On Tue, Oct 10, 2023 at 04:40:20PM -0400, Josef Bacik wrote:
> diff --git a/include/linux/blk-crypto-profile.h
> b/include/linux/blk-crypto-profile.h
> index 90ab33cb5d0e..3c002e85631a 100644
> --- a/include/linux/blk-crypto-profile.h
> +++ b/include/linux/blk-crypto-profile.h
> @@ -100,6 +100,13 @@ struct blk_crypto_profile {
> */
> struct device *dev;
>
> + /**
> + * @process_bio_supported: Some things, like btrfs, require the
> + * encrypted data for checksumming. Drivers set this to true if they can
> + * handle the process_bio() callback.
> + */
> + bool process_bio_supported;
> +
Is there any reason to think that real inline encryption hardware could support
this? For the encryption case it seems impossible, since Linux never gets
access to the ciphertext unless it reads it back from disk.
So, given the dependency on blk-crypto-fallback, I'm wondering if we should just
do something like 'profile == blk_crypto_fallback_profile' instead of having
this bool in the struct blk_crypto_profile.
- Eric
