You should address your concern to Linus and the kernel developer mailing list. :)

However hugetlb_shm_group already address this problem giving access to a specific group.

It is a start.

Fabrizio

________________________________

Da: Alexei_Roudnev@(protected)]
Inviato: mer 25/01/2006 21.34
A: Magni Fabrizio
Cc: Alexei_Roudnev; Drapala, Jason; suse-oracle@(protected)
Oggetto: Re: R: [suse-oracle] hugepage on 64-bit question



It is the whole thing - approach is WRONG.

No any resource should be owned by root (root-owning cause extra processes to have ful root privileges). Resources should be owned by groups, so that you can always assign resounce control to the user by assigning group (it is not only about hugetlb).

Old and well known design flow.

Alex Roudnev
Exigen Group
415 806 3741





"Magni Fabrizio" <Fabrizio.Magni@(protected)>

01/25/2006 12:16 PM

To
"Alexei_Roudnev" <Alexei_Roudnev@(protected)>
cc
Subject
R: [suse-oracle] hugepage on 64-bit question  

   




Alexei,
the problem was known (but I didn't understood you were hitting it) but it is not really kernel related.

The hugepages are "owned" by root and can be used only by the superuser.
The mcap was a trick to permit other users to exploit them. So the suse kernel is giving you an option which on a normal 2.6.5 kernel you don't have.

On more adavnced kernel versions (probably we will see them with SLES10) the disable_cap_mlock should not be there anymore.

hugetlb_shm_group is going to take its place.

Fabrizio


________________________________

Da: Alexei_Roudnev [mailto:Alexei_Roudnev@(protected)]
Inviato: mer 25/01/2006 21.10
A: Magni Fabrizio; Drapala, Jason; suse-oracle@(protected)
Oggetto: Re: [suse-oracle] hugepage on 64-bit question


So, as we know now, problem depend of the kernel, and can be resolbed by
vm.mcap* variable setting.

----- Original Message -----
From: "Magni Fabrizio" <Fabrizio.Magni@(protected)>
To: "Alexei_Roudnev" <Alexei_Roudnev@(protected)"
<jdrapala@(protected)>
Sent: Tuesday, January 17, 2006 9:44 PM
Subject: RE: [suse-oracle] hugepage on 64-bit question



> I turned off hugepage (one of scripts do it) - it was the
> only installation problem. May be you should configure
> hugepages in the kernel (I advice agains, except if you have
> > 16 GB of memory). SLES9 have some inconsistency here (with
> hugepages).
>

Just fo asking Alexei,
Which problem did you find?
I haven't seen any with 10gR2 so far.

Can you even describe the inconsistences?

Thanks
Fabrizio

--
To unsubscribe, email: suse-oracle-unsubscribe@(protected)
For additional commands, email: suse-oracle-help@(protected)
Please see http://www.suse.com/oracle/ <http://www.suse.com/oracle/> before posting