RFR 8226543 : Reduce GC pressure during message digest calculations in password-based encryption

Sean Mullan sean.mullan at oracle.com
Tue Jun 25 17:55:24 UTC 2019


Hi Ivan,

This fix looks good.

--Sean

On 6/20/19 10:22 PM, Ivan Gerasimov wrote:
> Hello!
> 
> In PBES1Core.deriveCipherKey() there are loops that look like following:
> 
>              for (int i = 0; i < iCount; i++) {
>                  md.update(toBeHashed);
>                  toBeHashed = md.digest(); // this resets the digest
>              }
> 
> Calling md.digest() each time allocates a new byte[] array.
> It would be more GC friendly, if the same array were reused.
> 
> In the fix, a first call to md.digest() allocates a properly sized array.
> Subsequent iterations reuse this array.
> 
> Would you please help review?
> 
> BUGURL: https://bugs.openjdk.java.net/browse/JDK-8226543
> WEBREV: http://cr.openjdk.java.net/~igerasim/8226543/00/webrev/
> 
> No new regression test, as the behavior was not changed.
> Mach5 control job is green (tiers 1-4 on all supported platforms).
> 


More information about the security-dev mailing list