[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230329072553.GH4253@hirez.programming.kicks-ass.net>
Date: Wed, 29 Mar 2023 09:25:53 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Josh Poimboeuf <jpoimboe@...nel.org>
Cc: x86@...nel.org, linux-kernel@...r.kernel.org,
Miroslav Benes <mbenes@...e.cz>
Subject: Re: [PATCH 2/5] objtool: Combine '--backtrace' with '--verbose'
On Tue, Mar 28, 2023 at 01:19:21PM -0700, Josh Poimboeuf wrote:
> On Tue, Mar 28, 2023 at 10:07:43AM +0200, Peter Zijlstra wrote:
> > On Mon, Mar 27, 2023 at 09:00:45AM -0700, Josh Poimboeuf wrote:
> > > Get rid of the '--backtrace' option, instead including that
> > > functionality in '--verbose'. This makes it easy to gather all the
> > > information needed for diagnosing objtool warnings.
> >
> > Hurmm.. can't we have verbose imply backtrace but keep the separate
> > option? I'm not sure if I always want the objdump thing -- esp with
> > multiple warnings on vmlinux that's going to be really slow -- better to
> > dump the whole of vmlinux.o once at the end.
>
> That's a good point, vmlinux would be unbearable for multiple warnings.
>
> We could accumulate a list of affected functions and then supply that to
> objdump-func.sh at the end and dump them all at the same time.
>
> objdump-func.sh would need to be changed to look for multiple funcs at
> once.
>
> If I do that, do you still want the separate backtrace option?
For now, lets keep the orthogonal options -- we can always remove it
later if we find they go unused. This automagic stuff is going to take a
bit of getting used to I recon ;-)
Powered by blists - more mailing lists