netdev - Re: [PATCH bpf-next] bpftool: Exit early when it's not possible to dump a REUSEPORT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <240116cf-0e74-e9fb-5100-00965c578e72@netronome.com>
Date:   Thu, 11 Apr 2019 12:20:02 +0100
From:   Quentin Monnet <quentin.monnet@...ronome.com>
To:     Benjamin Poirier <bpoirier@...e.com>,
        Daniel Borkmann <daniel@...earbox.net>
Cc:     netdev@...r.kernel.org, bpf@...r.kernel.org
Subject: Re: [PATCH bpf-next] bpftool: Exit early when it's not possible to
 dump a REUSEPORT_SOCKARRAY map

2019-04-11 17:27 UTC+0900 ~ Benjamin Poirier <bpoirier@...e.com>
> avoids outputting a series of
> 	value:
> 	No space left on device
> 
> The value itself is not wrong but bpf_fd_reuseport_array_lookup_elem() can
> only return it if the map was created with value_size = 8. There's nothing
> bpftool can do about it. Instead of repeating this error for every key in
> the map, print the error once, print an explanatory message and exit.
> 
> Signed-off-by: Benjamin Poirier <bpoirier@...e.com>
> ---
> 
> Note that this will lead to a merge conflict if the patch "bpftool: Fix
> errno variable usage" is merged in the bpf branch:
> 
> <<<<<<< HEAD
>                 if (lookup_errno == ENOENT)
> =======
>                 if (errno == ENOENT) {
>>>>>>>> bpf-next
> 
> The resolution should be trivial:
>                 if (lookup_errno == ENOENT) {
> 
>  tools/bpf/bpftool/map.c | 21 +++++++++++++++------
>  1 file changed, 15 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/bpf/bpftool/map.c b/tools/bpf/bpftool/map.c
> index e96903078991..bb648e42923c 100644
> --- a/tools/bpf/bpftool/map.c
> +++ b/tools/bpf/bpftool/map.c
> @@ -684,7 +684,6 @@ static int dump_map_elem(int fd, void *key, void *value,
>  			 struct bpf_map_info *map_info, struct btf *btf,
>  			 json_writer_t *btf_wtr)
>  {
> -	int num_elems = 0;
>  	int lookup_errno;
>  
>  	if (!bpf_map_lookup_elem(fd, key, value)) {
> @@ -702,9 +701,8 @@ static int dump_map_elem(int fd, void *key, void *value,
>  			} else {
>  				print_entry_plain(map_info, key, value);
>  			}
> -			num_elems++;
>  		}
> -		return num_elems;
> +		return 1;
>  	}
>  
>  	/* lookup error handling */
> @@ -722,11 +720,13 @@ static int dump_map_elem(int fd, void *key, void *value,
>  		jsonw_string_field(json_wtr, "error", strerror(lookup_errno));
>  		jsonw_end_object(json_wtr);
>  	} else {
> -		if (errno == ENOENT)
> +		if (errno == ENOENT) {
>  			print_entry_plain(map_info, key, NULL);
> -		else
> +		} else {
>  			print_entry_error(map_info, key,
>  					  strerror(lookup_errno));
> +			return -lookup_errno;

Hi, thanks! One comment: if we return -lookup_errno here...

> +		}
>  	}
>  
>  	return 0;
> @@ -787,7 +787,16 @@ static int do_dump(int argc, char **argv)
>  				err = 0;
>  			break;
>  		}
> -		num_elems += dump_map_elem(fd, key, value, &info, btf, btf_wtr);
> +		err = dump_map_elem(fd, key, value, &info, btf, btf_wtr);
> +		/* bpf_fd_reuseport_array_lookup_elem() can only return a
> +		 * value if the map's value_size == 8
> +		 */
> +		if (info.type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY &&
> +		    info.value_size != 8 && err == -ENOSPC) {
> +			p_err("cannot dump REUSEPORT_SOCKARRAY map with value_size != 8");
> +			goto exit_free;
> +		}
> +		num_elems += err;

... I think we can have a negative err value here (if we're not in the
case of reuseport_sockarray with wrong map size)? In that case, the
function could print something like "Found -1 elements"?

So maybe update num_elems only if err is positive? Although I'd like
even better if we could somehow find a way to move this error handling
along with the rest of it in dump_map_elem(), and avoid making that
function return either a number of elements or an error depending on the
result from the lookup.

Best regards,
Quentin