[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190430153932.7d741d4d@cakuba>
Date: Tue, 30 Apr 2019 15:39:32 -0400
From: Jakub Kicinski <jakub.kicinski@...ronome.com>
To: Y Song <ys114321@...il.com>
Cc: Quentin Monnet <quentin.monnet@...ronome.com>,
Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
bpf <bpf@...r.kernel.org>, netdev <netdev@...r.kernel.org>,
oss-drivers@...ronome.com
Subject: Re: [PATCH bpf-next 1/6] tools: bpftool: add --log-libbpf option to
get debug info from libbpf
On Tue, 30 Apr 2019 08:31:53 -0700, Y Song wrote:
> On Tue, Apr 30, 2019 at 2:34 AM Quentin Monnet
> <quentin.monnet@...ronome.com> wrote:
> >
> > Hi Yonghong,
> >
> > 2019-04-29 16:32 UTC-0700 ~ Y Song <ys114321@...il.com>
> > > On Mon, Apr 29, 2019 at 2:53 AM Quentin Monnet
> > > <quentin.monnet@...ronome.com> wrote:
> > >>
> > >> libbpf has three levels of priority for output: warn, info, debug. By
> > >> default, debug output is not printed to stderr.
> > >>
> > >> Add a new "--log-libbpf LOG_LEVEL" option to bpftool to provide more
> > >> flexibility on the log level for libbpf. LOG_LEVEL is a comma-separated
> > >> list of levels of log to print ("warn", "info", "debug"). The value
> > >> corresponding to the default behaviour would be "warn,info".
> > >
> > > Do you think option like "warn,debug" will be useful for bpftool users?
> > > Maybe at bpftool level, we could allow user only to supply minimum level
> > > for log output, e.g., "info" will output "warn,info"?
> > I've been pondering this, too. Since we allow to combine all levels for
> > the verifier logs it feels a bit odd to be less flexible for libbpf. And
> > we could imagine a user who wants verifier logs (so libbpf "debug") but
> > prefers to limit libbpf output (so no "info")... Although I admit this
> > might be a bit far-fetched.
> >
> > I can resend a version with the option taking only the minimal log
> > level, as you describe, if you think this is best.
>
> Thanks, I think providing a single minimum level for output probably
> better.
I have a weak preference for what we have here, because it's similar to
the kernel bit opt in (log level, stats etc)..
Powered by blists - more mailing lists