RFR: 8132330: Ineffective SecureRandom usage in RSA encoding with PKCS1Padding

[David Schlosnagle]

What happens in the case Where there are not enough non-zero random bytes
in the new version?

On Wednesday, August 5, 2015, Sergey Kuksenko <sergey.kuksenko at oracle.com>
wrote:

> Hi All,
>
> Please review changes to the following performance improvement:
>
> https://bugs.openjdk.java.net/browse/JDK-8132330
>
> Webrev: http://cr.openjdk.java.net/~skuksenko/crypto/8132330/webrev.02/
>
> Sponsorship is required.
>
> ------
> Currently sun.security.rsa.RSAPadding::padV15 uses fixed size (64 bytes)
> buffer for obtaining random bytes from SecureRandom.
> Here we got two sources of inefficiency:
> - when length of required padding is greater than 64 -> several calls of
> SecureRandom.nextBytes are required that increases contention blocking on
> SecureRandom
> - when length of required padding is less than 64 it causes unnecessary
> overhead (e.g. nextBytes for 16 bytes is 4x times faster than for 64 bytes)
>
> Suggested performance improvement shows RSA encoding (public key) speedup:
> - 10%-18% when NativePRNG SecureRandom is used,
> - 5%-7% when SHA1PRNG is used.
>
>
> --
> Best regards,
> Sergey Kuksenko
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/security-dev/attachments/20150807/9d20267f/attachment.htm>