On Wed, Oct 30, 2024 at 05:22:35PM +0100, Harald Freudenberger wrote:
>
> +static int s390_phmac_sha2_clone_tfm(struct crypto_shash *dst,
> + struct crypto_shash *src)
> +{
> + struct s390_phmac_ctx *dst_ctx = crypto_shash_ctx(dst);
> + struct s390_phmac_ctx *src_ctx = crypto_shash_ctx(src);
> + int rc;
> +
> + rc = s390_phmac_sha2_init_tfm(dst);
> + if (rc)
> + return rc;
> +
> + if (src_ctx->key && src_ctx->keylen) {
> + dst_ctx->key = kmemdup(src_ctx->key, src_ctx->keylen,
> + GFP_KERNEL);
Any allocations in clone_tfm must be GFP_ATOMIC since this is
supposed to be used on paths where we can't sleep (otherwise
you could just allocate a new tfm).
Cheers,
--
Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
