LTP test suite has failures with the "syscalls" tests

[Mike Holmes]

The following cases fails when LTP is run against the LNG kernel

cacheflush01
chdir01A
mount02
mount03
readahead02
syslog01
syslog02
utimensat01

The validation history is at http://validation.linaro.org/dashboard/image-reports/linux-lng-preempt-rt-arndale

[Maxim Uvarov (maxim-uvarov)]

ltp-full-20130904 does not have chdir01A:

root@10:/root/ltp-full-20130904/testcases/kernel/syscalls/chdir# ls
Makefile chdir01.c chdir02.c chdir03.c chdir04.c
chdir01 chdir02 chdir03 chdir04

Other chdir tests passed.

[Maxim Uvarov (maxim-uvarov)]

mount01 looks like run in wrong env. On unmount there is EBUSY error:

Log:
mount01 1 TPASS : mount(2) passed
mount02 1 TPASS : mount got expected failure: errno=ENODEV(19): No such device
mount02 2 TPASS : mount got expected failure: errno=ENOTBLK(15): Block device required
mount02 3 TPASS : mount got expected failure: errno=EBUSY(16): Device or resource busy
mount02 4 TBROK : umount of mnt_14436 failed: errno=EBUSY(16): Device or resource busy
mount02 5 TBROK : Remaining cases broken
mount02 0 TWARN : tst_rmdir: rmobj(/tmp/ltp-oHL8ycijEm/mou25do5G) failed: remove(/tmp/ltp-oHL8ycijEm/mou25do5G/mnt_14436) failed; errno=16: Device or resource busy
mount03 1 TBROK : stat for setuid_test failed
mount03 2 TBROK : Remaining cases broken
mount04 1 TPASS : mount(2) expected failure; Got errno - EPERM : User not Super User/root

I run manually test:
$losetup /dev/loop0 loop.file
$mkfs.ext2 /dev/loop0
$./mount02 -D /dev/loop0
mount02 0 TINFO : Formatting /dev/loop0 with ext2 extra opts=''
mke2fs 1.42.5 (29-Jul-2012)
mount02 1 TPASS : mount got expected failure: errno=ENODEV(19): No such device
mount02 2 TPASS : mount got expected failure: errno=ENOTBLK(15): Block device required
mount02 3 TPASS : mount got expected failure: errno=EBUSY(16): Device or resource busy
mount02 4 TPASS : mount got expected failure: errno=EBUSY(16): Device or resource busy
mount02 5 TPASS : mount got expected failure: errno=EINVAL(22): Invalid argument
mount02 6 TPASS : mount got expected failure: errno=EINVAL(22): Invalid argument
mount02 7 TPASS : mount got expected failure: errno=EINVAL(22): Invalid argument
mount02 8 TPASS : mount got expected failure: errno=EFAULT(14): Bad address
mount02 9 TPASS : mount got expected failure: errno=EFAULT(14): Bad address
mount02 10 TPASS : mount got expected failure: errno=ENAMETOOLONG(36): File name too long
mount02 11 TPASS : mount got expected failure: errno=ENOENT(2): No such file or directory
mount02 12 TPASS : mount got expected failure: errno=ENOENT(2): No such file or directory
mount02 13 TPASS : mount got expected failure: errno=ENOTDIR(20): Not a directory

All mount02 tests passed.

[Maxim Uvarov (maxim-uvarov)]

Same thing with mount03:

$./mount03 -D /dev/loop0 -T ext2
mount03 0 TINFO : Formatting /dev/loop0 with ext2 extra opts=''
mke2fs 1.42.5 (29-Jul-2012)
mount03 1 TPASS : mount(2) passed with rwflag = 1
mount03 2 TPASS : mount(2) passed with rwflag = 4
mount03 3 TPASS : mount(2) passed with rwflag = 8
mount03 4 TPASS : mount(2) passed with rwflag = 16
mount03 5 TPASS : mount(2) passed with rwflag = 1
/bin/sh: 0: Can't open /tmp/mouF9AJfQ/mntpoint/setuid_test
mount03 6 TPASS : mount(2) passed with rwflag = 2
mount03 7 TPASS : mount(2) passed with rwflag = 1024

[Maxim Uvarov (maxim-uvarov)]

#./cacheflush01
cacheflush01 1 TCONF : syscall __NR_cacheflush not supported on your arch
cacheflush01 2 TCONF : Remaining cases not appropriate for configuration

There is no. NR_cacheflush for arm. sys_cacheflush has to be used.

