linux-kernel - Re: SLUB: purpose of sysfs events on cache creation/removal

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: <20191127162400.GT20912@dhcp22.suse.cz>
Date:   Wed, 27 Nov 2019 17:24:00 +0100
From:   Michal Hocko <mhocko@...nel.org>
To:     Christopher Lameter <cl@...ux.com>
Cc:     LKML <linux-kernel@...r.kernel.org>, linux-mm@...ck.org
Subject: Re: SLUB: purpose of sysfs events on cache creation/removal

On Wed 27-11-19 15:40:19, Cristopher Lameter wrote:
> On Tue, 26 Nov 2019, Michal Hocko wrote:
> 
> > > I have no idea about what this is.
> >
> > It seems to be there since the initial merge. I suspect this is just
> > following a generic sysfs rule that each file has to provide those
> > events?
> 
> I have never heard of anyone using this.
> 
> > > There have been many people who
> > > reworked the sysfs support and this has been the cause for a lot of
> > > breakage over the years.
> >
> > Remember any specifics?
> 
> The sequencing of setup / teardown of sysfs entries has frequently been
> a problem and that caused numerous issues with slab initialization as well
> as kmem cache creation. Initially kmalloc DMA caches were created on
> demand which caused some issues. Then there was the back and forth with
> cache aliasing during kmem_cache_create() that caused another set of
> instabilities.
> 
> > I am mostly interested in potential users. In other words I am thinking
> > to suppress those events. There is already ke knob to control existence
> > of memcg caches but I do not see anything like this for root caches.
> >
> 
> I am not aware of any users but the deployments of Linux are so diverse
> these days that I am not sure that there are no users.

Would you mind a patch that would add a kernel command line parameter
that would work like memcg_sysfs_enabled? The default for the config
would be on. Or it would be preferrable to simply drop only events?
-- 
Michal Hocko
SUSE Labs