Re: Bug in net/ipv6/ip6_fib.c:fib6_dump_table()

[Josh Hunt <johunt@akamai.com>]

On 06/12/2012 12:22 PM, Debabrata Banerjee wrote:
> Looks like commit 2bec5a369ee79576a3eea2c23863325089785a2c "ipv6: fib:
> fix crash when changing large fib while dumping" is the culprit. The
> result of this code is that if there is a tree addition while a dump
> has suspended because the netlink skb is full, it will simply go back
> to the top of the tree and you end up with duplicate/triplicate/etc
> routes. It looks like the code attempts to count nodes, but it's a
> linear count and the data structure is a tree so that's a big problem.
> The net result is potentially DOSable, since if route table updates
> happen often enough in proportion to table size, a dump will attempt
> to return an infinite amount of routes (observed). So this commit
> should be reverted. However I am interested in the problem that commit
> tried to solve, if anyone has more information on that. My assumption
> is the fib tree gets corrupted and eventually it crashes in
> fib6_dump_table(), which I assume can still happen.
> 
> I can easily demonstrate the bug by adding cloned/cache routes while I
> check the results of fib6_dump_table:
> 
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 593
> 189
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 884
> 16
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 888
> 78
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 507
> 507
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 533
> 533
> root@a172-25-43-12.deploy.akamaitechnologies.com:~# ip -6 -o route
> show table cache |tee tmp | wc -l; sort tmp | uniq -u | wc -l
> 571
> 571
> 
> Thanks,
> Debabrata

Ping?

Can anyone provide details of the crash which was intended to be fixed
by 2bec5a369ee79576a3eea2c23863325089785a2c? With this patch in and
doing concurrent adds/deletes and dumping the table via netlink causes
duplicate entries to be reported. Reverting this patch causes those
problems to go away. We can provide a more detailed test if that is
needed, but so far our testing has been unable to reproduce the crash
mentioned in the above commit with it reverted.

Thanks
Josh