On Tue, Apr 11, 2017 at 11:34:56PM +0100, Ben Hutchings wrote: > On Tue, 2017-04-11 at 22:55 +0200, Fernando Santagata wrote: > > On Mon, Apr 10, 2017 at 05:23:26PM +0100, Ben Hutchings wrote: > > > Control: tag -1 moreinfo > > > > > > On Sun, 2017-04-09 at 12:09 +0200, Fernando Santagata wrote: > > > > I think this is related to this thread in the linux-mm mailing list, > > > > dating > > > > back to kernel version 4.7, the first one that exibits this behavior: > > > > > > > > https://lists.gt.net/linux/kernel/2528084 > > > > > > > > This error shows even when using concurrent programming under Perl6, so > > > > it > > > > seems to be really related to sharing memory. > > > > > > > > The last usable kernel in this respect is version 4.6. > > > > > > The warning message tells you how to work around this (add kernel > > > parameter 'ignore_rlimit_data'). Doesn't that work? > > > > I only tried fiddling with ulimit, but ignore_rlimit_data works fine. > > > > I suggest adding either a warning in some doc file, or making that > > option the default, lest you receive dozens of stupid bug reports :-) > > I certainly don't want to disable a resource limit globally - the > failure to apply the data size limit was itself a bug. And I hope that > it is not very common to set data size limits that are too low, so we > don't have to document yet another change in the release notes that > will mostly be ignored. > > Can you please check whether the data limit is already set in a shell > (ulimit -d)?
Right now is at its default state: $ ulimit -a -t: cpu time (seconds) unlimited -f: file size (blocks) unlimited -d: data seg size (kbytes) 131072 -s: stack size (kbytes) 8192 -c: core file size (blocks) 0 -m: resident set size (kbytes) unlimited -u: processes 31447 -n: file descriptors 1024 -l: locked-in-memory size (kbytes) 64 -v: address space (kbytes) unlimited -x: file locks unlimited -i: pending signals 31447 -q: bytes in POSIX msg queues 819200 -e: max nice 0 -r: max rt priority 0 -N 15: unlimited previously I tried to increase it to 200000, but it didn't work. I guess the problem is with the heap, not the data segment. -- Fernando Santagata