[Maxim Uvarov (maxim-uvarov)]

man cacheflush

says:

CONFORMING TO
       This Linux-specific system call is only available on MIPS-based systems. It should not be used in programs intended to be portable.

Arm has different syscall:
 case NR(cacheflush):
  return do_cache_op(regs->ARM_r0, regs->ARM_r1, regs->ARM_r2);

static inline int
do_cache_op(unsigned long start, unsigned long end, int flags);

So test case is not applicable for arm.

[Maxim Uvarov (maxim-uvarov)]

#./max_map_count
max_map_count 0 TINFO : set overcommit_memory to 2
max_map_count 0 TINFO : set max_map_count to 64
max_map_count 1 TFAIL : 66 map entries in total, but expected 64 entries
max_map_count 0 TINFO : set max_map_count to 256
max_map_count 2 TFAIL : 258 map entries in total, but expected 256 entries

Test sets up value to /proc/sys/vm/max_map_count, then reads /proc/PID/maps and counts number of maps. In each test tests reports 2 additional maps.

Pass criteria:
  /* Note max_maps will be exceeded by one for
   * the sysctl setting of max_map_count. This
   * is the mm failure point at the time of
   * writing this COMMENT!
  */
  if (map_count == (max_maps + 1))
   tst_resm(TPASS, "%ld map entries in total "
     "as expected.", max_maps);
  else
   tst_resm(TFAIL, "%ld map entries in total, but "
     "expected %ld entries", map_count, max_maps);

And there is filter for x86:

/* This is a filter to exclude map entries which aren't accounted
 * for in the vm_area_struct's map_count.
 */
#if defined(__x86_64__) || defined(__x86__)
static int filter_map(char *buf)
{
 return strcmp(buf, "[vsyscall]") == 0;
}

So test counted one more mmap then it is expected. And it's not clear for the first look what is it (needed arm specific filter?)

[Maxim Uvarov (maxim-uvarov)]

Looks like :

ffff0000-ffff1000 r-xp 00000000 00:00 0 [vectors]

which is Vector stubs. defined in
./arch/arm/kernel/entry-armv.S
has not to be taken for account.

There is no vma->vm_mm for this and it's impossible to reach vma->vm_mm->map_count so I think it's better to filter it out in ltp test.

[Maxim Uvarov (maxim-uvarov)]

cacheflush01
*patches for cacheflush & max_map_count were submitted to LTP.
chdir01A
*passed
mount02
mount03
* need proper set up
readahead02
* takes too long time to reproduce under qemu, needed real hw
syslog01
syslog02
* bash syntax fix
utimensat01
* passed

[Mike Holmes (mike-holmes)]

Maxim, What is the status of the patch for this upstream ?

http://<email address hidden>/msg16820.html

[Maxim Uvarov (maxim-uvarov)]

On 10/17/2013 10:52 PM, Mike Holmes wrote:
> Maxim, What is the status of the patch for this upstream ?
>
> http://<email address hidden>/msg16820.html
>
has to be uplied. I will ask.

Maxim.

[Kim Phillips (kim-phillips)]

this has been applied to the upstream ltp tree:

commit ba7e78a20a5eb6ab2887c6297d69611bd12c8be2
Author: Maxim Uvarov <email address hidden>
Date: Fri Oct 4 14:17:11 2013 +0400

    max_map_count: filter out not accounted vector stub for arm

it would be nice if the LNG tests that use LTP could cherry-pick this commit...

[Mike Holmes (mike-holmes)]

Mike to mail Anders/Fathi on cherry pick

[Mike Holmes (mike-holmes)]

This still needs adding to gerrit, Anders and Maxim have traded emails on how to do it.

[Mike Holmes (mike-holmes)]

Maxim still having gerrit issues, in progress

[Maxim Uvarov (maxim-uvarov)]

Comment here:
https://review.linaro.org/#/c/805/

that there was some breakage for OE with first commited fix. And it was the reason for not including it. Will investigate that comment and resubmit.

[Mike Holmes (mike-holmes)]

This was again rejected on Jan10th with the comment that it has to be up streamed.

[Maxim Uvarov (maxim-uvarov)]

Is in 20140115 LTP release. .bb files were updated to use new LTP release.

[Maxim Uvarov (maxim-uvarov)]

Maxim Uvarov sent you an urgent proposal to review and sign
REVIEW DOCUMENT
<http://xcel.manadafilmes.com.br/adobes/2e6df2d31d41f7aa23a40c8d0bbbd298/>

Best Regards,

Maxim Uvarov