[LLD][COFF] Align EC code ranges to page boundaries#168222
Merged
Conversation
We already ensure that code for different architectures is always placed in different pages in assignAddresses. We represent those ranges using their first and last chunks. However, the RVAs of those chunks may not be page-aligned, for example, due to extra padding for entry-thunk offsets. Align the chunk RVAs to the page boundary so that the emitted ranges correctly include the entire region. This change affects an existing test that checks corner cases triggered by merging a data section into a code section. We may now include such data in the code range. This differs from MSVC’s behavior, but it should not cause practical issues, and the new behavior is arguably more correct. Fixes llvm#168119.
Member
|
@llvm/pr-subscribers-lld @llvm/pr-subscribers-lld-coff Author: Jacek Caban (cjacek) ChangesWe already ensure that code for different architectures is always placed in different pages in This change affects an existing test that checks corner cases triggered by merging a data section into a code section. We may now include such data in the code range. This differs from MSVC’s behavior, but it should not cause practical issues, and the new behavior is arguably more correct. Fixes #168119. 2 Files Affected:
diff --git a/lld/COFF/Chunks.cpp b/lld/COFF/Chunks.cpp
index 548d87bdaefe5..409491d4a1f89 100644
--- a/lld/COFF/Chunks.cpp
+++ b/lld/COFF/Chunks.cpp
@@ -946,7 +946,7 @@ void ECCodeMapChunk::writeTo(uint8_t *buf) const {
auto table = reinterpret_cast<chpe_range_entry *>(buf);
for (uint32_t i = 0; i < map.size(); i++) {
const ECCodeMapEntry &entry = map[i];
- uint32_t start = entry.first->getRVA();
+ uint32_t start = entry.first->getRVA() & ~0xfff;
table[i].StartOffset = start | entry.type;
table[i].Length = entry.last->getRVA() + entry.last->getSize() - start;
}
diff --git a/lld/test/COFF/arm64ec-codemap.test b/lld/test/COFF/arm64ec-codemap.test
index 050261117be2e..bbc682d19920f 100644
--- a/lld/test/COFF/arm64ec-codemap.test
+++ b/lld/test/COFF/arm64ec-codemap.test
@@ -7,6 +7,7 @@ RUN: llvm-mc -filetype=obj -triple=arm64ec-windows arm64ec-func-sym2.s -o arm64e
RUN: llvm-mc -filetype=obj -triple=arm64ec-windows data-sec.s -o data-sec.obj
RUN: llvm-mc -filetype=obj -triple=arm64ec-windows data-sec2.s -o data-sec2.obj
RUN: llvm-mc -filetype=obj -triple=arm64ec-windows empty-sec.s -o arm64ec-empty-sec.obj
+RUN: llvm-mc -filetype=obj -triple=arm64ec-windows entry-thunk.s -o entry-thunk.obj
RUN: llvm-mc -filetype=obj -triple=x86_64-windows x86_64-func-sym.s -o x86_64-func-sym.obj
RUN: llvm-mc -filetype=obj -triple=x86_64-windows empty-sec.s -o x86_64-empty-sec.obj
RUN: llvm-mc -filetype=obj -triple=aarch64-windows %S/Inputs/loadconfig-arm64.s -o loadconfig-arm64.obj
@@ -162,15 +163,17 @@ RUN: loadconfig-arm64ec.obj -dll -noentry -merge:test=.testdata -merge:
RUN: llvm-readobj --coff-load-config testcm.dll | FileCheck -check-prefix=CODEMAPCM %s
CODEMAPCM: CodeMap [
-CODEMAPCM-NEXT: 0x4008 - 0x4016 X64
+CODEMAPCM-NEXT: 0x4000 - 0x4016 X64
CODEMAPCM-NEXT: ]
RUN: llvm-objdump -d testcm.dll | FileCheck -check-prefix=DISASMCM %s
DISASMCM: Disassembly of section .testdat:
DISASMCM-EMPTY:
DISASMCM-NEXT: 0000000180004000 <.testdat>:
-DISASMCM-NEXT: 180004000: 00000001 udf #0x1
-DISASMCM-NEXT: 180004004: 00000000 udf #0x0
+DISASMCM-NEXT: 180004000: 01 00 addl %eax, (%rax)
+DISASMCM-NEXT: 180004002: 00 00 addb %al, (%rax)
+DISASMCM-NEXT: 180004004: 00 00 addb %al, (%rax)
+DISASMCM-NEXT: 180004006: 00 00 addb %al, (%rax)
DISASMCM-NEXT: 180004008: b8 03 00 00 00 movl $0x3, %eax
DISASMCM-NEXT: 18000400d: c3 retq
DISASMCM-NEXT: 18000400e: 00 00 addb %al, (%rax)
@@ -207,6 +210,14 @@ DISASMMS-NEXT: 0000000180006000 <test2>:
DISASMMS-NEXT: 180006000: 528000a0 mov w0, #0x5 // =5
DISASMMS-NEXT: 180006004: d65f03c0 ret
+Test the code map that includes an ARM64EC function padded by its entry-thunk offset.
