As you know, VFS layer has introduced non-block aio flag IOCB_NOWAIT, which informs kernel to bail out if an AIO request will block for reasons such as file allocations, or a writeback triggered, or would block while allocating requests while performing direct I/O. Subsequent, pwritev2/preadv2 also can leverage this part kernel code. So far, ext4/xfs/btrfs have supported this feature, I'd like to add the related code for ocfs2 file system. Gang He (3): ocfs2: add ocfs2_try_rw_lock and ocfs2_try_inode_lock ocfs2: add ocfs2_overwrite_io function ocfs2: nowait aio support fs/ocfs2/dir.c | 2 +- fs/ocfs2/dlmglue.c | 42 ++++++++++++++++++++++++---- fs/ocfs2/dlmglue.h | 6 +++- fs/ocfs2/extent_map.c | 67 +++++++++++++++++++++++++++++++++++++++++++++ fs/ocfs2/extent_map.h | 3 ++ fs/ocfs2/file.c | 74 +++++++++++++++++++++++++++++++++++++------------- fs/ocfs2/mmap.c | 2 +- fs/ocfs2/ocfs2_trace.h | 10 ++++--- 8 files changed, 175 insertions(+), 31 deletions(-) -- 1.8.5.6
Gang He
2017-Nov-27 09:46 UTC
[Ocfs2-devel] [PATCH 1/3] ocfs2: add ocfs2_try_rw_lock and ocfs2_try_inode_lock
Add ocfs2_try_rw_lock and ocfs2_try_inode_lock functions, which
will be used in non-block IO scenarios.
Signed-off-by: Gang He <ghe at suse.com>
---
fs/ocfs2/dlmglue.c | 22 ++++++++++++++++++++++
fs/ocfs2/dlmglue.h | 4 ++++
2 files changed, 26 insertions(+)
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 4689940..5cfbd04 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -1742,6 +1742,28 @@ int ocfs2_rw_lock(struct inode *inode, int write)
return status;
}
+int ocfs2_try_rw_lock(struct inode *inode, int write)
+{
+ int status, level;
+ struct ocfs2_lock_res *lockres;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+ mlog(0, "inode %llu try to take %s RW lock\n",
+ (unsigned long long)OCFS2_I(inode)->ip_blkno,
+ write ? "EXMODE" : "PRMODE");
+
+ if (ocfs2_mount_local(osb))
+ return 0;
+
+ lockres = &OCFS2_I(inode)->ip_rw_lockres;
+
+ level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
+
+ status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level,
+ DLM_LKF_NOQUEUE, 0);
+ return status;
+}
+
void ocfs2_rw_unlock(struct inode *inode, int write)
{
int level = write ? DLM_LOCK_EX : DLM_LOCK_PR;
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
index a7fc18b..05910fc 100644
--- a/fs/ocfs2/dlmglue.h
+++ b/fs/ocfs2/dlmglue.h
@@ -116,6 +116,7 @@ void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res
*lockres,
int ocfs2_create_new_inode_locks(struct inode *inode);
int ocfs2_drop_inode_locks(struct inode *inode);
int ocfs2_rw_lock(struct inode *inode, int write);
+int ocfs2_try_rw_lock(struct inode *inode, int write);
void ocfs2_rw_unlock(struct inode *inode, int write);
int ocfs2_open_lock(struct inode *inode);
int ocfs2_try_open_lock(struct inode *inode, int write);
@@ -140,6 +141,9 @@ int ocfs2_inode_lock_with_page(struct inode *inode,
/* 99% of the time we don't want to supply any additional flags --
* those are for very specific cases only. */
#define ocfs2_inode_lock(i, b, e) ocfs2_inode_lock_full_nested(i, b, e, 0,
OI_LS_NORMAL)
+#define ocfs2_try_inode_lock(i, b, e)\
+ ocfs2_inode_lock_full_nested(i, b, e, OCFS2_META_LOCK_NOQUEUE,\
+ OI_LS_NORMAL)
void ocfs2_inode_unlock(struct inode *inode,
int ex);
int ocfs2_super_lock(struct ocfs2_super *osb,
--
1.8.5.6
Gang He
2017-Nov-27 09:46 UTC
[Ocfs2-devel] [PATCH 2/3] ocfs2: add ocfs2_overwrite_io function
Add ocfs2_overwrite_io function, which is used to judge if
overwrite allocated blocks, otherwise, the write will bring extra
block allocation overhead.
