netdev - RE: [PATCH net-next] net/tls: Mark the end in scatterlist table

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: <DB7PR04MB4252535BC972A7555E74A0DF8B2C0@DB7PR04MB4252.eurprd04.prod.outlook.com>
Date:   Thu, 2 Aug 2018 17:23:27 +0000
From:   Vakul Garg <vakul.garg@....com>
To:     Dave Watson <davejwatson@...com>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "borisp@...lanox.com" <borisp@...lanox.com>,
        "aviadye@...lanox.com" <aviadye@...lanox.com>,
        "davem@...emloft.net" <davem@...emloft.net>
Subject: RE: [PATCH net-next] net/tls: Mark the end in scatterlist table



> -----Original Message-----
> From: Dave Watson [mailto:davejwatson@...com]
> Sent: Thursday, August 2, 2018 10:47 PM
> To: Vakul Garg <vakul.garg@....com>
> Cc: netdev@...r.kernel.org; borisp@...lanox.com;
> aviadye@...lanox.com; davem@...emloft.net
> Subject: Re: [PATCH net-next] net/tls: Mark the end in scatterlist table
> 
> On 08/02/18 05:05 PM, Vakul Garg wrote:
> > In case zerocopy_from_iter() fails, 'end' won't get marked.
> > So fallback path is fine.
> >
> > > Which codepath is calling sg_nents()?
> >
> > While testing my WIP implementation of combined dynamic memory
> > allocation for (aead_req || sgin || sgout || aad || iv), I was getting
> random kernel crashes.
> > To debug it I had inserted sg_nents() in my code. The KASAN then
> > immediately complained that sg_nents() went beyond the allocated
> memory for scatterlist.
> > This led me to find that scatterlist table end has not been marked.
> >
> 
> If this isn't causing KASAN issues for the existing code, it probably makes
> more sense to put in a future series with that WIP work then.

Isn't using a sgtable with unmarked end already bad and should be fixed?
Crypto hardware accelerator drivers could be broken while using sg lists with
unmarked end.