Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: ATF X86 race with Debug Registers



On 10.03.2018 22:47, Andreas Gustafsson wrote:
> Kamil Rytarowski wrote:
>> Sometimes we catch a bug with Debug Registers in the ptrace(2) tests:
>>
>> NetBSD/amd64
>> http://releng.netbsd.org/b5reports/amd64/2018/2018.03.06.11.21.31/test.html#lib_libc_sys_t_ptrace_wait6_dbregs_dr3_trap_variable_readwrite_write_2bytes
>>
>> NetBSD/i386
>> http://releng.netbsd.org/b5reports/i386/2017/2017.12.06.17.54.58/test.html#lib_libc_sys_t_ptrace_wait3_dbregs_dr2_trap_variable_readwrite_read_2bytes
>>
>> I was trying to reproduce this locally on a real hardware and in qemu
>> (softemu), including concurrent execution, but after merely 100M
>> attempts to reproduce the test-case, I've not encountered a single failure.
>>
>> Are there some special properties of these releng machines? qemu+kvm?
> 
> Nothing special about it AFAIK, it's just qemu from pkgsrc running on
> babylon5.netbsd.org, a dual Xeon L5630, under NetBSD 7.1_STABLE.
> Typical load is 6-8 qemu processes and a niced -j 16 build.  There's
> no kvm - that's only supported under Linux.
> 
> If you look at the test.log file corresponding to the test.html, you
> can find the qemu version and full qemu command line near the top.
> 

I've reproduced this failure locally in qemu, after 50-100k iterations
of a test using Debug Register and trapping a variable I can see that
instead of SIGTRAP, I'm receiving SIGSTOP.

Reproducing can take like 2h, but I will try to investigate this.

Attachment: signature.asc
Description: OpenPGP digital signature



Home | Main Index | Thread Index | Old Index