Signed-off-by: Gang He <ghe at suse.com>
---
fs/ocfs2/extent_map.c | 67 +++++++++++++++++++++++++++++++++++++++++++++++++++
fs/ocfs2/extent_map.h | 3 +++
2 files changed, 70 insertions(+)
diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c
index e4719e0..98bf325 100644
--- a/fs/ocfs2/extent_map.c
+++ b/fs/ocfs2/extent_map.c
@@ -832,6 +832,73 @@ int ocfs2_fiemap(struct inode *inode, struct
fiemap_extent_info *fieinfo,
return ret;
}
+/* Is IO overwriting allocated blocks? */
+int ocfs2_overwrite_io(struct inode *inode, u64 map_start, u64 map_len,
+ int wait)
+{
+ int ret = 0, is_last;
+ u32 mapping_end, cpos;
+ struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+ struct buffer_head *di_bh = NULL;
+ struct ocfs2_extent_rec rec;
+
+ if (wait)
+ ret = ocfs2_inode_lock(inode, &di_bh, 0);
+ else
+ ret = ocfs2_try_inode_lock(inode, &di_bh, 0);
+ if (ret)
+ goto out;
+
+ if (wait)
+ down_read(&OCFS2_I(inode)->ip_alloc_sem);
+ else {
+ if (!down_read_trylock(&OCFS2_I(inode)->ip_alloc_sem)) {
+ ret = -EAGAIN;
+ goto out_unlock1;
+ }
+ }
+
+ if ((OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) &&
+ ((map_start + map_len) <= i_size_read(inode)))
+ goto out_unlock2;
+
+ cpos = map_start >> osb->s_clustersize_bits;
+ mapping_end = ocfs2_clusters_for_bytes(inode->i_sb,
+ map_start + map_len);
+ is_last = 0;
+ while (cpos < mapping_end && !is_last) {
+ ret = ocfs2_get_clusters_nocache(inode, di_bh, cpos,
+ NULL, &rec, &is_last);
+ if (ret) {
+ mlog_errno(ret);
+ goto out_unlock2;
+ }
+
+ if (rec.e_blkno == 0ULL)
+ break;
+
+ if (rec.e_flags & OCFS2_EXT_REFCOUNTED)
+ break;
+
+ cpos = le32_to_cpu(rec.e_cpos) +
+ le16_to_cpu(rec.e_leaf_clusters);
+ }
+
+ if (cpos < mapping_end)
+ ret = 1;
+
+out_unlock2:
+ brelse(di_bh);
+
+ up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+out_unlock1:
+ ocfs2_inode_unlock(inode, 0);
+
+out:
+ return (ret ? 0 : 1);
+}
+
int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int whence)
{
struct inode *inode = file->f_mapping->host;
diff --git a/fs/ocfs2/extent_map.h b/fs/ocfs2/extent_map.h
index 67ea57d..fd9e86a 100644
--- a/fs/ocfs2/extent_map.h
+++ b/fs/ocfs2/extent_map.h
@@ -53,6 +53,9 @@ int ocfs2_extent_map_get_blocks(struct inode *inode, u64
v_blkno, u64 *p_blkno,
int ocfs2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
u64 map_start, u64 map_len);
+int ocfs2_overwrite_io(struct inode *inode, u64 map_start, u64 map_len,
+ int wait);
+
int ocfs2_seek_data_hole_offset(struct file *file, loff_t *offset, int origin);
int ocfs2_xattr_get_clusters(struct inode *inode, u32 v_cluster,
--
1.8.5.6
Return EAGAIN if any of the following checks fail for direct I/O:
Can not get the related locks immediately,
Blocks are not allocated at the write location, it will trigger
block allocation and block IO operations.
