Hi,
I've been away 1.5 days *grin*. Thanks for the replies.
Let's display some anti-social behaviour and reply to all your
mails at once :P
--
> Alan Cox <[EMAIL PROTECTED]>
> You can do what you like. If its under the gdb debugger it wont be
> setuid
I suspected that. Perhaps I should have tried before asking ;)
> No. And a setuid program won't dump core unless run as the id it would
> setuid to anyway
?
In the system state I referred to *any* program dumped core; I will
recheck this for setuid programs once I complete this mail (sorry for
garbling your reply).
--
> Joe <[EMAIL PROTECTED]>
>
> what does your code look like? something like that shown
> below?
>
> for (i=1;i>0; i++) {
> somevar = (char*)malloc(1024);
> }
Yes. Including the fork in your next mail.
>
> if you have users doing this then you have other problems.. I
The only one causing trouble is me :)
Hmm. A spammer could perhaps easily 'freeze' your system by addressing
an unconfigured sendmail (unconfigured = no throttle, no
maxdaemonchildren). But then again, I'm not sure.
> think ANY system/OS would have problems as it is equivalant to a
> memory leak ...
Few mails ago someone wanted to use 1/2 Gig for IPC as non-root. I'd
like to have /proc access to user resource limits (not necessarily on
a per user basis). Rik van Riel wrote a patch to kill processes when
running out of memory; quota would suffice for limiting disk resources,
I guess...
--
> Jesse Pollard <[EMAIL PROTECTED]>
> This is also a call for per/user and per/process limits. You can also
You're one step ahead ;) That would be really nifty. Any ideas how
efficient/fast it would be compared to current implementation?
> hang the system by using too many processes/open file ids ... which in
Well, I think the number of processes is limited 255 by default; if you
leave 5% of that (and some mem) to root-only (as Robert proposes below), I
would be able to fix the problem.
--
> Robert Sander <[EMAIL PROTECTED]>
> IMHO he wants free "process-numbers" for root like the 5% free space
> on the filesystem. Would that be possible?
That's about what I meant. Ideally I would like to be able to login
as root no matter how high the load (i.e. keep some resources left)
Perhaps this could be managed through a cute /proc interface?
--
Thanks again,
xander van wiggen
PS. you _can_ send mail to this address as I'm testing on another system
:)
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
