[U-Boot] [PATCH] Add bootscript support to esbc_validate.

York Sun yorksun at freescale.com
Wed Mar 11 19:47:25 CET 2015



On 03/11/2015 11:44 AM, Scott Wood wrote:
> On Wed, 2015-03-11 at 10:50 -0700, York Sun wrote:
>>
>> On 03/11/2015 03:39 AM, Gupta Ruchika-R66431 wrote:
>>> Hi York,
>>>
>>>> -----Original Message-----
>>>> From: Sun York-R58495
>>>> Sent: Tuesday, March 10, 2015 10:03 PM
>>>> To: Gupta Ruchika-R66431; Rana Gaurav-B46163; u-boot at lists.denx.de
>>>> Cc: Wood Scott-B07421; Bansal Aneesh-B39320
>>>> Subject: Re: [PATCH] Add bootscript support to esbc_validate.
>>>>
>>>> On 03/10/2015 09:25 AM, Gupta Ruchika-R66431 wrote:
>>>>> Hi York,
>>>>>
>>>>>> -----Original Message-----
>>>>>> From: Sun York-R58495
>>>>>> Sent: Tuesday, March 10, 2015 9:45 PM
>>>>>> To: Rana Gaurav-B46163; u-boot at lists.denx.de
>>>>>> Cc: Wood Scott-B07421; Gupta Ruchika-R66431; Bansal Aneesh-B39320
>>>>>> Subject: Re: [PATCH] Add bootscript support to esbc_validate.
>>>>>>
>>>>>>
>>>>>>
>>>>>> On 03/10/2015 01:38 AM, Gaurav Rana wrote:
>>>>>>> 1. Default environment will be used for secure boot flow  which
>>>>>>> can't be edited or saved.
>>>>>>> 2. Command for secure boot is predefined in the default  environment
>>>>>>> which will run on autoboot (and autoboot is  the only option allowed
>>>>>>> in case of secure boot) and it  looks like this:
>>>>>>>  #define CONFIG_SECBOOT \
>>>>>>>  "setenv bs_hdraddr 0xe8e00000;"                 \
>>>>>>>  "esbc_validate $bs_hdraddr;"                    \
>>>>>>>  "source $img_addr;"                             \
>>>>>>>  "esbc_halt;"
>>>>>>>  #endif
>>>>>>> 3. Boot Script can contain esbc_validate commands and bootm command.
>>>>>>>  Uboot source command used in default secure boot command will  run
>>>>>>> the bootscript.
>>>>>>> 4. Command esbc_halt added to ensure either bootm executes  after
>>>>>>> validation of images or core should just spin.
>>>>>>>
>>>>>> What's the purpose of "esbc_halt"? Once it enters the spin, how to
>>>>>> get it out?
>>>>> The purpose of bootscript is to validate the next level images and then
>>>> pass control to it, so bootscript must contain a bootm command. We don't
>>>> expect control to return back to u-boot. Hence a command esbc_halt is
>>>> introduced which would make the core spin and not provide uboot prompt in
>>>> case bootscript doesn't pass control to next level image.
>>>>> For secure chain of trust, only validated bootscript should be allowed to
>>>> execute and be responsible for passing control to next level image.
>>>>>
>>>>
>>>> Ruchika,
>>>>
>>>> Do you expect secure boot to run automatically once u-boot reaches the prompt
>>>> and the "source $img_addr" to actually boot the OS? You put "esbc_halt" as a
>>>> fall-back to catch failure above? It doesn't sounds very secure to me.
>>>
>>> The bootscript is first validated. Only an authenticated user, who has the private key can sign the bootscript. Thus validating bootscript is important in secure boot chain of trust. 
>>>
>>> You are right regarding fallback as esbc_halt. In the esbc_halt implementation, we will add code to clear security secrets on the chip, and issue a reset. We will send a separate patch for that.
>>>
>>
>> Wouldn't it be possible to call a reset/hang/panic when the validation fails,
>> before "source $img_addr"?
> 
> I'd assume it already has that, but it's still good to have something to
> deal with the case where the script returns due to some failure.
> 

If that's the case, I am OK with the addition of "esbc_halt" command.

York



More information about the U-Boot mailing list