linux-kernel - Re: [PATCH net-next v1] netdevsim: Forbid devlink reload when adding or deleting ports

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: <20210805121203.1ac615d4@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Date:   Thu, 5 Aug 2021 12:12:03 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Leon Romanovsky <leon@...nel.org>
Cc:     "David S . Miller" <davem@...emloft.net>,
        linux-kernel@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: [PATCH net-next v1] netdevsim: Forbid devlink reload when
 adding or deleting ports

On Thu, 5 Aug 2021 21:02:23 +0300 Leon Romanovsky wrote:
> > > As it should, given add/delete ports takes the port_list_lock which is
> > > destroyed by down but not (due to the forced failure) re-initialized by
> > > up.
> > > 
> > > If we want to handle adding ports while down we can just bump port
> > > count and return, although I don't think there's a practical need
> > > to support that.  
> > 
> > Sorry, but for me netdevsim looks like complete dumpster. 

I worry that netdevsim's gone unwieldy as a reflection of the quality of
the devlink APIs that got added, not by itself :/

> > It was intended for fast prototyping, but ended to be huge pile of
> > debugfs entries and selftest to execute random flows.

It's for selftests, IDK what fast prototyping is in terms of driver
APIs. Fast prototyping makes me think of the "it works" attitude which
is not sufficiently high bar for core APIs IMO, I'm sure you'll agree.

netdevsim was written specifically to be able to exercise HW APIs which
are implemented by small fraction of drivers. Especially offload APIs
as those can easily be broken by people changing the SW implementation
without capable HW at hand.

BTW I wonder if there is a term in human science of situation like when
a recent contributor tells the guy who wrote the code what the code was
intended for :)

> > Do you want me to move in_reload = false line to be after if (nsim_dev->fail_reload)
> > check?  
> 
> BTW, the current implementation where in_reload before if, actually
> preserves same behaviour as was with devlink_reload_enable() implementation.

Right, but I think as you rightly pointed out the current protection
of reload is broken. I'm not saying you must make it perfect or else..
just pointing out a gap you could address if you so choose.