Skip to contentSkip to conversation

Final io_uring updates for 6.4-rc1

GIT PULL Merged Jens Axboe 4 participants 10 messages May 7, 2023 – May 9, 2023

Subsystems (3)
lore
Share & tools
lore

Jens Axboe

author maintainer
#0 lore raw

Hi Linus,

Nothing major in here, just two different parts:

- Small series from Breno that enables passing the full SQE down for ->uring_cmd(). This is a prerequisite for enabling full network socket operations. Queued up a bit late because of some stylistic concerns that got resolved, would be nice to have this in 6.4-rc1 so the dependent work will be easier to handle for 6.5.

- Fix for the huge page coalescing, which was a regression introduced in the 6.3 kernel release (Tobias).

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Please pull!

The following changes since commit 3c85cc43c8e7855d202da184baf00c7b8eeacf71:

Revert "io_uring/rsrc: disallow multi-source reg buffers" (2023-04-20 06:51:48 -0600)

are available in the Git repository at:

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

for you to fetch changes up to d2b7fa6174bc4260e496cbf84375c73636914641:

io_uring: Remove unnecessary BUILD_BUG_ON (2023-05-04 08:19:05 -0600)

----------------------------------------------------------------
for-6.4/io_uring-2023-05-07

----------------------------------------------------------------
Breno Leitao (3):
io_uring: Create a helper to return the SQE size
io_uring: Pass whole sqe to commands
io_uring: Remove unnecessary BUILD_BUG_ON

Tobias Holl (1):
io_uring/rsrc: check for nonconsecutive pages

 drivers/block/ublk_drv.c  | 26 +++++++++++++-------------
 drivers/nvme/host/ioctl.c |  2 +-
 include/linux/io_uring.h  |  7 ++++++-
 io_uring/io_uring.h       | 10 ++++++++++
 io_uring/opdef.c          |  2 +-
 io_uring/rsrc.c           |  7 ++++++-
 io_uring/uring_cmd.c      | 12 +++---------
 io_uring/uring_cmd.h      |  8 --------
 8 files changed, 40 insertions(+), 34 deletions(-)

Merge resolution:

commit 775e045e380626ce769d95badc79ea08edc1b15d
Merge: efd141da30bb d2b7fa6174bc
Author: Jens Axboe <[email protected]>
Date: Thu May 4 08:22:24 2023 -0600

Merge branch 'for-6.4/io_uring' into test

* for-6.4/io_uring:
io_uring: Remove unnecessary BUILD_BUG_ON
io_uring: Pass whole sqe to commands
io_uring: Create a helper to return the SQE size io_uring/rsrc: check for nonconsecutive pages

drivers/block/ublk_drv.c +0 -0
diff --cc drivers/block/ublk_drv.c
index 72a5cde9a5af,42f4d7ca962e..0af6a41f92b2
@@@ -1397,23 -1362,6 +1397,23 @@@ static int __ublk_ch_uring_cmd(struct i
  	return -EIOCBQUEUED;
  }
  
 +static int ublk_ch_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags)
 +{
- 	struct ublksrv_io_cmd *ub_src = (struct ublksrv_io_cmd *) cmd->cmd;
++	const struct ublksrv_io_cmd *ub_src = io_uring_sqe_cmd(cmd->sqe);
 +	struct ublksrv_io_cmd ub_cmd;
 +
 +	/*
 +	 * Not necessary for async retry, but let's keep it simple and always
 +	 * copy the values to avoid any potential reuse.
 +	 */
 +	ub_cmd.q_id = READ_ONCE(ub_src->q_id);
 +	ub_cmd.tag = READ_ONCE(ub_src->tag);
 +	ub_cmd.result = READ_ONCE(ub_src->result);
 +	ub_cmd.addr = READ_ONCE(ub_src->addr);
 +
 +	return __ublk_ch_uring_cmd(cmd, issue_flags, &ub_cmd);
 +}
 +
  static const struct file_operations ublk_ch_fops = {
  	.owner = THIS_MODULE,
  	.open = ublk_ch_open,
@@@ -2240,9 -2171,8 +2240,9 @@@ exit
  static int ublk_ctrl_uring_cmd(struct io_uring_cmd *cmd,
  		unsigned int issue_flags)
  {
- 	struct ublksrv_ctrl_cmd *header = (struct ublksrv_ctrl_cmd *)cmd->cmd;
+ 	const struct ublksrv_ctrl_cmd *header = io_uring_sqe_cmd(cmd->sqe);
  	struct ublk_device *ub = NULL;
 +	u32 cmd_op = cmd->cmd_op;
  	int ret = -EINVAL;
  
  	if (issue_flags & IO_URING_F_NONBLOCK)
Signature · 1 line
Jens Axboe
#1 lore raw
Quoted 5 lines · Jens Axboe

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Mine is somewhat different - I think the "ub_cmd" argument to __ublk_ch_uring_cmd() should also have been made a 'const' pointer.

