RFR: 8189170: Add option to disable stack overflow checking in primordial thread for use with JNI_CreateJavaJVM

Tomas Kalibera tomas.kalibera at gmail.com
Mon Nov 20 14:37:21 UTC 2017


I have tested with R and it seems to be working fine.

Specifically, I've tested with jdk/hs 47868 as baseline, with R-devel 
73756, rJava 0.9-9 (modified to print VM args). I've run R recursive 
calls after initializing the jvm via .jinit(), using 
RJAVA_JVM_STACK_WORKAROUND=1 (detect inserted guard pages, but only 
print a message when they appear).

The baseline inserted guard pages as expected following -Xss (default 
and when specified 1m..9m) and following the implied cap when rlimit 
stack is unlimited, and R consequenlty crashed in the baseline. The 
modified version with the fix activated did not insert guard pages. 
Tested with rlimit stack of 8m (Ubuntu default) and unlimited.

 From R, I activated the fix via
options(java.parameters=c("-XX:+UnlockExperimentalVMOptions", 
"-XX:+DisablePrimordialThreadGuardPages"))

Tested on Ubuntu 17.04, x86_64. jdk/hs built with gcc 4.9.4 (but still 
needed --disable-warnings-as-errors)

Thanks a lot for the new option!
Tomas

On 11/20/2017 02:06 PM, David Holmes wrote:
> Thank you both.
>
> David
>
> On 20/11/2017 11:05 PM, Daniel D. Daugherty wrote:
>> I'm also good with this change.
>>
>> Dan
>>
>>
>> On 11/20/17 4:04 AM, Thomas Stüfe wrote:
>>> Hi David,
>>>
>>> looks fine.
>>>
>>> Thanks, Thomas
>>>
>>> On Mon, Nov 20, 2017 at 4:44 AM, David Holmes 
>>> <david.holmes at oracle.com <mailto:david.holmes at oracle.com>> wrote:
>>>
>>>     Hi Dan,
>>>
>>>     Thanks for the review. I've addressed all of yours and Thomas's
>>>     comments.
>>>
>>>     http://cr.openjdk.java.net/~dholmes/8189170/webrev.v2/
>>> <http://cr.openjdk.java.net/%7Edholmes/8189170/webrev.v2/>
>>>
>>>     The main change is:
>>>
>>>     +   static bool is_primordial_thread(void)
>>>     + #if defined(_WINDOWS) || defined(BSD)
>>>     +     // No way to identify the primordial thread.
>>>     +     { return false; }
>>>     + #else
>>>     +   ;
>>>     + #endif
>>>
>>>
>>>     Also to clarify this is logically a "product" flag for the systems
>>>     that need it, not a diagnostic one. But as we don't want this flag
>>>     to be in common use and it is of such narrow applicability, we
>>>     have made it experimental. This was discussed previously in the
>>>     CSR review thread for this:
>>>
>>> http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2017-October/024859.html
>>> <http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/2017-October/024859.html>
>>>
>>>     Once Tomas K. has been able to test this, and you and Thomas
>>>     sign-off on the updates, I'll get this pushed. But as I'm away
>>>     from Thursday it may not be until next week.
>>>
>>>     Thanks,
>>>     David
>>>     -----
>>>
>>>
>>>
>>>     On 18/11/2017 1:40 AM, Daniel D. Daugherty wrote:
>>>
>>>         On 11/14/17 5:39 AM, David Holmes wrote:
>>>
>>>             Bug: https://bugs.openjdk.java.net/browse/JDK-8189170
>>> <https://bugs.openjdk.java.net/browse/JDK-8189170>
>>>
>>>             webrev:
>>> http://cr.openjdk.java.net/~dholmes/8189170/webrev/
>>> <http://cr.openjdk.java.net/%7Edholmes/8189170/webrev/>
>>>
>>>
>>>         src/hotspot/os/aix/os_aix.cpp
>>>              No comments.
>>>
>>>         src/hotspot/os/aix/os_aix.hpp
>>>              No comments.
>>>
>>>         src/hotspot/os/bsd/os_bsd.cpp
>>>              L880: // thread in any special way - so just report 
>>> 'false'
>>>                  Nit - needs a period after 'false' to make it a 
>>> sentence.
>>>
>>>              L882:     return false;
>>>                  Nit - indent should be 2 spaces.
>>>
>>>         src/hotspot/os/bsd/os_bsd.hpp
>>>              No comments.
>>>
>>>         src/hotspot/os/linux/os_linux.cpp
>>>              L947:   if (stack_size >= (size_t)(3 * page_size()))
>>>                  Nit - needs '{' and '}' on the if-statement body.
>>>
>>>              L3070: // always place it right after end of the mapped
>>>         region
>>>                  Nit - needs a period after 'region' to make it a
>>>         sentence.
>>>                  (Not your problem, but you did update the sentence...)
>>>
>>>         src/hotspot/os/linux/os_linux.hpp
>>>              No comments.
>>>
>>>         src/hotspot/os/solaris/os_solaris.cpp
>>>              No comments other than I learned something new about
>>>         thr_main() :-)
>>>              Thanks for cleaning up that mess.
>>>
>>>         src/hotspot/os/windows/os_windows.cpp
>>>              L452: // thread in any special way - so just report 
>>> 'false'
>>>                  Nit - needs a period after 'false' to make it a 
>>> sentence.
>>>
>>>              L454:     return false;
>>>                  Nit - indent should be 2 spaces.
>>>
>>>         src/hotspot/share/gc/shared/threadLocalAllocBuffer.cpp
>>>              No comments.
>>>
>>>         src/hotspot/share/runtime/globals.hpp
>>>              L1181:   experimental(bool,
>>>         DisablePrimordialThreadGuardPages, false,
