[1/2] perf tests: Fix incorrect free and false TEST_OK result

Commit cc1121ab9687d660cc02f50b1a4974112f87a8e6 ('perf machine: Fix
machine.vmlinux_maps to make sure to clear the old one') reveals a bug
in 'perf test' that in all test cases which use setup_fake_machine()
incorrectly call free() when failure, because all users of
setup_fake_machine() use static allocated 'machines' structure, but
setup_fake_machine() calls machine__delete() which try to free() it.

If a normal user try those test cases this problem can be seen:
 $ cat /proc/sys/kernel/kptr_restrict
 1
 $ ./perf test 'hist'
 15: Test matching and linking multiple hists                 :*** Error in `./perf': munmap_chunk(): invalid pointer: 0x00007ffd6e900090 ***
 ======= Backtrace: =========
 /lib64/libc.so.6(+0x6eeef)[0x7fcec97e1eef]
 /lib64/libc.so.6(+0x78cae)[0x7fcec97ebcae]
 ./perf(setup_fake_machine+0x1cd)[0x4721ad]
 ./perf(test__hists_link+0xbf)[0x472d3f]
 ./perf[0x4648df]
 ./perf(cmd_test+0x589)[0x464ec9]
 ./perf[0x47fd11]
 ./perf(main+0x5f6)[0x432b96]
 /lib64/libc.so.6(__libc_start_main+0xf5)[0x7fcec9794bd5]
 ./perf[0x432cc5]
 ======= Memory map: ========
 ...
 25: Test filtering hist entries                              :*** Error in `./perf': munmap_chunk(): invalid pointer: 0x00007ffd6e900080 ***
 ======= Backtrace: =========
 /lib64/libc.so.6(+0x6eeef)[0x7fcec97e1eef]
 /lib64/libc.so.6(+0x78cae)[0x7fcec97ebcae]
 ./perf(setup_fake_machine+0x1cd)[0x4721ad]
 ./perf(test__hists_filter+0xc6)[0x4730f6]
 ./perf[0x4648df]
 ./perf(cmd_test+0x589)[0x464ec9]
 ./perf[0x47fd11]
 ./perf(main+0x5f6)[0x432b96]
 /lib64/libc.so.6(__libc_start_main+0xf5)[0x7fcec9794bd5]
 ./perf[0x432cc5]
 ======= Memory map: ========
 ...
 ...

Actually, all users of 'machines__init()' should avoid calling
machine__delete(). We can further add a flag in machine structure to
enforce this restriction.

After changing machine__delete() to machine__exit() another problem
arises:

 $ ./perf test 'hist'
 15: Test matching and linking multiple hists                 : Ok
 25: Test filtering hist entries                              : Ok
 28: Test output sorting of hist entries                      : Ok
 29: Test cumulation of child hist entries                    : Ok

The result is not true:

 $ ./perf test -v 'hist'
 15: Test matching and linking multiple hists                 :
 --- start ---
 test child forked, pid 17657
 Not enough memory for machine setup
 Not enough memory for machine setup
 test child finished with 0
 ---- end ----
 Test matching and linking multiple hists: Ok
 25: Test filtering hist entries                              :
 --- start ---
 test child forked, pid 17658
 Not enough memory for machine setup
 Not enough memory for machine setup
 test child finished with 0
 ---- end ----
 Test filtering hist entries: Ok
 28: Test output sorting of hist entries                      :
 --- start ---
 test child forked, pid 17659
 Not enough memory for machine setup
 Not enough memory for machine setup
 test child finished with 0
 ---- end ----
 Test output sorting of hist entries: Ok
 29: Test cumulation of child hist entries                    :
 --- start ---
 test child forked, pid 17660
 Not enough memory for machine setup
 Not enough memory for machine setup
 test child finished with 0
 ---- end ----
 Test cumulation of child hist entries: Ok

Because the test body is not executed at all.

The reason is that *ALL* hists test cases forget to reset err after
using it to hold an error code:

  err = TEST_FAIL;
  ...
  err = parse_events(evlist, "cpu-clock", NULL);
  if (err)
      goto out;
  /* err is already 0 here */
  ...
  machine = setup_fake_machine(&machines);
  if (!machine)
      goto out;
  ...
 out:
  ...
  return err;

This patch ensure err is reset.

In case when kptr_restrict prevent normal user get kernel address, this
test should be skipped, not fail. This patch use linux/err.h to store
error code in return value of setup_fake_machine(), and let 'EACCES' to
indicate this problem.

Also, the debug message 'Not enough memory for machine setup' is not
true and should be fixed.

Here is the final result:
 $ ./perf test 'hist'
 15: Test matching and linking multiple hists                 : Skip
 25: Test filtering hist entries                              : Skip
 28: Test output sorting of hist entries                      : Skip
 29: Test cumulation of child hist entries                    : Skip

 $ ./perf test -v 'hist'
 15: Test matching and linking multiple hists                 :
 --- start ---
 test child forked, pid 20177
 Failed to create kernel maps
 Hint: Check /proc/sys/kernel/kptr_restrict.
 Failed for machine setup
 test child finished with -2
 ---- end ----
 Test matching and linking multiple hists: Skip
 25: Test filtering hist entries
 ...

 $ sudo ./perf test hist
 15: Test matching and linking multiple hists                 : Ok
 25: Test filtering hist entries                              : Ok
 28: Test output sorting of hist entries                      : Ok
 29: Test cumulation of child hist entries                    : Ok

Signed-off-by: Wang Nan <wangnan0@huawei.com>

Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/tests/hists_common.c   | 15 ++++++++++-----
 tools/perf/tests/hists_common.h   |  1 +
 tools/perf/tests/hists_cumulate.c |  6 +++++-
 tools/perf/tests/hists_filter.c   |  6 +++++-
 tools/perf/tests/hists_link.c     |  6 +++++-
 tools/perf/tests/hists_output.c   |  6 +++++-
 6 files changed, 31 insertions(+), 9 deletions(-)

-- 
1.8.3.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/