Subject: Re: increase default for number of files descriptors per process (was Re: Speeding up "pstat -T")
To: der Mouse <mouse@Rodents.Montreal.QC.CA>
From: Andrew Brown <atatat@atatdot.net>
List: tech-kern
Date: 10/05/2003 22:41:04
On Sun, Oct 05, 2003 at 08:24:41PM -0400, der Mouse wrote:
>[atatat]
>>> allow me to rephrase:
>[perry]
>> Your rephrasing doesn't actually make it any easier for me to
>> understand what you intended...
>
>There are too many "hard limit"s floating around.  I think I understand
>what atatat meant (check me?):  The "hard limit" that is putatively to
>be "eliminate"d is not the RLIMIT_STACK rlim_max, but rather the cap
>above which not even root is allowed to raise RLIMIT_STACK's rlim_max.

yeah.  that.  there could be a sysctl limit that said "independent of
who thou art or whom thou knowest, thou mayest not cross this
threshold", i don't know if that would be worth all that much.

>>> this is not to say that there would not be a limit to the hard
>>> limit; only that it would not be set at kernel compile time.
>
>"That cap will still exist; it just will be set at boot time, not
>kernel build time."  This actually prompts me to ask, will it be
>modifiable without rebooting, such as with a sysctl?  (Obviously the
>new value would apply only to execs that happen after the change,
>unless you want to get fancy.)

if it was there, it would be sysctlable.  if it was not, it would
probably be something like SSIZE_MAX or something.  or maybe INT_MAX.

-- 
|-----< "CODE WARRIOR" >-----|
codewarrior@daemon.org             * "ah!  i see you have the internet
twofsonet@graffiti.com (Andrew Brown)                that goes *ping*!"
werdna@squooshy.com       * "information is power -- share the wealth."