Re: [PATCH v4 RESEND] f2fs: fix out-of-free problem caused by atomic write

From: Yunlong Song
Date: Tue Nov 21 2017 - 22:51:50 EST


ping again...

On 2017/11/17 8:54, Yunlong Song wrote:
f2fs_balance_fs only actives once in the commit_inmem_pages, but there
are more than one page to commit, so all the other pages will miss the
check. This will lead to out-of-free problem when commit a very large
file. However, we cannot do f2fs_balance_fs for each inmem page, since
this will break atomicity. As a result, we should do f2fs_balance_fs
for all the inmem pages together.

Signed-off-by: Yunlong Song <yunlong.song@xxxxxxxxxx>
---
fs/f2fs/debug.c | 5 +++--
fs/f2fs/f2fs.h | 26 ++++++++++++++++++++++++--
fs/f2fs/segment.c | 30 ++++++++++++++++++++++++------
fs/f2fs/segment.h | 4 +++-
4 files changed, 54 insertions(+), 11 deletions(-)

diff --git a/fs/f2fs/debug.c b/fs/f2fs/debug.c
index f7eec50..41c47c4 100644
--- a/fs/f2fs/debug.c
+++ b/fs/f2fs/debug.c
@@ -50,6 +50,7 @@ static void update_general_status(struct f2fs_sb_info *sbi)
si->ndirty_files = sbi->ndirty_inode[FILE_INODE];
si->ndirty_all = sbi->ndirty_inode[DIRTY_META];
si->inmem_pages = get_pages(sbi, F2FS_INMEM_PAGES);
+ si->inmem_commit_pages = get_pages(sbi, F2FS_INMEM_COMMIT_PAGES);
si->aw_cnt = atomic_read(&sbi->aw_cnt);
si->vw_cnt = atomic_read(&sbi->vw_cnt);
si->max_aw_cnt = atomic_read(&sbi->max_aw_cnt);
@@ -360,9 +361,9 @@ static int stat_show(struct seq_file *s, void *v)
si->nr_discarding, si->nr_discarded,
si->nr_discard_cmd, si->undiscard_blks);
seq_printf(s, " - inmem: %4d, atomic IO: %4d (Max. %4d), "
- "volatile IO: %4d (Max. %4d)\n",
+ "volatile IO: %4d (Max. %4d), commit: %4d\n",
si->inmem_pages, si->aw_cnt, si->max_aw_cnt,
- si->vw_cnt, si->max_vw_cnt);
+ si->vw_cnt, si->max_vw_cnt, si->inmem_commit_pages);
seq_printf(s, " - nodes: %4d in %4d\n",
si->ndirty_node, si->node_pages);
seq_printf(s, " - dents: %4d in dirs:%4d (%4d)\n",
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 13a96b8..749bdb6 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -610,6 +610,7 @@ struct f2fs_inode_info {
struct list_head inmem_pages; /* inmemory pages managed by f2fs */
struct task_struct *inmem_task; /* store inmemory task */
struct mutex inmem_lock; /* lock for inmemory pages */
+ unsigned long inmem_blocks; /* inmemory blocks */
struct extent_tree *extent_tree; /* cached extent_tree entry */
struct rw_semaphore dio_rwsem[2];/* avoid racing between dio and gc */
struct rw_semaphore i_mmap_sem;
@@ -863,6 +864,7 @@ enum count_type {
F2FS_DIRTY_NODES,
F2FS_DIRTY_META,
F2FS_INMEM_PAGES,
+ F2FS_INMEM_COMMIT_PAGES,
F2FS_DIRTY_IMETA,
F2FS_WB_CP_DATA,
F2FS_WB_DATA,
@@ -1600,7 +1602,21 @@ static inline void inc_page_count(struct f2fs_sb_info *sbi, int count_type)
atomic_inc(&sbi->nr_pages[count_type]);
if (count_type == F2FS_DIRTY_DATA || count_type == F2FS_INMEM_PAGES ||
- count_type == F2FS_WB_CP_DATA || count_type == F2FS_WB_DATA)
+ count_type == F2FS_WB_CP_DATA || count_type == F2FS_WB_DATA ||
+ count_type == F2FS_INMEM_COMMIT_PAGES)
+ return;
+
+ set_sbi_flag(sbi, SBI_IS_DIRTY);
+}
+
+static inline void inc_pages_count(struct f2fs_sb_info *sbi, int count_type,
+ int pages)
+{
+ atomic_add(pages, &sbi->nr_pages[count_type]);
+
+ if (count_type == F2FS_DIRTY_DATA || count_type == F2FS_INMEM_PAGES ||
+ count_type == F2FS_WB_CP_DATA || count_type == F2FS_WB_DATA ||
+ count_type == F2FS_INMEM_COMMIT_PAGES)
return;
set_sbi_flag(sbi, SBI_IS_DIRTY);
@@ -1618,6 +1634,12 @@ static inline void dec_page_count(struct f2fs_sb_info *sbi, int count_type)
atomic_dec(&sbi->nr_pages[count_type]);
}
+static inline void dec_pages_count(struct f2fs_sb_info *sbi, int count_type,
+ int pages)
+{
+ atomic_sub(pages, &sbi->nr_pages[count_type]);
+}
+
static inline void inode_dec_dirty_pages(struct inode *inode)
{
if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) &&
@@ -2716,7 +2738,7 @@ struct f2fs_stat_info {
unsigned long long hit_total, total_ext;
int ext_tree, zombie_tree, ext_node;
int ndirty_node, ndirty_dent, ndirty_meta, ndirty_data, ndirty_imeta;
- int inmem_pages;
+ int inmem_pages, inmem_commit_pages;
unsigned int ndirty_dirs, ndirty_files, ndirty_all;
