Re: [v2 PATCH] Revert "crypto: run initcalls for generic implementations earlier"

[Herbert Xu <herbert@gondor.apana.org.au>]

On Thu, Jun 05, 2025 at 04:29:36PM +0200, Lukas Wunner wrote:
>
> I just wanted to provide a heads-up and an opportunity for early feedback
> to this patch.

Thanks for the heads up.

Ideally the PCI enumeration should happen later, but I could
certainly live with moving the required algorithms earlier.
 
> diff --git a/crypto/algboss.c b/crypto/algboss.c
> index 846f586..74517f2 100644
> --- a/crypto/algboss.c
> +++ b/crypto/algboss.c
> @@ -247,7 +247,7 @@ static void __exit cryptomgr_exit(void)
>  	BUG_ON(err);
>  }
>  
> -module_init(cryptomgr_init);
> +subsys_initcall(cryptomgr_init);

Shouldn't this be one earlier than subsys_initcall as otherwise
you're at the whim of linker ordering?

> diff --git a/crypto/sha256.c b/crypto/sha256.c
> index 4aeb213..a20c920 100644
> --- a/crypto/sha256.c
> +++ b/crypto/sha256.c
> @@ -264,7 +264,7 @@ static int __init crypto_sha256_mod_init(void)
>  		num_algs -= 2;
>  	return crypto_register_shashes(algs, ARRAY_SIZE(algs));
>  }
> -module_init(crypto_sha256_mod_init);
> +subsys_initcall(crypto_sha256_mod_init);

This also depends on lib/crypto/sha256 which would need to be moved
as well.  However, I think rather than moving lib/crypto/sha256 even
earlier, we should simply register the Crypto API sha256-arch later.

Cheers,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt