netdev - Re: Fw: [Bug 216195] New: Maxlinear GPY115 UBSAN: shift-out-of-bounds

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:   Fri, 1 Jul 2022 19:34:58 -0700
From:   Florian Fainelli <f.fainelli@...il.com>
To:     Stephen Hemminger <stephen@...workplumber.org>,
        Giuseppe Cavallaro <peppe.cavallaro@...com>,
        Alexandre Torgue <alexandre.torgue@...s.st.com>,
        Jose Abreu <joabreu@...opsys.com>,
        "Ling, Pei Lee" <pei.lee.ling@...el.com>,
        Voon Weifeng <weifeng.voon@...el.com>,
        Wong Vee Khee <vee.khee.wong@...ux.intel.com>,
        Mohammad Athari Bin Ismail <mohammad.athari.ismail@...el.com>,
        Joao.Pinto@...opsys.com
Cc:     netdev@...r.kernel.org
Subject: Re: Fw: [Bug 216195] New: Maxlinear GPY115 UBSAN: shift-out-of-bounds

+Intel folks, Joao,

On 7/1/2022 4:45 PM, Stephen Hemminger wrote:
> 
> 
> Begin forwarded message:
> 
> Date: Fri, 01 Jul 2022 08:26:11 +0000
> From: bugzilla-daemon@...nel.org
> To: stephen@...workplumber.org
> Subject: [Bug 216195] New: Maxlinear GPY115 UBSAN: shift-out-of-bounds
> 
> 
> https://bugzilla.kernel.org/show_bug.cgi?id=216195
> 
>              Bug ID: 216195
>             Summary: Maxlinear GPY115 UBSAN: shift-out-of-bounds
>             Product: Networking
>             Version: 2.5
>      Kernel Version: 5.15-5.17
>            Hardware: Intel
>                  OS: Linux
>                Tree: Mainline
>              Status: NEW
>            Severity: normal
>            Priority: P1
>           Component: Other
>            Assignee: stephen@...workplumber.org
>            Reporter: cedric@...espeed.nl
>          Regression: No
> 
> This is related to specific (wired) Ethernet IC: Maxlinear Ethernet GPY115B
> The network interface does show up in ifconfig but is not functioning, no
> ip-address is assigned.
> 
> During the bootup process I get the following warning:
> 
>      UBSAN: shift-out-of-bounds in
>      /build/linux-WLUive/linux-
>      5.15.0/drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:224:12
>      Shift exponent 40 is too large for 32-bit type 'unsigned int'
> 
> This results in the following error:
> 
>      Maxlinear Ethernet GPY115B stmmac-3:01: gpy_config_aneg failed -110
> 
> The specific kernel I am running: 5.15.0-40-generic on Ubuntu but also tested
> this with Fedora Server 36 running kernel 5.17.
> Config: CONFIG_MAXLINEAR_GPHY=m
> The module is installed here:
> /usr/lib/modules/5.15.0-generic/kernel/drivers/net/phy/mxl-gpy.ko
> 
> Additional information:
> https://askubuntu.com/questions/1416068/how-to-enable-maxlinear-phy-gpy115-drivers
> 
> Datasheet:
> https://www.maxlinear.com/product/connectivity/wired/ethernet/ethernet-transceivers-phy/gpy115
> 

I do not know how or if the undefined behavior relates to the Ethernet 
PHY driver reporting a -ETIMEDOUT during auto-negotiation however the 
undefined behavior looks legit to me. Line 224 is:

value |= MTL_RXQ_DMA_QXMDMACH(chan, queue);

with:

#define MTL_RXQ_DMA_QXMDMACH(chan, q)  ((chan) << (8 * (q)))

So for the shift to exceed 32-bits we would need q to be >=4, or q = 3 
and chan >= 256, more on that below. The report does not indicate which 
platform is used but googling around shows this askubuntu bug report 
from the same person:

https://askubuntu.com/questions/1416068/how-to-enable-maxlinear-phy-gpy115-drivers

and indicates that the platform is an Onlogic HX310 which is Intel 
Elkhart SoC. This means that dwmac-intel.c is likely the glue driver 
being used for registration and there we have several paths through 
which plat->rx_queues_to_use is used as an iterator to initialize 
plat->rx_queues_cfg[i].chan = i.

Given there can be 1, 6 or 8 RX queues according to that file, then the 
latter 2 combinations will trigger undefined behaviors by shifting too 
much to the left.

Not having access to the DMWAC4 datasheet however I do not know how to 
fix it.

Fixes tag for that bug would be:

Fixes: d43042f4da3e ("net: stmmac: mapping mtl rx to dma channel")
-- 
Florian