mbox series

[v7,00/11] io_uring: Initial support for {s,g}etsockopt commands

Message ID 20231016134750.1381153-1-leitao@debian.org (mailing list archive)
Headers show
Series io_uring: Initial support for {s,g}etsockopt commands | expand

Message

Breno Leitao Oct. 16, 2023, 1:47 p.m. UTC
This patchset adds support for getsockopt (SOCKET_URING_OP_GETSOCKOPT)
and setsockopt (SOCKET_URING_OP_SETSOCKOPT) in io_uring commands.
SOCKET_URING_OP_SETSOCKOPT implements generic case, covering all levels
and optnames. SOCKET_URING_OP_GETSOCKOPT is limited, for now, to
SOL_SOCKET level, which seems to be the most common level parameter for
get/setsockopt(2).

In order to keep the implementation (and tests) simple, some refactors
were done prior to the changes, as follows:

Patches 1-2: Make BPF cgroup filters sockptr aware

Patches 3-4: Remove the core {s,g}etsockopt() core function from
__sys_{g,s}etsockopt, so, the code could be reused by other callers, such as
io_uring.

Patch 5: Pass compat mode to the file/socket callbacks

Patch 6-7: Move io_uring helpers from io_uring_zerocopy_tx to a generic
io_uring headers. This simplify the test case (last patch). Also copy the
io_uring UAPI to the tests directory.

Patch 8: Protect io_uring_cmd_sock() to not be called if CONFIG_NET is
disabled.

These changes were tested with a new test[1] in liburing, LTP sockopt*
tests, as also with bpf/progs/sockopt test case, which is now adapted to
run using both system calls and io_uring commands.

[1] Link: https://github.com/leitao/liburing/blob/getsock/test/socket-getsetsock-cmd.c

RFC -> V1:
	* Copy user memory at io_uring subsystem, and call proto_ops
	  callbacks using kernel memory
	* Implement all the cases for SOCKET_URING_OP_SETSOCKOPT

V1 -> V2
	* Implemented the BPF part
	* Using user pointers from optval to avoid kmalloc in io_uring part.

V2 -> V3:
	* Break down __sys_setsockopt and reuse the core code, avoiding
	  duplicated code. This removed the requirement to expose
	  sock_use_custom_sol_socket().
	* Added io_uring test to selftests/bpf/sockopt.
	* Fixed compat argument, by passing it to the issue_flags.

V3 -> V4:
	* Rebase on top of commit 1ded5e5a5931b ("net: annotate data-races around sock->ops")
	* Also broke down __sys_setsockopt() to reuse the core function
	  from io_uring.
	* Create a new patch to return -EOPNOTSUPP if CONFIG_NET is
	  disabled.
	* Added two SOL_SOCKET tests in bpf/prog_tests/sockopt.

V4 -> V5:
	* Do not use sockptr anymore, by changing the optlen getsock argument
	  to be a user pointer (instead of a kernel pointer). This change also drop
	  the limitation on getsockopt from previous versions, and now all
	  levels are supported.
	* Simplified the BPF sockopt test, since there is no more limitation on
	  the io_uring commands.
	* No more changes in the BPF subsystem.
	* Moved the optlen field in the SQE struct. It is now a pointer instead
	  of u32.

V5 -> V6:
	* Removed the need for #ifdef CONFIG_NET as suggested by Gabriel
	  Krisman.
	* Changed the variable declaration order to respect the reverse
	  xmas declaration as suggested by Paolo Abeni.

