Re: [PATCH RESEND v4] mmc: Avoid reprogram all keys to Inline Crypto Engine for MMC runtime suspend resume

Eric Biggers <[email protected]> Thu, 23 Jul 2026 23:01:04 +0000
Newsgroups org.kernel.vger.linux-mmc,org.kernel.vger.linux-arm-msm,org.kernel.vger.linux-kernel
Message-ID <[email protected]>
On Fri, May 29, 2026 at 04:05:52PM +0530, Neeraj Soni wrote:
> Yes for cold boot this is true and Crypto reprogram is not needed. This patch addreses the use cases like
> runtime suspend resume, deep sleep / Quick boot (DS/QB) and suspend to disk (aka hibernation) use cases
> where the runtime context is preserved either in dynamic memory (DS/QB) or in secondry memory (hibernation).
> This is tested internally on SAR1250P chipset for hibernation and DS/QB.

But sdhci_msm_gcc_reset() is called only from the driver probe function
(sdhci_msm_probe()), before MMC_CAP2_CRYPTO is set.  So Jie's comment is
correct: the new key reprogramming path is dead code.  And as per the
following comment, key reprogramming is still needed in some cases
(https://lore.kernel.org/linux-mmc/[email protected]/).
So doesn't this patch break the sdhci-msm driver?

I'm also skeptical of the following change, which seems unrelated to the
rest of the patch:

> -#ifdef CONFIG_MMC_CRYPTO
>  #define MMC_CAP2_CRYPTO		(1 << 27)	/* Host supports inline encryption */
> -#else
> -#define MMC_CAP2_CRYPTO		0
> -#endif

Because of that, mtk-sd sets MMC_CAP2_CRYPTO even when
CONFIG_MMC_CRYPTO=n.  That makes CQHCI_CRYPTO_GENERAL_ENABLE be used
even when CONFIG_MMC_CRYPTO=n.

- Eric