48657 - NSS doesn't work with PR_AF_INET6 (IPv6) sockets

Status: RESOLVED FIXED

(NSS :: Libraries, defect, P3)

Description

[John G. Myers]

The session nonce cache is keyed by the IPv4 address, so attempting to use NSS 
with IPv6 sockets results in an assertion in ssl_GetPeerInfo().

The session cache needs to be changed to be keyed by PRIPv6Addr, with v4 
addresses being converted to v4-mapped v6 addresses using NSPR's 
PR_ConvertIPv4AddrToIPv6().

Comment 1

[John G. Myers]

Attachment: Proposed fix (also fixes some warnings)

Comment 2

[Nelson Bolyard (seldom reads bugmail)]

John,  Although bug 51436 hasn't been marked as resolved/fixed yet,
the fix is checked in to the tip and working.  So, you should be
able to resume testing of your proposed changes.  Note, however,
that other changes to SSL code have been checked in, so you'll 
need to do another update to merge those changes into your sources
before testing.

Comment 3

[John G. Myers]

I still can't get selfserv to work.  With an unmodifed NSS checked out today, 
the handshake completes, but the next packet fails the MAC check.  From the 
trace info, it looks like it didn't decrypt properly.

Comment 4

[Nelson Bolyard (seldom reads bugmail)]

Seems to be working for me.  What platform are you testing on?

Comment 5

[John G. Myers]

I found my problem, it's bug 52192.  I'm back to testing.

Comment 6

[John G. Myers]

I've verified that the patch works with selfserv, though I wasted a fair amount 
of time due to the fact that header dependencies are broken in the build system.

Comment 7

[John G. Myers]

Attachment: Updated fix

Comment 8

[John G. Myers]

Fix checked in

Comment 9

[John G. Myers]

Please consider back-merging this to NSS 3.0.1.

Comment 10

[John G. Myers]

Applying this patch to 3.0.1 will allow the patch for PSM to support IPv6 to go 
in before PSM switches to NSS 3.1.

Comment 11

[Nelson Bolyard (seldom reads bugmail)]

John,  I'm marking this fixed for NSS 3.1 (Again) so this won't 
hold up NSS 3.1.  
 
We're considering your request to carry it back to 3.0.1.  
If we decide to do it, you or I can open another bug with a 
different target fix release.