On Sat, 7 Mar 2026 04:16:14 GMT, Yasumasa Suenaga <[email protected]> wrote:
>> We saw the failure in serviceability/sa/TestJhsdbJstackMixedWithXComp.java >> on Valhalla repo. `jhsdb jstack --mixed` could not unwind continuation call >> frames as following: >> >> >> * LingeredAppWithVirtualThread.run() bci:15 line:69 >> (Interpreted frame) >> * java.lang.Thread.runWith(java.lang.Object, >> java.lang.Runnable) bci:5 line:1540 (Compiled frame [deoptimized]; >> information may be imprecise) >> * java.lang.VirtualThread.run(java.lang.Runnable) bci:62 >> line:472 (Compiled frame [deoptimized]; information may be imprecise) >> 0x00007fdad773cf98 <StubRoutines (continuation stubs)> >> 0xfefefefefefefefe ???????? >> >> >> I found that `frame::sender_for_compiled_frame()` in frame_x86.inline.hpp >> has a special case if sender PC has return barrier entry, but SA does not >> handle it. >> >> This is not only a problem on Valhalla. Same problem exists on JDK. So I >> want to fix on JDK. >> This PR passed serviceability/sa tests on Linux, and also >> TestJhsdbJstackMixedWithXComp.java on Valhalla passed 100 times. >> >> This PR is assembled by following commits: >> >> * Follows continuation-related code in HotSpot, and use it on AMD64 SA code >> * >> https://github.com/openjdk/jdk/pull/30107/changes/4af559ee0dbc0bb61e0bcd91bb17459a5abf50ad >> * Fix for AArch64 >> * >> https://github.com/openjdk/jdk/pull/30107/changes/2cff0fe40c52b88dd8d79ad1534e73b7b0f88f8d >> * Fix for RISC-V >> * >> https://github.com/openjdk/jdk/pull/30107/changes/fdef0384577bb71d34371f11bc5878494f276857 >> * Fix for PPC64 >> * >> https://github.com/openjdk/jdk/pull/30107/changes/5d9774d82c7e8c4d92eae8995ab014232ce9590e > > Yasumasa Suenaga has updated the pull request incrementally with one > additional commit since the last revision: > > Make setSP() to abstract method src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/riscv64/RISCV64Frame.java line 272: > 270: if (cb.isUpcallStub()) { > 271: return senderForUpcallStub(map, (UpcallStub)cb); > 272: } else { Hi, I tried this on `linux-riscv64` platform and I witnessed test failure when running `serviceability/sa/TestJhsdbJstackWithVirtualThread.java`: java.lang.RuntimeException: 'must have non-zero frame size' found in stdout at jdk.test.lib.process.OutputAnalyzer.shouldNotContain(OutputAnalyzer.java:299) at TestJhsdbJstackWithVirtualThread.runJstack(TestJhsdbJstackWithVirtualThread.java:63) at TestJhsdbJstackWithVirtualThread.main(TestJhsdbJstackWithVirtualThread.java:74) at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104) at java.base/java.lang.reflect.Method.invoke(Method.java:565) at com.sun.javatest.regtest.agent.MainActionHelper$AgentVMRunnable.run(MainActionHelper.java:335) at java.base/java.lang.Thread.run(Thread.java:1527) JavaTest Message: Test threw exception: java.lang.RuntimeException JavaTest Message: shutting down test I guess you might want to add following add-on change. I see aarch64 and amd64 has a similar frame size check before invoking `senderForCompiledFrame` method. diff --git a/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/riscv64/RISCV64Frame.java b/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/riscv64/RISCV64Frame.java index 67b4314a3c7..a35c0735979 100644 --- a/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/riscv64/RISCV64Frame.java +++ b/src/jdk.hotspot.agent/share/classes/sun/jvm/hotspot/runtime/riscv64/RISCV64Frame.java @@ -269,7 +269,7 @@ public Frame sender(RegisterMap regMap, CodeBlob cb) { if (cb != null) { if (cb.isUpcallStub()) { return senderForUpcallStub(map, (UpcallStub)cb); - } else { + } else if (cb.getFrameSize() > 0) { return senderForCompiledFrame(map, cb); } } ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/30107#discussion_r2901107049
