfs2_dio_end_io, NULL, 0); - else - return ocfs2_direct_IO_write(iocb, iter, offset); -} - static void ocfs2_figure_cluster_boundaries(struct ocfs2_super *osb, u32 cpos, unsigned int *start, @@ -1326,14 +817,14 @@ static void ocfs2_free_unwritten_list(struct inode *inode, struct list_head *head) { struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_unwritten_extent *dz = NULL, *tmp = NULL; + struct ocfs2_unwritten_extent *ue = NULL, *tmp = NULL; - list_for_each_entry_safe(dz, tmp, head, ue_node) { - list_del(&dz->ue_node); + list_for_each_entry_safe(ue, tmp, head, ue_node) { + list_del(&ue->ue_node); spin_lock(&oi->ip_lock); - list_del(&dz->ue_ip_node); + list_del(&ue->ue_ip_node); spin_unlock(&oi->ip_lock); - kfree(dz); + kfree(ue); } } @@ -1834,7 +1325,7 @@ static int ocfs2_unwritten_check(struct inode *inode, struct ocfs2_write_cluster_desc *desc) { struct ocfs2_inode_info *oi = OCFS2_I(inode); - struct ocfs2_unwritten_extent *dz = NULL, *new = NULL; + struct ocfs2_unwritten_extent *ue = NULL, *new = NULL; int ret = 0; if (!desc->c_needs_zero) @@ -1845,8 +1336,8 @@ retry: /* Needs not to zero no metter buffer or direct. The one who is zero * the cluster is doing zero. And he will clear unwritten after all * cluster io finished. */ - list_for_each_entry(dz, &oi->ip_unwritten_list, ue_ip_node) { - if (desc->c_cpos == dz->ue_cpos) { + list_for_each_entry(ue, &oi->ip_unwritten_list, ue_ip_node) { + if (desc->c_cpos == ue->ue_cpos) { BUG_ON(desc->c_new); desc->c_needs_zero = 0; desc->c_clear_unwritten = 0; @@ -2608,6 +2099,344 @@ static int ocfs2_write_end(struct file *file, struct address_space *mapping, return ret; } +struct ocfs2_dio_write_ctxt { + struct list_head dw_zero_list; + unsigned dw_zero_count; + int dw_orphaned; + pid_t dw_writer_pid; +}; + +static struct ocfs2_dio_write_ctxt * +ocfs2_dio_alloc_write_ctx(struct buffer_head *bh, int *alloc) +{ + struct ocfs2_dio_write_ctxt *dwc = NULL; + + if (bh->b_private) + return bh->b_private; + + dwc = kmalloc(sizeof(struct ocfs2_dio_write_ctxt), GFP_NOFS); + if (dwc == NULL) + return NULL; + INIT_LIST_HEAD(&dwc->dw_zero_list); + dwc->dw_zero_count = 0; + dwc->dw_orphaned = 0; + dwc->dw_writer_pid = task_pid_nr(current); + bh->b_private = dwc; + *alloc = 1; + + return dwc; +} + +static void ocfs2_dio_free_write_ctx(struct inode *inode, + struct ocfs2_dio_write_ctxt *dwc) +{ + ocfs2_free_unwritten_list(inode, &dwc->dw_zero_list); + kfree(dwc); +} + +/* + * TODO: Make this into a generic get_blocks function. + * + * From do_direct_io in direct-io.c: + * "So what we do is to permit the ->get_blocks function to populate + * bh.b_size with the size of IO which is permitted at this offset and + * this i_blkbits." + * + * This function is called directly from get_more_blocks in direct-io.c. + * + * called like this: dio->get_blocks(dio->inode, fs_startblk, + * fs_count, map_bh, dio->rw == WRITE); + */ +static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, + struct buffer_head *bh_result, int create) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_write_ctxt *wc; + struct ocfs2_write_cluster_desc *desc = NULL; + struct ocfs2_dio_write_ctxt *dwc = NULL; + struct buffer_head *di_bh = NULL; + u64 p_blkno; + loff_t pos = iblock << inode->i_sb->s_blocksize_bits; + unsigned len, total_len = bh_result->b_size; + int ret = 0, first_get_block = 0; + + len = osb->s_clustersize - (pos & (osb->s_clustersize - 1)); + len = min(total_len, len); + + mlog(0, "get block of %lu at %llu:%u req %u\n", + inode->i_ino, pos, len, total_len); + + /* This is the fast path for re-write. */ + ret = ocfs2_get_block(inode, iblock, bh_result, create); + + if (buffer_mapped(bh_result) && + !buffer_new(bh_result) && + ret == 0) + goto out; + + /* Clear state set by ocfs2_get_block. */ + bh_result->b_state = 0; + + dwc = ocfs2_dio_alloc_write_ctx(bh_result, &first_get_block); + if (unlikely(dwc == NULL)) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + if (ocfs2_clusters_for_bytes(inode->i_sb, pos + total_len) > + ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)) && + !dwc->dw_orphaned) { + /* + * when we are going to alloc extents beyond file size, add the + * inode to orphan dir, so we can recall those spaces when + * system crashed during write. + */ + ret = ocfs2_add_inode_to_orphan(osb, inode); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + dwc->dw_orphaned = 1; + } + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret) { + mlog_errno(ret); + goto out; + } + + if (first_get_block) { + if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + ret = ocfs2_zero_tail(inode, di_bh, pos); + else + ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, + total_len, NULL); + if (ret < 0) { + mlog_errno(ret); + goto unlock; + } + } + + ret = ocfs2_write_begin_nolock(inode->i_mapping, pos, len, + OCFS2_WRITE_DIRECT, NULL, + (void **)&wc, di_bh, NULL); + if (ret) { + mlog_errno(ret); + goto unlock; + } + + desc = &wc->w_desc[0]; + + p_blkno = ocfs2_clusters_to_blocks(inode->i_sb, desc->c_phys); + BUG_ON(p_blkno == 0); + p_blkno += iblock & (u64)(ocfs2_clusters_to_blocks(inode->i_sb, 1) - 1); + + map_bh(bh_result, inode->i_sb, p_blkno); + bh_result->b_size = len; + if (desc->c_needs_zero) + set_buffer_new(bh_result); + + /* May sleep in end_io. It should not happen in a irq context. So defer + * it to dio work queue. */ + set_buffer_defer_completion(bh_result); + + if (!list_empty(&wc->w_unwritten_list)) { + struct ocfs2_unwritten_extent *ue = NULL; + + ue = list_first_entry(&wc->w_unwritten_list, + struct ocfs2_unwritten_extent, + ue_node); + BUG_ON(ue->ue_cpos != desc->c_cpos); + /* The physical address may be 0, fill it. */ + ue->ue_phys = desc->c_phys; + + list_splice_tail_init(&wc->w_unwritten_list, &dwc->dw_zero_list); + dwc->dw_zero_count++; + } + + ret = ocfs2_write_end_nolock(inode->i_mapping, pos, len, len, NULL, wc); + BUG_ON(ret != len); + ret = 0; +unlock: + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); +out: + if (ret < 0) + ret = -EIO; + return ret; +} + +static void ocfs2_dio_end_io_write(struct inode *inode, + struct ocfs2_dio_write_ctxt *dwc, + loff_t offset, + ssize_t bytes) +{ + struct ocfs2_cached_dealloc_ctxt dealloc; + struct ocfs2_extent_tree et; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_unwritten_extent *ue = NULL; + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + struct ocfs2_alloc_context *data_ac = NULL; + struct ocfs2_alloc_context *meta_ac = NULL; + handle_t *handle = NULL; + loff_t end = offset + bytes; + int ret = 0, credits = 0, locked = 0; + + ocfs2_init_dealloc_ctxt(&dealloc); + + /* We do clear unwritten, delete orphan, change i_size here. If neither + * of these happen, we can skip all this. */ + if (list_empty(&dwc->dw_zero_list) && + end <= i_size_read(inode) && + !dwc->dw_orphaned) + goto out; + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + /* ocfs2_file_write_iter will get i_mutex, so we need not lock if we + * are in that context. */ + if (dwc->dw_writer_pid != task_pid_nr(current)) { + mutex_lock(&inode->i_mutex); + locked = 1; + } + + /* Delete orphan before acquire i_mutex. */ + if (dwc->dw_orphaned) { + BUG_ON(dwc->dw_writer_pid != task_pid_nr(current)); + + end = end > i_size_read(inode) ? end : 0; + + ret = ocfs2_del_inode_from_orphan(osb, inode, di_bh, + !!end, end); + if (ret < 0) + mlog_errno(ret); + } + + di = (struct ocfs2_dinode *)di_bh; + + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh); + + ret = ocfs2_lock_allocators(inode, &et, 0, dwc->dw_zero_count*2, + &data_ac, &meta_ac); + + credits = ocfs2_calc_extend_credits(inode->i_sb, &di->id2.i_list); + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto unlock; + } + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto commit; + } + + list_for_each_entry(ue, &dwc->dw_zero_list, ue_node) { + ret = ocfs2_mark_extent_written(inode, &et, handle, + ue->ue_cpos, 1, + ue->ue_phys, + meta_ac, &dealloc); + if (ret < 0) { + mlog_errno(ret); + break; + } + } + + if (end > i_size_read(inode)) { + ret = ocfs2_set_inode_size(handle, inode, di_bh, end); + if (ret < 0) + mlog_errno(ret); + } +commit: + ocfs2_commit_trans(osb, handle); +unlock: + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); +out: + ocfs2_run_deallocs(osb, &dealloc); + if (locked) + mutex_unlock(&inode->i_mutex); + ocfs2_dio_free_write_ctx(inode, dwc); + if (data_ac) + ocfs2_free_alloc_context(data_ac); + if (meta_ac) + ocfs2_free_alloc_context(meta_ac); +} + +/* + * ocfs2_dio_end_io is called by the dio core when a dio is finished. We're + * particularly interested in the aio/dio case. We use the rw_lock DLM lock + * to protect io on one node from truncation on another. + */ +static int ocfs2_dio_end_io(struct kiocb *iocb, + loff_t offset, + ssize_t bytes, + void *private) +{ + struct inode *inode = file_inode(iocb->ki_filp); + int level; + + if (bytes <= 0) + return 0; + + /* this io's submitter should not have unlocked this before we could */ + BUG_ON(!ocfs2_iocb_is_rw_locked(iocb)); + + if (ocfs2_iocb_is_unaligned_aio(iocb)) { + ocfs2_iocb_clear_unaligned_aio(iocb); + + mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); + } + + if (private) + ocfs2_dio_end_io_write(inode, private, offset, bytes); + + ocfs2_iocb_clear_rw_locked(iocb); + + level = ocfs2_iocb_rw_locked_level(iocb); + ocfs2_rw_unlock(inode, level); + return 0; +} + +static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter, + loff_t offset) +{ + struct file *file = iocb->ki_filp; + struct inode *inode = file_inode(file)->i_mapping->host; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + loff_t end = offset + iter->count; + get_block_t *get_block; + + /* + * Fallback to buffered I/O if we see an inode without + * extents. + */ + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return 0; + + /* Fallback to buffered I/O if we do not support append dio. */ + if (end > i_size_read(inode) && !ocfs2_supports_append_dio(osb)) + return 0; + + if (iov_iter_rw(iter) == READ) + get_block = ocfs2_get_block; + else + get_block = ocfs2_dio_get_block; + + return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, + iter, offset, get_block, + ocfs2_dio_end_io, NULL, 0); +} + const struct address_space_operations ocfs2_aops = { .readpage = ocfs2_readpage, .readpages = ocfs2_readpages, -- cgit v1.2.3 From f1f973ffce96a47c2b3f142e91eccef5bf22f699 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:12 -0700 Subject: ocfs2: code clean up for direct io Clean up ocfs2_file_write_iter & ocfs2_prepare_inode_for_write: * remove append dio check: it will be checked in ocfs2_direct_IO() * remove file hole check: file hole is supported for now * remove inline data check: it will be checked in ocfs2_direct_IO() * remove the full_coherence check when append dio: we will get the inode_lock in ocfs2_dio_get_block, there is no need to fall back to buffer io to ensure the coherence semantics. Now the drop dio procedure is gone. :) [akpm@linux-foundation.org: remove unused label] Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/file.c | 138 +++---------------------------------------------- fs/ocfs2/ocfs2_trace.h | 16 ++---- 2 files changed, 10 insertions(+), 144 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 7cb38fdca229..1ab182321b18 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -1381,44 +1381,6 @@ out: return ret; } -/* - * Will look for holes and unwritten extents in the range starting at - * pos for count bytes (inclusive). - */ -static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos, - size_t count) -{ - int ret = 0; - unsigned int extent_flags; - u32 cpos, clusters, extent_len, phys_cpos; - struct super_block *sb = inode->i_sb; - - cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits; - clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos; - - while (clusters) { - ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len, - &extent_flags); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - - if (phys_cpos == 0 || (extent_flags & OCFS2_EXT_UNWRITTEN)) { - ret = 1; - break; - } - - if (extent_len > clusters) - extent_len = clusters; - - clusters -= extent_len; - cpos += extent_len; - } -out: - return ret; -} - static int ocfs2_write_remove_suid(struct inode *inode) { int ret; @@ -2129,18 +2091,12 @@ out: static int ocfs2_prepare_inode_for_write(struct file *file, loff_t pos, - size_t count, - int appending, - int *direct_io, - int *has_refcount) + size_t count) { int ret = 0, meta_level = 0; struct dentry *dentry = file->f_path.dentry; struct inode *inode = d_inode(dentry); loff_t end; - struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); - int full_coherency = !(osb->s_mount_opt & - OCFS2_MOUNT_COHERENCY_BUFFERED); /* * We start with a read level meta lock and only jump to an ex @@ -2189,10 +2145,6 @@ static int ocfs2_prepare_inode_for_write(struct file *file, pos, count, &meta_level); - if (has_refcount) - *has_refcount = 1; - if (direct_io) - *direct_io = 0; } if (ret < 0) { @@ -2200,67 +2152,12 @@ static int ocfs2_prepare_inode_for_write(struct file *file, goto out_unlock; } - /* - * Skip the O_DIRECT checks if we don't need - * them. - */ - if (!direct_io || !