Re: [PATCH 12/23] bpf: add IOURING program type

[Pavel Begunkov <asml.silence@gmail.com>]

On 5/21/21 12:34 AM, Song Liu wrote:
>> On May 19, 2021, at 7:13 AM, Pavel Begunkov <asml.silence@gmail.com> wrote:
>>
>> Draft a new program type BPF_PROG_TYPE_IOURING, which will be used by
>> io_uring to execute BPF-based requests.
>>
>> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
>> ---
>> fs/io_uring.c             | 21 +++++++++++++++++++++
>> include/linux/bpf_types.h |  2 ++
>> include/uapi/linux/bpf.h  |  1 +
>> kernel/bpf/syscall.c      |  1 +
>> kernel/bpf/verifier.c     |  5 ++++-
>> 5 files changed, 29 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/io_uring.c b/fs/io_uring.c
>> index 1a4c9e513ac9..882b16b5e5eb 100644
[...]
>> +BPF_PROG_TYPE(BPF_PROG_TYPE_IOURING, bpf_io_uring,
>> +	      void *, void *)
>>
>> BPF_MAP_TYPE(BPF_MAP_TYPE_ARRAY, array_map_ops)
>> BPF_MAP_TYPE(BPF_MAP_TYPE_PERCPU_ARRAY, percpu_array_map_ops)
>> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> index 4ba4ef0ff63a..de544f0fbeef 100644
>> --- a/include/uapi/linux/bpf.h
>> +++ b/include/uapi/linux/bpf.h
>> @@ -206,6 +206,7 @@ enum bpf_prog_type {
>> 	BPF_PROG_TYPE_EXT,
>> 	BPF_PROG_TYPE_LSM,
>> 	BPF_PROG_TYPE_SK_LOOKUP,
>> +	BPF_PROG_TYPE_IOURING,
>> };
>>
>> enum bpf_attach_type {
>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> index 250503482cda..6ef7a26f4dc3 100644
>> --- a/kernel/bpf/syscall.c
>> +++ b/kernel/bpf/syscall.c
>> @@ -2041,6 +2041,7 @@ static bool is_net_admin_prog_type(enum bpf_prog_type prog_type)
>> 	case BPF_PROG_TYPE_CGROUP_SOCKOPT:
>> 	case BPF_PROG_TYPE_CGROUP_SYSCTL:
>> 	case BPF_PROG_TYPE_SOCK_OPS:
>> +	case BPF_PROG_TYPE_IOURING:
>> 	case BPF_PROG_TYPE_EXT: /* extends any prog */
>> 		return true;
>> 	case BPF_PROG_TYPE_CGROUP_SKB:
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index 0399ac092b36..2a53f44618a7 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -8558,6 +8558,9 @@ static int check_return_code(struct bpf_verifier_env *env)
>> 	case BPF_PROG_TYPE_SK_LOOKUP:
>> 		range = tnum_range(SK_DROP, SK_PASS);
>> 		break;
>> +	case BPF_PROG_TYPE_IOURING:
>> +		range = tnum_const(0);
>> +		break;
>> 	case BPF_PROG_TYPE_EXT:
>> 		/* freplace program can return anything as its return value
>> 		 * depends on the to-be-replaced kernel func or bpf program.
>> @@ -12560,7 +12563,7 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
>> 	u64 key;
>>
>> 	if (prog->aux->sleepable && prog->type != BPF_PROG_TYPE_TRACING &&
>> -	    prog->type != BPF_PROG_TYPE_LSM) {
>> +	    prog->type != BPF_PROG_TYPE_LSM && prog->type != BPF_PROG_TYPE_IOURING) {
> 
> Is IOURING program sleepable? If so, please highlight that in the commit log 

Supposed to work with both, sleepable and not, but with different set
of helpers. e.g. can't submit requests nor do bpf_copy_from_user() if
can't sleep. The only other difference in handling is rcu around
non-sleepable, but please shut out if I forgot anything

> and update the warning below. 

Sure

> 
>> 		verbose(env, "Only fentry/fexit/fmod_ret and lsm programs can be sleepable\n");
>> 		return -EINVAL;
>> 	}
>> -- 
>> 2.31.1
>>
> 

-- 
Pavel Begunkov