openbsd/src/lib/libcrypto/evp/evp_digest.c, branch OPENBSD_7_9A mirror of https://github.com/libressl/openbsd.git
https://git.lua4.win/openbsd/atom?h=OPENBSD_7_92025-05-10T05:54:39+00:00Use err_local.h rather than err.h in most places2025-05-10T05:54:39+00:00tb2025-05-10T05:54:39+00:00urn:sha1:41e8f99dd1625a9f0c80ce9d4383e95b18e85709
ok jsing
Hide deprecated functions in evp.h2024-04-10T15:00:38+00:00beck2024-04-10T15:00:38+00:00urn:sha1:20ec75270d82cdbe18bc78b00516a72ae16b122e
use LCRYPTO_UNUSED and remove the LIBRESSL_INTERNAL guard around them.
ok tb@
Hide public symbols in evp.h2024-04-09T13:52:42+00:00beck2024-04-09T13:52:42+00:00urn:sha1:9835bb3c82adc630df1f3aaa7dfbf7f0506b0bc4
largely mechanically done by the guentherizer 9000
ok tb@
Remove EVP_MD_meth* again2024-03-02T09:59:56+00:00tb2024-03-02T09:59:56+00:00urn:sha1:1894c9da2aed32c69be2c297ff92c86f701c6ddd
Erlang upstream disabled the otp_test_engine for LibreSSL >= 3.5 without
explanation. It was the only reason we added this garbage API in the first
place. Meanwhile libfido2 started using it for a mock up of OpenSSL 3's
broken fetch design with old API. This is pointless, so all this garbage
goes away again (in particular we can remove the absolutely horrifying
EVP_MD_meth_set_app_datasize() again).
ok jsing
Fix signature and semantics of EVP_{CIPHER,MD}_CTX_init()2024-03-02T09:55:30+00:00tb2024-03-02T09:55:30+00:00urn:sha1:4083c1a64fc21730cb59ac92344eaeeda5fd688b
When the EVP_CIPHER_CTX and the EVP_MD_CTX were still expected to live
on the stack, these initialization APIs were wrappers around memset.
In OpenSSL 1.1, somebody removed them and carelessly made _init() an
alias of _reset() aka _cleanup(). As a consequence, both signature and
semantics changed.
Unsurprisingly, there is now code out there that actually uses the new
semantics, which causes leaks on LibreSSL and older OpenSSL. This aligns
our _init() with OpenSSL 1.1 semantics.
ok jsing
Use EVP_MD_CTX_legacy_clear() internally2024-02-18T15:45:42+00:00tb2024-02-18T15:45:42+00:00urn:sha1:ff3e386de60f6149938edb124d31bc2fd7ae04f6
ok jsing
Add EVP_MD_CTX_legacy_clear()2024-02-18T15:43:21+00:00tb2024-02-18T15:43:21+00:00urn:sha1:57ee559fa828c2d7f5c21fd1bee340fec067a508
This is analogous to EVP_CIPHER_CTX_legacy_clear() and will serve as an
internal replacement for EVP_MD_CTX_init() until the conversion to heap
allocated ctx is completed. This way EVP_MD_CTX_init() can be changed to
match the OpenSSL 1.1 API.
ok jsing
Make EVP_{CIPHER,MD}_CTX_{cleanup,reset}() NULL-safe2024-01-30T17:41:01+00:00tb2024-01-30T17:41:01+00:00urn:sha1:abcb4251e0970a4a64b4764962fd5fa357070770
We have a bunch of code that relies on this. Surely there is code out
there in the wider ecosystem that relies on these being NULL-safe by
now since upstream sprinkles NULL checks wherever they can.
ok beck joshua
Move the EVP_MD block size accessor down2023-12-29T07:22:47+00:00tb2023-12-29T07:22:47+00:00urn:sha1:290b2556a46fdf4db11544123a9f095ed17d6e22
This way all the EVP_MD accessors are in the order of the struct fields.
Well, arguably the EVP_MD_meth* should come first, but they are scheduled
to go meet the dodo.
Move the EVP_MD_CTX flag accessors up a bit2023-12-29T07:14:02+00:00tb2023-12-29T07:14:02+00:00urn:sha1:0523322302fbb4c568b70aa19f3f3012143af56a
This way the accessors are sorted the same way as the struct.