Re: [PATCH] bpftool: Query only cgroup-related attach types

[Quentin Monnet <qmo@kernel.org>]

On 29/05/2024 14:10, Kenta Tada wrote:
> When CONFIG_NETKIT=y,
> bpftool-cgroup shows error even if the cgroup's path is correct:
> 
> $ bpftool cgroup tree /sys/fs/cgroup
> CgroupPath
> ID       AttachType      AttachFlags     Name
> Error: can't query bpf programs attached to /sys/fs/cgroup: No such device or address
> 
> From strace and kernel tracing, I found netkit returned ENXIO and this command failed.
> I think this AttachType(BPF_NETKIT_PRIMARY) is not relevant to cgroup.
> 
> bpftool-cgroup should query just only cgroup-related attach types.
> 
> Signed-off-by: Kenta Tada <tadakentaso@gmail.com>
> ---
>  tools/bpf/bpftool/cgroup.c | 47 +++++++++++++++++++++++++++++++++-----
>  1 file changed, 41 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/bpf/bpftool/cgroup.c b/tools/bpf/bpftool/cgroup.c
> index af6898c0f388..bb2703aa4756 100644
> --- a/tools/bpf/bpftool/cgroup.c
> +++ b/tools/bpf/bpftool/cgroup.c
> @@ -19,6 +19,39 @@
>  
>  #include "main.h"
>  
> +static const bool cgroup_attach_types[] = {
> +	[BPF_CGROUP_INET_INGRESS] = true,
> +	[BPF_CGROUP_INET_EGRESS] = true,
> +	[BPF_CGROUP_INET_SOCK_CREATE] = true,
> +	[BPF_CGROUP_INET_SOCK_RELEASE] = true,
> +	[BPF_CGROUP_INET4_BIND] = true,
> +	[BPF_CGROUP_INET6_BIND] = true,
> +	[BPF_CGROUP_INET4_POST_BIND] = true,
> +	[BPF_CGROUP_INET6_POST_BIND] = true,
> +	[BPF_CGROUP_INET4_CONNECT] = true,
> +	[BPF_CGROUP_INET6_CONNECT] = true,
> +	[BPF_CGROUP_UNIX_CONNECT] = true,
> +	[BPF_CGROUP_INET4_GETPEERNAME] = true,
> +	[BPF_CGROUP_INET6_GETPEERNAME] = true,
> +	[BPF_CGROUP_UNIX_GETPEERNAME] = true,
> +	[BPF_CGROUP_INET4_GETSOCKNAME] = true,
> +	[BPF_CGROUP_INET6_GETSOCKNAME] = true,
> +	[BPF_CGROUP_UNIX_GETSOCKNAME] = true,
> +	[BPF_CGROUP_UDP4_SENDMSG] = true,
> +	[BPF_CGROUP_UDP6_SENDMSG] = true,
> +	[BPF_CGROUP_UNIX_SENDMSG] = true,
> +	[BPF_CGROUP_UDP4_RECVMSG] = true,
> +	[BPF_CGROUP_UDP6_RECVMSG] = true,
> +	[BPF_CGROUP_UNIX_RECVMSG] = true,
> +	[BPF_CGROUP_SOCK_OPS] = true,
> +	[BPF_CGROUP_DEVICE] = true,
> +	[BPF_CGROUP_SYSCTL] = true,
> +	[BPF_CGROUP_GETSOCKOPT] = true,
> +	[BPF_CGROUP_SETSOCKOPT] = true,
> +	[BPF_LSM_CGROUP] = true,
> +	[__MAX_BPF_ATTACH_TYPE] = false,
> +};


Thanks for this!

I can't say I'm glad to see another version of the list of
cgroup-related attach types (in addition to HELP_SPEC_ATTACH_TYPES and
to the manual page). But the alternative would be to explicitly skip
BPF_NETKIT_PRIMARY, which is not great, either. Too bad we don't have a
way to check whether the type is cgroup-related in libbpf or from the
bpf.h headers; but I don't think there's much interest to add it there,
so we'll probably have the array. We should account for it in
tools/testing/selftests/bpf/test_bpftool_synctypes.py, but I can do this
as a follow-up if you don't feel like messing up with the Python script.


> +
>  #define HELP_SPEC_ATTACH_FLAGS						\
>  	"ATTACH_FLAGS := { multi | override }"
>  
> @@ -187,14 +220,16 @@ static int cgroup_has_attached_progs(int cgroup_fd)
>  	bool no_prog = true;
>  
>  	for (type = 0; type < __MAX_BPF_ATTACH_TYPE; type++) {
> -		int count = count_attached_bpf_progs(cgroup_fd, type);
> +		if (cgroup_attach_types[type]) {


Please change here:

                int count;

                if (!cgroup_attach_types[type])
                        continue;

And no need to further indent the rest of the block.


> +			int count = count_attached_bpf_progs(cgroup_fd, type);
>  
> -		if (count < 0 && errno != EINVAL)
> -			return -1;
> +			if (count < 0 && errno != EINVAL)
> +				return -1;
>  
> -		if (count > 0) {
> -			no_prog = false;
> -			break;
> +			if (count > 0) {
> +				no_prog = false;
> +				break;
> +			}
>  		}
>  	}
>