And then using an actual initializer allows us to do the same thing for the copy in ublk_ch_uring_cmd() (and also makes it clear that it initializes the whole struct - which it did, but still...)

So my conflict resolution looks a bit more complicated, but I think it's the RightThing(tm) to do, and is consistent with the constification in commit fd9b8547bc5c ("io_uring: Pass whole sqe to commands").

Linus

author maintainer
#2 lore raw
Quoted 10 lines · Linus Torvalds
Quoted 5 lines · On Sun, May 7, 2023 at 5:00?AM Jens Axboe <[email protected]> wrote:

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Mine is somewhat different - I think the "ub_cmd" argument to __ublk_ch_uring_cmd() should also have been made a 'const' pointer.

Quoted 8 lines

And then using an actual initializer allows us to do the same thing for the copy in ublk_ch_uring_cmd() (and also makes it clear that it initializes the whole struct - which it did, but still...)

So my conflict resolution looks a bit more complicated, but I think it's the RightThing(tm) to do, and is consistent with the constification in commit fd9b8547bc5c ("io_uring: Pass whole sqe to commands").

Your resolution is obviously fine, and I did ponder pushing the constification further. Thanks!

Signature · 1 line
Jens Axboe
merge bot Merged
#3 lore raw
In reply to Jens Axboe
The pull request you sent on Sun, 7 May 2023 06:00:48 -0600:

The pull request you sent on Sun, 7 May 2023 06:00:48 -0600:

Quoted 1 line

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/03e5cb7b50feb687508946a702febaba24c77f0b

Thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html
#4 lore raw
In reply to Jens Axboe

Hi,

Quoted 33 lines · Jens Axboe

Hi Linus,

Nothing major in here, just two different parts:

- Small series from Breno that enables passing the full SQE down for ->uring_cmd(). This is a prerequisite for enabling full network socket operations. Queued up a bit late because of some stylistic concerns that got resolved, would be nice to have this in 6.4-rc1 so the dependent work will be easier to handle for 6.5.

- Fix for the huge page coalescing, which was a regression introduced in the 6.3 kernel release (Tobias).

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Please pull!

The following changes since commit 3c85cc43c8e7855d202da184baf00c7b8eeacf71:

Revert "io_uring/rsrc: disallow multi-source reg buffers" (2023-04-20 06:51:48 -0600)

are available in the Git repository at:

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

for you to fetch changes up to d2b7fa6174bc4260e496cbf84375c73636914641:

io_uring: Remove unnecessary BUILD_BUG_ON (2023-05-04 08:19:05 -0600)

Quoted 7 lines

----------------------------------------------------------------
for-6.4/io_uring-2023-05-07

----------------------------------------------------------------
Breno Leitao (3):
io_uring: Create a helper to return the SQE size
io_uring: Pass whole sqe to commands

This commit causes broken builds when IO_URING=n and NVME_CORE=y, as io_uring_sqe_cmd(), called in drivers/nvme/host/ioctl.c, ends up being undefined. This was also reported [1] by 0-day bot on your branch yesterday, but it's worse now that Linus merged the pull request.

Not sure what the better fix would be. Move io_uring_sqe_cmd() outside of the "#if defined(CONFIG_IO_URING)" block?

ChenYu

[1] https://lore.kernel.org/all/[email protected]/

Quoted 77 lines

io_uring: Remove unnecessary BUILD_BUG_ON

Tobias Holl (1):
io_uring/rsrc: check for nonconsecutive pages

 drivers/block/ublk_drv.c  | 26 +++++++++++++-------------
 drivers/nvme/host/ioctl.c |  2 +-
 include/linux/io_uring.h  |  7 ++++++-
 io_uring/io_uring.h       | 10 ++++++++++
 io_uring/opdef.c          |  2 +-
 io_uring/rsrc.c           |  7 ++++++-
 io_uring/uring_cmd.c      | 12 +++---------
 io_uring/uring_cmd.h      |  8 --------
 8 files changed, 40 insertions(+), 34 deletions(-)

Merge resolution:

commit 775e045e380626ce769d95badc79ea08edc1b15d
Merge: efd141da30bb d2b7fa6174bc
Author: Jens Axboe <[email protected]>
Date: Thu May 4 08:22:24 2023 -0600

Merge branch 'for-6.4/io_uring' into test

* for-6.4/io_uring:
io_uring: Remove unnecessary BUILD_BUG_ON
io_uring: Pass whole sqe to commands
io_uring: Create a helper to return the SQE size io_uring/rsrc: check for nonconsecutive pages

drivers/block/ublk_drv.c +2 -2
diff --cc drivers/block/ublk_drv.c
index 72a5cde9a5af,42f4d7ca962e..0af6a41f92b2
@@@ -1397,23 -1362,6 +1397,23 @@@ static int __ublk_ch_uring_cmd(struct i
  	return -EIOCBQUEUED;
  }
 
