netdev - Re: traceability of wifi packet drops

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: <d21a2b2cd2e83d91173dc8952871fc734fa40155.camel@sipsolutions.net>
Date:   Tue, 28 Mar 2023 10:18:51 +0200
From:   Johannes Berg <johannes@...solutions.net>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     Jakub Kicinski <kuba@...nel.org>, netdev@...r.kernel.org,
        linux-wireless@...r.kernel.org
Subject: Re: traceability of wifi packet drops

On Tue, 2023-03-28 at 01:16 -0700, Eric Dumazet wrote:
> 
> About visibility: Even before 'drop reasons' developers can always use
> the call graph .
> 
> perf record -a -g -e skb:kfree_skb ...

Yeah, except right now it doesn't work in mac80211 because we have these
"RX handlers" that just return a status, and then a single place drops
the skb.

> Really drop reasons are nice when you want filtering and convenience.
> But they are a lot of work to add/maintain.
> This all comes at a cost (both maintenance but also run time cost
> because we need to propagate reasons )

Right ... I don't think the runtime cost would be much of an issue here
(we just get them as a return value from one function to pass to
kfree_skb), but the maintenance is hard.

I guess we could also restructure the code to call kfree_skb() in all
those places that want drop it and then we have the call trace, but I
doubt that would actually be cheaper both in the maintenance sense
(always need to pay attention to it) or in the runtime sense.

johannes