Hi,
Please find a problem reported by Mark Ver for the mincore01 test case
in LTP. He has also provided fix for the same. Kindly review this patch:
=======================================================================
Mark Reports:
=======================================================================
mincore01 testcase fails with:
t75ticltst:~/ltp-full-20071031/testcases/kernel/syscalls/mincore # ./mincore01
mincore01 1 PASS : expected failure: errno = 22 (Invalid argument)
mincore01 2 FAIL : call succeeded unexpectedly
mincore01 3 PASS : expected failure: errno = 12 (Cannot allocate memory)
t75ticltst:~/ltp-full-20071031/testcases/kernel/syscalls/mincore #
It only happens on 31bit and seems to work on on SLES-9 SP3.
If this is not an installation problem,
Provide output from "uname -a", if possible:
Linux t75ticltst 2.6.5-7.305-s390 #1 SMP Tue Nov 13 23:19:08 UTC 2007 s390 s390
s390 GNU/Linux
Hardware Environment
Machine type (p650, x235, SF2, etc.): Tested on System z9 EC and z800
Cpu type (Power4, Power5, IA-64, etc.):
Describe any special hardware you think might be relevant to this problem:
Is this reproducible? Yes
If so, how long does it (did it) take to reproduce it? 20sec - 30min
Describe the steps:
- Install SLES-9 SP4 31bit system
- Get, unpack, and compile the ltp test suite
- Run the mincore01 testcase
===============================================================================
Investigation from Anoop:
===============================================================================
This turned out to be a test case problem. The only way how mincore can
generate an -EFAULT on s390 is if the target buffer is an area that is not
contained in a vma. mmap()-ing with a NULL
address gives you a pointer to a piece of memory which will be the highest
address in the anonymous mapping area, if no other mmap() is done. The testcase
tries to create an invalid pointer by unmap()-ing it. But, if the stack is
unlimited (ulimit -s unlimited) and because stack vma is VM_GROWSDOWN, the
unmap()-ed pointer is part of the stack and is a valid pointer which makes the
syscall succeed. The testcase can be fixed by setting a stack limit by "ulimit
-s <value>" before its execution.
Attached is the patch which should fix this.
===============================================================================
===============================================================================
Regards--
Subrata,
Signed-off-by: Anoop Vijayan <[EMAIL PROTECTED]>
--- ltp-full-20071031/testcases/kernel/syscalls/mincore/mincore01.c.orig 2007-12-10 15:01:32.995206304 -0800
+++ ltp-full-20071031/testcases/kernel/syscalls/mincore/mincore01.c 2007-12-10 15:08:53.524235712 -0800
@@ -60,12 +60,14 @@
#include <sys/mman.h>
#include <stdlib.h>
#include <unistd.h>
+#include <sys/resource.h>
#include <sys/types.h>
#include <sys/stat.h>
#include "test.h"
#include "usctest.h"
static int PAGESIZE;
+static rlim_t STACK_LIMIT = 10485760;
static void cleanup(void);
static void setup(void);
@@ -167,13 +169,22 @@ void
setup2()
{
unsigned char *t;
-
+ struct rlimit limit;
+
/* Create pointer to invalid address */
if( MAP_FAILED == (t = mmap(0,global_len,PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,0,0)) ) {
tst_brkm(TBROK,cleanup,"mmaping anonymous memory failed: %s",strerror(errno));
}
munmap(t,global_len);
-
+
+ /* set stack limit so that the unmaped pointer is invalid for architectures like s390 */
+ limit.rlim_cur = STACK_LIMIT;
+ limit.rlim_max = STACK_LIMIT;
+
+ if (setrlimit(RLIMIT_STACK, &limit) != 0) {
+ tst_brkm(TBROK,cleanup,"setrlimit failed: %s",strerror(errno));
+ }
+
TC[1].addr = global_pointer;
TC[1].len = global_len;
TC[1].vector = t;
-------------------------------------------------------------------------
SF.Net email is sponsored by:
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
