RFR: 8323681: SA PointerFinder code should support G1

Chris Plummer cjplummer at openjdk.org
Fri Feb 2 23:28:09 UTC 2024


This PR adds G1 support to PointerFinder/PointerLocation. Previously we only had SerialGC support. Previously for G1 addresses SA would only report that the address is "In unknown section of Java heap" with no other details. Now it provides details for G1 addresses. Here are some examples for clhsdb `threadcontext` output. `threadcontext` dumps the contents of the thread's registers, some of which are often in the java heap. In the output below the first line is without verbose output and the 2nd is with:


rbp: 0x00000000a0004080: In G1 heap region
rbp: 0x00000000a0004080: In G1 heap Region: 0x00000000a0000000,0x00000000a0018a30,0x00000000a1000000:Old


I also added an improvement to how SA deals with addresses in the TLAB. It used to only report that the address is in a TLAB and provide details about the TLAB in verbose mode. Now if verbose mode is used, the heap region information is included after the TLAB information. Here again is an example without and with verbose output:


rsi: 0x0000000166000000: In TLAB for thread "main" sun.jvm.hotspot.runtime.JavaThread at 0x00007f11c8029250
rsi: 0x0000000166000000: In TLAB for thread ("main" #1 prio=5 tid=0x00007f11c8029250 nid=1503 runnable [0x0000000000000000]
   java.lang.Thread.State: RUNNABLE
   JavaThread state: _thread_in_java
)  [0x0000000166000000,0x00000001662d0c90,0x00000001667ffdc0,{0x0000000166800000})
In G1 heap Region: 0x0000000166000000,0x0000000166800000,0x0000000167000000:Eden


Note at the end it indicates the address is in the Eden space, which is probably always the case for G1 TLAB addresses. For SerialGC is shows something similar.


rsi: 0x0000000088ff99e0: In TLAB for thread "main" sun.jvm.hotspot.runtime.JavaThread at 0x00007f0544029110
rsi: 0x0000000088ff99e0: In TLAB for thread ("main" #1 prio=5 tid=0x00007f0544029110 nid=3098 runnable [0x0000000000000000]
   java.lang.Thread.State: RUNNABLE
   JavaThread state: _thread_in_java
)  [0x0000000088ff99e0,0x000000008978c090,0x0000000089ae54b0,{0x0000000089ae56f0})
In heap new generation:  eden [0x0000000080200000,0x0000000089ae56f0,0x00000000a2420000) space capacity = 572653568, 27.99656213789626 used
  from [0x00000000a6860000,0x00000000a6860030,0x00000000aaca0000) space capacity = 71565312, 6.707160027472528E-5 used
  to   [0x00000000a2420000,0x00000000a2420000,0x00000000a6860000) space capacity = 71565312, 0.0 used


Testing all svc test in tier1, tier2, and tier5. Currently in progress.

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

Commit messages:
 - Add G1 support to PointerFinder and PointerLocation.

Changes: https://git.openjdk.org/jdk/pull/17691/files
 Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=17691&range=00
  Issue: https://bugs.openjdk.org/browse/JDK-8323681
  Stats: 70 lines in 4 files changed: 54 ins; 6 del; 10 mod
  Patch: https://git.openjdk.org/jdk/pull/17691.diff
  Fetch: git fetch https://git.openjdk.org/jdk.git pull/17691/head:pull/17691

PR: https://git.openjdk.org/jdk/pull/17691


More information about the serviceability-dev mailing list