8367325: [s390x] build failure due to JDK-8361376 #27213

offamitkumar · 2025-09-11T09:51:29Z

Fixes the SIGLL caused after JDK-8361376.

Issue was with cs instruction, which requires the address to be aligned. And after this change address at which operating was not aligned.

Wisdom from Principle of Z Operations:

Sytax: CS R₁,R₃,D₂(B₂)

Sytax: CSY R₁,R₃,D₂(B₂)

The second operand of COMPARE AND SWAP (CS, CSY) must be designated on a word boundary.

=> 0x3fffc3149ba <_ZN17BarrierSetNMethod15set_guard_valueEP7nmethodii+266>: cs %r1,%r4,0(%r3)
(gdb) i r r3
r3             0x3ffe500017a       4397593526650
(gdb) p ($r3 % 8) 
$5 = 2
(gdb) si 

Thread 16 "C1 CompilerThre" received signal SIGILL, Illegal instruction.
NativeMethodBarrier::set_guard_value (bit_mask=2147483647, value=1, this=0x3ffe5000166)
    at /home/amit/jdk/src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp:74
74         if (v == old_value) break;

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8367325: [s390x] build failure due to JDK-8361376 (Bug - P1)

Reviewers

Martin Doerr (@TheRealMDoerr - Reviewer)
Dean Long (@dean-long - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/27213/head:pull/27213
$ git checkout pull/27213

Update a local copy of the PR:
$ git checkout pull/27213
$ git pull https://git.openjdk.org/jdk.git pull/27213/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 27213

View PR using the GUI difftool:
$ git pr show -t 27213

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/27213.diff

Using Webrev

Link to Webrev Comment

bridgekeeper · 2025-09-11T09:53:42Z

👋 Welcome back amitkumar! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2025-09-11T09:53:57Z

@offamitkumar This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8367325: [s390x] build failure due to JDK-8361376

Reviewed-by: mdoerr, dlong

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 111 new commits pushed to the master branch:

6e4e966: 8365792: GenShen: assertion "Generations aren't reconciled"
e4cb86d: 8367473: Shenandoah: Make the detailed evacuation metrics a runtime diagnostic option
4be4826: 8367450: Shenandoah: Log the composition of the collection set
... and 108 more: https://git.openjdk.org/jdk/compare/3d679087b0376c221d536780cee387dc2dd8019e...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

➡️ To integrate this PR with the above commit message to the master branch, type /integrate in a new comment.

openjdk · 2025-09-11T09:54:56Z

@offamitkumar The following label will be automatically applied to this pull request:

hotspot

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2025-09-11T10:17:54Z

Webrevs

TheRealMDoerr · 2025-09-11T10:39:20Z

I think align(4) would be sufficient.
The offset of the immediate field of cfi is 3*6+2 = 20 which is already a multiple of 4. Aligning the start of the sequence to 4 should be the right thing.
Please add a comment explaining that the immediate field of cfi needs to be 4-Byte aligned!

dean-long · 2025-09-11T20:17:24Z

src/hotspot/cpu/s390/gc/shared/barrierSetAssembler_s390.cpp


 void BarrierSetAssembler::nmethod_entry_barrier(MacroAssembler* masm) {
  BarrierSetNMethod* bs_nm = BarrierSet::barrier_set()->barrier_set_nmethod();
+  __ align(8); // must align the following block which requires atomic updates


Thanks for fixing this. Rather than aligning at the start, wouldn't it be more maintainable to do
__ align(4, offset() + 2);
before the z_cfi?

That would require much more changes. The barrier sequence is expected to have a fixed size.

OK, then how about we align the beginning using
__ align(4, offset() + 3*6 + 2);
to make it more self-documenting?

Or better: define a constant for 3*6 + 2 and reuse it in class NativeMethodBarrier.

I have updated the code, please have a look now

TheRealMDoerr · 2025-09-15T10:49:20Z

src/hotspot/cpu/s390/gc/shared/barrierSetAssembler_s390.hpp

                                OptoReg::Name opto_reg) const;
 #endif // COMPILER2
+
+  static const int PATCHABLE_INSTRUCTION_OFFSET = 3 * 6 + 2;


Maybe better call it PATCHABLE_BARRIER_VALUE_OFFSET?
It's no longer the instruction start address.
Or even better: define 2 constants. The version without the +2 is still needed below.

I have given name as PATCHABLE_SEQ_START_OFFSET and PATCHABLE_BARRIER_VALUE_OFFSET. Please let me know if there is better option :)

TheRealMDoerr · 2025-09-15T10:49:37Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp

 #include "code/nativeInst.hpp"
 #include "code/nmethod.hpp"
 #include "gc/shared/barrierSetNMethod.hpp"
+#include "gc/shared/barrierSetAssembler.hpp"


Should be sorted alphabetically.