int nats, dirty_nats, sits, dirty_sits;
int free_nids, avail_nids, alloc_nids;
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 46dfbca..2ff1bba4 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -174,6 +174,8 @@ bool need_SSR(struct f2fs_sb_info *sbi)
int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
+ int inmem_commit_secs = get_blocktype_secs(sbi,
+ F2FS_INMEM_COMMIT_PAGES);
if (test_opt(sbi, LFS))
return false;
@@ -181,7 +183,7 @@ bool need_SSR(struct f2fs_sb_info *sbi)
return true;
return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
- 2 * reserved_sections(sbi));
+ inmem_commit_secs + 2 * reserved_sections(sbi));
}
void register_inmem_page(struct inode *inode, struct page *page)
@@ -210,6 +212,7 @@ void register_inmem_page(struct inode *inode, struct page *page)
list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
+ fi->inmem_blocks++;
mutex_unlock(&fi->inmem_lock);
trace_f2fs_register_inmem_page(page, INMEM);
@@ -221,6 +224,7 @@ static int __revoke_inmem_pages(struct inode *inode,
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
struct inmem_pages *cur, *tmp;
int err = 0;
+ struct f2fs_inode_info *fi = F2FS_I(inode);
list_for_each_entry_safe(cur, tmp, head, list) {
struct page *page = cur->page;
@@ -263,6 +267,7 @@ static int __revoke_inmem_pages(struct inode *inode,
list_del(&cur->list);
kmem_cache_free(inmem_entry_slab, cur);
dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
+ fi->inmem_blocks--;
}
return err;
}
@@ -302,6 +307,10 @@ void drop_inmem_pages(struct inode *inode)
if (!list_empty(&fi->inmem_ilist))
list_del_init(&fi->inmem_ilist);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
+ if (fi->inmem_blocks) {
+ f2fs_bug_on(sbi, 1);
+ fi->inmem_blocks = 0;
+ }
mutex_unlock(&fi->inmem_lock);
clear_inode_flag(inode, FI_ATOMIC_FILE);
@@ -326,6 +335,7 @@ void drop_inmem_page(struct inode *inode, struct page *page)
f2fs_bug_on(sbi, !cur || cur->page != page);
list_del(&cur->list);
+ fi->inmem_blocks--;
mutex_unlock(&fi->inmem_lock);
dec_page_count(sbi, F2FS_INMEM_PAGES);
@@ -354,7 +364,7 @@ static int __commit_inmem_pages(struct inode *inode,
.io_type = FS_DATA_IO,
};
pgoff_t last_idx = ULONG_MAX;
- int err = 0;
+ int err = 0, inmem_blocks = fi->inmem_blocks;
list_for_each_entry_safe(cur, tmp, &fi->inmem_pages, list) {
struct page *page = cur->page;
@@ -390,6 +400,8 @@ static int __commit_inmem_pages(struct inode *inode,
}
unlock_page(page);
list_move_tail(&cur->list, revoke_list);
+ dec_page_count(sbi, F2FS_INMEM_COMMIT_PAGES);
+ inmem_blocks--;
}
if (last_idx != ULONG_MAX)
@@ -397,6 +409,8 @@ static int __commit_inmem_pages(struct inode *inode,
if (!err)
__revoke_inmem_pages(inode, revoke_list, false, false);
+ else
+ dec_pages_count(sbi, F2FS_INMEM_COMMIT_PAGES, inmem_blocks);
return err;
}
@@ -409,12 +423,12 @@ int commit_inmem_pages(struct inode *inode)
int err;
INIT_LIST_HEAD(&revoke_list);
+ set_inode_flag(inode, FI_ATOMIC_COMMIT);
+ mutex_lock(&fi->inmem_lock);
+ inc_pages_count(sbi, F2FS_INMEM_COMMIT_PAGES, fi->inmem_blocks);
f2fs_balance_fs(sbi, true);
f2fs_lock_op(sbi);
- set_inode_flag(inode, FI_ATOMIC_COMMIT);
-
- mutex_lock(&fi->inmem_lock);
err = __commit_inmem_pages(inode, &revoke_list);
if (err) {
int ret;
@@ -437,11 +451,15 @@ int commit_inmem_pages(struct inode *inode)
if (!list_empty(&fi->inmem_ilist))
list_del_init(&fi->inmem_ilist);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
+ if (fi->inmem_blocks) {
+ f2fs_bug_on(sbi, 1);
+ fi->inmem_blocks = 0;
+ }
+ f2fs_unlock_op(sbi);
mutex_unlock(&fi->inmem_lock);
clear_inode_flag(inode, FI_ATOMIC_COMMIT);
- f2fs_unlock_op(sbi);
return err;
}
diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
index 8d93652..f3885de 100644
--- a/fs/f2fs/segment.h
+++ b/fs/f2fs/segment.h
@@ -503,12 +503,14 @@ static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
+ int inmem_commit_secs = get_blocktype_secs(sbi,
+ F2FS_INMEM_COMMIT_PAGES);
if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
return false;
return (free_sections(sbi) + freed) <=
- (node_secs + 2 * dent_secs + imeta_secs +
+ (node_secs + 2 * dent_secs + imeta_secs + inmem_commit_secs +
reserved_sections(sbi) + needed);
}

--
Thanks,
Yunlong Song