setrlimit() does not check for excess "cur" values

[Lavrentiev, Anton (NIH/NLM/NCBI) [C]]

Hi,

It looks like Cygwin's implementation of setrlimit() does not check
whether a "cur" value being set for a resource has been requested
higher than the "max" value.  It should have returned EINVAL.

From the documentation (Linux):
EINVAL resource is not valid; or, for setrlimit(): rlim->rlim_cur was greater 
than rlim->rlim_max.

Also noted for Linux in the "BUGS" section:
Kernels before 2.4.22 did not diagnose the error EINVAL  for  setrlimit()  when 
 rlim->rlim_cur  was  greater  than rlim->rlim_max.

Test case:

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/resource.h>

int main(void)
{
    struct rlimit rl;

    memset(&rl, 0, sizeof(rl));

    rl.rlim_cur = 10;
    rl.rlim_max = 1;

    if (setrlimit(RLIMIT_CORE, &rl) != 0)
        perror("RLIMIT_CORE");

    rl.rlim_cur = RLIM_INFINITY;
    if (setrlimit(RLIMIT_NOFILE, &rl) != 0)
        perror("RLIMIT_NOFILE");

    return 0;
}

Compilation:
gcc -Wall test.c

Execution in Cygwin:
> ./a.exe
(no output)

Execution on Linux:
> ./a.out
RLIMIT_CORE: Invalid argument
RLIMIT_NOFILE: Invalid argument

The following patch is suggested for cygwin/resource.cc:

*** resource.cc Wed May  1 14:10:28 2013
--- resource.cc.orig    Mon Apr  8 21:04:52 2013
*************** setrlimit (int resource, const struct rl
*** 173,181 ****
    if (getrlimit (resource, &oldlimits) < 0)
      return -1;
  
!   if (rlp->rlim_cur > rlp->rlim_max)
!     resource = RLIMIT_NLIMITS; // Catch default below, EINVAL
!   else if (oldlimits.rlim_cur == rlp->rlim_cur &&
        oldlimits.rlim_max == rlp->rlim_max)
      // No change in resource requirements, succeed immediately
      return 0;
--- 173,179 ----
    if (getrlimit (resource, &oldlimits) < 0)
      return -1;
  
!   if (oldlimits.rlim_cur == rlp->rlim_cur &&
        oldlimits.rlim_max == rlp->rlim_max)
      // No change in resource requirements, succeed immediately
      return 0;

TIA,

Anton Lavrentiev
Contractor NIH/NLM/NCBI


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple