Hi everybody,

Does anybody know if Solaris 8 has a kernel parameter that limits the
maximum stack size per process?
I am referring to the 'hard' limit. HP-UX has a kernel parameter called
maxssiz (maximum stack size).
The only parameter I've seen in the docs is lwp_default_stksize but it
sets the default minimum.

Thanks a lot in advance,

Max

Hi Max, see man limit, ulimit and  rlim_fd_max in /etc/system

http://docsun.cites.uiuc.edu/sun_docs/C/solaris_9/SUNWaadm/SOLTUNEPARAMREF/p13.html

Mariano

max ha escrito:

> Hi everybody,
>
> Does anybody know if Solaris 8 has a kernel parameter that limits the
> maximum stack size per process?
> I am referring to the 'hard' limit. HP-UX has a kernel parameter called
> maxssiz (maximum stack size).
> The only parameter I've seen in the docs is lwp_default_stksize but it
> sets the default minimum.
> 
> Thanks a lot in advance,
> 
> Max

"max" <maxw_cc@yahoo.com> writes:

>Does anybody know if Solaris 8 has a kernel parameter that limits the
>maximum stack size per process?

There's a "ulimit" setting which you could use.

I don't think it's settable in the kernel.

>I am referring to the 'hard' limit. HP-UX has a kernel parameter called
>maxssiz (maximum stack size).
>The only parameter I've seen in the docs is lwp_default_stksize but it
>sets the default minimum.

It doesn't set the minimum size; it sets the *kernel*'s stack
size.

You should leave it alone.

Casper
-- 
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

First of all thank you very much to all for taking the time to help me
in this matter.

Casper,

>Casper H.S. Dik wrote:
> "max" <maxw_cc@yahoo.com> writes:
>
> >Does anybody know if Solaris 8 has a kernel parameter that limits the
> >maximum stack size per process?
>
> There's a "ulimit" setting which you could use.
>
> I don't think it's settable in the kernel.

So there's only a 'soft limit' and the process can setrlimit() its
rlim_max of its stack to RLIM_INFINITY?

>
> >I am referring to the 'hard' limit. HP-UX has a kernel parameter called
> >maxssiz (maximum stack size).
> >The only parameter I've seen in the docs is lwp_default_stksize but it
> >sets the default minimum.
>
> It doesn't set the minimum size; it sets the *kernel*'s stack
> size.
>
> You should leave it alone.
>

Thank you for the information about the lwp_default_stksize parameter.
I know I may be digressing a little bit but could you please explain me
basically how it is in terms of LWPs and kernel threads, I mean LWPs
are user threads running under a kernel thread? Could you please give
me some information about it? That would help me to understand more the
kernel parameters related to memory.


Thank you so much in advance,

Max

>So there's only a 'soft limit' and the process can setrlimit() its
>rlim_max of its stack to RLIM_INFINITY?

No; because a hardlimit could conceivably be set.  E.g., by
setting the process.max-stack-size rctl through assigning the
user a project.

>Thank you for the information about the lwp_default_stksize parameter.
>I know I may be digressing a little bit but could you please explain me
>basically how it is in terms of LWPs and kernel threads, I mean LWPs
>are user threads running under a kernel thread? Could you please give
>me some information about it? That would help me to understand more the
>kernel parameters related to memory.


The lwp_default_stksize controls the size of the *kernel stack*
for all *kernel* threads.  Each "LWP" in userland has an associated
kernel thread stack which is used to execute system calls.
And then there are many kernel threads doing things like NFS
or asynchronous bits.

The reason it is tunable is that some vendors of software allocate
too much stuff on the stack and this causes kernel stack
overflows; so it's there to work around bugs in device drivers.

Casper
-- 
Expressed in this posting are my opinions.  They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.

Thank you very much Casper,

> >So there's only a 'soft limit' and the process can setrlimit() its
> >rlim_max of its stack to RLIM_INFINITY?
>
> No; because a hardlimit could conceivably be set.  E.g., by
> setting the process.max-stack-size rctl through assigning the
> user a project.

Let me see if I understand you, so  if the administrator hasn't
properly yet configured the project database entry that corresponds to
the user  with the proper limits, there are no default system-wide
enforced constraints to this situation?

Thanks a lot,

Max

"max" <maxw_cc@yahoo.com> writes:

>Thank you very much Casper,

>> >So there's only a 'soft limit' and the process can setrlimit() its
>> >rlim_max of its stack to RLIM_INFINITY?
>>
>> No; because a hardlimit could conceivably be set.  E.g., by
>> setting the process.max-stack-size rctl through assigning the
>> user a project.

>Let me see if I understand you, so  if the administrator hasn't
>properly yet configured the project database entry that corresponds to
>the user  with the proper limits, there are no default system-wide
>enforced constraints to this situation?


Correct; there's a 8MB softlimit and no hard limit.

Casper