[Openswan dev] IPComp

[Dominique Blas]

Le vendredi 2 Juillet 2004 17:57, Paul Wouters a écrit :
> On Fri, 2 Jul 2004, D. Hugh Redelmeier wrote:

Dear all,

Firsteval, I'd should say thank you both for the time you spent on my question.
On second hand I should apologize for a point I forgot to mention (see therafter [*]).

Well, I've made several tests with different kernels (the one this use now and for witch I produce the barf you know) and a 
stock 2.6.7 with a few patches from netfilter (pom-ng-20040621).
Both passed successfully the tests. 

There was no failed memory allocation, no leak of memory (as far as I can see under 1GB of memory and within a few minutes)
whatever there was compression or not and whatever the ESP algorithm
was (3DES or AES). Moreover, I didn't observe any << shrinkage >> to a 348 bytes packet when using compression.
pings of large packets (> 1500) work well.

So, I'll ask monday the guys to upgrade the machine to 128 M and I think, thus, the problem will vanish and OS2 definitively adopted.


The conclusion is that OS2 is definitively greedier than racoon : racoon works well under the same condition but has other major drawbacks for me as I explained.

Again, thank you to all, sincerely

Best,

db

[*]In fact, I have been so accustomed to work this way for several years I completely ignored that the machine for which I send you
the barf is not a conventional machine ... with a hard disk.
that is a machine that boots on a flash memory and so ... the RAM is shared between the live memory and a emulation of mass memory.
Currently the ratio is (I've just check) 29M for the disk and thus 35 M for the RAM.
Paul, that is what is hidden behind buffers size : 30 M is about the same as 29 M.
And thus there is only 1892 kB left not enough probably for OS2 to work in good condition.