io_uring lseek

From: Hao Xu <[email protected]>

The two functions share almost same code, merge them together. No functional change in this patch.

Quoted 4 lines · Hao Xu

From: Hao Xu <[email protected]>

The two functions share almost same code, merge them together. No functional change in this patch.

No, please don't. seek data and seek hole have subtly different semantics and return values, and we've explicitly kept them separate because it's much easier to maintain them as separate functions with separate semantics than combine them into a single function with lots of non-obvious conditional behaviour.

The fact that the new iomap_seek() API requires a boolean "SEEK_DATA or SEEK_HOLE" field to indicate that the caller wants makes it clear that this isn't actually an improvement over explicit iomap_seek_data() and iomap_seek_hole() API calls.

-Dave.

From: Hao Xu <[email protected]>

To support nowait llseek(), IOMAP_NOWAIT semantics should be respected. In xfs, xfs_seek_iomap_begin() is the only place which may be blocked by ilock and extent loading. Let's turn it into trylock logic just like what we've done in xfs_readdir().

Signed-off-by Hao Xu <[email protected]>

---
 fs/xfs/xfs_iomap.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

fs/xfs/xfs_iomap.c +0 -0

index 18c8f168b153..bbd7c6b27701 100644

Quoted 20 lines · Hao Xu

From: Hao Xu <[email protected]>

To support nowait llseek(), IOMAP_NOWAIT semantics should be respected. In xfs, xfs_seek_iomap_begin() is the only place which may be blocked by ilock and extent loading. Let's turn it into trylock logic just like what we've done in xfs_readdir().

Signed-off-by Hao Xu <[email protected]>

---
 fs/xfs/xfs_iomap.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

fs/xfs/xfs_iomap.c +0 -0

index 18c8f168b153..bbd7c6b27701 100644
@@ -1294,7 +1294,9 @@ xfs_seek_iomap_begin(
 	if (xfs_is_shutdown(mp))
 		return -EIO;
 

fs/xfs/xfs_iomap.c xfs_seek_iomap_begin( · line 1297

Quoted 2 lines

fs/xfs/xfs_iomap.c +1 -1

-	lockmode = xfs_ilock_data_map_shared(ip);
+	lockmode = xfs_ilock_data_map_shared_generic(ip, flags & IOMAP_NOWAIT);

What does this magic XFS function I can't find anywhere in this patch set do?

-Dave.

Signature · 2 lines

Dave Chinner
[email protected]

Quoted 21 lines · Dave Chinner

Quoted 20 lines · Hao Xu

From: Hao Xu <[email protected]>

To support nowait llseek(), IOMAP_NOWAIT semantics should be respected. In xfs, xfs_seek_iomap_begin() is the only place which may be blocked by ilock and extent loading. Let's turn it into trylock logic just like what we've done in xfs_readdir().

Signed-off-by Hao Xu <[email protected]>

---
  fs/xfs/xfs_iomap.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)

fs/xfs/xfs_iomap.c +0 -0

index 18c8f168b153..bbd7c6b27701 100644
@@ -1294,7 +1294,9 @@ xfs_seek_iomap_begin(
  	if (xfs_is_shutdown(mp))
  		return -EIO;
 

Quoted 5 lines

fs/xfs/xfs_iomap.c xfs_seek_iomap_begin( · line 1297

Quoted 2 lines

fs/xfs/xfs_iomap.c +1 -1

-	lockmode = xfs_ilock_data_map_shared(ip);
+	lockmode = xfs_ilock_data_map_shared_generic(ip, flags & IOMAP_NOWAIT);

What does this magic XFS function I can't find anywhere in this patch set do?

It's in (iirc) the io_uring getdents patchset that wasn't cc'd to linux-xfs and that I haven't looked at yet.

--D

Quoted 4 lines

-Dave.

Signature · 2 lines

Dave Chinner
[email protected]

From: Hao Xu <[email protected]>

Add a nowait parameter for iomap_seek(), later IOMAP_NOWAIT is set according to this parameter's value.

From: Hao Xu <[email protected]>

Add llseek_nowait() operation for xfs, it acts just like llseek(). The thing different is it delivers nowait parameter to iomap layer.