Signed-off-by: Gang He <ghe at suse.com>
---
fs/ocfs2/dir.c | 2 +-
fs/ocfs2/dlmglue.c | 20 ++++++++++----
fs/ocfs2/dlmglue.h | 2 +-
fs/ocfs2/file.c | 74 +++++++++++++++++++++++++++++++++++++-------------
fs/ocfs2/mmap.c | 2 +-
fs/ocfs2/ocfs2_trace.h | 10 ++++---
6 files changed, 79 insertions(+), 31 deletions(-)
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index febe631..ea50901 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -1957,7 +1957,7 @@ int ocfs2_readdir(struct file *file, struct dir_context
*ctx)
trace_ocfs2_readdir((unsigned long long)OCFS2_I(inode)->ip_blkno);
- error = ocfs2_inode_lock_atime(inode, file->f_path.mnt, &lock_level);
+ error = ocfs2_inode_lock_atime(inode, file->f_path.mnt, &lock_level,
1);
if (lock_level && error >= 0) {
/* We release EX lock which used to update atime
* and get PR lock again to reduce contention
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index 5cfbd04..feb8dbe 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -2516,13 +2516,18 @@ int ocfs2_inode_lock_with_page(struct inode *inode,
int ocfs2_inode_lock_atime(struct inode *inode,
struct vfsmount *vfsmnt,
- int *level)
+ int *level, int wait)
{
int ret;
- ret = ocfs2_inode_lock(inode, NULL, 0);
+ if (wait)
+ ret = ocfs2_inode_lock(inode, NULL, 0);
+ else
+ ret = ocfs2_try_inode_lock(inode, NULL, 0);
+
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
return ret;
}
@@ -2534,9 +2539,14 @@ int ocfs2_inode_lock_atime(struct inode *inode,
struct buffer_head *bh = NULL;
ocfs2_inode_unlock(inode, 0);
- ret = ocfs2_inode_lock(inode, &bh, 1);
+ if (wait)
+ ret = ocfs2_inode_lock(inode, &bh, 1);
+ else
+ ret = ocfs2_try_inode_lock(inode, &bh, 1);
+
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
return ret;
}
*level = 1;
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
index 05910fc..c83dbb5 100644
--- a/fs/ocfs2/dlmglue.h
+++ b/fs/ocfs2/dlmglue.h
@@ -123,7 +123,7 @@ void ocfs2_refcount_lock_res_init(struct ocfs2_lock_res
*lockres,
void ocfs2_open_unlock(struct inode *inode);
int ocfs2_inode_lock_atime(struct inode *inode,
struct vfsmount *vfsmnt,
- int *level);
+ int *level, int wait);
int ocfs2_inode_lock_full_nested(struct inode *inode,
struct buffer_head **ret_bh,
int ex,
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index dc455d4..900f04e 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -140,6 +140,8 @@ static int ocfs2_file_open(struct inode *inode, struct file
*file)
spin_unlock(&oi->ip_lock);
}
+ file->f_mode |= FMODE_NOWAIT;
+
leave:
return status;
}
@@ -2132,8 +2134,7 @@ static int ocfs2_prepare_inode_for_refcount(struct inode
*inode,
}
static int ocfs2_prepare_inode_for_write(struct file *file,
- loff_t pos,
- size_t count)
+ loff_t pos, size_t count, int wait)
{
int ret = 0, meta_level = 0;
struct dentry *dentry = file->f_path.dentry;
@@ -2145,10 +2146,14 @@ static int ocfs2_prepare_inode_for_write(struct file
*file,
* if we need to make modifications here.
*/
for(;;) {
- ret = ocfs2_inode_lock(inode, NULL, meta_level);
+ if (wait)
+ ret = ocfs2_inode_lock(inode, NULL, meta_level);
+ else
+ ret = ocfs2_try_inode_lock(inode, NULL, meta_level);
if (ret < 0) {
meta_level = -1;
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto out;
}
@@ -2199,7 +2204,7 @@ static int ocfs2_prepare_inode_for_write(struct file
*file,
out_unlock:
trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno,
- pos, count);
+ pos, count, wait);
if (meta_level >= 0)
ocfs2_inode_unlock(inode, meta_level);
@@ -2211,7 +2216,7 @@ static int ocfs2_prepare_inode_for_write(struct file
*file,
static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
struct iov_iter *from)
{
- int direct_io, rw_level;
+ int rw_level;
ssize_t written = 0;
ssize_t ret;
size_t count = iov_iter_count(from);
@@ -2223,6 +2228,8 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
void *saved_ki_complete = NULL;
int append_write = ((iocb->ki_pos + count) > i_size_read(inode) ? 1 :
0);
+ int direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
+ int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0;
trace_ocfs2_file_aio_write(inode, file, file->f_path.dentry,
(unsigned long long)OCFS2_I(inode)->ip_blkno,
@@ -2230,12 +2237,17 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
file->f_path.dentry->d_name.name,
(unsigned int)from->nr_segs); /* GRRRRR */
+ if (!direct_io && nowait)
+ return -EOPNOTSUPP;
+
if (count == 0)
return 0;
- direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
-
- inode_lock(inode);
+ if (direct_io && nowait) {
+ if (!inode_trylock(inode))
+ return -EAGAIN;
+ } else
+ inode_lock(inode);
/*
* Concurrent O_DIRECT writes are allowed with
@@ -2244,9 +2256,13 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
*/
rw_level = (!direct_io || full_coherency || append_write);
- ret = ocfs2_rw_lock(inode, rw_level);
+ if (direct_io && nowait)
+ ret = ocfs2_try_rw_lock(inode, rw_level);
+ else
+ ret = ocfs2_rw_lock(inode, rw_level);
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto out_mutex;
}
@@ -2260,9 +2276,13 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
* other nodes to drop their caches. Buffered I/O
* already does this in write_begin().
