Re: ulimit -d/-m do not actually work

[Greg Troxel <gdt%lexort.com@localhost>]

mlelstv%serpens.de@localhost (Michael van Elst) writes:

> malloc doesn't use the "data segment", so that limit doesn't apply to it.
>
>             -d          the data segment size of a process (kilobytes)
>
>      RLIMIT_DATA     The maximum size (in bytes) of the data segment for a
>                      process; this defines how far a program may extend its
>                      break with the sbrk(2) system call.

Thanks.  So that solves why "ulimit -d" doesn't do what someone would
expect it to do.

> You can try sbrk(2), it should still obey the data segment size (and
> may fail if you use it concurrently with malloc()).

Sure, but my real goal is to have a memory limit that works, so that I
can test that software isn't unreasonably piggy, and to protect the
system.

> "max memory size" is about physical memory usage and isn't a hard limit.
>
>             -m          the total physical memory that can be in use by a
>                         process (kilobytes)
>
>      RLIMIT_RSS      The maximum size (in bytes) to which a process's resident
>                      set size may grow.  This imposes a limit on the amount of
>                      physical memory to be given to a process; if memory is
>                      tight, the system will prefer to take memory from
>                      processes that are exceeding their declared resident set
>                      size.

Interesting, but that seems to be a NetBSD extension.   Looking at
POSIX, -m is not defined:

  https://pubs.opengroup.org/onlinepubs/9799919799/

but setrlimit defines RLIMIT_AS (and not RLIMIT_RSS):

  https://pubs.opengroup.org/onlinepubs/9799919799/functions/getrlimit.html

Our shell has -m and -v, both not specified by POSIX, and these
therefore must be about RLIMIT_RSS and RLIMIT_AS respectively.

> Things were much easier when these limits were invented, in particular
> without multithreading and shared libraries, a single heap (the
> "data segment") was sufficient for all memory allocations.

Sure, but it's really surprising that POSIX does not define a way to
limit total usage.

My test program can allocate 53548 kB, when total vsize is 200000 kB

$ ulimit -v 200000
$ ./test-limits
53548

so this does limit memory usage.

Thanks for making me realize that "memory" in -m is pages allocated to
real memory.