>>>                  Experimental or diagnostic?
>>>
>>>         src/hotspot/share/runtime/os.hpp
>>>              L450:   // that loads/creates the JVM via JNI_CreateJavaVM
>>>                  Nit - needs a period after 'JNI_CreateJavaVM'.
>>>
>>>              L453:   // launcher nevers uses the primordial thread as
>>>         the main thread, but
>>>                  Typo - 'nevers' -> 'never'
>>>
>>>              L455:   // have to special-case handling of the
>>>         primordial thread if it attaches
>>>                  Typo - 'have to' -> 'have'
>>>
>>>         src/hotspot/share/runtime/thread.cpp
>>>              I like the new log message.
>>>
>>>         src/hotspot/share/runtime/thread.inline.hpp
>>>              L159:   if (os::uses_stack_guard_pages() &&
>>>              L160:       !(DisablePrimordialThreadGuardPages &&
>>>         os::is_primordial_thread())) {
>>>                  I'm not sure why, but I had to read these two lines
>>>         several
>>>                  times before I convinced myself the logic is right.
>>>
>>>         I'm good with the code changes. Your call on whether to fix the
>>>         bits or not.
>>>
>>>         Regarding this part of Thomas' review:
>>>
>>>                 Or, something like this:
>>>                 static bool is_primordial_thread(void)
>>>                 WINDOWS_ONLY({return false;}) BSD_ONLY({return false;})
>>>
>>>
>>>             I can do that. I'll see if anyone else has any comments on
>>>             this.
>>>
>>>
>>>         I'm assuming that Thomas means for the above to be at the
>>>         platform independent layer (runtime/os.hpp?) so it is more
>>>         obvious to the reader that the function doesn't do anything
>>>         on those two platforms... I know we dislike such things at
>>>         the platform independent layer, but it does feel appropriate
>>>         to make this limitation more obvious.
>>>
>>>         I would be okay if you made the change that Thomas is 
>>> suggesting.
>>>
>>>         Thumbs up!
>>>
>>>         Dan
>>>
>>>
>>>
>>>
>>>             There are three parts to this which made it somewhat more
>>>             complicated than I had envisaged:
>>>
>>>             1. Add the flag and disable the guards.
>>>
>>>             This is relatively straight-forward:
>>>
>>>              void JavaThread::create_stack_guard_pages() {
>>>                if (!os::uses_stack_guard_pages() ||
>>>                    _stack_guard_state != stack_guard_unused ||
>>>                    (DisablePrimordialThreadGuardPages &&
>>>             os::is_primordial_thread())) {
>>>                    log_info(os, thread)("Stack guard page creation for
>>>             thread "
>>>                                         UINTX_FORMAT " disabled",
>>>             os::current_thread_id());
>>>                  return;
>>>
>>>             with a tweak in JavaThread::stack_guards_enabled as well.
>>>
>>>             2. Promote os::XXX::is_initial_thread to
>>>             os::is_primordial_thread()
>>>
>>>             We have three platforms that already implement this
>>>             functionality: Linux, Solaris and AIX. For BSD/OSX and
>>>             Windows we don't attempt to do anything different for the
>>>             primordial thread, and we don't have a way to detect the
>>>             primordial thread - so is_primordial_thread() just returns
>>>             false.
>>>
>>>             I also tidied up some comments regarding terminology.
>>>
>>>             3. Thomas noticed that we unconditionally subtract
>>>             2*page_size() from the rlimit stack size, without checking
>>>             it was bigger than 2*page_size() to start with. I was
>>>             unsure how to tackle this. It's no good leaving stack_size
>>>             at zero so I opted to ensure we would have at least one
>>>             page left. Of course in such cases we would hit the bug in
>>>             libc (if it still exists, which seems unlikely but time
>>>             consuming to establish).
>>>
>>>             Testing:
>>>               - Tier 1 (JPRT) testing with a modified launcher that
>>>             uses the primordial thread
>>>                 - with guard pages enabled
>>>                 - with guard pages disabled
>>>               - Tier 1 (JPRT) normal JDK build (which should be
>>>             unaffected by this change)
>>>
>>>             The testing was primarily to ensure that disabling the
>>>             stack guards on the primordial thread wasn't totally
>>>             broken. A number of tests fail:
>>>             - setting -Xss32m fails for the primordial thread when
>>>             guards are enabled and the rlimit is 8m
>>>             - tests that would hit StackOverflowError crash the VM
>>>             when guards are disabled (as you would expect).
>>>             - runtime/execstack/Testexecstack.java crashes when the
>>>             guards are disabled
>>>
>>>             Thanks,
>>>             David
>>>
>>>
>>>
>>



More information about the hotspot-runtime-dev mailing list