[LTP] [PATCH] syscalls/cma: test case should return TCONF if syscall doesn't exist

Jason Xing kerneljasonxing@linux.alibaba.com
Tue Dec 4 03:49:21 CET 2018



On 2018/12/3 下午5:58, Jan Stancek wrote:
>
>
> ----- Original Message -----
>> The original code returns TFAIL when process_vm_readv/_writev syscalls
>> do not exist, while we are expecting a TCONF. During investigation, we
>> find it is due to '#if defined(__NR_process_vm_readv/_writev)' fails to
>> work as expected.  We fix this issue by using ltp_syscall() function in
>> setup function. BTW, we don't use new tst_syscall() interface due to
>> compatibility issue exists between old apis and new ones.
>
> Fine by me, but then you should also add syscall numbers to
> include/lapi/syscalls:
>

Thanks.
I have TWO questions:

1. If I use this command #./strip_syscall.awk 
/usr/include/asm/unistd_64.h > x86_64.in, it will update and add more 
new syscall numbers into x86_64.in file which includes many other 
syscall numbers not only process_vm_readv/_writev. To solve this 
problem, can I just add the process_vm_readv/_writev syscall numbers and 
let go of other numbers?

2. I don't have so many types of machines that support different 
architectures like hppa and s390x. So what should I do to know this 
syscall number on other architectures. I'm running the tests on x86_64 arch.

Jason

> process_vm01.c: In function ‘setup’:
> process_vm01.c:97: warning: unused parameter ‘argv’
> process_vm01.c: In function ‘cma_test_params_read’:
> process_vm01.c:127: error: ‘__NR_process_vm_readv’ undeclared (first use in this function)
> process_vm01.c:127: error: (Each undeclared identifier is reported only once
> process_vm01.c:127: error: for each function it appears in.)
> process_vm01.c: In function ‘cma_test_params_write’:
> process_vm01.c:136: error: ‘__NR_process_vm_writev’ undeclared (first use in this function)
>
> Regards,
> Jan
>


More information about the ltp mailing list