V6 -> V7:
	* Changed the optlen back to a value in the SQE instead of
	  user-pointer. This is similar to version 4.
	  [https://lore.kernel.org/all/20231009095518.288a5573@kernel.org/]
	* Imported the io_uring.h into tools/include/uapi/linux to be able to
	  run the tests in machines without liburing.
	  [https://lore.kernel.org/all/77405214-ae42-d58b-1d40-c639683a0cb1@linux.dev/]

Breno Leitao (11):
  bpf: Leverage sockptr_t in BPF getsockopt hook
  bpf: Leverage sockptr_t in BPF setsockopt hook
  net/socket: Break down __sys_setsockopt
  net/socket: Break down __sys_getsockopt
  io_uring/cmd: Pass compat mode in issue_flags
  tools headers: Grab copy of io_uring.h
  selftests/net: Extract uring helpers to be reusable
  io_uring/cmd: return -EOPNOTSUPP if net is disabled
  io_uring/cmd: Introduce SOCKET_URING_OP_GETSOCKOPT
  io_uring/cmd: Introduce SOCKET_URING_OP_SETSOCKOPT
  selftests/bpf/sockopt: Add io_uring support
Breno Leitao (11):
  bpf: Add sockptr support for getsockopt
  bpf: Add sockptr support for setsockopt
  net/socket: Break down __sys_setsockopt
  net/socket: Break down __sys_getsockopt
  io_uring/cmd: Pass compat mode in issue_flags
  tools headers: Grab copy of io_uring.h
  selftests/net: Extract uring helpers to be reusable
  io_uring/cmd: return -EOPNOTSUPP if net is disabled
  io_uring/cmd: Introduce SOCKET_URING_OP_GETSOCKOPT
  io_uring/cmd: Introduce SOCKET_URING_OP_SETSOCKOPT
  selftests/bpf/sockopt: Add io_uring support

 include/linux/bpf-cgroup.h                    |   9 +-
 include/linux/io_uring.h                      |   1 +
 include/net/sock.h                            |   6 +-
 include/uapi/linux/io_uring.h                 |   8 +
 io_uring/uring_cmd.c                          |  53 ++
 kernel/bpf/cgroup.c                           |  25 +-
 net/core/sock.c                               |   8 -
 net/socket.c                                  | 103 ++-
 tools/include/io_uring/mini_liburing.h        | 282 +++++++
 tools/include/uapi/linux/io_uring.h           | 757 ++++++++++++++++++
 .../selftests/bpf/prog_tests/sockopt.c        | 113 ++-
 tools/testing/selftests/net/Makefile          |   1 +
 .../selftests/net/io_uring_zerocopy_tx.c      | 268 +------
 13 files changed, 1300 insertions(+), 334 deletions(-)
 create mode 100644 tools/include/io_uring/mini_liburing.h
 create mode 100644 tools/include/uapi/linux/io_uring.h

Comments

Jens Axboe Oct. 19, 2023, 2:58 p.m. UTC | #1
On 10/16/23 7:47 AM, Breno Leitao wrote:
> This patchset adds support for getsockopt (SOCKET_URING_OP_GETSOCKOPT)
> and setsockopt (SOCKET_URING_OP_SETSOCKOPT) in io_uring commands.
> SOCKET_URING_OP_SETSOCKOPT implements generic case, covering all levels
> and optnames. SOCKET_URING_OP_GETSOCKOPT is limited, for now, to
> SOL_SOCKET level, which seems to be the most common level parameter for
> get/setsockopt(2).
> 
> In order to keep the implementation (and tests) simple, some refactors
> were done prior to the changes, as follows:

Looks like folks are mostly happy with this now, so the next question is
how to stage it?
Jakub Kicinski Oct. 19, 2023, 3:33 p.m. UTC | #2
On Thu, 19 Oct 2023 08:58:59 -0600 Jens Axboe wrote:
> On 10/16/23 7:47 AM, Breno Leitao wrote:
> > This patchset adds support for getsockopt (SOCKET_URING_OP_GETSOCKOPT)
> > and setsockopt (SOCKET_URING_OP_SETSOCKOPT) in io_uring commands.
> > SOCKET_URING_OP_SETSOCKOPT implements generic case, covering all levels
> > and optnames. SOCKET_URING_OP_GETSOCKOPT is limited, for now, to
> > SOL_SOCKET level, which seems to be the most common level parameter for
> > get/setsockopt(2).
> > 
> > In order to keep the implementation (and tests) simple, some refactors
> > were done prior to the changes, as follows:  
> 
> Looks like folks are mostly happy with this now, so the next question is
> how to stage it?