TheRealMDoerr · 2025-09-15T10:50:34Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp

-      address inst_addr = get_barrier_start_address() + PATCHABLE_INSTRUCTION_OFFSET;
+      address inst_addr = get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_INSTRUCTION_OFFSET;

      DEBUG_ONLY(Assembler::is_z_cfi(*((long*)inst_addr)));


The cfi check needs an adjustment to point to the instruction start, right?

Now It is pointing to z_cfi:

(gdb) x/i inst_addr 0x3ffe500017a: cfi %r0,0

This line looks broken (already before this PR). Shouldn't it be something like assert(Assembler::is_z_cfi...)?

Will it be good to do something like this and get rid of PATCHABLE_SEQ_START_OFFSET ?:

diff --git a/src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp b/src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp index 4dc50232c17..807d3cdd899 100644 --- a/src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp +++ b/src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp @@ -38,10 +38,11 @@ class NativeMethodBarrier: public NativeInstruction { } address get_patchable_data_address() const { +#ifdef ASSERT address inst_addr = get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_SEQ_START_OFFSET; - - DEBUG_ONLY(Assembler::is_z_cfi(*((long*)inst_addr))); - return inst_addr + 2; + assert(Assembler::is_z_cfi(*((long*)inst_addr)), "should be"); +#endif // ASSERT + return get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_BARRIER_VALUE_OFFSET; } public:

TheRealMDoerr · 2025-09-15T10:52:39Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp


  public:
-    static const int BARRIER_TOTAL_LENGTH = PATCHABLE_INSTRUCTION_OFFSET + 2*6 + 2; // bytes
+    static const int BARRIER_TOTAL_LENGTH = BarrierSetAssembler::PATCHABLE_INSTRUCTION_OFFSET + 2*6; // bytes


TheRealMDoerr · 2025-09-17T10:02:10Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp


  public:
-    static const int BARRIER_TOTAL_LENGTH = PATCHABLE_INSTRUCTION_OFFSET + 2*6 + 2; // bytes
+    static const int BARRIER_TOTAL_LENGTH = BarrierSetAssembler::PATCHABLE_BARRIER_VALUE_OFFSET + 2*6; // bytes


This is very confusing. There are 2 instructions with 6 Bytes and one instruction with 2 Bytes after PATCHABLE_SEQ_START_OFFSET.

TheRealMDoerr · 2025-09-17T10:06:37Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp

+      address inst_addr = get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_SEQ_START_OFFSET;

      DEBUG_ONLY(Assembler::is_z_cfi(*((long*)inst_addr)));
      return inst_addr + 2;


Maybe use get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_BARRIER_VALUE_OFFSET? inst_addr could be removed.

TheRealMDoerr · 2025-09-17T10:07:55Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp

-      address inst_addr = get_barrier_start_address() + PATCHABLE_INSTRUCTION_OFFSET;
+      address inst_addr = get_barrier_start_address() + BarrierSetAssembler::PATCHABLE_INSTRUCTION_OFFSET;

      DEBUG_ONLY(Assembler::is_z_cfi(*((long*)inst_addr)));


This line looks broken (already before this PR). Shouldn't it be something like assert(Assembler::is_z_cfi...)?

TheRealMDoerr · 2025-09-18T10:25:34Z

src/hotspot/cpu/s390/gc/shared/barrierSetAssembler_s390.cpp

 void BarrierSetAssembler::nmethod_entry_barrier(MacroAssembler* masm) {
  BarrierSetNMethod* bs_nm = BarrierSet::barrier_set()->barrier_set_nmethod();
-  __ align(4, __ offset() + PATCHABLE_BARRIER_VALUE_OFFSET); // must align the following block which requires atomic updates
+  __ align(4, __ offset() + BARRIER_TOTAL_LENGTH); // must align the following block which requires atomic updates


Now, this is confusing. We don't want to align the end of the barrier. We need to align the patchable field which is the immediate field of the cfi instruction.

I couldn't find out better way to keep the main code free from magic numbers, so now there are three constants:

holding offset to the instruction (cfi)

holding the offset the the patchable data

total barrier length.

I guess explanation in .hpp is enough to explain the usecase.

TheRealMDoerr

Looks correct to me. I only have some minor comments.

TheRealMDoerr · 2025-09-18T12:24:05Z

src/hotspot/cpu/s390/gc/shared/barrierSetAssembler_s390.hpp

+
+  // first 2 bytes are for cfi instruction opcode and next 4 bytes will be the value/data to be patched,
+  // so we are skipping first 2 bytes and returning the address of value/data field
+  static const int OFFSET_TO_PATCHABLE_DATA = 6 + 6 + 6 + 2; // iihf(6) + iilf(6) + lg(6) + CFI_OPCODE(2)


OFFSET_TO_PATCHABLE_DATA_INSTRUCTION could be used, here.

