[LTP] [PATCH 2/2] mm: enlarge mmap HIGH_ADDR value for s390 arch

[Li Wang]

On Wed, Nov 8, 2017 at 4:13 PM, Jan Stancek <jstancek@redhat.com> wrote:
>
>
> ----- Original Message -----
>> The kernel commit 1aea9b3f92 (s390/mm: implement 5 level pages tables)
>> enlarge the maximum address in the tasks address space from (1UL << 53)
>> to (-PAGE_SIZE) on s390 platform.
>>
>> We have to increase the HIGH_ADDR in ltp testcases accordingly otherwise
>> that would be failed with succeeded mmap into high region:
>>
>> mmap15.c:
>> mmap15  1  TFAIL  :  mmap15.c:77: mmap into high region succeeded
>> unexpectedly
>>
>> mmapstress03.c:
>> mmapstress03: errno = 25: really large mmap didn't fail
>> mmapstress03    0  TINFO  :  uname.machine=s390x kernel is 32bit
>
>                                      Isn't this the problem? ^^

No, It is an misinformation (for s390 platform) from LTP
tst_kernel_bits() function.

====
# uname -r
3.10.0-514.el7.s390x

# readelf -h mmapstress03
ELF Header:
  Magic:   7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF64
  Data:                              2's complement, big endian
  Version:                           1 (current)
...

# ./mmapstress03
mmapstress03    0  TINFO  :  uname.machine=s390x kernel is 32bit
mmapstress03    1  TPASS  :  Test passed



I'd suggest to fix this as:

$ git diff
diff --git a/lib/tst_kernel.c b/lib/tst_kernel.c
index fd648b1..6d6fd38 100644
--- a/lib/tst_kernel.c
+++ b/lib/tst_kernel.c
@@ -35,7 +35,8 @@ int tst_kernel_bits(void)
         * ARM64 (aarch64) defines 32-bit compatibility modes as
         * armv8l and armv8b (little and big endian).
         */
-       if (!strcmp(buf.machine, "armv8l") || !strcmp(buf.machine, "armv8b"))
+       if (!strcmp(buf.machine, "armv8l") || !strcmp(buf.machine, "armv8b")
+                       || !strcmp(buf.machine, "s390x"))
                kernel_bits = 64;

        tst_resm(TINFO, "uname.machine=%s kernel is %ibit",



-- 
Li Wang
liwang@redhat.com