linux-kernel - Re: RFC: Using regmap in ARM64 for EL3 register access

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, 4 Apr 2018 13:18:02 -0500
From:   Thor Thayer <thor.thayer@...ux.intel.com>
To:     Sudeep Holla <sudeep.holla@....com>,
        Mark Brown <broonie@...nel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>
Cc:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: RFC: Using regmap in ARM64 for EL3 register access

Hi Sudeep,

On 04/04/2018 05:47 AM, Sudeep Holla wrote:
> 
> 
> On 30/03/18 00:00, Thor Thayer wrote:
>> Hi,
>>
>> I'm working on an ARM64 architecture that needs to manipulate some
>> protected registers that are only accessible in EL3. Linux is running at
>> EL1 which doesn't have the proper permissions for these registers.
>>
> If the hardware/IP registers are designed not to provide access to EL1,
> then providing one in software by some means is simply wrong approach to
> solve whatever issue you are trying to address here.
> 
Thank you for the reply!

Sorry, I didn't describe this very well. This mechanism is similar to 
PSCI. We have System Management registers that could be needed by more 
than 1 virtual machine. So arbitration is needed and we use U-Boot for this.

>> Since U-Boot is running at the higher EL3, we communicate to the U-Boot
>> functions through a SMC mechanism.
>>
> Please follow SMCCC to add any SMC interface.
> 
Yes, the existing implementation is doing this and implements a SiP 
service framework.

>> The regmap framework seems like a good match for accessing these
>> registers. We need the same functionality as I2C and SPI regmaps - read,
>> write, and update registers.
>>
> No, providing register access to EL1 using regmap+SMC just defeats the
> hardware security restrictions and may provide ways to exploit.
> Why not abstract to the level of services you need ?
> 
I'm not following about abstracting the level of services. For accessing 
the registers, I thought a regmap implementation would be nice because 
of the read/write/update functionality that would wrap around the SMC 
SiP service functions.

>> Any comments or suggestions about using regmap for this purpose? Is
>> there a better method?
>>
> If you are looking for power management features, then PSCI and SCMI are
> couple of specifications to look at. I would really like to know more
> details on your use case to provide any suggestions.
> 
Thank you for those pointers. I vaguely knew about PSCI but I wasn't 
aware of SCMI. Let me study these interfaces, particularly SCMI, before 
I reply further because this may answer many of my questions.

Thank you again,

Thor