Hi, I think there maybe some bugs caused by the recent crypto patches that got merged to master. ref: https://lore.kernel.org/qemu-devel/cafeaca-e_1wflun2hpttt2bszxksmbnxkak_uzuhwrh_fb6...@mail.gmail.com/T/#t
I think before these patches the "qcrypto_hash_bytes" or "qcrypto_hash_bytesv" apis used to behave such that a user of the apis could pass his own allocated buffers. In that case, the passed buffers would be used to fill in the hash result instead of allocating a new buffer. But I think in the master branch now, these apis always allocate the result buffer regardless of it's users passing their own buffers or not. So this is problematic for wherever the users of the apis are passing their own allocated buffers. For example, I see target/i386/sev.c is one such api user. Am I missing something here or does this look like it's a clear bug? If I am correct, I think it makes sense to keep the old behavior of the apis where new buffers are not allocated if the user supplies one (I think it probably also makes sense if we force the users to always provide the bufffer instead of the apis themselves allocating them). I noticed this bug in the nitro-enclave work I am doing where rebasing the branch builds but the actual behavior is buggy because of this new change of the api implementations. Regards, Dorjoy