(*direct_io)) - break; - - /* - * There's no sane way to do direct writes to an inode - * with inline data. - */ - if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { - *direct_io = 0; - break; - } - - /* - * Allowing concurrent direct writes means - * i_size changes wouldn't be synchronized, so - * one node could wind up truncating another - * nodes writes. - */ - if (end > i_size_read(inode) && !full_coherency) { - *direct_io = 0; - break; - } - - /* - * Fallback to old way if the feature bit is not set. - */ - if (end > i_size_read(inode) && - !ocfs2_supports_append_dio(osb)) { - *direct_io = 0; - break; - } - - /* - * We don't fill holes during direct io, so - * check for them here. If any are found, the - * caller will have to retake some cluster - * locks and initiate the io as buffered. - */ - ret = ocfs2_check_range_for_holes(inode, pos, count); - if (ret == 1) { - /* - * Fallback to old way if the feature bit is not set. - * Otherwise try dio first and then complete the rest - * request through buffer io. - */ - if (!ocfs2_supports_append_dio(osb)) - *direct_io = 0; - ret = 0; - } else if (ret < 0) - mlog_errno(ret); break; } out_unlock: trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno, - pos, appending, count, - direct_io, has_refcount); + pos, count); if (meta_level >= 0) ocfs2_inode_unlock(inode, meta_level); @@ -2272,18 +2169,16 @@ out: static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, struct iov_iter *from) { - int direct_io, appending, rw_level; - int can_do_direct, has_refcount = 0; + int direct_io, rw_level; ssize_t written = 0; ssize_t ret; - size_t count = iov_iter_count(from), orig_count; + size_t count = iov_iter_count(from); struct file *file = iocb->ki_filp; struct inode *inode = file_inode(file); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); int full_coherency = !(osb->s_mount_opt & OCFS2_MOUNT_COHERENCY_BUFFERED); int unaligned_dio = 0; - int dropped_dio = 0; int append_write = ((iocb->ki_pos + count) >= i_size_read(inode) ? 1 : 0); @@ -2296,12 +2191,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, if (count == 0) return 0; - appending = iocb->ki_flags & IOCB_APPEND ? 1 : 0; direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0; inode_lock(inode); -relock: /* * Concurrent O_DIRECT writes are allowed with * mount_option "coherency=buffered". @@ -2334,7 +2227,6 @@ relock: ocfs2_inode_unlock(inode, 1); } - orig_count = iov_iter_count(from); ret = generic_write_checks(iocb, from); if (ret <= 0) { if (ret) @@ -2343,9 +2235,7 @@ relock: } count = ret; - can_do_direct = direct_io; - ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count, appending, - &can_do_direct, &has_refcount); + ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count); if (ret < 0) { mlog_errno(ret); goto out; @@ -2354,22 +2244,6 @@ relock: if (direct_io && !is_sync_kiocb(iocb)) unaligned_dio = ocfs2_is_io_unaligned(inode, count, iocb->ki_pos); - /* - * We can't complete the direct I/O as requested, fall back to - * buffered I/O. - */ - if (direct_io && !can_do_direct) { - ocfs2_rw_unlock(inode, rw_level); - - rw_level = -1; - - direct_io = 0; - iocb->ki_flags &= ~IOCB_DIRECT; - iov_iter_reexpand(from, orig_count); - dropped_dio = 1; - goto relock; - } - if (unaligned_dio) { /* * Wait on previous unaligned aio to complete before @@ -2405,7 +2279,7 @@ relock: goto no_sync; if (((file->f_flags & O_DSYNC) && !direct_io) || - IS_SYNC(inode) || dropped_dio) { + IS_SYNC(inode)) { ret = filemap_fdatawrite_range(file->f_mapping, iocb->ki_pos - written, iocb->ki_pos - 1); diff --git a/fs/ocfs2/ocfs2_trace.h b/fs/ocfs2/ocfs2_trace.h index 24b7e7f591dc..f8f5fc5e6c05 100644 --- a/fs/ocfs2/ocfs2_trace.h +++ b/fs/ocfs2/ocfs2_trace.h @@ -1450,28 +1450,20 @@ DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_remove_inode_range); TRACE_EVENT(ocfs2_prepare_inode_for_write, TP_PROTO(unsigned long long ino, unsigned long long saved_pos, - int appending, unsigned long count, - int *direct_io, int *has_refcount), - TP_ARGS(ino, saved_pos, appending, count, direct_io, has_refcount), + unsigned long count), + TP_ARGS(ino, saved_pos, count), TP_STRUCT__entry( __field(unsigned long long, ino) __field(unsigned long long, saved_pos) - __field(int, appending) __field(unsigned long, count) - __field(int, direct_io) - __field(int, has_refcount) ), TP_fast_assign( __entry->ino = ino; __entry->saved_pos = saved_pos; - __entry->appending = appending; __entry->count = count; - __entry->direct_io = direct_io ? *direct_io : -1; - __entry->has_refcount = has_refcount ? *has_refcount : -1; ), - TP_printk("%llu %llu %d %lu %d %d", __entry->ino, - __entry->saved_pos, __entry->appending, __entry->count, - __entry->direct_io, __entry->has_refcount) + TP_printk("%llu %llu %lu", __entry->ino, + __entry->saved_pos, __entry->count) ); DEFINE_OCFS2_INT_EVENT(generic_file_aio_read_ret); -- cgit v1.2.3 From e63890f38ade9497b5609ddeb7f52df0fe55ea15 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:15 -0700 Subject: ocfs2: fix ip_unaligned_aio deadlock with dio work queue In the current implementation of unaligned aio+dio, lock order behave as follow: in user process context: -> call io_submit() -> get i_mutex <== window1 -> get ip_unaligned_aio -> submit direct io to block device -> release i_mutex -> io_submit() return in dio work queue context(the work queue is created in __blockdev_direct_IO): -> release ip_unaligned_aio <== window2 -> get i_mutex -> clear unwritten flag & change i_size -> release i_mutex There is a limitation to the thread number of dio work queue. 256 at default. If all 256 thread are in the above 'window2' stage, and there is a user process in the 'window1' stage, the system will became deadlock. Since the user process hold i_mutex to wait ip_unaligned_aio lock, while there is a direct bio hold ip_unaligned_aio mutex who is waiting for a dio work queue thread to be schedule. But all the dio work queue thread is waiting for i_mutex lock in 'window2'. This case only happened in a test which send a large number(more than 256) of aio at one io_submit() call. My design is to remove ip_unaligned_aio lock. Change it to a sync io instead. Just like ip_unaligned_aio lock, serialize the unaligned aio dio. [akpm@linux-foundation.org: remove OCFS2_IOCB_UNALIGNED_IO, per Junxiao Bi] Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 6 ------ fs/ocfs2/aops.h | 8 -------- fs/ocfs2/file.c | 27 +++++++++------------------ fs/ocfs2/inode.h | 3 --- fs/ocfs2/super.c | 1 - 5 files changed, 9 insertions(+), 36 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 578ce225cf83..843475793504 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2391,12 +2391,6 @@ static int ocfs2_dio_end_io(struct kiocb *iocb, /* this io's submitter should not have unlocked this before we could */ BUG_ON(!ocfs2_iocb_is_rw_locked(iocb)); - if (ocfs2_iocb_is_unaligned_aio(iocb)) { - ocfs2_iocb_clear_unaligned_aio(iocb); - - mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); - } - if (private) ocfs2_dio_end_io_write(inode, private, offset, bytes); diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index d06b80f58f83..b1c9f28a57b1 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -84,7 +84,6 @@ static inline void ocfs2_iocb_set_rw_locked(struct kiocb *iocb, int level) enum ocfs2_iocb_lock_bits { OCFS2_IOCB_RW_LOCK = 0, OCFS2_IOCB_RW_LOCK_LEVEL, - OCFS2_IOCB_UNALIGNED_IO, OCFS2_IOCB_NUM_LOCKS }; @@ -93,11 +92,4 @@ enum ocfs2_iocb_lock_bits { #define ocfs2_iocb_rw_locked_level(iocb) \ test_bit(OCFS2_IOCB_RW_LOCK_LEVEL, (unsigned long *)&iocb->private) -#define ocfs2_iocb_set_unaligned_aio(iocb) \ - set_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) -#define ocfs2_iocb_clear_unaligned_aio(iocb) \ - clear_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) -#define ocfs2_iocb_is_unaligned_aio(iocb) \ - test_bit(OCFS2_IOCB_UNALIGNED_IO, (unsigned long *)&iocb->private) - #endif /* OCFS2_FILE_H */ diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 1ab182321b18..c18ab45f8d21 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -2178,7 +2178,7 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); int full_coherency = !