crypto: Remove uses of CRYPTO_F_DONE

[markj]

Summary

Previously OCF set CRYPTO_F_DONE prior to invoking the completion
callback, even if the request failed. This isn't particularly useful
and leads to bugs when consumers retry a failed request, since OCF also
asserts that CRYPTO_F_DONE is clear in crypto_dispatch(). (Really, OCF
should retry requests that fail with EAGAIN, but that's a larger
change.)

For now, just stop setting CRYPTO_F_DONE to simplify consumers (and fix
those which fail to clear the flag before retrying a request). In krb5,
keep using it, but set it in the callback. Fix buggy handling of EAGAIN
there as well. Panic if the request fails for a different reason, since
we can't signal it to upper layers.

PR: 286321

[jhb]

sys/kgssapi/krb5/kcrypto_aes.c

121–123

Doesn't this case still need to set CRYPTO_F_DONE?

Another way this could be handled btw would be to clear crp_opaque to NULL when a request is completed rather than using this flag. That would let you axe the flag entirely I think?

[markj]

Apply John's suggestion

[markj]

sys/kgssapi/krb5/kcrypto_aes.c

121–123

I don't think we need to set the flag in the synchronous case, as the dispatch code only checks it in the async case.

Clearing the pointer instead seems fine.

[jhb]

So maybe one commit to update the krb5 code to use crp_opaque = NULL and then you can just remove the flag entirely in a second commit? The changes look ok to me though (I'm just also fine with now removing the flag entirely)

[markj]

Apply John's suggestion of splitting the diff into two.

[markj]

In D50104#1145336, @jhb wrote:

So maybe one commit to update the krb5 code to use crp_opaque = NULL and then you can just remove the flag entirely in a second commit? The changes look ok to me though (I'm just also fine with now removing the flag entirely)

That is better, thanks. See D50238 for the krb5 code.

[jhb]

sys/opencrypto/cryptodev.h

436

Any reason to not just remove it outright?

[markj]

sys/opencrypto/cryptodev.h

436

Out-of-tree openzfs still refers to it. I'll patch it and eventually remove this definition.

[jhb]

sys/opencrypto/cryptodev.h

436

Ok, that's fair.