RFR (s) 8085865: hs_err improvement: Printing /proc/cpuinfo makes too long hs_err files

[Daniel D. Daugherty]

On 6/17/15 6:15 AM, Coleen Phillimore wrote:
>
>
> On 6/17/15 12:08 AM, David Holmes wrote:
>> Hi Coleen,
>>
>> On 16/06/2015 10:10 AM, Coleen Phillimore wrote:
>>>
>>> Hi, I've added Chris's requested flags line but this made me change 
>>> more
>>> lines of code because the flags line can be pretty long and we don't
>>> want to allocate more stack space during error handling.
>>>
>>> http://cr.openjdk.java.net/~coleenp/8085865.03/
>>
>> The logic here is a little odd to me:
>>
>> 2227         if (strstr(buf, "model name") != NULL) {
>> 2228           if (!model_name_printed) {
>> 2229             st->print_raw("\nCPU Model and flags from 
>> /proc/cpuinfo:\n");
>> 2230             st->print_raw(buf);
>> 2231             model_name_printed = true;
>> 2232           } else {
>> 2233             // model name printed but not flags?  Odd, just exit.
>> 2234             return true;
>> 2235           }
>> 2236         }
>>
>> You'll only hit 2233 if "model name" appears twice (with no flag line 
>> in between). Is that assumed to be the "model name" from a second 
>> processor section?

That's what this comment implies:

+        // Assume model name comes before flags


> Yes, it is assumed.  It is a little odd but I didn't want to add more 
> logic for waiting to find the 'flags' section, and potentially 
> printing model name 32 times.

Perhaps a comment like this:

// Print the first "model name" line and the first "flags" line
// that we find and nothing more. We assume "model name" comes
// before "flags" so if we find a second "model name", then the
// "flags" field is considered missing.

Dan

>
>>
>> Otherwise seems okay.
>
> Thanks!
> Coleen
>>
>> Thanks,
>> David
>>
>>> Tested on my machine.  It's a little bit longer but nowhere nearly as
>>> long as printing the 4 processors (or 32 on some other people's 
>>> machines).
>>>
>>> CPU Model and flags from /proc/cpuinfo:
>>> model name      : Intel(R) Core(TM) i5-3550 CPU @ 3.30GHz
>>> flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
>>> mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe
>>> syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl
>>> xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor
>>> ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic
>>> popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm ida arat 
>>> epb
>>> xsaveopt pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase
>>> smep erms
>>>
>>> Also tested with JPRT.
>>>
>>> Thanks,
>>> Coleen
>>>
>>> On 6/15/15 1:39 PM, Coleen Phillimore wrote:
>>>>
>>>>
>>>> On 6/15/15 12:04 PM, Christian Thalinger wrote:
>>>>>> On Jun 12, 2015, at 3:07 AM, Coleen Phillimore
>>>>>> <coleen.phillimore at oracle.com> wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>> On 6/12/15 4:17 AM, Mikael Gerdin wrote:
>>>>>>> Hi Coleen,
>>>>>>>
>>>>>>> On 2015-06-11 18:06, Coleen Phillimore wrote:
>>>>>>>> Summary: summarize information from linux-x86; it's too long and
>>>>>>>> redundant
>>>>>>>>
>>>>>>>> The cpuinfo on one linux/x86 machine now prints:
>>>>>>>>
>>>>>>>> /proc/cpuinfo:
>>>>>>>> model name      : Intel(R) Core(TM) i5-3550 CPU @ 3.30GHz
>>>>>>>>
>>>>>>>> on 32 bit machine:
>>>>>>>> *
>>>>>>>> */proc/cpuinfo:
>>>>>>>> model name      : Intel(R) Xeon(R) CPU L5520  @ 2.27GHz
>>>>>>>> **
>>>>>>>> Other linux platforms are unaffected because they don't print 
>>>>>>>> as much
>>>>>>>> and information seems not redundant.
>>>>>>> Would you be ok with printing one of the "flags" lines as well?
>>>>>>>
>>>>>>> flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr
>>>>>>> pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm
>>>>>>> pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts
>>>>>>> rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni
>>>>>>> pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr
>>>>>>> pdcm pcid dca sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes
>>>>>>> xsave avx lahf_lm ida arat epb pln pts dtherm tpr_shadow vnmi
>>>>>>> flexpriority ept vpid xsaveopt
>>>>>>>
>>>>>>> It would give some quick insight into what CPU features were
>>>>>>> available during runtime.
>>>>>> We already print
>>>>>>
>>>>>> CPU:total 4 (4 cores per cpu, 1 threads per core) family 6 model 58
>>>>>> stepping 9, cmov, cx8, fxsr, mmx, sse, sse2, sse3, ssse3, sse4.1,
>>>>>> sse4.2, popcnt, avx, aes, clmul, erms, tsc, tscinvbit, tscinv
>>>>>>
>>>>>> For the flags we use.   This isn't enough?
>>>>> I’d prefer to print the flags the CPU supports too since these are
>>>>> not the same.
>>>>
>>>> Okay, sure, I'll add that line.
>>>> Coleen
>>>>
>>>>>
>>>>>> thanks,
>>>>>> Coleen
>>>>>>> /Mikael
>>>>>>>
>>>>>>>> Rather than duplicating the pd_print_cpu_info function for all the
>>>>>>>> os_cpu/linux* combinations, I added an ifdef.   In os_aix.cpp
>>>>>>>> there's a
>>>>>>>> pd_print_cpu_info, which is called from os::print_cpu_info that
>>>>>>>> duplicates the information but I didn't touch it.
>>>>>>>>
>>>>>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8085865.01/
>>>>>>>> bug link https://bugs.openjdk.java.net/browse/JDK-8085865
>>>>>>>>
>>>>>>>> Thanks,
>>>>>>>> Coleen
>>>>>>>>
>>>>>>>>
>>>>
>>>
>