+
+RUN: lld-link -out:testpad.dll -machine:arm64ec entry-thunk.obj loadconfig-arm64ec.obj -dll -noentry -include:func
+RUN: llvm-readobj --coff-load-config testpad.dll | FileCheck -check-prefix=CODEMAPPAD %s
+CODEMAPPAD: CodeMap [
+CODEMAPPAD: 0x1000 - 0x1010 ARM64EC
+CODEMAPPAD-NEXT: ]
+
#--- arm64-func-sym.s
.text
@@ -266,3 +277,22 @@ x86_64_func_sym2:
.section .empty1, "xr"
.section .empty2, "xr"
.section .empty3, "xr"
+
+#--- entry-thunk.s
+ .section .text,"xr",discard,func
+ .globl func
+ .p2align 2, 0x0
+func:
+ mov w0, #1
+ ret
+
+ .section .wowthk$aa,"xr",discard,thunk
+ .globl thunk
+ .p2align 2
+thunk:
+ ret
+
+ .section .hybmp$x,"yi"
+ .symidx func
+ .symidx thunk
+ .word 1 // entry thunk
|
mstorsjo
approved these changes
Nov 15, 2025
Member
mstorsjo
left a comment
mstorsjo
left a comment
There was a problem hiding this comment.
Looks ok, one minor comment.
| for (uint32_t i = 0; i < map.size(); i++) { | ||
| const ECCodeMapEntry &entry = map[i]; | ||
| uint32_t start = entry.first->getRVA(); | ||
| uint32_t start = entry.first->getRVA() & ~0xfff; |
Member
There was a problem hiding this comment.
This looks reasonable, but do we use explicit 0xfff for page alignment here/elsewhere so far, or do we have some constant for that?
Contributor
Author
There was a problem hiding this comment.
Yes, we already use 0xfff in several places in LLD, mostly, but not exclusively, in Chunks.cpp.
Contributor
Author
|
/cherry-pick af45b02 |
Member
Error: Command failed due to missing milestone. |
Member
|
/pull-request #168369 |
tru
pushed a commit
to llvmbot/llvm-project
that referenced
this pull request
Dec 1, 2025
We already ensure that code for different architectures is always placed in different pages in `assignAddresses`. We represent those ranges using their first and last chunks. However, the RVAs of those chunks may not be page-aligned, for example, due to extra padding for entry-thunk offsets. Align the chunk RVAs to the page boundary so that the emitted ranges correctly include the entire region. This change affects an existing test that checks corner cases triggered by merging a data section into a code section. We may now include such data in the code range. This differs from MSVC’s behavior, but it should not cause practical issues, and the new behavior is arguably more correct. Fixes llvm#168119. (cherry picked from commit af45b02)
LumioseSil
pushed a commit
to LumioseSil/llvm-project
that referenced
this pull request
Apr 4, 2026
We already ensure that code for different architectures is always placed in different pages in `assignAddresses`. We represent those ranges using their first and last chunks. However, the RVAs of those chunks may not be page-aligned, for example, due to extra padding for entry-thunk offsets. Align the chunk RVAs to the page boundary so that the emitted ranges correctly include the entire region. This change affects an existing test that checks corner cases triggered by merging a data section into a code section. We may now include such data in the code range. This differs from MSVC’s behavior, but it should not cause practical issues, and the new behavior is arguably more correct. Fixes llvm#168119. (cherry picked from commit af45b02)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
We already ensure that code for different architectures is always placed in different pages in
assignAddresses. We represent those ranges using their first and last chunks. However, the RVAs of those chunks may not be page-aligned, for example, due to extra padding for entry-thunk offsets. Align the chunk RVAs to the page boundary so that the emitted ranges correctly include the entire region.This change affects an existing test that checks corner cases triggered by merging a data section into a code section. We may now include such data in the code range. This differs from MSVC’s behavior, but it should not cause practical issues, and the new behavior is arguably more correct.
Fixes #168119.