linux-kernel - Re: [PATCH v2 0/6] Improve visibility of writeback

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]

Date: Thu, 28 Mar 2024 15:36:42 -0400
From: Kent Overstreet <kent.overstreet@...ux.dev>
To: Andrew Morton <akpm@...ux-foundation.org>
Cc: Kemeng Shi <shikemeng@...weicloud.com>, willy@...radead.org, 
	jack@...e.cz, bfoster@...hat.com, tj@...nel.org, dsterba@...e.com, 
	mjguzik@...il.com, dhowells@...hat.com, linux-kernel@...r.kernel.org, 
	linux-mm@...ck.org, linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH v2 0/6] Improve visibility of writeback

On Thu, Mar 28, 2024 at 12:23:52PM -0700, Andrew Morton wrote:
> On Thu, 28 Mar 2024 15:15:03 -0400 Kent Overstreet <kent.overstreet@...ux.dev> wrote:
> 
> > On Wed, Mar 27, 2024 at 10:40:10AM -0700, Andrew Morton wrote:
> > > On Wed, 27 Mar 2024 23:57:45 +0800 Kemeng Shi <shikemeng@...weicloud.com> wrote:
> > > 
> > > > This series tries to improve visilibity of writeback.
> > > 
> > > Well...  why?  Is anyone usefully using the existing instrumentation? 
> > > What is to be gained by expanding it further?  What is the case for
> > > adding this code?
> > > 
> > > I don't recall hearing of anyone using the existing debug
> > > instrumentation so perhaps we should remove it!
> > 
> > Remove debug instrumentation!? Surely you just?
> 
> Absolutely not.  Any code in the kernel should have ongoing
> justification for remaining there.  If no such justification exists,
> out it goes.

Certainly, but this isn't remotely a case where I'd expect to be getting
that kind of feedback. Debugging instrumentation is very much a case
where no one notices it 99% of the time, but when you need it you
_really_ need it.

Not having it can turn a 10 minute "oh, that thing is acting wonky -
it's because your system is overloaded/your drive is wonky/x subsystem
sucks, we know about it and we're working on it" into a weeklong
bughunt, burning up expensive engineer time pointlessly.

To debug complex systems efficiently, in production, in the wild, we
need to be able to see what's going on - we need more of this stuff.

Not to say that this couldn't use more work - perhaps additional focus
on what kinds of issues we expect to need to debug with this, what the
numbers mean and are useful for, documentation on how this relates to
writeback internals, etc.