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: <BY5PR12MB40344989C3CB5881729F660CDEFF0@BY5PR12MB4034.namprd12.prod.outlook.com>
Date:   Tue, 10 Mar 2020 12:18:15 +0000
From:   Eugeniy Paltsev <Eugeniy.Paltsev@...opsys.com>
To:     Vineet Gupta <Vineet.Gupta1@...opsys.com>,
        "linux-snps-arc@...ts.infradead.org" 
        <linux-snps-arc@...ts.infradead.org>
CC:     Alexey Brodkin <Alexey.Brodkin@...opsys.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 2/4] ARC: handle DSP presence in HW

>From: Vineet Gupta <vgupta@...opsys.com>
>Sent: Saturday, March 7, 2020 03:12
>To: Eugeniy Paltsev; linux-snps-arc@...ts.infradead.org
>Cc: Alexey Brodkin; linux-kernel@...r.kernel.org
>Subject: Re: [PATCH v2 2/4] ARC: handle DSP presence in HW
>
>On 3/5/20 12:02 PM, Eugeniy Paltsev wrote:
>> In case of DSP extension presence in HW some instructions
>> (related to integer multiply, multiply-accumulate, and divide
>> operation) executes on this DSP execution unit. So their
>> execution will depend on dsp configuration register (DSP_CTRL)
>>
>> As we want these instructions to execute the same way regardless
>> of DSP presence we need to set DSP_CTRL properly. However this
>> register can be modified bu any usersace app therefore any
>> usersace may break kernel execution.
>>
>> Fix that by configure DSP_CTRL in CPU early code and in IRQs
>> entries.
>
> How about below ....
> 
> "When DSP extensions are present, some of the regular integer instructions such as
> DIV, MACD etc are executed in the DSP unit with semantics alterable by flags in
> DSP_CTRL aux register. This register is writable by userspace and thus can
> potentially affect corresponding instructions in kernel code, intentionally or
> otherwise. So safegaurd kernel by effectively disabling DSP_CTRL upon bootup and
> every entry to kernel.
> 
> Do note that for this config we simply zero out the DSP_CTRL reg assuming
> userspace doesn't really care about DSP. The next patch caters to the DSP aware
> userspace which this actually saved/restored upon kernel entry."
>From: Vineet Gupta <vgupta@...opsys.com>
>Sent: Saturday, March 7, 2020 03:12
>To: Eugeniy Paltsev; linux-snps-arc@...ts.infradead.org
>Cc: Alexey Brodkin; linux-kernel@...r.kernel.org
>Subject: Re: [PATCH v2 2/4] ARC: handle DSP presence in HW
>
>On 3/5/20 12:02 PM, Eugeniy Paltsev wrote:
>> In case of DSP extension presence in HW some instructions
>> (related to integer multiply, multiply-accumulate, and divide
>> operation) executes on this DSP execution unit. So their
>> execution will depend on dsp configuration register (DSP_CTRL)
>>
>> As we want these instructions to execute the same way regardless
>> of DSP presence we need to set DSP_CTRL properly. However this
>> register can be modified bu any usersace app therefore any
>> usersace may break kernel execution.
>>
>> Fix that by configure DSP_CTRL in CPU early code and in IRQs
>> entries.
>
> How about below ....

> 
> "When DSP extensions are present, some of the regular integer instructions such as
> DIV, MACD etc are executed in the DSP unit with semantics alterable by flags in
> DSP_CTRL aux register. This register is writable by userspace and thus can
> potentially affect corresponding instructions in kernel code, intentionally or
> otherwise. So safegaurd kernel by effectively disabling DSP_CTRL upon bootup and
> every entry to kernel.
> 
> Do note that for this config we simply zero out the DSP_CTRL reg assuming
> userspace doesn't really care about DSP. The next patch caters to the DSP aware
> userspace which this actually saved/restored upon kernel entry."

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