linux-kernel - Re: ARM64 kernel crash

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] [day] [month] [year] [list]

Date:   Wed, 15 Dec 2021 12:29:25 +0000
From:   Robin Murphy <robin.murphy@....com>
To:     Mark Rutland <mark.rutland@....com>, tjcw@...tab.net
Cc:     linux-kernel@...r.kernel.org, Chris Ward <tjcw01@...il.com>,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: ARM64 kernel crash

On 2021-12-15 10:53, Mark Rutland wrote:
> Hi Chris,
> 
> On Wed, Dec 15, 2021 at 07:36:34AM +0000, Chris Ward wrote:
>> Please personally cc me on answers/comments as I am not currently
>> subscribed to the LKML.
>> Trying again because I messed up the format badly the first time
>> round. Apologies for that.
> 
> For arm/arm64 stuff, it's best to also Cc LAKML
> (linux-arm-kernel@...ts.infradead.org). I've added that here, and for the
> benefit of others I've kept your entire mail as context. In future replies I
> will trim this.
> 
> Please also note that kernel mailing list style is to reply inline (as I've don
> with this mail) rather than to top-post. That helps to keep discussions
> legible, and makes it possible to trim irrelevant context. Please keep any
> replies in kind with that style.
> 
>> My team has a problem which is being bounced between Canonical support
>> and Xilinx support.
>> We are using kernel 5.4.0-xilinx-v2020.2 built from sources under
>> https://github.com/Xilinx/linux-xlnx with a Ubuntu 20.04 userland on
>> an ARM64 embedded linux machine (i.e. not x86-64).
> 
> Can you share the config used? e.g. was that `defconfig`, or something else?
> 
>> When trying to set up a file system on a ramdisk, we get a kernel crash for
>> sizes of ramdisk larger than 2GB while trying to 'dd if=/dev/zero ...'  in
>> preparation for issuing mkfs.

FWIW "larger than 2GB" immediately stands out as a massive red flag for 
signed integer overflow leading to memory corruption *somewhere*, and 
all the splats in the log below do look like they could simply be 
fallout from that.

Building the vendor kernel with KASAN enabled might be another 
worthwhile avenue of investigation - if that can flag up a wild memory 
access at the point where it actually occurs, you've got more chance of 
working backwards from there to the root cause.

Robin.