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: <YHWthzF+l0iacMuw@alley>
References: <20210410015300.3764485-1-swboyd@chromium.org> <20210410015300.3764485-5-swboyd@chromium.org> <YHWthzF+l0iacMuw@alley>
Subject: Re: [PATCH v4 04/13] dump_stack: Add vmlinux build ID to stack traces
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>,
        Steven Rostedt <rostedt@goodmis.org>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Matthew Wilcox <willy@infradead.org>
To:     Petr Mladek <pmladek@suse.com>
Date:   Tue, 13 Apr 2021 13:36:48 -0700
Message-ID: <161834620810.3764895.16783352875023260799@swboyd.mtv.corp.google.com>
User-Agent: alot/0.9.1
Precedence: bulk

Quoting Petr Mladek (2021-04-13 07:41:11)
> > diff --git a/lib/dump_stack.c b/lib/dump_stack.c
> > index f5a33b6f773f..d685331b065f 100644
> > --- a/lib/dump_stack.c
> > +++ b/lib/dump_stack.c
> > @@ -5,6 +5,7 @@
> >   */
> > =20
> >  #include <linux/kernel.h>
> > +#include <linux/buildid.h>
> >  #include <linux/export.h>
> >  #include <linux/sched.h>
> >  #include <linux/sched/debug.h>
> > @@ -36,6 +37,14 @@ void __init dump_stack_set_arch_desc(const char *fmt=
, ...)
> >       va_end(args);
> >  }
> > =20
> > +#if IS_ENABLED(CONFIG_STACKTRACE_BUILD_ID)
> > +#define BUILD_ID_FMT " %20phN"
> > +#define BUILD_ID_VAL vmlinux_build_id
> > +#else
> > +#define BUILD_ID_FMT "%s"
> > +#define BUILD_ID_VAL ""
> > +#endif
>=20
> 3rd patch always defines and initializes vmlinux_build_id. But it is
> used only when CONFIG_STACKTRACE_BUILD_ID is enabled.

It is also used for crash code.

> Is it intentional, please?

Yes, mostly for simplicity with the other user.

>=20
> It is not a big deal for vmlinux_build_id. But it is more questionable
> for the per-module id. I am going to open this question for 5th patch
> as well.
>=20

Right, for the vmlinux_build_id symbol it is not exported, and the whole
buildid.c file is part of lib-y, so if the symbol isn't used the linker
should drop it during link phase. I can drop the early init call if the
config is disabled and crash kernel code isn't enabled, and then rely on
the linker to drop the vmlinux_build_id symbol. Let me see if that can
work so that we don't have to parse it at boot if it is never used.