(osb->s_mount_opt & OCFS2_MOUNT_COHERENCY_BUFFERED); - int unaligned_dio = 0; + void *saved_ki_complete = NULL; int append_write = ((iocb->ki_pos + count) >= i_size_read(inode) ? 1 : 0); @@ -2241,17 +2241,12 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, goto out; } - if (direct_io && !is_sync_kiocb(iocb)) - unaligned_dio = ocfs2_is_io_unaligned(inode, count, iocb->ki_pos); - - if (unaligned_dio) { + if (direct_io && !is_sync_kiocb(iocb) && + ocfs2_is_io_unaligned(inode, count, iocb->ki_pos)) { /* - * Wait on previous unaligned aio to complete before - * proceeding. + * Make it a sync io if it's an unaligned aio. */ - mutex_lock(&OCFS2_I(inode)->ip_unaligned_aio); - /* Mark the iocb as needing an unlock in ocfs2_dio_end_io */ - ocfs2_iocb_set_unaligned_aio(iocb); + saved_ki_complete = xchg(&iocb->ki_complete, NULL); } /* communicate with ocfs2_dio_end_io */ @@ -2272,11 +2267,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, */ if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) { rw_level = -1; - unaligned_dio = 0; } if (unlikely(written <= 0)) - goto no_sync; + goto out; if (((file->f_flags & O_DSYNC) && !direct_io) || IS_SYNC(inode)) { @@ -2298,13 +2292,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, iocb->ki_pos - 1); } -no_sync: - if (unaligned_dio && ocfs2_iocb_is_unaligned_aio(iocb)) { - ocfs2_iocb_clear_unaligned_aio(iocb); - mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio); - } - out: + if (saved_ki_complete) + xchg(&iocb->ki_complete, saved_ki_complete); + if (rw_level != -1) ocfs2_rw_unlock(inode, rw_level); diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index 68e8cf9bda97..d8f3fc8d2551 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -43,9 +43,6 @@ struct ocfs2_inode_info /* protects extended attribute changes on this inode */ struct rw_semaphore ip_xattr_sem; - /* Number of outstanding AIO's which are not page aligned */ - struct mutex ip_unaligned_aio; - /* These fields are protected by ip_lock */ spinlock_t ip_lock; u32 ip_open_count; diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index d142fb75740e..e2c4e3839d7b 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -1747,7 +1747,6 @@ static void ocfs2_inode_init_once(void *data) INIT_LIST_HEAD(&oi->ip_io_markers); INIT_LIST_HEAD(&oi->ip_unwritten_list); oi->ip_dir_start_lookup = 0; - mutex_init(&oi->ip_unaligned_aio); init_rwsem(&oi->ip_alloc_sem); init_rwsem(&oi->ip_xattr_sem); mutex_init(&oi->ip_io_mutex); -- cgit v1.2.3 From a86a72a4a4e0ec109a98e2737948864ed6794bf7 Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:18 -0700 Subject: ocfs2: take ip_alloc_sem in ocfs2_dio_get_block & ocfs2_dio_end_io_write Take ip_alloc_sem to prevent concurrent access to extent tree, which may cause the extent tree in an unstable state. Signed-off-by: Ryan Ding Reviewed-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 843475793504..0f3816325808 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2151,6 +2151,7 @@ static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) { struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_write_ctxt *wc; struct ocfs2_write_cluster_desc *desc = NULL; struct ocfs2_dio_write_ctxt *dwc = NULL; @@ -2166,9 +2167,12 @@ static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, mlog(0, "get block of %lu at %llu:%u req %u\n", inode->i_ino, pos, len, total_len); + down_read(&oi->ip_alloc_sem); /* This is the fast path for re-write. */ ret = ocfs2_get_block(inode, iblock, bh_result, create); + up_read(&oi->ip_alloc_sem); + if (buffer_mapped(bh_result) && !buffer_new(bh_result) && ret == 0) @@ -2206,6 +2210,8 @@ static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, goto out; } + down_write(&oi->ip_alloc_sem); + if (first_get_block) { if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) ret = ocfs2_zero_tail(inode, di_bh, pos); @@ -2259,6 +2265,7 @@ static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, BUG_ON(ret != len); ret = 0; unlock: + up_write(&oi->ip_alloc_sem); ocfs2_inode_unlock(inode, 1); brelse(di_bh); out: @@ -2275,6 +2282,7 @@ static void ocfs2_dio_end_io_write(struct inode *inode, struct ocfs2_cached_dealloc_ctxt dealloc; struct ocfs2_extent_tree et; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_unwritten_extent *ue = NULL; struct buffer_head *di_bh = NULL; struct ocfs2_dinode *di; @@ -2293,12 +2301,6 @@ static void ocfs2_dio_end_io_write(struct inode *inode, !dwc->dw_orphaned) goto out; - ret = ocfs2_inode_lock(inode, &di_bh, 1); - if (ret < 0) { - mlog_errno(ret); - goto out; - } - /* ocfs2_file_write_iter will get i_mutex, so we need not lock if we * are in that context. */ if (dwc->dw_writer_pid != task_pid_nr(current)) { @@ -2306,6 +2308,14 @@ static void ocfs2_dio_end_io_write(struct inode *inode, locked = 1; } + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + down_write(&oi->ip_alloc_sem); + /* Delete orphan before acquire i_mutex. */ if (dwc->dw_orphaned) { BUG_ON(dwc->dw_writer_pid != task_pid_nr(current)); @@ -2359,6 +2369,7 @@ static void ocfs2_dio_end_io_write(struct inode *inode, commit: ocfs2_commit_trans(osb, handle); unlock: + up_write(&oi->ip_alloc_sem); ocfs2_inode_unlock(inode, 1); brelse(di_bh); out: -- cgit v1.2.3 From ce170828e24959c69e7a40364731edc0535c550f Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:20 -0700 Subject: ocfs2: fix disk file size and memory file size mismatch When doing append direct write in an already allocated cluster, and fast path in ocfs2_dio_get_block() is triggered, function ocfs2_dio_end_io_write() will be skipped as there is no context allocated. As a result, the disk file size will not be changed as it should be. The solution is to skip fast path when we are about to change file size. Fixes: af1310367f41 ("ocfs2: fix sparse file & data ordering issue in direct io.") Signed-off-by: Ryan Ding Acked-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 27 +++++++++++++++++---------- 1 file changed, 17 insertions(+), 10 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 0f3816325808..328ac7f99d52 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2167,19 +2167,26 @@ static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock, mlog(0, "get block of %lu at %llu:%u req %u\n", inode->i_ino, pos, len, total_len); - down_read(&oi->ip_alloc_sem); - /* This is the fast path for re-write. */ - ret = ocfs2_get_block(inode, iblock, bh_result, create); + /* + * Because we need to change file size in ocfs2_dio_end_io_write(), or + * we may need to add it to orphan dir. So can not fall to fast path + * while file size will be changed. + */ + if (pos + total_len <= i_size_read(inode)) { + down_read(&oi->ip_alloc_sem); + /* This is the fast path for re-write. */ + ret = ocfs2_get_block(inode, iblock, bh_result, create); - up_read(&oi->ip_alloc_sem); + up_read(&oi->ip_alloc_sem); - if (buffer_mapped(bh_result) && - !buffer_new(bh_result) && - ret == 0) - goto out; + if (buffer_mapped(bh_result) && + !buffer_new(bh_result) && + ret == 0) + goto out; - /* Clear state set by ocfs2_get_block. */ - bh_result->b_state = 0; + /* Clear state set by ocfs2_get_block. */ + bh_result->b_state = 0; + } dwc = ocfs2_dio_alloc_write_ctx(bh_result, &first_get_block); if (unlikely(dwc == NULL)) { -- cgit v1.2.3 From 28888681b4f641ce9a96478ce4683577cd3abbff Mon Sep 17 00:00:00 2001 From: Ryan Ding Date: Fri, 25 Mar 2016 14:21:23 -0700 Subject: ocfs2: fix a deadlock issue in ocfs2_dio_end_io_write() The code should call ocfs2_free_alloc_context() to free meta_ac & data_ac before calling ocfs2_run_deallocs(). Because ocfs2_run_deallocs() will acquire the system inode's i_mutex hold by meta_ac. So try to release the lock before ocfs2_run_deallocs(). Fixes: af1310367f41 ("ocfs2: fix sparse file & data ordering issue in direct io.") Signed-off-by: Ryan Ding Acked-by: Junxiao Bi Cc: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/aops.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 328ac7f99d52..1581240a7ca0 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -2341,6 +2341,10 @@ static void ocfs2_dio_end_io_write(struct inode *inode, ret = ocfs2_lock_allocators(inode, &et, 0, dwc->dw_zero_count*2, &data_ac, &meta_ac); + if (ret) { + mlog_errno(ret); + goto unlock; + } credits = ocfs2_calc_extend_credits(inode->i_sb, &di->id2.i_list); @@ -2380,14 +2384,14 @@ unlock: ocfs2_inode_unlock(inode, 1); brelse(di_bh); out: - ocfs2_run_deallocs(osb, &dealloc); - if (locked) - mutex_unlock(&inode->i_mutex); - ocfs2_dio_free_write_ctx(inode, dwc); if (data_ac) ocfs2_free_alloc_context(data_ac); if (meta_ac) ocfs2_free_alloc_context(meta_ac); + ocfs2_run_deallocs(osb, &dealloc); + if (locked) + mutex_unlock(&inode->i_mutex); + ocfs2_dio_free_write_ctx(inode, dwc); } /* -- cgit v1.2.3 From ac7cf246dfdbec3d8fed296c7bf30e16f5099dac Mon Sep 17 00:00:00 2001 From: Joseph Qi Date: Fri, 25 Mar 2016 14:21:26 -0700 Subject: ocfs2/dlm: fix race between convert and recovery There is a race window between dlmconvert_remote and dlm_move_lockres_to_recovery_list, which will cause a lock with OCFS2_LOCK_BUSY in grant list, thus system hangs. dlmconvert_remote { spin_lock(&res->spinlock); list_move_tail(&lock->list, &res->converting); lock->convert_pending = 1; spin_unlock(&res->spinlock); status = dlm_send_remote_convert_request(); >>>>>> race window, master has queued ast and return DLM_NORMAL, and then down before sending ast. this node detects master down and calls dlm_move_lockres_to_recovery_list, which will revert the lock to grant list. Then OCFS2_LOCK_BUSY won't be cleared as new master won't send ast any more because it thinks already be authorized. spin_lock(&res->spinlock); lock->convert_pending = 0; if (status != DLM_NORMAL) dlm_revert_pending_convert(res, lock); spin_unlock(&res->spinlock); } In this case, check if res->state has DLM_LOCK_RES_RECOVERING bit set (res is still in recovering) or res master changed (new master has finished recovery), reset the status to DLM_RECOVERING, then it will retry convert. Signed-off-by: Joseph Qi Reported-by: Yiwen Jiang Reviewed-by: Junxiao Bi Cc: Mark Fasheh Cc: Joel Becker Cc: Tariq Saeed Cc: Junxiao Bi Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/dlm/dlmconvert.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c index e36d63ff1783..84de55ed865a 100644 --- a/fs/ocfs2/dlm/dlmconvert.c +++ b/fs/ocfs2/dlm/dlmconvert.c @@ -262,6 +262,7 @@ enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm, struct dlm_lock *lock, int flags, int type) { enum dlm_status status; + u8 old_owner = res->owner; mlog(0, "type=%d, convert_type=%d, busy=%d\n", lock->ml.type, lock->ml.convert_type, res->state & DLM_LOCK_RES_IN_PROGRESS); @@ -316,11 +317,19 @@ enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm, spin_lock(&res->spinlock); res->state &= ~DLM_LOCK_RES_IN_PROGRESS; lock->convert_pending = 0; - /* if it failed, move it back to granted queue */ + /* if it failed, move it back to granted queue. + * if master returns DLM_NORMAL and then down before sending ast, + * it may have already been moved to granted queue, reset to + * DLM_RECOVERING and retry convert */ if (status != DLM_NORMAL) { if (status != DLM_NOTQUEUED) dlm_error(status); dlm_revert_pending_convert(res, lock); + } else if ((res->state & DLM_LOCK_RES_RECOVERING) || + (old_owner != res->owner)) { + mlog(0, "res %.*s is in recovering or has been recovered.\n", + res->lockname.len, res->lockname.name); + status = DLM_RECOVERING; } bail: spin_unlock(&res->spinlock); -- cgit v1.2.3 From be12b299a83fc807bbaccd2bcb8ec50cbb0cb55c Mon Sep 17 00:00:00 2001 From: Joseph Qi Date: Fri, 25 Mar 2016 14:21:29 -0700 Subject: ocfs2/dlm: fix BUG in dlm_move_lockres_to_recovery_list When master handles convert request, it queues ast first and then returns status. This may happen that the ast is sent before the request status because the above two messages are sent by two threads. And right after the ast is sent, if master down, it may trigger BUG in dlm_move_lockres_to_recovery_list in the requested node because ast handler moves it to grant list without clear lock->convert_pending. So remove BUG_ON statement and check if the ast is processed in dlmconvert_remote. Signed-off-by: Joseph Qi Reported-by: Yiwen Jiang Cc: Junxiao Bi Cc: Mark Fasheh Cc: Joel Becker Cc: Tariq Saeed Cc: Junxiao Bi Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/dlm/dlmconvert.c | 13 +++++++++++++ fs/ocfs2/dlm/dlmrecovery.c | 1 - 2 files changed, 13 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c index 84de55ed865a..f90931335c6b 100644 --- a/fs/ocfs2/dlm/dlmconvert.c +++ b/fs/ocfs2/dlm/dlmconvert.c @@ -288,6 +288,19 @@ enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm, status = DLM_DENIED; goto bail; } + + if (lock->ml.type == type && lock->ml.convert_type == LKM_IVMODE) { + mlog(0, "last convert request returned DLM_RECOVERING, but " + "owner has already queued and sent ast to me. res %.*s, " + "(cookie=%u:%llu, type=%d, conv=%d)\n", + res->lockname.len, res->lockname.name, + dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)), + dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)), + lock->ml.type, lock->ml.convert_type); + status = DLM_NORMAL; + goto bail; + } + res->state |= DLM_LOCK_RES_IN_PROGRESS; /* move lock to local convert queue */ /* do not alter lock refcount. switching lists. */ diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c index cd38488a10fc..f6b313898763 100644 --- a/fs/ocfs2/dlm/dlmrecovery.c +++ b/fs/ocfs2/dlm/dlmrecovery.c @@ -2083,7 +2083,6 @@ void dlm_move_lockres_to_recovery_list(struct dlm_ctxt *dlm, dlm_lock_get(lock); if (lock->convert_pending) { /* move converting lock back to granted */ - BUG_ON(i != DLM_CONVERTING_LIST); mlog(0, "node died with convert pending " "on %.*s. move back to granted list.\n", res->lockname.len, res->lockname.name); -- cgit v1.2.3 From 35ddf78e41852cbbd216948425a091ab9815678f Mon Sep 17 00:00:00 2001 From: jiangyiwen Date: Fri, 25 Mar 2016 14:21:32 -0700 Subject: ocfs2: fix occurring deadlock by changing ocfs2_wq from global to local This patch fixes a deadlock, as follows: Node 1 Node 2 Node 3 1)volume a and b are only mount vol a only mount vol b mounted 2) start to mount b start to mount a 3) check hb of Node 3 check hb of Node 2 in vol a, qs_holds++ in vol b, qs_holds++ 4) -------------------- all nodes' network down -------------------- 5) progress of mount b the same situation as failed, and then call Node 2 ocfs2_dismount_volume. but the process is hung, since there is a work in ocfs2_wq cannot beo completed. This work is about vol a, because ocfs2_wq is global wq. BTW, this work which is scheduled in ocfs2_wq is ocfs2_orphan_scan_work, and the context in this work needs to take inode lock of orphan_dir, because lockres owner are Node 1 and all nodes' nework has been down at the same time, so it can't get the inode lock. 6) Why can't this node be fenced when network disconnected? Because the process of mount is hung what caused qs_holds is not equal 0. Because all works in the ocfs2_wq are relative to the super block. The solution is to change the ocfs2_wq from global to local. In other words, move it into struct ocfs2_super. Signed-off-by: Yiwen Jiang Reviewed-by: Joseph Qi Cc: Xue jiufei Cc: Mark Fasheh Cc: Joel Becker Cc: Cc: Junxiao Bi Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/alloc.c | 4 ++-- fs/ocfs2/journal.c | 8 ++++---- fs/ocfs2/localalloc.c | 4 ++-- fs/ocfs2/ocfs2.h | 8 ++++++++ fs/ocfs2/quota_global.c | 2 +- fs/ocfs2/super.c | 37 +++++++++++++++---------------------- fs/ocfs2/super.h | 2 -- 7 files changed, 32 insertions(+), 33 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index d002579c6f2b..1748cea1c9d7 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -6079,7 +6079,7 @@ void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, if (cancel) cancel_delayed_work(&osb->osb_truncate_log_wq); - queue_delayed_work(ocfs2_wq, &osb->osb_truncate_log_wq, + queue_delayed_work(osb->ocfs2_wq, &osb->osb_truncate_log_wq, OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL); } } @@ -6253,7 +6253,7 @@ void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb) if (tl_inode) { cancel_delayed_work(&osb->osb_truncate_log_wq); - flush_workqueue(ocfs2_wq); + flush_workqueue(osb->ocfs2_wq); status = ocfs2_flush_truncate_log(osb); if (status < 0) diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index 61b833b721d8..e607419cdfa4 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c @@ -231,7 +231,7 @@ void ocfs2_recovery_exit(struct ocfs2_super *osb) /* At this point, we know that no more recovery threads can be * launched, so wait for any recovery completion work to * complete. */ - flush_workqueue(ocfs2_wq); + flush_workqueue(osb->ocfs2_wq); /* * Now that recovery is shut down, and the osb is about to be @@ -1326,7 +1326,7 @@ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal, spin_lock(&journal->j_lock); list_add_tail(&item->lri_list, &journal->j_la_cleanups); - queue_work(ocfs2_wq, &journal->j_recovery_work); + queue_work(journal->j_osb->ocfs2_wq, &journal->j_recovery_work); spin_unlock(&journal->j_lock); } @@ -1968,7 +1968,7 @@ static void ocfs2_orphan_scan_work(struct work_struct *work) mutex_lock(&os->os_lock); ocfs2_queue_orphan_scan(osb); if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE) - queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work, + queue_delayed_work(osb->ocfs2_wq, &os->os_orphan_scan_work, ocfs2_orphan_scan_timeout()); mutex_unlock(&os->os_lock); } @@ -2008,7 +2008,7 @@ void ocfs2_orphan_scan_start(struct ocfs2_super *osb) atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE); else { atomic_set(&os->os_state, ORPHAN_SCAN_ACTIVE); - queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work, + queue_delayed_work(osb->ocfs2_wq, &os->os_orphan_scan_work, ocfs2_orphan_scan_timeout()); } } diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c index 7d62c43a2c3e..fe0d1f9571bb 100644 --- a/fs/ocfs2/localalloc.c +++ b/fs/ocfs2/localalloc.c @@ -386,7 +386,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb) struct ocfs2_dinode *alloc = NULL; cancel_delayed_work(&osb->la_enable_wq); - flush_workqueue(ocfs2_wq); + flush_workqueue(osb->ocfs2_wq); if (osb->local_alloc_state == OCFS2_LA_UNUSED) goto out; @@ -1085,7 +1085,7 @@ static int ocfs2_recalc_la_window(struct ocfs2_super *osb, } else { osb->local_alloc_state = OCFS2_LA_DISABLED; } - queue_delayed_work(ocfs2_wq, &osb->la_enable_wq, + queue_delayed_work(osb->ocfs2_wq, &osb->la_enable_wq, OCFS2_LA_ENABLE_INTERVAL); goto out_unlock; } diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h index 7a0126267847..6cf6538a0651 100644 --- a/fs/ocfs2/ocfs2.h +++ b/fs/ocfs2/ocfs2.h @@ -464,6 +464,14 @@ struct ocfs2_super struct ocfs2_refcount_tree *osb_ref_tree_lru; struct mutex system_file_mutex; + + /* + * OCFS2 needs to schedule several different types of work which + * require cluster locking, disk I/O, recovery waits, etc. Since these + * types of work tend to be heavy we avoid using the kernel events + * workqueue and schedule on our own. + */ + struct workqueue_struct *ocfs2_wq; }; #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c index 91bc674203ed..3892f3c079ca 100644 --- a/fs/ocfs2/quota_global.c +++ b/fs/ocfs2/quota_global.c @@ -726,7 +726,7 @@ static int ocfs2_release_dquot(struct dquot *dquot) dqgrab(dquot); /* First entry on list -> queue work */ if (llist_add(&OCFS2_DQUOT(dquot)->list, &osb->dquot_drop_list)) - queue_work(ocfs2_wq, &osb->dquot_drop_work); + queue_work(osb->ocfs2_wq, &osb->dquot_drop_work); goto