Signed-off-by Hao Xu <[email protected]>

---
 fs/xfs/xfs_file.c | 29 +++++++++++++++++++++++++++--
 1 file changed, 27 insertions(+), 2 deletions(-)

fs/xfs/xfs_file.c +26 -2

index 73adc0aee2ff..cba82264221d 100644
@@ -1257,10 +1257,11 @@ xfs_file_readdir(
 }
 
 STATIC loff_t
-xfs_file_llseek(
+__xfs_file_llseek(
 	struct file	*file,
 	loff_t		offset,
-	int		whence)
+	int		whence,
+	bool		nowait)
 {
 	struct inode		*inode = file->f_mapping->host;
 

@@ -1282,6 +1283,28 @@ xfs_file_llseek(
 	return vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
 }
 
+STATIC loff_t
+xfs_file_llseek(
+	struct file	*file,
+	loff_t		offset,
+	int		whence)
+{
+	return __xfs_file_llseek(file, offset, whence, false);
+}
+
+STATIC loff_t
+xfs_file_llseek_nowait(
+	struct file	*file,
+	loff_t		offset,
+	int		whence,
+	bool		nowait)
+{
+	if (file->f_op == &xfs_file_operations)
+		return __xfs_file_llseek(file, offset, whence, nowait);
+	else
+		return generic_file_llseek(file, offset, whence);
+}
+
 #ifdef CONFIG_FS_DAX
 static inline vm_fault_t
 xfs_dax_fault(

@@ -1442,6 +1465,7 @@ xfs_file_mmap(
 
 const struct file_operations xfs_file_operations = {
 	.llseek		= xfs_file_llseek,
+	.llseek_nowait	= xfs_file_llseek_nowait,
 	.read_iter	= xfs_file_read_iter,
 	.write_iter	= xfs_file_write_iter,
 	.splice_read	= xfs_file_splice_read,

Quoted 66 lines · Hao Xu

From: Hao Xu <[email protected]>

Add llseek_nowait() operation for xfs, it acts just like llseek(). The thing different is it delivers nowait parameter to iomap layer.

Signed-off-by Hao Xu <[email protected]>

---
  fs/xfs/xfs_file.c | 29 +++++++++++++++++++++++++++--
  1 file changed, 27 insertions(+), 2 deletions(-)

fs/xfs/xfs_file.c +26 -2

index 73adc0aee2ff..cba82264221d 100644
@@ -1257,10 +1257,11 @@ xfs_file_readdir(
  }
 
  STATIC loff_t
-xfs_file_llseek(
+__xfs_file_llseek(
  	struct file	*file,
  	loff_t		offset,
-	int		whence)
+	int		whence,
+	bool		nowait)
  {
  	struct inode		*inode = file->f_mapping->host;
 

@@ -1282,6 +1283,28 @@ xfs_file_llseek(
  	return vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
  }
 
+STATIC loff_t
+xfs_file_llseek(
+	struct file	*file,
+	loff_t		offset,
+	int		whence)
+{
+	return __xfs_file_llseek(file, offset, whence, false);
+}
+
+STATIC loff_t
+xfs_file_llseek_nowait(
+	struct file	*file,
+	loff_t		offset,
+	int		whence,
+	bool		nowait)
+{
+	if (file->f_op == &xfs_file_operations)
+		return __xfs_file_llseek(file, offset, whence, nowait);
+	else
+		return generic_file_llseek(file, offset, whence);
+}
+
  #ifdef CONFIG_FS_DAX
  static inline vm_fault_t
  xfs_dax_fault(

@@ -1442,6 +1465,7 @@ xfs_file_mmap(
 
  const struct file_operations xfs_file_operations = {
  	.llseek		= xfs_file_llseek,
+	.llseek_nowait	= xfs_file_llseek_nowait,
  	.read_iter	= xfs_file_read_iter,
  	.write_iter	= xfs_file_write_iter,
  	.splice_read	= xfs_file_splice_read,

fs/xfs/xfs_file.c const struct file_operations xfs_dir_file_operations = { · line 1497

Quoted 7 lines

fs/xfs/xfs_file.c +1 -0

@@ -1467,6 +1491,7 @@ const struct file_operations xfs_dir_file_operations = {
  	.read		= generic_read_dir,
  	.iterate_shared	= xfs_file_readdir,
  	.llseek		= generic_file_llseek,
+	.llseek_nowait	= xfs_file_llseek_nowait,
  	.unlocked_ioctl	= xfs_file_ioctl,
  #ifdef CONFIG_COMPAT
  	.compat_ioctl	= xfs_file_compat_ioctl,

This is pretty nasty. It would be far better just to change the .llseek method than to inflict this on every filesystem for the forseeable future.

Not that I'm a fan of passing "nowait" booleans all through the file operations methods - that way lies madness. We use a control structure for the IO path operations (kiocb) to hold per-call context information, perhaps we need something similar for these other methods that people are wanting to hook up to io_uring (e.g. readdir) so taht we don't have to play whack-a-mole with every new io_uring method that people want and then end up with a different nowait solution for every method.

-Dave.

