[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20231004092205.02c8eb0b@kernel.org>
Date: Wed, 4 Oct 2023 09:22:05 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Johannes Berg <johannes@...solutions.net>
Cc: linux-kernel@...r.kernel.org, linux-trace-kernel@...r.kernel.org,
netdev@...r.kernel.org, linux-wireless@...r.kernel.org
Subject: Re: [PATCH 0/4] tracing: improve symbolic printing
On Thu, 21 Sep 2023 10:51:30 +0200 Johannes Berg wrote:
> So I was frustrated with not seeing the names of SKB dropreasons
> for all but the core reasons, and then while looking into this
> all, realized, that the current __print_symbolic() is pretty bad
> anyway.
>
> So I came up with a new approach, using a separate declaration
> of the symbols, and __print_sym() in there, but to userspace it
> all doesn't matter, it shows it the same way, just dyamically
> instead of munging with the strings all the time.
>
> This is a huge .data savings as far as I can tell, with a modest
> amount (~4k) of .text addition, while making it all dynamic and
> in the SKB dropreason case even reusing the existing list that
> dropmonitor uses today. Surely patch 3 isn't needed here, but it
> felt right.
>
> Anyway, I think it's a pretty reasonable approach overall, and
> it does works.
>
> I've listed a number of open questions in the first patch since
> that's where the real changes for this are.
Potentially naive question - the trace point holds enum skb_drop_reason.
The user space can get the names from BTF. Can we not teach user space
to generically look up names of enums in BTF?
Powered by blists - more mailing lists