netdev - Re: Experimental fix for MSI-X issue on r8169

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 22 Aug 2018 07:56:19 +0200
From:   Heiner Kallweit <hkallweit1@...il.com>
To:     David Miller <davem@...emloft.net>, jian-hong@...lessm.com
Cc:     steved424@...il.com, gogen@...root.org, netdev@...r.kernel.org,
        linux@...lessm.com
Subject: Re: Experimental fix for MSI-X issue on r8169

On 22.08.2018 06:24, David Miller wrote:
> From: Jian-Hong Pan <jian-hong@...lessm.com>
> Date: Wed, 22 Aug 2018 11:01:02 +0800
> 
>  ...
>> [   56.462464] r8169 0000:02:00.0: MSI-X entry: context resume:
>> ffffffff ffffffff ffffffff ffffffff
>  ...
>> uh!  The MSI-X entry seems missed after resume on this laptop!
> 
> Yeah, having all of the MSI-X entry values be all-1's is not a good
> sign.
> 
all-1's seems to indicate that PCI access to the MSI-X table
BAR/region fails. Because falling back to MSI helps, accessing
the other BAR/region with the memory-mapped registers works.
I'll check with Realtek whether this symptom rings any bell.


> But this is quite a curious set of debugging traces we now have.
> 
> In the working case, the vector number in the DATA field seems
> to change, which suggests that something is assigning new values
> and programming them into these fields at resume time.
> 
> But in the failing cases, all of the values are garbage.
> 
> I would expect, given what the working trace looks like, that in the
> failing case some values would be wrong and the DATA value would have
> some new yet valid value.  But that is not what we are seeing here.
> 
> Weird.
> 
>