[PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[Jakub Kicinski]

When prog array is updated with bpftool users often refer
to the map via the ID.  Unfortunately, that's likely
to lead to confusion because prog arrays get flushed when
the last user reference is gone.  If there is no other
reference bpftool will create one, update successfully
just to close the map again and have it flushed.

Warn about this case in non-JSON mode.

If the problem continues causing confusion we can remove
the support for referring to a map by ID for prog array
update completely.  For now it seems like the potential
inconvenience to users who know what they're doing outweighs
the benefit.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
---
 tools/bpf/bpftool/map.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/bpf/bpftool/map.c b/tools/bpf/bpftool/map.c
index 8cb0e26907ff..6f33818bb6b6 100644
--- a/tools/bpf/bpftool/map.c
+++ b/tools/bpf/bpftool/map.c
@@ -426,6 +426,9 @@ static int parse_elem(char **argv, struct bpf_map_info *info,
 				p_err("not enough value arguments for map of progs");
 				return -1;
 			}
+			if (is_prefix(*argv, "id"))
+				p_info("Warning: updating program array via MAP_ID, make sure this map is kept open\n"
+				       "         by some process or pinned otherwise update will be lost");
 
 			fd = prog_parse_fd(&argc, &argv);
 			if (fd < 0)
-- 
2.19.2

Re: [oss-drivers] [PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[Quentin Monnet]

On Mon, 28 Jan 2019 at 18:29, Jakub Kicinski
<jakub.kicinski@netronome.com> wrote:
>
> When prog array is updated with bpftool users often refer
> to the map via the ID.  Unfortunately, that's likely
> to lead to confusion because prog arrays get flushed when
> the last user reference is gone.  If there is no other
> reference bpftool will create one, update successfully
> just to close the map again and have it flushed.
>
> Warn about this case in non-JSON mode.
>
> If the problem continues causing confusion we can remove
> the support for referring to a map by ID for prog array
> update completely.  For now it seems like the potential
> inconvenience to users who know what they're doing outweighs
> the benefit.
>
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>

All good for me!
Quentin

Re: [oss-drivers] [PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[Quentin Monnet]

2019-01-28 18:51 UTC+0000 ~ Quentin Monnet <quentin.monnet@netronome.com>
> On Mon, 28 Jan 2019 at 18:29, Jakub Kicinski
> <jakub.kicinski@netronome.com> wrote:
>>
>> When prog array is updated with bpftool users often refer
>> to the map via the ID.  Unfortunately, that's likely
>> to lead to confusion because prog arrays get flushed when
>> the last user reference is gone.  If there is no other
>> reference bpftool will create one, update successfully
>> just to close the map again and have it flushed.
>>
>> Warn about this case in non-JSON mode.
>>
>> If the problem continues causing confusion we can remove
>> the support for referring to a map by ID for prog array
>> update completely.  For now it seems like the potential
>> inconvenience to users who know what they're doing outweighs
>> the benefit.
>>
>> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
>> Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
> 
> All good for me!
> Quentin
> 

[Hmm. I read that on GMail interface and thought it was a repost on our 
internal ML. Sorry for the noise.]

Re: [PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[Song Liu]

On Mon, Jan 28, 2019 at 10:30 AM Jakub Kicinski
<jakub.kicinski@netronome.com> wrote:
>
> When prog array is updated with bpftool users often refer
> to the map via the ID.  Unfortunately, that's likely
> to lead to confusion because prog arrays get flushed when
> the last user reference is gone.  If there is no other
> reference bpftool will create one, update successfully
> just to close the map again and have it flushed.
>
> Warn about this case in non-JSON mode.
>
> If the problem continues causing confusion we can remove
> the support for referring to a map by ID for prog array
> update completely.  For now it seems like the potential
> inconvenience to users who know what they're doing outweighs
> the benefit.
>
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>

Acked-by: Song Liu <songliubraving@fb.com>

> ---
>  tools/bpf/bpftool/map.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/tools/bpf/bpftool/map.c b/tools/bpf/bpftool/map.c
> index 8cb0e26907ff..6f33818bb6b6 100644
> --- a/tools/bpf/bpftool/map.c
> +++ b/tools/bpf/bpftool/map.c
> @@ -426,6 +426,9 @@ static int parse_elem(char **argv, struct bpf_map_info *info,
>                                 p_err("not enough value arguments for map of progs");
>                                 return -1;
>                         }
> +                       if (is_prefix(*argv, "id"))
> +                               p_info("Warning: updating program array via MAP_ID, make sure this map is kept open\n"
> +                                      "         by some process or pinned otherwise update will be lost");
>
>                         fd = prog_parse_fd(&argc, &argv);
>                         if (fd < 0)
> --
> 2.19.2
>

Re: [PATCH bpf-next] tools: bpftool: warn about risky prog array updates

[Daniel Borkmann]

On 01/28/2019 07:29 PM, Jakub Kicinski wrote:
> When prog array is updated with bpftool users often refer
> to the map via the ID.  Unfortunately, that's likely
> to lead to confusion because prog arrays get flushed when
> the last user reference is gone.  If there is no other
> reference bpftool will create one, update successfully
> just to close the map again and have it flushed.
> 
> Warn about this case in non-JSON mode.
> 
> If the problem continues causing confusion we can remove
> the support for referring to a map by ID for prog array
> update completely.  For now it seems like the potential
> inconvenience to users who know what they're doing outweighs
> the benefit.
> 
> Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
> Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>

Applied, thanks!