OpenSSL version 3.3.0-dev: OpenSSL::ASN1::ASN1Error: utctime/generalizedtime is too short

[junaruga]

I got the following errors only in the CI openssl-head case when running the latest master commit 1e8e246 on openssl-head on my forked repository. The used OpenSSL is the latest master branch openssl/openssl@1f03d33 according to the log.

https://github.com/junaruga/ruby-openssl/actions/runs/8233851472/job/22514261811#step:12:613

1) Error: test_generalizedtime(OpenSSL::TestASN1): OpenSSL::ASN1::ASN1Error: generalizedtime is too short
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:698:in `decode'
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:698:in `decode_test'
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:433:in `test_generalizedtime'
     430:       OpenSSL::ASN1::GeneralizedTime.new(Time.utc(9999, 9, 8, 23, 43, 39))
     431:     # LibreSSL 3.6.0 requires the seconds element
     432:     return if libressl?
  => 433:     decode_test B(%w{ 18 0D }) + "201612081934Z".b,
     434:       OpenSSL::ASN1::GeneralizedTime.new(Time.utc(2016, 12, 8, 19, 34, 0))
     435:     # not implemented
     436:     # decode_test B(%w{ 18 13 }) + "20161208193439+0930".b,

2) Error: test_utctime(OpenSSL::TestASN1): OpenSSL::ASN1::ASN1Error: utctime is too short
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:698:in `decode'
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:698:in `decode_test'
/home/runner/work/ruby-openssl/ruby-openssl/test/openssl/test_asn1.rb:411:in `test_utctime'
     408:     end
     409:     # Seconds is omitted. LibreSSL 3.6.0 requires it
     410:     return if libressl?
  => 411:     decode_test B(%w{ 17 0B }) + "1609082343Z".b,
     412:       OpenSSL::ASN1::UTCTime.new(Time.utc(2016, 9, 8, 23, 43, 0))
     413:     # not implemented
     414:     # decode_test B(%w{ 17 11 }) + "500908234339+0930".b,

[rhenium]

Grep with "generalizedtime is too short" lead me to this OpenSSL PR: openssl/openssl#23483.

I think we can remove these assertions.

[junaruga]

Thanks for finding the commit. Did you mean that we remove the decode_test lines?

For the 2nd error test, I debugged. Below is a minimal Ruby reproducer.

$ ruby -I ./lib -r openssl -e 'der = [%w{ 17 0B }.join].pack("H*") + "1609082343Z".b; OpenSSL::ASN1.decode(der)'
-e:1:in `decode': utctime is too short (OpenSSL::ASN1::ASN1Error)
  from -e:1:in `<main>'

Below is the gdb debugging and backtrace.

What do you think about that we adjust the value of the der for the OpenSSL::ASN1.decode(der) to pass the tests? I just don't understand the meaning of the 2 assertions that caused the errors.

$ gdb --args ruby -I ./lib -r openssl -e 'der = [%w{ 17 0B }.join].pack("H*") + "1609082343Z".b; OpenSSL::ASN1.decode(der)'
...
(gdb) l
436	    VALUE ret;
437	    int status = 0;
438	
439	    p = der;
440	    if(!(time = d2i_ASN1_TIME(NULL, &p, length)))
441		ossl_raise(eASN1Error, NULL);
442	    ret = rb_protect(asn1time_to_time_i,
443			     (VALUE)time, &status);
444	    ASN1_TIME_free(time);
445	    if(status) rb_jump_tag(status);
(gdb) f
#0  decode_time (der=0x7fffe86ac520 "\027\v1609082343Z", length=13)
    at ../../../../ext/openssl/ossl_asn1.c:441
441		ossl_raise(eASN1Error, NULL);
(gdb) p time
$7 = (ASN1_TIME *) 0x0
(gdb) p length
$8 = 13
(gdb) bt
#0  decode_time (der=0x7fffe86ac520 "\027\v1609082343Z", length=13)
    at ../../../../ext/openssl/ossl_asn1.c:441
#1  0x00007fffe8744716 in int_ossl_asn1_decode0_prim (pp=0x7fffffffc648, length=11, 
    hlen=2, tag=23, tc=20667148, num_read=0x7fffffffc5b0)
    at ../../../../ext/openssl/ossl_asn1.c:778
#2  0x00007fffe8744de9 in ossl_asn1_decode0 (pp=0x7fffffffc648, length=13, 
    offset=0x7fffffffc630, depth=0, yield=0, num_read=0x7fffffffc638)
    at ../../../../ext/openssl/ossl_asn1.c:919
#3  0x00007fffe874501f in ossl_asn1_decode (self=140737093809160, obj=140737092699520)
    at ../../../../ext/openssl/ossl_asn1.c:1006
...

[junaruga]

By the way, there are 2 kinds of the errors: OpenSSL::ASN1::ASN1Error: generalizedtime is too short and OpenSSL::ASN1::ASN1Error: utctime is too short.

[junaruga]

I debugged the 2nd error test again into the asn1_ex_c2i function changing the logic.