 +static int ublk_ch_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags)
 +{
- 	struct ublksrv_io_cmd *ub_src = (struct ublksrv_io_cmd *) cmd->cmd;
++	const struct ublksrv_io_cmd *ub_src = io_uring_sqe_cmd(cmd->sqe);
 +	struct ublksrv_io_cmd ub_cmd;
 +
 +	/*
 +	 * Not necessary for async retry, but let's keep it simple and always
 +	 * copy the values to avoid any potential reuse.
 +	 */
 +	ub_cmd.q_id = READ_ONCE(ub_src->q_id);
 +	ub_cmd.tag = READ_ONCE(ub_src->tag);
 +	ub_cmd.result = READ_ONCE(ub_src->result);
 +	ub_cmd.addr = READ_ONCE(ub_src->addr);
 +
 +	return __ublk_ch_uring_cmd(cmd, issue_flags, &ub_cmd);
 +}
 +
  static const struct file_operations ublk_ch_fops = {
  	.owner = THIS_MODULE,
  	.open = ublk_ch_open,
@@@ -2240,9 -2171,8 +2240,9 @@@ exit
  static int ublk_ctrl_uring_cmd(struct io_uring_cmd *cmd,
  		unsigned int issue_flags)
  {
- 	struct ublksrv_ctrl_cmd *header = (struct ublksrv_ctrl_cmd *)cmd->cmd;
+ 	const struct ublksrv_ctrl_cmd *header = io_uring_sqe_cmd(cmd->sqe);
  	struct ublk_device *ub = NULL;
 +	u32 cmd_op = cmd->cmd_op;
  	int ret = -EINVAL;
 
  	if (issue_flags & IO_URING_F_NONBLOCK)
Signature · 2 lines
-- 
Jens Axboe
author maintainer
#5 lore raw
Quoted 44 lines · Chen-Yu Tsai

Hi,

Quoted 33 lines · Jens Axboe

Hi Linus,

Nothing major in here, just two different parts:

- Small series from Breno that enables passing the full SQE down for ->uring_cmd(). This is a prerequisite for enabling full network socket operations. Queued up a bit late because of some stylistic concerns that got resolved, would be nice to have this in 6.4-rc1 so the dependent work will be easier to handle for 6.5.

- Fix for the huge page coalescing, which was a regression introduced in the 6.3 kernel release (Tobias).

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Please pull!

The following changes since commit 3c85cc43c8e7855d202da184baf00c7b8eeacf71:

Revert "io_uring/rsrc: disallow multi-source reg buffers" (2023-04-20 06:51:48 -0600)

are available in the Git repository at:

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

for you to fetch changes up to d2b7fa6174bc4260e496cbf84375c73636914641:

io_uring: Remove unnecessary BUILD_BUG_ON (2023-05-04 08:19:05 -0600)

Quoted 7 lines

----------------------------------------------------------------
for-6.4/io_uring-2023-05-07

----------------------------------------------------------------
Breno Leitao (3):
io_uring: Create a helper to return the SQE size
io_uring: Pass whole sqe to commands

Quoted 7 lines

This commit causes broken builds when IO_URING=n and NVME_CORE=y, as io_uring_sqe_cmd(), called in drivers/nvme/host/ioctl.c, ends up being undefined. This was also reported [1] by 0-day bot on your branch yesterday, but it's worse now that Linus merged the pull request.

Not sure what the better fix would be. Move io_uring_sqe_cmd() outside of the "#if defined(CONFIG_IO_URING)" block?

Queued up a patch for this:

https://git.kernel.dk/cgit/linux/commit/?h=io_uring-6.4&id=5d371b2f2b0d1a047582563ee36af8ffb5022847

Signature · 1 line
Jens Axboe
Tested-by
#6 lore raw
Quoted 51 lines · Jens Axboe
Quoted 49 lines · Chen-Yu Tsai

Hi,

Quoted 33 lines · Jens Axboe

Hi Linus,

Nothing major in here, just two different parts:

- Small series from Breno that enables passing the full SQE down for ->uring_cmd(). This is a prerequisite for enabling full network socket operations. Queued up a bit late because of some stylistic concerns that got resolved, would be nice to have this in 6.4-rc1 so the dependent work will be easier to handle for 6.5.

- Fix for the huge page coalescing, which was a regression introduced in the 6.3 kernel release (Tobias).

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Please pull!

