Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
In-Reply-To: <20210301214319.7e54c66f@oasis.local.home>
References: <20210301174749.1269154-1-swboyd@chromium.org> <20210301174749.1269154-6-swboyd@chromium.org> <20210301214319.7e54c66f@oasis.local.home>
Subject: Re: [PATCH 5/7] printk: Make %pS and friends print module build ID
From:   Stephen Boyd <swboyd@chromium.org>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        linux-kernel@vger.kernel.org, Jiri Olsa <jolsa@kernel.org>,
        Alexei Starovoitov <ast@kernel.org>,
        Jessica Yu <jeyu@kernel.org>,
        Evan Green <evgreen@chromium.org>,
        Hsin-Yi Wang <hsinyi@chromium.org>,
        Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        linux-doc@vger.kernel.org
To:     Steven Rostedt <rostedt@goodmis.org>
Date:   Tue, 02 Mar 2021 15:24:46 -0800
Message-ID: <161472748610.1478170.16522080242572382702@swboyd.mtv.corp.google.com>
User-Agent: alot/0.9.1
Precedence: bulk

Quoting Steven Rostedt (2021-03-01 18:43:19)
> On Mon,  1 Mar 2021 09:47:47 -0800
> Stephen Boyd <swboyd@chromium.org> wrote:
>=20
> > The %pS printk format (among some others) is used to print kernel
> > addresses symbolically. When the kernel prints an address inside of a
> > module, the kernel prints the addresses' symbol name along with the
> > module's name that contains the address. Let's make kernel stacktraces
> > easier to identify on KALLSYMS builds by including the build ID of a
> > module when we print the address.
>=20
> Please no!
>=20
> This kills the output of tracing with offset, and can possibly break
> scripts. I don't want to look at traces like this!
>=20
>           <idle>-0       [004] ..s1   353.842577: ipv4_conntrack_in+0x0/0=
x10 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_hook_slo=
w+0x40/0xb0
>           <idle>-0       [004] ..s1   353.842577: nf_conntrack_in+0x0/0x5=
c0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_hook_slow=
+0x40/0xb0
>           <idle>-0       [004] ..s1   353.842577: get_l4proto+0x0/0x190 [=
nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_conntrack_in+=
0x92/0x5c0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051)
>           <idle>-0       [004] ..s1   353.842577: nf_ct_get_tuple+0x0/0x2=
40 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_conntrack=
_in+0xec/0x5c0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051)
>           <idle>-0       [004] ..s1   353.842577: hash_conntrack_raw+0x0/=
0x170 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_conntr=
ack_in+0x28c/0x5c0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051)
>           <idle>-0       [004] ..s1   353.842578: __nf_conntrack_find_get=
.isra.0+0x0/0x2f0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051)=
 <-nf_conntrack_in+0x29d/0x5c0 [nf_conntrack] (3b39eb771b2566331887f671c741=
f90bfba0b051)
>           <idle>-0       [004] ..s1   353.842578: nf_conntrack_tcp_packet=
+0x0/0x1760 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_=
conntrack_in+0x3c8/0x5c0 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfb=
a0b051)
>           <idle>-0       [004] ..s2   353.842578: nf_ct_seq_offset+0x0/0x=
40 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_conntrack=
_tcp_packet+0x26d/0x1760 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfb=
a0b051)
>           <idle>-0       [004] ..s1   353.842578: __nf_ct_refresh_acct+0x=
0/0x50 [nf_conntrack] (3b39eb771b2566331887f671c741f90bfba0b051) <-nf_connt=
rack_tcp_packet+0x558/0x1760 [nf_conntrack] (3b39eb771b2566331887f671c741f9=
0bfba0b051)
>=20
>  NACK!
>=20

Heh, OK. Would adding a new printk format work then? From after the cut:

> Yet another alternative is to add a printk format like %pSB for
> backtrace prints. This would require a handful of architecture updates
> and I'm not sure it's worth the effort for that.
>

Or maybe take the approach of putting all the linked in module build IDs
in the "Modules linked in:" section? But as I said in the commit text
that can become quite verbose. Looking forward to your suggestions here.