*/
- ret = ocfs2_inode_lock(inode, NULL, 1);
+ if (nowait)
+ ret = ocfs2_try_inode_lock(inode, NULL, 1);
+ else
+ ret = ocfs2_inode_lock(inode, NULL, 1);
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto out;
}
@@ -2277,9 +2297,17 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
}
count = ret;
- ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count);
+ if (direct_io && nowait) {
+ if (!ocfs2_overwrite_io(inode, iocb->ki_pos, count, 0)) {
+ ret = -EAGAIN;
+ goto out;
+ }
+ }
+
+ ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count, !nowait);
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto out;
}
@@ -2355,6 +2383,7 @@ static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
int ret = 0, rw_level = -1, lock_level = 0;
struct file *filp = iocb->ki_filp;
struct inode *inode = file_inode(filp);
+ int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0;
trace_ocfs2_file_aio_read(inode, filp, filp->f_path.dentry,
(unsigned long long)OCFS2_I(inode)->ip_blkno,
@@ -2374,9 +2403,14 @@ static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
* need locks to protect pending reads from racing with truncate.
*/
if (iocb->ki_flags & IOCB_DIRECT) {
- ret = ocfs2_rw_lock(inode, 0);
+ if (nowait)
+ ret = ocfs2_try_rw_lock(inode, 0);
+ else
+ ret = ocfs2_rw_lock(inode, 0);
+
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto bail;
}
rw_level = 0;
@@ -2393,9 +2427,11 @@ static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
* like i_size. This allows the checks down below
* generic_file_aio_read() a chance of actually working.
*/
- ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level);
+ ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level,
+ !nowait);
if (ret < 0) {
- mlog_errno(ret);
+ if (ret != -EAGAIN)
+ mlog_errno(ret);
goto bail;
}
ocfs2_inode_unlock(inode, lock_level);
diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
index 098f5c7..fb9a20e 100644
--- a/fs/ocfs2/mmap.c
+++ b/fs/ocfs2/mmap.c
@@ -184,7 +184,7 @@ int ocfs2_mmap(struct file *file, struct vm_area_struct
*vma)
int ret = 0, lock_level = 0;
ret = ocfs2_inode_lock_atime(file_inode(file),
- file->f_path.mnt, &lock_level);
+ file->f_path.mnt, &lock_level, 1);
if (ret < 0) {
mlog_errno(ret);
goto out;
diff --git a/fs/ocfs2/ocfs2_trace.h b/fs/ocfs2/ocfs2_trace.h
index a0b5d00..e2a11aa 100644
--- a/fs/ocfs2/ocfs2_trace.h
+++ b/fs/ocfs2/ocfs2_trace.h
@@ -1449,20 +1449,22 @@
TRACE_EVENT(ocfs2_prepare_inode_for_write,
TP_PROTO(unsigned long long ino, unsigned long long saved_pos,
- unsigned long count),
- TP_ARGS(ino, saved_pos, count),
+ unsigned long count, int wait),
+ TP_ARGS(ino, saved_pos, count, wait),
TP_STRUCT__entry(
__field(unsigned long long, ino)
__field(unsigned long long, saved_pos)
__field(unsigned long, count)
+ __field(int, wait)
),
TP_fast_assign(
__entry->ino = ino;
__entry->saved_pos = saved_pos;
__entry->count = count;
+ __entry->wait = wait;
),
- TP_printk("%llu %llu %lu", __entry->ino,
- __entry->saved_pos, __entry->count)
+ TP_printk("%llu %llu %lu %d", __entry->ino,
+ __entry->saved_pos, __entry->count, __entry->wait)
);
DEFINE_OCFS2_INT_EVENT(generic_file_aio_read_ret);
--
1.8.5.6