out; } status = ocfs2_lock_global_qf(oinfo, 1); diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index e2c4e3839d7b..7db631e1c8b0 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -80,12 +80,6 @@ static struct kmem_cache *ocfs2_inode_cachep; struct kmem_cache *ocfs2_dquot_cachep; struct kmem_cache *ocfs2_qf_chunk_cachep; -/* OCFS2 needs to schedule several different types of work which - * require cluster locking, disk I/O, recovery waits, etc. Since these - * types of work tend to be heavy we avoid using the kernel events - * workqueue and schedule on our own. */ -struct workqueue_struct *ocfs2_wq = NULL; - static struct dentry *ocfs2_debugfs_root; MODULE_AUTHOR("Oracle"); @@ -1613,33 +1607,25 @@ static int __init ocfs2_init(void) if (status < 0) goto out2; - ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); - if (!ocfs2_wq) { - status = -ENOMEM; - goto out3; - } - ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); if (!ocfs2_debugfs_root) { status = -ENOMEM; mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); - goto out4; + goto out3; } ocfs2_set_locking_protocol(); status = register_quota_format(&ocfs2_quota_format); if (status < 0) - goto out4; + goto out3; status = register_filesystem(&ocfs2_fs_type); if (!status) return 0; unregister_quota_format(&ocfs2_quota_format); -out4: - destroy_workqueue(ocfs2_wq); - debugfs_remove(ocfs2_debugfs_root); out3: + debugfs_remove(ocfs2_debugfs_root); ocfs2_free_mem_caches(); out2: exit_ocfs2_uptodate_cache(); @@ -1650,11 +1636,6 @@ out1: static void __exit ocfs2_exit(void) { - if (ocfs2_wq) { - flush_workqueue(ocfs2_wq); - destroy_workqueue(ocfs2_wq); - } - unregister_quota_format(&ocfs2_quota_format); debugfs_remove(ocfs2_debugfs_root); @@ -2349,6 +2330,12 @@ static int ocfs2_initialize_super(struct super_block *sb, } cleancache_init_shared_fs(sb); + osb->ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); + if (!osb->ocfs2_wq) { + status = -ENOMEM; + mlog_errno(status); + } + bail: return status; } @@ -2536,6 +2523,12 @@ static void ocfs2_delete_osb(struct ocfs2_super *osb) { /* This function assumes that the caller has the main osb resource */ + /* ocfs2_initializer_super have already created this workqueue */ + if (osb->ocfs2_wq) { + flush_workqueue(osb->ocfs2_wq); + destroy_workqueue(osb->ocfs2_wq); + } + ocfs2_free_slot_info(osb); kfree(osb->osb_orphan_wipes); diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h index b477d0b1c7b6..b023e4f3d740 100644 --- a/fs/ocfs2/super.h +++ b/fs/ocfs2/super.h @@ -26,8 +26,6 @@ #ifndef OCFS2_SUPER_H #define OCFS2_SUPER_H -extern struct workqueue_struct *ocfs2_wq; - int ocfs2_publish_get_mount_state(struct ocfs2_super *osb, int node_num); -- cgit v1.2.3 From 584dca3440732afa84fbca07567bb66e1453936a Mon Sep 17 00:00:00 2001 From: jiangyiwen Date: Fri, 25 Mar 2016 14:21:35 -0700 Subject: ocfs2: solve a problem of crossing the boundary in updating backups In update_backups() there exists a problem of crossing the boundary as follows: we assume that lun will be resized to 1TB(cluster_size is 32kb), it will include 0~33554431 cluster, in update_backups func, it will backup super block in location of 1TB which is the 33554432th cluster, so the phenomenon of crossing the boundary happens. Signed-off-by: Yiwen Jiang Reviewed-by: Joseph Qi Cc: Xue jiufei Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/resize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c index 576b9a04873f..18451e0fab81 100644 --- a/fs/ocfs2/resize.c +++ b/fs/ocfs2/resize.c @@ -196,7 +196,7 @@ static int update_backups(struct inode * inode, u32 clusters, char *data) for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) { blkno = ocfs2_backup_super_blkno(inode->i_sb, i); cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno); - if (cluster > clusters) + if (cluster >= clusters) break; ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup); -- cgit v1.2.3 From e5054c9aefd26ac2401f4b17eda82fab71796dca Mon Sep 17 00:00:00 2001 From: xuejiufei Date: Fri, 25 Mar 2016 14:21:38 -0700 Subject: ocfs2/dlm: move lock to the tail of grant queue while doing in-place convert We have found a bug when two nodes doing umount one after another. 1) Node 1 migrate a lockres that has 3 locks in grant queue such as N2(PR)<->N3(NL)<->N4(PR) to N2. After migration, lvb of the lock N3(NL) and N4(PR) are empty on node 2 because migration target do not copy lvb to these two lock. 2) Node 3 want to convert to PR, it can be granted in __dlmconvert_master(), and the order of these locks is unchanged. The lvb of the lock N3(PR) on node 2 is copyed from lockres in function dlm_update_lvb() while the lvb of lock N4(PR) is still empty. 3) Node 2 want to leave domain, it will migrate this lockres to node 3. Then node 2 will trigger the BUG in dlm_prepare_lvb_for_migration() when adding the lock N4(PR) to mres with the following message because the lvb of mres is already copied from lock N3(PR), but the lvb of lock N4(PR) is empty. "Mismatched lvb in lock cookie=%u:%llu, name=%.*s, node=%u" [akpm@linux-foundation.org: tweak comment] Signed-off-by: xuejiufei Acked-by: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Cc: Junxiao Bi Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/dlm/dlmconvert.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs') diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c index f90931335c6b..cdeafb4e7ed6 100644 --- a/fs/ocfs2/dlm/dlmconvert.c +++ b/fs/ocfs2/dlm/dlmconvert.c @@ -212,6 +212,12 @@ grant: if (lock->lksb->flags & DLM_LKSB_PUT_LVB) memcpy(res->lvb, lock->lksb->lvb, DLM_LVB_LEN); + /* + * Move the lock to the tail because it may be the only lock which has + * an invalid lvb. + */ + list_move_tail(&lock->list, &res->granted); + status = DLM_NORMAL; *call_ast = 1; goto unlock_exit; -- cgit v1.2.3 From 1721598985dc3b92bc3ff45ad0c287bd1814417a Mon Sep 17 00:00:00 2001 From: Xue jiufei Date: Fri, 25 Mar 2016 14:21:41 -0700 Subject: ocfs2: extend transaction for ocfs2_remove_rightmost_path() and ocfs2_update_edge_lengths() before to avoid inconsistency between inode and et I found that jbd2_journal_restart() is called in some places without keeping things consistently before. However, jbd2_journal_restart() may commit the handle's transaction and restart another one. If the first transaction is committed successfully while another not, it may cause filesystem inconsistency or read only. This is an effort to fix this kind of problems. This patch (of 3): The following functions will be called while truncating an extent: ocfs2_remove_btree_range -> ocfs2_start_trans -> ocfs2_remove_extent -> ocfs2_truncate_rec -> ocfs2_extend_rotate_transaction -> jbd2_journal_restart if jbd2_journal_extend fail -> ocfs2_rotate_tree_left -> ocfs2_remove_rightmost_path -> ocfs2_extend_rotate_transaction -> ocfs2_unlink_subtree -> ocfs2_update_edge_lengths -> ocfs2_extend_trans -> jbd2_journal_restart if jbd2_journal_extend fail -> ocfs2_et_update_clusters -> ocfs2_commit_trans jbd2_journal_restart() may be called and it may happened that the buffers dirtied in ocfs2_truncate_rec() are committed while buffers dirtied in ocfs2_et_update_clusters() are not, the total clusters on extent tree and i_clusters in ocfs2_dinode is inconsistency. So the clusters got from ocfs2_dinode is incorrect, and it also cause read-only problem when call ocfs2_commit_truncate() with the error message: "Inode %llu has empty extent block at %llu". We should extend enough credits for function ocfs2_remove_rightmost_path and ocfs2_update_edge_lengths to avoid this inconsistency. Signed-off-by: joyce.xue Acked-by: Joseph Qi Cc: Mark Fasheh Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/alloc.c | 82 +++++++++++++++++++++++++++++++++++++------------------- 1 file changed, 54 insertions(+), 28 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index 1748cea1c9d7..f9ddb5a440a6 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -2516,21 +2516,6 @@ static int ocfs2_update_edge_lengths(handle_t *handle, struct ocfs2_extent_block *eb; u32 range; - /* - * In normal tree rotation process, we will never touch the - * tree branch above subtree_index and ocfs2_extend_rotate_transaction - * doesn't reserve the credits for them either. - * - * But we do have a special case here which will update the rightmost - * records for all the bh in the path. - * So we have to allocate extra credits and access them. - */ - ret = ocfs2_extend_trans(handle, subtree_index); - if (ret) { - mlog_errno(ret); - goto out; - } - ret = ocfs2_journal_access_path(et->et_ci, handle, path); if (ret) { mlog_errno(ret); @@ -2956,7 +2941,7 @@ static int __ocfs2_rotate_tree_left(handle_t *handle, right_path->p_node[subtree_root].bh->b_blocknr, right_path->p_tree_depth); - ret = ocfs2_extend_rotate_transaction(handle, subtree_root, + ret = ocfs2_extend_rotate_transaction(handle, 0, orig_credits, left_path); if (ret) { mlog_errno(ret); @@ -3029,21 +3014,9 @@ static int ocfs2_remove_rightmost_path(handle_t *handle, struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; - ret = ocfs2_et_sanity_check(et); if (ret) goto out; - /* - * There's two ways we handle this depending on - * whether path is the only existing one. - */ - ret = ocfs2_extend_rotate_transaction(handle, 0, - handle->h_buffer_credits, - path); - if (ret) { - mlog_errno(ret); - goto out; - } ret = ocfs2_journal_access_path(et->et_ci, handle, path); if (ret) { @@ -3641,6 +3614,14 @@ static int ocfs2_merge_rec_left(struct ocfs2_path *right_path, */ if (le16_to_cpu(right_rec->e_leaf_clusters) == 0 && le16_to_cpu(el->l_next_free_rec) == 1) { + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + right_path); + if (ret) { + mlog_errno(ret); + goto out; + } ret = ocfs2_remove_rightmost_path(handle, et, right_path, @@ -3679,6 +3660,14 @@ static int ocfs2_try_to_merge_extent(handle_t *handle, BUG_ON(ctxt->c_contig_type == CONTIG_NONE); if (ctxt->c_split_covers_rec && ctxt->c_has_empty_extent) { + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + path); + if (ret) { + mlog_errno(ret); + goto out; + } /* * The merge code will need to create an empty * extent to take the place of the newly @@ -3727,6 +3716,15 @@ static int ocfs2_try_to_merge_extent(handle_t *handle, */ BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0])); + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + path); + if (ret) { + mlog_errno(ret); + goto out; + } + /* The merge left us with an empty extent, remove it. */ ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { @@ -3748,6 +3746,15 @@ static int ocfs2_try_to_merge_extent(handle_t *handle, goto out; } + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + path); + if (ret) { + mlog_errno(ret); + goto out; + } + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); /* * Error from this last rotate is not critical, so @@ -3783,6 +3790,16 @@ static int ocfs2_try_to_merge_extent(handle_t *handle, } if (ctxt->c_split_covers_rec) { + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + path); + if (ret) { + mlog_errno(ret); + ret = 0; + goto out; + } + /* * The merge may have left an empty extent in * our leaf. Try to rotate it away. @@ -5342,6 +5359,15 @@ static int ocfs2_truncate_rec(handle_t *handle, struct ocfs2_extent_block *eb; if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) { + /* extend credit for ocfs2_remove_rightmost_path */ + ret = ocfs2_extend_rotate_transaction(handle, 0, + handle->h_buffer_credits, + path); + if (ret) { + mlog_errno(ret); + goto out; + } + ret = ocfs2_rotate_tree_left(handle, et, path, dealloc); if (ret) { mlog_errno(ret); -- cgit v1.2.3 From 102c2595aa193f598c0f4b1bf2037d168c80e551 Mon Sep 17 00:00:00 2001 From: Xue jiufei Date: Fri, 25 Mar 2016 14:21:44 -0700 Subject: ocfs2: extend enough credits for freeing one truncate record while replaying truncate records Now function ocfs2_replay_truncate_records() first modifies tl_used, then calls ocfs2_extend_trans() to extend transactions for gd and alloc inode used for freeing clusters. jbd2_journal_restart() may be called and it may happen that tl_used in truncate log is decreased but the clusters are not freed, which means these clusters are lost. So we should avoid extending transactions in these two operations. Signed-off-by: joyce.xue Reviewed-by: Mark Fasheh Acked-by: Joseph Qi Cc: Joel Becker Cc: Junxiao Bi Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/ocfs2/alloc.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index f9ddb5a440a6..70907d638b60 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -5954,16 +5954,6 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, ocfs2_journal_dirty(handle, tl_bh); - /* TODO: Perhaps we can calculate the bulk of the - * credits up front rather than extending like - * this. */ - status = ocfs2_extend_trans(handle, - OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC); - if (status < 0) { - mlog_errno(status); - goto bail; - } - rec = tl->tl_recs[i]; start_blk = ocfs2_clusters_to_blocks(data_alloc_inode->i_sb, le32_to_cpu(rec.t_start)); @@ -5984,6 +5974,13 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb, goto bail; } } + + status = ocfs2_extend_trans(handle, + OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC); + if (status < 0) { + mlog_errno(status); + goto bail; + } i--; } @@ -6042,7 +6039,7 @@ int __ocfs2_flush_truncate_log(struct ocfs2_super *osb) goto out_mutex; } - handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE); + handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC); if (IS_ERR(handle)) { status = PTR_ERR(handle); mlog_errno(status); -- cgit v1.2.3