The following changes since commit 3c85cc43c8e7855d202da184baf00c7b8eeacf71:

Revert "io_uring/rsrc: disallow multi-source reg buffers" (2023-04-20 06:51:48 -0600)

are available in the Git repository at:

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

for you to fetch changes up to d2b7fa6174bc4260e496cbf84375c73636914641:

io_uring: Remove unnecessary BUILD_BUG_ON (2023-05-04 08:19:05 -0600)

Quoted 7 lines

----------------------------------------------------------------
for-6.4/io_uring-2023-05-07

----------------------------------------------------------------
Breno Leitao (3):
io_uring: Create a helper to return the SQE size
io_uring: Pass whole sqe to commands

This commit causes broken builds when IO_URING=n and NVME_CORE=y, as io_uring_sqe_cmd(), called in drivers/nvme/host/ioctl.c, ends up being undefined. This was also reported [1] by 0-day bot on your branch yesterday, but it's worse now that Linus merged the pull request.

Quoted 6 lines
Quoted 2 lines

Not sure what the better fix would be. Move io_uring_sqe_cmd() outside of the "#if defined(CONFIG_IO_URING)" block?

Queued up a patch for this:

https://git.kernel.dk/cgit/linux/commit/?h=io_uring-6.4&id=5d371b2f2b0d1a047582563ee36af8ffb5022847

Thanks! Looks like the Reported-by line for the test bot is missing a right angle bracket?

Also, consider it

author maintainer
#7 lore raw
Quoted 57 lines · Chen-Yu Tsai
Quoted 56 lines · On Tue, May 9, 2023 at 2:42?AM Jens Axboe <[email protected]> wrote:
Quoted 44 lines · Chen-Yu Tsai

Hi,

Quoted 33 lines · Jens Axboe

Hi Linus,

Nothing major in here, just two different parts:

- Small series from Breno that enables passing the full SQE down for ->uring_cmd(). This is a prerequisite for enabling full network socket operations. Queued up a bit late because of some stylistic concerns that got resolved, would be nice to have this in 6.4-rc1 so the dependent work will be easier to handle for 6.5.

- Fix for the huge page coalescing, which was a regression introduced in the 6.3 kernel release (Tobias).

Note that this will throw a merge conflict in the ublk_drv code, due to this branch still being based off the original for-6.4/io_uring branch. Resolution is pretty straight forward, I'm including it below for reference.

Please pull!

The following changes since commit 3c85cc43c8e7855d202da184baf00c7b8eeacf71:

Revert "io_uring/rsrc: disallow multi-source reg buffers" (2023-04-20 06:51:48 -0600)

are available in the Git repository at:

git://git.kernel.dk/linux.git tags/for-6.4/io_uring-2023-05-07

for you to fetch changes up to d2b7fa6174bc4260e496cbf84375c73636914641:

io_uring: Remove unnecessary BUILD_BUG_ON (2023-05-04 08:19:05 -0600)

Quoted 7 lines

----------------------------------------------------------------
for-6.4/io_uring-2023-05-07

----------------------------------------------------------------
Breno Leitao (3):
io_uring: Create a helper to return the SQE size
io_uring: Pass whole sqe to commands

Quoted 7 lines

This commit causes broken builds when IO_URING=n and NVME_CORE=y, as io_uring_sqe_cmd(), called in drivers/nvme/host/ioctl.c, ends up being undefined. This was also reported [1] by 0-day bot on your branch yesterday, but it's worse now that Linus merged the pull request.

Not sure what the better fix would be. Move io_uring_sqe_cmd() outside of the "#if defined(CONFIG_IO_URING)" block?

Queued up a patch for this:

Quoted 8 lines
Quoted 1 line

https://git.kernel.dk/cgit/linux/commit/?h=io_uring-6.4&id=5d371b2f2b0d1a047582563ee36af8ffb5022847

Thanks! Looks like the Reported-by line for the test bot is missing a right angle bracket?

Also, consider it

Oops yes, thanks for noticing. I'll correct that and add your tested-by.

Signature · 1 line
Jens Axboe
#8 lore raw
Quoted 2 lines · Jens Axboe

Oops yes, thanks for noticing. I'll correct that and add your tested-by.

.. this build problem isn't caught by Guenter's build testing, but can we please make sure that it's fixed in my tree by -rc2?

It would be lovely to have a release that doesn't end up having silly build problems pending for too long. Even if they might be unusual configs.

Linus

author maintainer
#9 lore raw
Quoted 10 lines · Linus Torvalds
Quoted 2 lines · On Tue, May 9, 2023 at 6:59?AM Jens Axboe <[email protected]> wrote:

Oops yes, thanks for noticing. I'll correct that and add your tested-by.

.. this build problem isn't caught by Guenter's build testing, but can we please make sure that it's fixed in my tree by -rc2?

It would be lovely to have a release that doesn't end up having silly build problems pending for too long. Even if they might be unusual configs.

It's in my tree, it'll go your way before -rc2 for sure.

Signature · 1 line
Jens Axboe