Considering again, I concluded the patch above should be applied.
The use of SALT in HKDF may be not secret and there are valid use cases with no last or shorter salt. It's different to the use case of HMAC, where KEY is secret.

I pushed the change needed for GnuPG to t5964 branch.
See: https://dev.gnupg.org/rGc281bd94349e4f7997a89927aaa2c2f45004b902

Added HKDF implementation to master.

Applied to 1.10 branch.

didn't seem to work with 1.9.x

Thanks! Interestingly didn't seem to work with 1.9.x but it does with 1.10x. Maybe I made some error when testing.

Patch applied to master with small changes.

I can only find this one: https://github.com/patrickfav/singlestep-kdf/wiki/NIST-SP-800-56C-Rev1:-Non-Official-Test-Vectors

Thanks @jukivili , Here is the changelog,

Thanks for updated patch. I'm travelling next week and have time to check it closely only after I'm back. On quick glance, it looks good. What is also needed is the changelog for git commit log.

Thanks @jukivili. I have never thought of interleaving with interger poly1305 operation and that's a good suggestion. Will think about that one.

I meant interleaving integer register based 1xPoly1305 with 8xChacha20 as is done for 4xChacha20 in cipher/chacha20-ppc.c (interleaved so that for each 4xChaCha20 processed, 4 blocks of 1xPoly1305 is executed). Quite often microarchitectures have separate execution units for integer registers and vector registers and then it makes sense to interleave integer-poly1305 with vector-chacha20 as algorithms do not end up competing for same execution resources. Interleaving vector-poly1305 and vector-chacha20 is not likely to give performance increase (and likely to run problems with running out of vector registers).

HI @jukivili , Thanks for the updates. For f14-f31 registers that was my mistake that did not think floating point will be used. Will correct that. For poly1305, it can be used on ARCH_3.0 so checking use_p10 doesn't seem to be necessary but I can include that as well.

Also applied to 1.10.

Applied and pushed.

I learned that it's now called "OneStep KDF" in SP 800-56Cr2.
It's "SSKDF" in OpenSSL (Single Step KDF, perhaps).

Problem is that new assembly is using VSX registers vs14-vs31 which overlap with floating-point registers f14-f31. f14-f31 are ABI callee saved, so those need to be stored and restored.

Tested patch with small change so that HWF_PPC_ARCH_3_00 is used instead of HWF_PPC_ARCH_3_10. Building bench-slope with "-O3 -flto" makes bug in new implementation visible. Without new implementations bench-slope is ok (testing with QEMU):

$ tests/bench-slope --disable-hwf ppc-arch_3_00 cipher chacha20
Cipher:
 CHACHA20       |  nanosecs/byte   mebibytes/sec   cycles/byte
     STREAM enc |      2.35 ns/B     405.0 MiB/s         - c/B
     STREAM dec |      2.32 ns/B     410.7 MiB/s         - c/B
   POLY1305 enc |      2.46 ns/B     388.0 MiB/s         - c/B
   POLY1305 dec |      2.34 ns/B     408.1 MiB/s         - c/B
  POLY1305 auth |     0.238 ns/B      4003 MiB/s         - c/B

-O2 problem with bench-slope seems strange. Does problem appear after this patch is applied?

In T5975#158113, @werner wrote:

I can imagine thar there are use cases for this. Thus I see no problems for the first part.

The second part is imho not a good idea. Libgcrypt is a building block for all kind of software and there are for sure legitimate reasons to use rsa512 (MCUs, short living keys, etc). Thus I think that the decision on the key size should be done by the software using libgcrypt.

Pushed the change (master and 1.10).

At first, we need to add/enhance new API for KDF in libgcrypt. Currently, the term "KDF" in libgcrypt is used with narrower focus, that is, only for password->key KDF.

I do not claim I understand anything of this assembler syntax :)

Lets implement it for 2.3

For the second, I wonder if newer xlclang++ compiler works with 1.9.

Thank you for the bug report.

Pushed the change.

Thanks for your confirmation.

Okay, confirmed: I was just wrong and the build failure was only ever with --disable-asm (i.e. the log in this bug is the only relevant one). Patch works.

Ok. Thank you for the clarification. I will drop the second part and keep only the FIPS change in the patch. Merge request already updated.

Thanks. Should be applied.

I can imagine thar there are use cases for this. Thus I see no problems for the first part.

Pushed the change. Also, it's backported to 1.10 branch.

Thanks for creating this ticket. I'll reply.

libgcrypt 1.10 is out with the API change for Windows, and we don't see any report (yet).

Pushed the change to master.

GCC 11.3 and GCC 12.1 are out with the fix.

When we implemented this first, Libgcrypt had no appropriate KDF support. I recall that I considered to change this but it turned out the for 2.2 the changes are too large. For 2.3 we will consider such a change.

Feedback from the lab is that they'd recommend returning a specific error code that indicates that the prime search failed and then relying on the caller to decide whether to loop or bubble up the error. I'm not sure who we would consider to be the "caller" of the relevant generation function in this case, though.

Full ack.

Here is my proposal patch:

diff --git a/random/random-drbg.c b/random/random-drbg.c
index 5a46fd92..f1cfe286 100644
--- a/random/random-drbg.c
+++ b/random/random-drbg.c
@@ -341,6 +341,9 @@ enum drbg_prefixes
  * Global variables
  ***************************************************************/

That sounds reasonable. The FIPS 186-5 draft (https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.186-5-draft.pdf) covers this in section A.1.3, although I'm not quite sure why a lower bound for p was chosen compared to q. The comment that seems to have triggered this change is published on page 68 of https://csrc.nist.gov/CSRC/media/Publications/fips/186/4/final/documents/comments-received-fips186-4-december-2015.pdf by Allen Roginsky. It only contains a suggestion of 20, presumably for both numbers.

I checked FIPS 186-4 (and FIPS 186-5-draft). It is Appendix A 1.3.

I was pointed by Daiki to the following patch in Fedora binutils, which allows listing the fdo packaging metadata, but it does not list any other unknown objects and unfortunately fails hard:

I just copied the value of 0xcafe2a8e and the name .note.fdo.integrity from Daiki's implementation. No other reason.

I have one follow-up is that the readelf chokes on the integrity note for some reason:

$ readelf -n /usr/lib64/libgcrypt.so.20.4.1
Displaying notes found in: .note.fdo.integrity
  Owner                Data size 	Description
  FDO                  0x00000020	Unknown note type: (0x8e2afeca)

I assume this is just because the readelf does not know this type. I see this type was initially proposed by Daiki, but I did not find any other sources for this magic number so before filling bugs for readelf, do we have some doc why the 0xcafe2a8e is used?

The set_bit is obvious but we should cross check with the specs. In the non-fips mode we also try w/o a limit.

I think that it is OK to loop forever until we find a prime.

Hi Jussi, yes for some reason, it went missing, I was checking performance numbers and found out the line went missing. Thanks.

Fixed in master. I rechecked that bulk implementation passes tests with qemu-ppc64le.

Looks like that line went missing in third/final version of AES-GCM patch at https://dev.gnupg.org/T5700

you also use the CPU cache size on GNU/Linux. Is it important to have that information on MS-Windows?

The attached patch implements getting the number of processors on MS-Windows.

see rC67b36154f88e for master.

Will add it. The reason I added Brainpool was due to a question on the performacne between Brainpool and other NIST.