linux-kernel - Re: [PATCH 1/2] i2c: virtio: disable timeout handling

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:   Tue, 19 Oct 2021 13:16:46 +0200
From:   Greg KH <gregkh@...uxfoundation.org>
To:     Viresh Kumar <viresh.kumar@...aro.org>
Cc:     Vincent Whitchurch <vincent.whitchurch@...s.com>, wsa@...nel.org,
        jie.deng@...el.com, virtualization@...ts.linux-foundation.org,
        linux-i2c@...r.kernel.org, linux-kernel@...r.kernel.org,
        kernel@...s.com
Subject: Re: [PATCH 1/2] i2c: virtio: disable timeout handling

On Tue, Oct 19, 2021 at 03:12:03PM +0530, Viresh Kumar wrote:
> On 19-10-21, 11:36, Greg KH wrote:
> > What is the "other side" here?  Is it something that you trust or not?
> 
> Other side can be a remote processor (for remoteproc over virtio or
> something similar), or traditionally it can be host OS or host
> firmware providing virtualisation to a Guest running Linux (this
> driver). Or something else..
> 
> I would incline towards "we trust the other side" here.

That's in contradition with what other people seem to think the virtio
drivers are for, see this crazy thread for details about that:
	https://lore.kernel.org/all/20211009003711.1390019-1-sathyanarayanan.kuppuswamy@linux.intel.com/

You can "trust" the hardware, but also handle things when hardware is
broken, which is most often the case in the real world.

So why is having a timeout a problem here?  If you have an overloaded
system, you want things to time out so that you can start to recover.

> > Usually we trust the hardware, but if you do not trust the hardware,
> > then yes, you need to have a timeout here.
> 
> The other side is the software that has access to the _Real_ hardware,
> and so we should trust it. So we can have a actually have a completion
> without timeout here, interesting.

And if that hardware stops working?  Timeouts are good to have, why not
just bump it up a bit if you are running into it in a real-world
situation?

thanks,

greg k-h