Re: [Qemu-devel] [PATCH v3 23/46] tests/tcg/x86_64: add Makefile.target

[Philippe Mathieu-Daudé]

On 04/25/2018 05:43 AM, Alex Bennée wrote:
> 
> Philippe Mathieu-Daudé <address@hidden> writes:
> 
>> On 04/24/2018 12:23 PM, Alex Bennée wrote:
>>> The sources for x86_64 are shared in the i386 directory which will be
>>> included thanks to TARGET_BASE_ARCH. However not all sources build so
>>> we need to filter out the ones we can't build in the 64 bit world.
>>>
>>> Signed-off-by: Alex Bennée <address@hidden>
>>> ---
>>>  tests/tcg/i386/Makefile.target   |  2 +-
>>>  tests/tcg/x86_64/Makefile.target | 15 +++++++++++++++
>>>  2 files changed, 16 insertions(+), 1 deletion(-)
>>>  create mode 100644 tests/tcg/x86_64/Makefile.target
>>>
>>> diff --git a/tests/tcg/i386/Makefile.target b/tests/tcg/i386/Makefile.target
>>> index 80ccad3d7b..27086c3f2b 100644
>>> --- a/tests/tcg/i386/Makefile.target
>>> +++ b/tests/tcg/i386/Makefile.target
>>> @@ -7,7 +7,7 @@ VPATH               += $(I386_SRC)
>>>
>>>  I386_SRCS=$(notdir $(wildcard $(I386_SRC)/*.c))
>>>  I386_TESTS=$(I386_SRCS:.c=)
>>> -
>>> +I386_ONLY_TESTS=$(filter-out test-i386-ssse3, $(I386_TESTS))
>>>  # Update TESTS
>>>  TESTS+=$(I386_ONLY_TESTS)
>>>
>>> diff --git a/tests/tcg/x86_64/Makefile.target 
>>> b/tests/tcg/x86_64/Makefile.target
>>> new file mode 100644
>>> index 0000000000..74f170b9ed
>>> --- /dev/null
>>> +++ b/tests/tcg/x86_64/Makefile.target
>>> @@ -0,0 +1,15 @@
>>> +# -*- Mode: makefile -*-
>>> +#
>>> +# x86_64 tests - included from tests/tcg/Makefile.target
>>> +#
>>> +# Currently we only build test-x86_64 and test-i386-ssse3 from
>>> +# $(SRC)/tests/tcg/i386/
>>> +#
>>> +
>>> +X86_64_TESTS=$(filter-out $(I386_ONLY_TESTS), $(TESTS))
>>> +X86_64_TESTS+=test-x86_64
>>
>> The linux-test is taking looooong... I canceled it after ~15min on a i7
>> 2GHz. Maybe it also belongs to the "BROKEN TEST" series on x86_64?
>> Currently it is only skipped if $(TARGET_NAME) == i386.
> 
> Hmm it runs pretty quickly for me:
> 
> time ./qemu-x86_64 ./tests/linux-test
> 0.02user 0.03system 0:00.08elapsed 72%CPU (0avgtext+0avgdata 7384maxresident)k
> 0inputs+0outputs (0major+1207minor)pagefaults 0swaps
> 
> I wonder if the failure mode you are seeing is the same as the others
> which basically hang, which is why I added them to the broken test patch.

The test_clone() hangs here:

    while (waitpid(pid1, &status1, 0) != pid1);

32107 wait4(32109, 0x7ffdd2814a50, 0, NULL) = -1 ECHILD (No child processes)
32107 wait4(32109,  <unfinished ...>
32109 nanosleep({tv_sec=0, tv_nsec=10000000},  <unfinished ...>
32107 <... wait4 resumed> 0x7ffdd2814a50, 0, NULL) = -1 ECHILD (No child
processes)
32107 wait4(32109, 0x7ffdd2814a50, 0, NULL) = -1 ECHILD (No child processes)

> 
>>
>> Reviewed-by: Philippe Mathieu-Daudé <address@hidden>
>>
>>> +TESTS:=$(X86_64_TESTS)
>>> +
>>> +test-x86_64: LDFLAGS+=-lm -lc
>>> +test-x86_64: test-i386.c test-i386.h test-i386-shift.h test-i386-muldiv.h
>>> +   $(CC) $(CFLAGS) $< -o $@ $(LDFLAGS)
>>>
>>
>> $ make run-tcg-tests-x86_64-linux-user
>>   RUN-TESTS for x86_64
>>   TEST    test-mmap (default) on x86_64
>>   TEST    test-mmap (8k pages) on x86_64
>> ASAN:DEADLYSIGNAL
>> =================================================================
>> ==5717==ERROR: AddressSanitizer: SEGV on unknown address 0x7fcba47fc000
>> (pc 0x7fcba8a3bf4d bp 0x7ffd19d74700 sp 0x7ffd19d73e88 T0)
>> ==5717==The signal is caused by a WRITE memory access.
>>     #0 0x7fcba8a3bf4c  (/lib/x86_64-linux-gnu/libc.so.6+0x159f4c)
>>     #1 0x7fcbab33f891  (/usr/lib/x86_64-linux-gnu/libasan.so.4+0x5c891)
>>     #2 0x55dd3ae4a7e0 in mmap_frag /source/qemu/linux-user/mmap.c:189
>>     #3 0x55dd3ae4bbce in target_mmap /source/qemu/linux-user/mmap.c:533
>>     #4 0x55dd3ae4b8cb in target_mmap /source/qemu/linux-user/mmap.c:501
>>     #5 0x55dd3ae3895c in do_syscall /source/qemu/linux-user/syscall.c:9448
>>     #6 0x55dd3adfaa57 in cpu_loop /source/qemu/linux-user/main.c:258
>>     #7 0x55dd3ae005f5 in main /source/qemu/linux-user/main.c:5147
>>     #8 0x7fcba8903a86 in __libc_start_main
>> (/lib/x86_64-linux-gnu/libc.so.6+0x21a86)
>>     #9 0x55dd3ad34a89 in _start
>> (/source/qemu/build/full/x86_64-linux-user/qemu-x86_64+0x350a89)
>> AddressSanitizer can not provide additional info.
>> SUMMARY: AddressSanitizer: SEGV (/lib/x86_64-linux-gnu/libc.so.6+0x159f4c)
>> ==5717==ABORTING
>> make[2]: ***
>> [/home/phil/source/qemu/tests/tcg/multiarch/Makefile.target:27:
>> run-test-mmap] Error 1
>>
>> Tested-by: Philippe Mathieu-Daudé <address@hidden>
> 
> This is only with ASAN enabled right?

Yes, but even without ASan test_clone() hangs.

Commenting test_clone():

$ make -j1 run-tcg-tests-x86_64-linux-user
  CROSS-BUILD x86_64 guest-tests with gcc-7
  CROSS-BUILD x86_64 guest-tests with gcc-7
  RUN-TESTS for x86_64
  TEST    test-mmap (default) on x86_64
  TEST    test-mmap (8k pages) on x86_64
  TEST    test-mmap (16k pages) on x86_64
  TEST    test-mmap (32k pages) on x86_64
  TEST    sha1 on x86_64
  TEST    linux-test on x86_64
  TEST    testthread on x86_64
  TEST    test-x86_64 on x86_64
$