Re: [PATCH v5 2/4] mmc: sdhci-msm: Reprogram crypto keys during host hardware initialization

Neeraj Soni <[email protected]> Fri, 31 Jul 2026 11:37:44 +0530
Newsgroups org.kernel.vger.linux-mmc,org.kernel.vger.linux-arm-msm,org.kernel.vger.linux-kernel
Message-ID <[email protected]>

On 7/31/2026 12:48 AM, Eric Biggers wrote:
> On Thu, Jul 30, 2026 at 07:38:20PM +0530, Neeraj Soni wrote:
>> Host looses the Inline Crypto Engine (ICE) keys during hardware reset.
>> Re-program the keys while intializing ICE to ensure that ICE is available
>> in the same state as before the hardware reset.
>>
>> Also set MMC_CAP2_CRYPTO_NO_REPROG to indicate that re-program of ICE keys
>> is not needed during MMC runtime suspend/resume.
>>
>> Signed-off-by: Seshu Madhavi Puppala <[email protected]>
>> Co-developed-by: Ram Prakash Gupta <[email protected]>
>> Signed-off-by: Ram Prakash Gupta <[email protected]>
>> Co-developed-by: Sarthak Garg <[email protected]>
>> Signed-off-by: Sarthak Garg <[email protected]>
>> Signed-off-by: Debraj Mukhopadhyay <[email protected]>
>> Signed-off-by: Neeraj Soni <[email protected]>
>> ---
>>  drivers/mmc/host/sdhci-msm.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c
>> index 4aff965f0e2e..7d71dc6f388d 100644
>> --- a/drivers/mmc/host/sdhci-msm.c
>> +++ b/drivers/mmc/host/sdhci-msm.c
>> @@ -1958,7 +1958,10 @@ static int sdhci_msm_ice_init(struct sdhci_msm_host *msm_host,
>>  				cap.sdus_mask * 512;
>>  	}
>>  
>> +	blk_crypto_reprogram_all_keys(&mmc->crypto_profile);
>> +
> 
> This still does nothing.  The crypto profile was just initialized, so no
> keyslots are in-use and blk_crypto_reprogram_all_keys() does nothing.
> 
This is needed to address those scenarios where device is running in a
stable state and only the host hardware goes for a reset, for example,
to recover from an error state or stall.

>>  	mmc->caps2 |= MMC_CAP2_CRYPTO;
>> +	mmc->caps2 |= MMC_CAP2_CRYPTO_NO_REPROG;
> 
> This seems to be the actual point of the patch, so let's simplify it to
> just this.  However, this is also broken because this is ordered before
> patch 3 that introduces the actually functional key reprogramming.
> Please drop the no-op call to blk_crypto_reprogram_all_keys(), then
> merge patches 2 and 3 into a single one.
> 
The call to blk_crypto_reprogram_all_keys() is not a no-op as clarified
above. Hence the changes were split in two.

> - Eric
> 
Regards,
Neeraj