$ gdb --args ruby -I ./lib -r openssl -e 'der = [%w{ 17 0B }.join].pack("H*") + "1609082343Z".b; OpenSSL::ASN1.decode(der)'
...
(gdb) f
#0  asn1_ex_c2i (pval=0x7fffffff9558, cont=0x7fffcfbe5a2a "1609082343Z", len=11, 
    utype=23, free_cont=0x7fffffff9365 "", it=0x7fffce3a4bc0 <local_it>)
    at crypto/asn1/tasn_dec.c:929
929	            ERR_raise(ERR_LIB_ASN1, ASN1_R_UTCTIME_IS_TOO_SHORT);

(gdb) l
924	        if (utype == V_ASN1_GENERALIZEDTIME && (len < 15)) {
925	            ERR_raise(ERR_LIB_ASN1, ASN1_R_GENERALIZEDTIME_IS_TOO_SHORT);
926	            goto err;
927	        }
928	        if (utype == V_ASN1_UTCTIME && (len < 13)) {
929	            ERR_raise(ERR_LIB_ASN1, ASN1_R_UTCTIME_IS_TOO_SHORT);
930	            goto err;
931	        }
932	        /* All based on ASN1_STRING and handled the same */
933	        if (*pval == NULL) {
(gdb) p len
$2 = 11
(gdb) bt
#0  asn1_ex_c2i (pval=0x7fffffff9558, cont=0x7fffcfbe5a2a "1609082343Z", len=11, 
    utype=23, free_cont=0x7fffffff9365 "", it=0x7fffce3a4bc0 <local_it>)
    at crypto/asn1/tasn_dec.c:929
#1  0x00007fffcdeec2a9 in asn1_d2i_ex_primitive (pval=0x7fffffff9558, 
    in=0x7fffffff9608, inlen=13, it=0x7fffce3a4bc0 <local_it>, tag=23, aclass=0, 
    opt=0 '\000', ctx=0x7fffffff9560) at crypto/asn1/tasn_dec.c:818
#2  0x00007fffcdeea82b in asn1_item_embed_d2i (pval=0x7fffffff9558, in=0x7fffffff9608, 
    len=13, it=0x7fffce3a4bc0 <local_it>, tag=-1, aclass=0, opt=0 '\000', 
    ctx=0x7fffffff9560, depth=1, libctx=0x0, propq=0x0) at crypto/asn1/tasn_dec.c:256
#3  0x00007fffcdeea238 in asn1_item_ex_d2i_intern (pval=0x7fffffff9558, 
    in=0x7fffffff9608, len=13, it=0x7fffce3a4bc0 <local_it>, tag=-1, aclass=0, 
    opt=0 '\000', ctx=0x7fffffff9560, libctx=0x0, propq=0x0)
    at crypto/asn1/tasn_dec.c:118
#4  0x00007fffcdeea321 in ASN1_item_d2i_ex (pval=0x7fffffff9558, in=0x7fffffff9608, 
    len=13, it=0x7fffce3a4bc0 <local_it>, libctx=0x0, propq=0x0)
    at crypto/asn1/tasn_dec.c:144
#5  0x00007fffcdeea375 in ASN1_item_d2i (pval=0x0, in=0x7fffffff9608, len=13, 
    it=0x7fffce3a4bc0 <local_it>) at crypto/asn1/tasn_dec.c:154
#6  0x00007fffcded8edc in d2i_ASN1_TIME (a=0x0, in=0x7fffffff9608, len=13)
    at crypto/asn1/a_time.c:27
#7  0x00007fffe8743cdd in decode_time (der=0x7fffcfbe5a28 "\027\v1609082343Z", 
    length=13) at ../../../../ext/openssl/ossl_asn1.c:440
#8  0x00007fffe8744716 in int_ossl_asn1_decode0_prim (pp=0x7fffffff9798, length=11, 
    hlen=2, tag=23, tc=18135820, num_read=0x7fffffff9700)
    at ../../../../ext/openssl/ossl_asn1.c:778
#9  0x00007fffe8744de9 in ossl_asn1_decode0 (pp=0x7fffffff9798, length=13, 
    offset=0x7fffffff9780, depth=0, yield=0, num_read=0x7fffffff9788)
    at ../../../../ext/openssl/ossl_asn1.c:919
#10 0x00007fffe874501f in ossl_asn1_decode (self=140737093953200, obj=140736678747360)
    at ../../../../ext/openssl/ossl_asn1.c:1006

[junaruga]

OK. It seems that the assertions getting the errors are testing the cases of returning the String format without the seconds.

According to the PDF file: ITU-T X.690 / ISO/IEC 8825-1 mentioned at openssl/openssl#23483 (comment), it seems that the document and/or OpenSSL project thinks that the 000000Z (HHMISSZ) is the minimal valid string, and thinks that the 0000Z (HHMIZ) is invalid.

• section 11.7: GeneralizedTime - 11.7.5 - YYYYMMDD000000Z
• section 11.8: UTCTime - 11.8.3 - YYMMDD000000Z

I think we can remove these assertions.

So, as rhenium mentioned, just dropping the assertions makes sense to me.

[junaruga]

I sent the PR #728.

[rhenium]

Yes, seconds in GeneralizedTime and UTCTime is required in DER, but is optional in BER, according to the document.

OpenSSL's parser hasn't been very consistent; it supports DER and partially BER. It still is more liberal than strictly DER by allowing timezone string, but I think the change itself is OK.

Either way, I think the current assertions were overly specific for the ruby/openssl test suite. The PR looks good to me.