openssl: Test failures on big endian ARMv9 with enable-asm

When building openssl 3.3.3 in a big endian Gentoo VM on an ARMv9 based computer (Radxa Orion O6), 14 tests fail when building with enable-asm. I also get failures in wget (with https) and ssh. If building with no-asm, then all the problems go away. So it seems there is some asm code which is not endian clean.

If I build on an ARMv8 machine instead (Nvidia Jetson TX2) then everything works even with enable-asm, so presumably it’s some asm that is enabled only for newer ARM.

cpu features on Orion O6 (where the tests fail):

fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm ssbs sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh bti ecv afp wfxt

cpu features on Jetson TX2 (where the tests succeed):

fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid

The tests that fail are the following:

Test Summary Report
-------------------
03-test_internal_chacha.t             (Wstat: 256 (exited 1) Tests: 1 Failed: 1)
  Failed test:  1
  Non-zero exit status: 1
05-test_cmac.t                        (Wstat: 256 (exited 1) Tests: 1 Failed: 1)
  Failed test:  1
  Non-zero exit status: 1
20-test_enc.t                         (Wstat: 1024 (exited 4) Tests: 99 Failed: 4)
  Failed tests:  87-88, 91-92
  Non-zero exit status: 4
20-test_enc_more.t                    (Wstat: 512 (exited 2) Tests: 141 Failed: 2)
  Failed tests:  136-137
  Non-zero exit status: 2
25-test_pkcs8.t                       (Wstat: 512 (exited 2) Tests: 15 Failed: 2)
  Failed tests:  14-15
  Non-zero exit status: 2
30-test_evp.t                         (Wstat: 512 (exited 2) Tests: 80 Failed: 2)
  Failed tests:  42, 50
  Non-zero exit status: 2
70-test_quic_record.t                 (Wstat: 256 (exited 1) Tests: 1 Failed: 1)
  Failed test:  1
  Non-zero exit status: 1
90-test_bio_enc.t                     (Wstat: 256 (exited 1) Tests: 1 Failed: 1)
  Failed test:  1
  Non-zero exit status: 1
Files=315, Tests=3694, 158 wallclock secs (11.21 usr  1.68 sys + 364.07 cusr 101.99 csys = 478.95 CPU)
Result: FAIL

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Comments: 56 (27 by maintainers)

Commits related to this issue

Most upvoted comments

Ok. Since the afalg test mentioned aes_cbc, I checked the kernel options for AES.

The kernel for the big endian VM has

CONFIG_CRYPTO_AES_ARM64=y

but the little endian L0 kernel (which is a vendor kernel) has

# CONFIG_CRYPTO_AES_ARM64 is not set

so maybe that is the reason for the test fail?

Ok, now V=1 output from test_internal_chacha is different: new_log.txt

Still FAIL though.

BTW, posting patches inline in github comments seems to mess up the whitespace which forces me to apply the patches by hand. Please attach the patches instead. Thanks.

I’m currently bisecting on test_internal_chacha and test_bio_enc to see where those stopped working. Then I’ll check whatever is left with V=1. 😄

Regarding master; as I wrote above master has the exact same set of fails as 3.3.3 and 3.2.0.