Review for fix to Bug 7078386

[Michael McMahon]

Brandon,

The fix looks good to me. Thanks for contributing it.

- Michael.

On 15/12/11 23:14, Brandon Passanisi wrote:
> Hello net-dev.  I was wondering if somebody could review the proposed 
> fix for the following bug:
>
>     Bug URL: http://monaco.sfbay.sun.com/detail.jsf?cr=7078386
>     Webrev : http://cr.openjdk.java.net/~michaelm/7078386/webrev.1/
>
>
> The bug is caused by an incorrect format string used in the fscanf 
> call within enumIPv6Interfaces of 
> jdk/src/solaris/native/java/net/NetworkInterface.c, which specifies a 
> two-digit hex number for the network interface index value.  It has 
> been shown that network interface index values can be more than 255, 
> thus requiring the need to obtain more than two digits.  I eventually 
> searched through the following linux kernel source file:
>
>     linux-source-2.6.32/net/ipv6/addrconf.c
>
> ... and it appears the interface index value is being written and an 
> integer within the function inet6_fill_ifinfo in the following block 
> of code starting on line 3785:
>
>         if (dev->ifindex != dev->iflink)
>             NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
>
> iflink is defined in linux-source-2.6.32/include/linux/netdevice.h as 
> an int on line 742:
>
>         /* Interface index. Unique device identifier    */
>         int            ifindex;
>         int            iflink;
>
> Because of this and because an integer is 32 bits, it appears that 
> updating the format string to use %08x for the interface index value 
> seems most correct.
>
> I have tested out this fix successfully in the following ways:
>
>     1. I ran the IBM-supplied test programs tapadd and Scope to create
>     over 255 address in order to test an interface index value of at
>     least three digits.  (These programs will be added to the bug
>     report if not present at this time.)  The tapadd program is a C
>     test requiring root access that uses tun/tap to create over 255
>     interfaces.  Scope is a java program that does a
>     NetworkInterface.getByName() on a known interface with an index
>     over 255.
>
>     2. I modified the Scope program to additionally test all created
>     interfaces, not just one interface with an index value over 255. 
>     (The original Scope test program just checks a single interface.) 
>     This is important to show there aren't regressions.
>
>     3. I created another test program using a script to run 'ip'
>     commands to create over 255 interfaces.  This script runs in a
>     loop using the commands "sudo ip tunnel add..." and "sudo ip -6
>     addr add..." to add the interfaces.  Then, I used the Scope
>     program to verify the created interfaces.
>
> I wasn't able to create a script that would create a really huge 
> amount of interfaces because using numbers like 65536 or higher in my 
> script caused my Ubuntu linux to hang.  So, I'm not quite sure how to 
> truly test large interface index values like 65536 or more.
>
> Thanks.
> -- 
> Oracle <http://www.oracle.com>
> Brandon Passanisi | Principle Member of Technical Staff
>
> Green Oracle <http://www.oracle.com/commitment> Oracle is committed to 
> developing practices and products that help protect the environment

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/net-dev/attachments/20111216/8e75c974/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 658 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/net-dev/attachments/20111216/8e75c974/attachment.gif 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 356 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/net-dev/attachments/20111216/8e75c974/attachment-0001.gif