TheRealMDoerr · 2025-09-18T12:26:09Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp


  public:
-    static const int BARRIER_TOTAL_LENGTH = PATCHABLE_INSTRUCTION_OFFSET + 2*6 + 2; // bytes
+    static const int BARRIER_TOTAL_LENGTH = BarrierSetAssembler::BARRIER_TOTAL_LENGTH;


Is it necessary to replicate it?

It is being used with NativeMethodBarrier, yes I can change it and technically it wouldn't affect anything, but just to keep code change bit shorter, I left it as it is.

TheRealMDoerr · 2025-09-18T12:26:42Z

src/hotspot/cpu/s390/gc/shared/barrierSetNMethod_s390.cpp

        offset += Assembler::instr_len(&start[offset]);

-        Assembler::is_z_cfi(*((long*)(start + offset)));
+        // it will be assignment operation, So it doesn't matter what value is already present in instr


I don't understand what you mean by "it will be assignment operation".

it is just a assignment being done in that method, i.e. instr variable will be just overwritten straightly. So currently whatever value instr holds, it wouldn't not affect the returned-value.

offamitkumar · 2025-09-18T12:41:03Z

@dean-long would it be possible for you to have another look at the latest changes ?
Thanks

dean-long

Looks reasonable to me.

offamitkumar · 2025-09-19T04:33:14Z

thank you so much for the reviews.
/integrate

openjdk · 2025-09-19T04:33:51Z

Going to push as commit 898fcff.
Since your change was applied there have been 115 commits pushed to the master branch:

5855fd2: 8367784: java/awt/Focus/InitialFocusTest/InitialFocusTest1.java failed with Wrong focus owner
c0815e4: 8367904: Test java/net/InetAddress/ptr/Lookup.java should throw SkippedException
7ec3fa5: 8367801: jtreg failure_handler - don't use the -L option for ps command
... and 112 more: https://git.openjdk.org/jdk/compare/3d679087b0376c221d536780cee387dc2dd8019e...master

Your commit was automatically rebased without conflicts.

openjdk · 2025-09-19T04:33:58Z

@offamitkumar Pushed as commit 898fcff.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

build fix

668caeb

openjdk bot changed the title ~~8367325~~ 8367325: [s390x] build failure due to JDK-8361376 Sep 11, 2025

openjdk bot added the hotspot hotspot-dev@openjdk.org label Sep 11, 2025

offamitkumar marked this pull request as ready for review September 11, 2025 10:12

openjdk bot added the rfr Pull request is ready for review label Sep 11, 2025

dean-long reviewed Sep 11, 2025

View reviewed changes

better fix

3545cf1

TheRealMDoerr reviewed Sep 15, 2025

View reviewed changes

update

d7d27bc

TheRealMDoerr reviewed Sep 17, 2025

View reviewed changes

refactor

e328dcc

TheRealMDoerr reviewed Sep 18, 2025

View reviewed changes

update

6684ea0

TheRealMDoerr approved these changes Sep 18, 2025

View reviewed changes

openjdk bot added the ready Pull request is ready to be integrated label Sep 18, 2025

dean-long approved these changes Sep 18, 2025

View reviewed changes

openjdk bot added the integrated Pull request has been integrated label Sep 19, 2025

openjdk bot closed this Sep 19, 2025

openjdk bot removed ready Pull request is ready to be integrated rfr Pull request is ready for review labels Sep 19, 2025

offamitkumar deleted the alignment_fix branch September 19, 2025 04:34

8367325: [s390x] build failure due to JDK-8361376 #27213

8367325: [s390x] build failure due to JDK-8361376 #27213

Uh oh!

Conversation

offamitkumar commented Sep 11, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewers

Reviewing

Uh oh!

bridgekeeper bot commented Sep 11, 2025

Uh oh!

openjdk bot commented Sep 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

openjdk bot commented Sep 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mlbridge bot commented Sep 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Webrevs

Uh oh!

TheRealMDoerr commented Sep 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TheRealMDoerr Sep 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

TheRealMDoerr left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

offamitkumar commented Sep 18, 2025

Uh oh!

dean-long left a comment

offamitkumar commented Sep 11, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Sep 11, 2025 •

edited

Loading

openjdk bot commented Sep 11, 2025 •

edited

Loading

mlbridge bot commented Sep 11, 2025 •

edited

Loading

TheRealMDoerr commented Sep 11, 2025 •

edited

Loading

TheRealMDoerr Sep 17, 2025 •

edited

Loading