RFR: 8324580: SIGFPE on THP initialization on kernels < 4.10 [v4]

[Zdenek Zambersky]

On Wed, 14 Feb 2024 06:51:57 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

>>> Yeah, I am almost convinced now that modifying `HugePages::thp_pagesize()` would have been better. Maybe with an extra boolean flag that means "We did guess the THP page size".
>> 
>> So far I have just modified logic. One more idea which come to my head was to just make fallback computation as method to HugePages (without logging). And then do something like this:
>> 
>> _large_page_size = HugePages::thp_pagesize();
>> if (_large_page_size == 0) {
>>     log_info(pagesize) ("Cannot determine THP page size (kernel < 4.10 ?)");
>>     ...
>>     _large_page_size = HugePages::thp_pagesize_fallback();
>> }
>
>> > Yeah, I am almost convinced now that modifying `HugePages::thp_pagesize()` would have been better. Maybe with an extra boolean flag that means "We did guess the THP page size".
>> 
>> So far I have just modified logic. One more idea which come to my head was to just make fallback computation as method to HugePages (without logging). And then do something like this:
>> 
>> ```
>> _large_page_size = HugePages::thp_pagesize();
>> if (_large_page_size == 0) {
>>     log_info(pagesize) ("Cannot determine THP page size (kernel < 4.10 ?)");
>>     ...
>>     _large_page_size = HugePages::thp_pagesize_fallback();
>> }
>> ```
> 
> Yes, we can do that too.

@tstuefe @kstefanj thank you

(macos build seems like infra issue)

-------------

PR Comment: https://git.openjdk.org/jdk/pull/17545#issuecomment-1946303531