Re: nessus on AMD64

On Wednesday 21 January 2004 13:26, Renaud Deraison wrote:
> You still did not tell me if standalone nasl was working. If it's not,
> then there's no point in debugging nessusd at this time.

Standalone nasl works just fine everything runs with it OK.

I am beginning to suspect some kind of bug in (dare I say it...) glibc or gcc!  
If that is the case, it is going to be hell to come up with a test case to 
prove my point.

I have been playing around adding printf statements in libnessus service.c and 
store.c where they do the mmap.  I then print the mallinfo structure.  The 
numbers look fine on i386 but seem to be a bit large on the amd64.

I have also been using the system monitor to look at the memory usage numbers 
for the various nessusd processes such as "serving" and "testing".  On the 
i386 these go from about 2.4MB to a bit less than 4MB.  However, on the amd64 
they go up to around 6MB and this is with lots of messages about being unable 
to allocate memory.

So I tried using mallopt in nessusd.c to set the M_MMAP_THRESHOLD to a very 
low number (64).  Memory usage quickly grows to about 125MB but a lot (not 
all) of the memory allocation error messages go away.  If I set the threshold 
to about 1024, then the mmap file mapping errors go away.

Running the low threshold on the i386 seems to have little effect in that 
memory usage stays below 4MB.

While pointer size increases from 4 bytes to 8 bytes, that does not explain 
this kind of behavior.

I will take a look at valgrind to see if it can help.
-- 
Gene

_______________________________________________
Nessus-devel mailing list
[email protected]
http://mail.nessus.org/mailman/listinfo/nessus-devel