[Openswan dev] OpenSwan 2.6.10-1 on OpenWrt 7.09 consistently hangs on large HTTP file transfer

[starlight at binnacle.cx]

I'd don't think it was ever a memory problem.  I'm just 
providing the requested info to help diagnose the issue(s).

Any chance one of the developers has a test-bed with a pair of 
WRT54Gs running back-to-back?  Seems to me the problem might be 
reproducible with a huge FTP through such a test setup, and much 
much faster.

I'm running another file transfer now.  Figured out that the 
other end has put QOS on a per-connection level, so I'm getting 
double the transfer rate by running two in parallel.


At 12:01 AM 12/5/2007 -0500, Paul Wouters wrote:
>On Tue, 4 Dec 2007, starlight at binnacle.cx wrote:
>
>> Ok, finally reproduced the problem.  Strangely this time
>> OpenSwan did not lock up, but instead the remote could not
>
>So this might be a different issue. Especially because it does
>not show any memory problems in your logs.
>
>> Anyway here is the failure log.  The blowout is at the 18:38
>> time interval.  Traffic was flowing from the 10.81.82.5.
>
>And no odd signs at that time either.
>
>So you start with:
>
>        total:    used:    free:  shared: buffers:  cached:
>Mem:  14647296 14008320   638976        0   999424  5206016
>Swap:        0        0        0
>MemTotal:        14304 kB
>MemFree:           624 kB
>
>and you end with:
>
>        total:    used:    free:  shared: buffers:  cached:
>Mem:  14647296 14032896   614400        0   999424  4964352
>Swap:        0        0        0
>MemTotal:        14304 kB
>MemFree:           600 kB
>
>At the beginning your're briefly losing 300kb, but then you stay
>stable on 600kb free. So memory isn't reall your problem, though
>I guess you are running a bit low, but you have a 4.9MB cache
>space to take memory from. No dip around 18:38 either
>
>Comparing the first and last slabinfo, and the one at 18:38, 
>there aren't
>any values that have suddenly changed beyond a few percent.
>
>I don't think you are running out of memory. But I also do not 
>know what
>your problem is.
>
>At least, not on this bug. Perhaps when it crashes, things are 
>different.
>
>Paul