linux-kernel - Re: [PATCH] kgdb: Avoid suspicious RCU usage warning

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20200603120051.dxpavvsxvsxnvuct@holly.lan>
Date:   Wed, 3 Jun 2020 13:00:51 +0100
From:   Daniel Thompson <daniel.thompson@...aro.org>
To:     Doug Anderson <dianders@...omium.org>
Cc:     Jason Wessel <jason.wessel@...driver.com>,
        Sumit Garg <sumit.garg@...aro.org>,
        kgdb-bugreport@...ts.sourceforge.net,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] kgdb: Avoid suspicious RCU usage warning

On Tue, Jun 02, 2020 at 03:56:33PM -0700, Doug Anderson wrote:
> > > 2. Perhaps remove the whole irq saving / restoring in kgdb_cpu_enter().
> >
> > Are you feeling lucky?
> >
> > I think there will come a time when bravery is called for but I'd rather
> > see this as part of a bigger rewrite instead of a single high risk
> > change.
> 
> Hrm, maybe.  I guess it depends on whether we want to take baby steps
> there or try to do it all at once.  If we take baby steps we will
> occasionally fall down but we'll slowly start getting things cleaned
> up.  If we wait for a full rewrite then we might be waiting for a long
> time.  It'll also be harder to figure out which of the big changes in
> the major rewrite broken someone.  ...or if the major rewrite comes in
> 20 small/bisectable patches it may be hard to revert patch 2 out of 20
> if the future patches all build upon it.  If we do one small high-risk
> change and then wait before building upon it then it'll be easy for
> someone to bisect and then yell for a revert.

My views are a bit too nuanced for me to agree or disagree with this.
I'm not against baby steps and I definitely *don't* want kgdb to
continue to be preserved in aspic.

However I'm still reluctant to start our baby steps with a "let's see
if this breaks something" patch given we know it could be a very large
number of kernel cycles before we get an answer. I would be much
happier if those baby steps started, for example, with refactoring to
decompose the beast into clearer (and dare I say better documented)
functions.

Or put another way, even if someone sent me 20 small bisectable patches
in a single kernel cycle I'd still want the high risk bits to be
towards the end of the patch set.

Daniel.