ioctl: get_log_page by nvme uring cmd

[996refuse]

hello libnvme master,

io_uring on linux is pretty mature now, for the system with liburing, do the batch nvme_get_log by nvme uring cmd.
This change could gain about 10% performance improvement for some large log pages on my lab.

i believe the performance can be better if implement producer and consumer in two separated threads, if it's worth and acceptable, i can spend some time on this later.

thanks

[igaw]

I am fine with using uring but would it be possible to have it fallback during runtime to the ioctl interface when uring is not available, like on older kernels?

[igaw]

This should be just one commit. I'll squash the changes into one when I am reviewing and testing it. We need also to update our CI build accordingly.

[996refuse]

thanks for your help !

there're 2 failed testcase. it requires mock liburing api

 8/34 ana                                              FAIL            0.01s   killed by signal 6 SIGABRT
10/34 discovery                                        FAIL            0.01s   killed by signal 6 SIGABRT

besides, it doesn't work on SPDK NVME CUSE devices. FYI

[igaw]

there're 2 failed testcase. it requires mock liburing api

Ah, I've missed this. Unfortunately, this is important, I really don't want the tests cases to break when liburing is used.

besides, it doesn't work on SPDK NVME CUSE devices. FYI

That is likely due the missing support of the CUSE devices.

[igaw]

I suppose, for an initial attempt the io_uring_opcode_supported(probe, IORING_OP_URING_CMD) should just fail.

io_uring_opcode_supported calls __sys_io_uring_register, this means you need to mock the io_uring_register:

https://github.com/axboe/liburing/blob/6aeac2041987649c32a8f15de104ec4339725c49/src/arch/syscall-defs.h#L64C1-L69C2

static inline int __sys_io_uring_register(unsigned int fd, unsigned int opcode,
					  const void *arg, unsigned int nr_args)
{
	return (int) __do_syscall4(__NR_io_uring_register, fd, opcode, arg,
				   nr_args);
}

I haven't looked into liburing at all, maybe it supports some sort of mocking for testing?

[996refuse]

io_uring_get_probe returning NULL could avoid breaking the cases perfectly. after a quick study, think there are no official mocking support of liburing.

[igaw]

Just squashed the changes into one patch. Thanks!