Re: kern/54188: uvm/pmap panic

[Patrick Welche <prlw1%cam.ac.uk@localhost>]

The following reply was made to PR kern/54188; it has been noted by GNATS.

From: Patrick Welche <prlw1%cam.ac.uk@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: kern/54188: uvm/pmap panic
Date: Fri, 10 May 2019 10:13:37 +0100

 /var/crash was out of space - I have a coredump!
 
 savecore: reboot after panic: [ 47842.1683276] panic: pmap_get_physpage: out of memory
 savecore: system went down at Fri May 10 03:21:27 2019
 
 so yes for a run of "daily".
 
 ps shows loads of g++:
 
 PID    LID S CPU     FLAGS       STRUCT LWP *               NAME WAIT
 3630     1 3   4   8020080   ffff99aeacd729e0                 as pipe_rd
 25629    1 3   0   8020000   ffff99addf88c560            cc1plus flt_noram5
 17341    1 3   4   8020080   ffff99aeb1b9ca00                g++ wait
 238      1 3   4   8020080   ffff99add4561240                 sh wait
 4294     1 3   2   8020080   ffff99aeb1b9c5c0                 as pipe_rd
 9653     1 3   1   8020000   ffff99af64ba5220            cc1plus flt_noram5
 18269    1 3   2   8020080   ffff99aeb1b9c180                g++ wait
 24735    1 3   2   8020080   ffff99ad51690600                 sh wait
 28406    1 3   2   8020080   ffff99af6e5106e0                 as pipe_rd
 5623     1 3   4   8020000   ffff99aeb02f46c0            cc1plus flt_noram5
 18465    1 3   2   8020080   ffff99af2ff34120                g++ wait
 8159     1 3   2   8020080   ffff99af5d2d5340                 sh wait
 18040    1 3   6   8020080   ffff99aceae39200                 as pipe_rd
 8094     1 3   2   8020000   ffff99addf88c9a0            cc1plus flt_noram5
 13491    1 3   6   8020080   ffff99ad5bab5740                g++ wait
 11886    1 3   6   8020080   ffff99af77c54480                 sh wait
 17974    1 3   3   8020080   ffff99ad50e880a0                 as pipe_rd
 ...
 
 so the poor box did simply run out of its 16GB memory
 plus 16GB swap as per the panic message. Presumably this
 can just be closed...
 
 (It seems all that work on debuggers is paying off - I could make
 use of the core in gdb and crash.)