James Morris [off-list ref] wrote:

I'd like to see some serious effort at code review and testing before this 
code is merged.

With regard to testing, I've run multiple simultaneous instances of a number
of test scripts against it continuously for the best part of a day:

 (1) A script to generate completely random data and attempt to stuff that
     into a key.  The completely random data blob is fed wholly and then
     partially in decreasing amounts to keyctl padd.  Then the script loops
     and starts again.

	./fuzz-x509.sh /tmp/data1

 (2) A script to generate random valid ASN.1:

	while :; do ./asn1random.pl | keyctl padd asymmetric vlad @s; done

 (3) A script to generate correctly formatted X.509 certificates filled with
     random data, including for the RSA key and signature fields.

	while :; do ./x509random.pl | keyctl padd asymmetric vlad @s; done

 (4) A variant of (3) that injects random bytes into the structure, whilst
     correctly maintaining the length counts outside of those.

	while :; do ./x509random.pl -i | keyctl padd asymmetric vlad @s; done

 (5) A script to repeatedly generate valid X.509 certificates and stuff those
     in, and then generate valid PKCS#7 signatures over random data and try to
     stuff those in too (which should fail).

	./x509-stuffer.sh

David