Re: -current kernel can't be complied for 3 days

From: Kris Kennaway (kris_at_obsecurity.org)
Date: 07/24/05

  • Next message: Michal Mertl: "Re: Recent fragility with if_wi, 802.11 adhoc/wep, and Tiger" 
    Date: Sun, 24 Jul 2005 15:25:30 -0400
To: Oliver Lehmann <lehmann@ans-netz.de>

    On Sat, Jul 23, 2005 at 09:13:49PM +0200, Oliver Lehmann wrote:
    > M. Warner Losh wrote:
    >
    > > Chances are you are running out of memory. Can you try the build w/o
    > > -jN? Large values of N can easily run a body out of swap space,
    > > yeilding a build failure when the tree is perfectly fine.
    > >
    > > Persoanally, I never do a -j larger then 4 or 8. top shows that the
    > > system that I have is giant bound when I do, so why add more people to
    > > the list of Giant waiters?
    >
    > Hi Warner,
    >
    > I don't think that i ran out of memory. At least swapper didn't sayed
    > sth. like that. Back in Dec 2001 I even had a -j100 running (for fun) and
    > it build successfully. This was on a dual PII 333 with 320MB of memoty
    > compiling a 4.XX on a 4.XX.
    > Why shouldn't a -j16 on a dual PIII 850 with 640MB of memory compiling a
    > 6.0 on a 5.4 doesn't work then? I know there are many differences between
    > 4.X and 5.X but why got it worse?
    > A -j4 worked btw (at least the one time I tried it)

    There is a default limit of 16 simultaneous execs that can be easily
    saturated on an SMP system (further execs cause aborts as you are
    seeing). Increase the value of the vm.exec_map_entries tunable.

    Kris

  • Next message: Michal Mertl: "Re: Recent fragility with if_wi, 802.11 adhoc/wep, and Tiger"