From: Hao Xu <[email protected]>

This series adds lseek for io_uring, the motivation to import this syscall is in previous io_uring getdents patchset, we lack a way to rewind the file cursor when it goes to the end of file. Another reason is lseek is a common syscall, it's good for coding consistency when users use io_uring as their main loop.

Patch 1 is code clean for iomap
Patch 2 adds IOMAP_NOWAIT logic for iomap lseek
Patch 3 adds a nowait parameter to for IOMAP_NOWAIT control Patch 4 adds llseek_nowait() for file_operations so that specific
filesystem can implement it for nowait lseek
Patch 5 adds llseek_nowait() implementation for xfs Patch 6 adds a new vfs wrapper for io_uring use
Patch 7 is the main io_uring lseek implementation

Note, this series depends on the previous io_uring getdents series.

This is marked RFC since there is (at least) an issue to be discussed:
The work in this series is mainly to reslove a problem that the current llseek() in struct file_operations doesn't have a place to deliver nowait info, and adding an argument to it results in update for llseek implementation of all filesystems (35 functions), so here I introduce a new llseek_nowait() as a workaround.

From: Hao Xu <[email protected]>

This series adds lseek for io_uring, the motivation to import this syscall is in previous io_uring getdents patchset, we lack a way to rewind the file cursor when it goes to the end of file. Another reason is lseek is a common syscall, it's good for coding consistency when users use io_uring as their main loop.

Patch 1 is code clean for iomap
Patch 2 adds IOMAP_NOWAIT logic for iomap lseek
Patch 3 adds a nowait parameter to for IOMAP_NOWAIT control Patch 4 adds llseek_nowait() for file_operations so that specific
filesystem can implement it for nowait lseek
Patch 5 adds llseek_nowait() implementation for xfs Patch 6 adds a new vfs wrapper for io_uring use
Patch 7 is the main io_uring lseek implementation

Note, this series depends on the previous io_uring getdents series.

This is marked RFC since there is (at least) an issue to be discussed:
The work in this series is mainly to reslove a problem that the current llseek() in struct file_operations doesn't have a place to deliver nowait info, and adding an argument to it results in update for llseek implementation of all filesystems (35 functions), so here I introduce a new llseek_nowait() as a workaround.

For performance, it has about 20%~30% improvement on iops. The test program is just like the one for io_uring getdents, here is the link to it: https://github.com/HowHsu/liburing/blob/llseek/test/lseek.c
- Each test runs about 30000 async requests/sync syscalls
- Each test runs 100 times and get the average value.
- offset is randomly generated value
- the file is a 1M all zero file

[howeyxu@~]$ python3 run_lseek.py
test args:  seek mode:SEEK_SET, offset: 334772
Average of  sync :  0.012300650000000002
Average of  iouring :  0.008528009999999999
30.67%

[howeyxu@~]$ python3 run_lseek.py
test args:  seek mode:SEEK_CUR, offset: 389292
Average of  sync :  0.012736129999999995
Average of  iouring :  0.00928725
27.08%

[howeyxu@~]$ python3 run_lseek.py
test args:  seek mode:SEEK_END, offset: 281141
Average of  sync :  0.01221595
Average of  iouring :  0.008442890000000003
30.89%

[howeyxu@~]$ python3 run_lseek.py
test args:  seek mode:SEEK_DATA, offset: 931103
Average of  sync :  0.015496230000000005
Average of  iouring :  0.012341509999999998
20.36%

[howeyxu@~]$ python3 run_lseek.py
test args:  seek mode:SEEK_HOLE, offset: 430194
Average of  sync :  0.01555663000000001
Average of  iouring :  0.012064940000000003
22.45%

Hao Xu (7):
iomap: merge iomap_seek_hole() and iomap_seek_data()
xfs: add nowait support for xfs_seek_iomap_begin() add nowait parameter for iomap_seek()
add llseek_nowait() for struct file_operations add llseek_nowait support for xfs
add vfs_lseek_nowait()
add lseek for io_uring

  fs/ext4/file.c                |  9 ++---
  fs/gfs2/inode.c               |  4 +--
  fs/iomap/seek.c               | 42 ++++++-----------------
  fs/read_write.c               | 18 ++++++++++
  fs/xfs/xfs_file.c             | 34 ++++++++++++++++---
  fs/xfs/xfs_iomap.c            |  4 ++-
  include/linux/fs.h            |  4 +++
  include/linux/iomap.h         |  6 ++--
  include/uapi/linux/io_uring.h |  1 +
  io_uring/fs.c                 | 63 +++++++++++++++++++++++++++++++++++
  io_uring/fs.h                 |  3 ++
  io_uring/opdef.c              |  8 +++++
  12 files changed, 145 insertions(+), 51 deletions(-)

base-commit: 4a4b046082eca8ae90b654d772fccc30e9f23f4d