Would be good to get acks from BPF folks but AFAICT first four patches
apply cleanly for us now. If they apply cleanly for you I reckon you
can take them directly with io-uring. It's -rc7 time, with a bit of
luck we'll get to the merge window without a conflict.
Jens Axboe Oct. 19, 2023, 3:40 p.m. UTC | #3
On 10/19/23 9:33 AM, Jakub Kicinski wrote:
> On Thu, 19 Oct 2023 08:58:59 -0600 Jens Axboe wrote:
>> On 10/16/23 7:47 AM, Breno Leitao wrote:
>>> This patchset adds support for getsockopt (SOCKET_URING_OP_GETSOCKOPT)
>>> and setsockopt (SOCKET_URING_OP_SETSOCKOPT) in io_uring commands.
>>> SOCKET_URING_OP_SETSOCKOPT implements generic case, covering all levels
>>> and optnames. SOCKET_URING_OP_GETSOCKOPT is limited, for now, to
>>> SOL_SOCKET level, which seems to be the most common level parameter for
>>> get/setsockopt(2).
>>>
>>> In order to keep the implementation (and tests) simple, some refactors
>>> were done prior to the changes, as follows:  
>>
>> Looks like folks are mostly happy with this now, so the next question is
>> how to stage it?
> 
> Would be good to get acks from BPF folks but AFAICT first four patches

Agree, those are still missing. BPF folks, do patches 1-2 look OK to
you?

> apply cleanly for us now. If they apply cleanly for you I reckon you
> can take them directly with io-uring. It's -rc7 time, with a bit of
> luck we'll get to the merge window without a conflict.

I'll tentatively setup a branch for this just to see if we run into
anything on the merge front. Depending on how the BPF side goes, I can
rebase/collect reviews/whatever as we go.
Jens Axboe Oct. 19, 2023, 3:41 p.m. UTC | #4
On Mon, 16 Oct 2023 06:47:38 -0700, Breno Leitao wrote:
> This patchset adds support for getsockopt (SOCKET_URING_OP_GETSOCKOPT)
> and setsockopt (SOCKET_URING_OP_SETSOCKOPT) in io_uring commands.
> SOCKET_URING_OP_SETSOCKOPT implements generic case, covering all levels
> and optnames. SOCKET_URING_OP_GETSOCKOPT is limited, for now, to
> SOL_SOCKET level, which seems to be the most common level parameter for
> get/setsockopt(2).
> 
> [...]

Applied, thanks!

[01/11] bpf: Add sockptr support for getsockopt
        commit: 7cb15cc7e081730df3392f136a8789f3d2c3fd66
[02/11] bpf: Add sockptr support for setsockopt
        commit: c028f6e54aa180747e384796760eee3bd78e0891
[03/11] net/socket: Break down __sys_setsockopt
        commit: e70464dcdcddb5128fe7956bf809683824c64de5
[04/11] net/socket: Break down __sys_getsockopt
        commit: 25f82732c8352bd0bec33c5a9989fd46cac5789f
[05/11] io_uring/cmd: Pass compat mode in issue_flags
        commit: 66c87d5639f2f80421b3a01f12dcb7718f996093
[06/11] tools headers: Grab copy of io_uring.h
        commit: c36507ed1a2c2cb05c4a2aad9acb39ca5d7c12fe
[07/11] selftests/net: Extract uring helpers to be reusable
        commit: 11336afdd4141bbbd144b118a8a559b1993dc5d2
[08/11] io_uring/cmd: return -EOPNOTSUPP if net is disabled
        commit: d807234143872e460cdf851f1b2bbda2b427f95d
[09/11] io_uring/cmd: Introduce SOCKET_URING_OP_GETSOCKOPT
        commit: c3199f61b896cdef3664dc12729a2beadf322783
[10/11] io_uring/cmd: Introduce SOCKET_URING_OP_SETSOCKOPT
        commit: 43ad652250d24e9496f4cd6a0d670417807ac9a0
[11/11] selftests/bpf/sockopt: Add io_uring support
        commit: d9710f1d12a99738ff168e252ab8e9ffdeb90ed5

Best regards,