Date: Wed, 22 Jan 1997 14:32:43 +1100 From: Bruce Evans <bde@zeta.org.au> To: imp@village.org, msmith@atrad.adelaide.edu.au Cc: hackers@FreeBSD.org Subject: Re: Cannot fork Message-ID: <199701220332.OAA32470@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>> Any ideas on how I might track down this problem? I'm rebuilding my >> kernel now with MAX_CHILDREN set to be 256 rather than whatever the MAX_CHILDREN is not an option. You probably mean CHILD_MAX. CHILD_MAX should never have been an option and now has no effect in -current. Limits are now set according to data in /etc/login.conf. Some of the limits are too small. I just hit the 8MB filesize limit when I tried to run `iozone 16'. This limit used to be broken (running a subprocess used to increase the limit to RLIM_INFINITY) so the problem wasn't evident. `su' to root to change the limits gave even smaller limits initially: `default' limits after login: cpu time (seconds, -t) unlimited file size (512-blocks, -f) 16384 <- too small data seg size (kbytes, -d) 16384 stack size (kbytes, -s) 2048 <- too small core file size (512-blocks, -c) 16384 max memory size (kbytes, -m) 30720 locked memory (kbytes, -l) 10240 max user processes (-u) 64 open files (-n) 64 `standard' limits after su: cpu time (seconds, -t) 5400 <- smaller file size (512-blocks, -f) 16384 data seg size (kbytes, -d) 8192 <- smaller stack size (kbytes, -s) 2048 core file size (512-blocks, -c) 16384 max memory size (kbytes, -m) 8192 <- smaller locked memory (kbytes, -l) 4096 <- smaller max user processes (-u) 32 <- smaller open files (-n) 24 <- smaller >What does 'limit maxproc' say? I usually raise it to 200 or so in >my .xsession; the default of 40 or so is very tight for heavy X work. This now only works for root, since login.conf specifies that the hard limits and the soft limits are set to the same (often small) values. Class `xuser' has slightly larger limits. I think that only the soft limits should be restricted for trusted interactive users, and most of the soft limits should be about twice as large (same as the old non-bogus kernel defaults for the non-X case). Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199701220332.OAA32470>