Skip to content

Conversation

mykouHW
Copy link

@mykouHW mykouHW commented Sep 1, 2025

fixes #156417

When the relocation section is placed before the relocated section and the relocated section is not defined in the linker script, an error will occur during the linking process.

Issue Cause:
In the judgment logic, addOrphanSections assumes that the RelocatedSection must be processed before the RelocationSection. Under this assumption, the OutputSection for the RelocatedSection has already been constructed, and the parent relationship associated with the InputSectionBase has been established.

If the RelocationSection is processed before the RelocatedSection, this assumption is violated. As a result, the condition rel->parent evaluates to null, causing add(relIS) to not execute. This skips the registration and construction process of the RelocatedSection, since its createOutputSection and recordSection methods have not yet been called at this point.

However, during the construction and registration of the RelocationSection in the addInputSec function, the RelocatedSection is accessed. Since the RelocatedSection has not been constructed yet, attempting to access it results in a null pointer error.

Solution:
Before processing the RelocationSection, ensure that the OutputSection for the RelocatedSection is created and registered. The creation and registration logic is protected by the add function, which prevents duplicate creation. However, it may result in duplicate establishment of the parent relationship, which does not affect correctness.

Copy link

github-actions bot commented Sep 1, 2025

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

@llvmbot
Copy link
Member

llvmbot commented Sep 1, 2025

@llvm/pr-subscribers-lld

Author: None (mykouHW)

Changes

Fix the error generated during the linking process when the relocation section is placed before the relocated section and the relocated section is not defined in the linker script.

Issue Cause:
In the judgment logic, addOrphanSections assumes that the RelocatedSection must be processed before the RelocationSection. Under this assumption, the OutputSection for the RelocatedSection has already been constructed, and the parent relationship associated with the InputSectionBase has been established.

If the RelocationSection is processed before the RelocatedSection, this assumption is violated. As a result, the condition rel->parent evaluates to null, causing add(relIS) to not execute. This skips the registration and construction process of the RelocatedSection, since its createOutputSection and recordSection methods have not yet been called at this point.

However, during the construction and registration of the RelocationSection in the addInputSec function, the RelocatedSection is accessed. Since the RelocatedSection has not been constructed yet, attempting to access it results in a null pointer error.

Solution:
Before processing the RelocationSection, ensure that the OutputSection for the RelocatedSection is created and registered. The creation and registration logic is protected by the add function, which prevents duplicate creation. However, it may result in duplicate establishment of the parent relationship, which does not affect correctness.


Full diff: https://github.com/llvm/llvm-project/pull/156354.diff

2 Files Affected:

  • (modified) lld/ELF/LinkerScript.cpp (+6-2)
  • (added) lld/test/ELF/linkerscript/orphan-sections-init.s (+30)
diff --git a/lld/ELF/LinkerScript.cpp b/lld/ELF/LinkerScript.cpp
index 921128dae2bdb..067abbc42a13d 100644
--- a/lld/ELF/LinkerScript.cpp
+++ b/lld/ELF/LinkerScript.cpp
@@ -1037,10 +1037,14 @@ void LinkerScript::addOrphanSections() {
     if (ctx.arg.relocatable && (isec->flags & SHF_LINK_ORDER))
       continue;
 
-    if (auto *sec = dyn_cast<InputSection>(isec))
-      if (InputSectionBase *rel = sec->getRelocatedSection())
+    if (auto *sec = dyn_cast<InputSection>(isec)){
+      if (InputSectionBase *rel = sec->getRelocatedSection()){
+        if (auto *relIS = dyn_cast_or_null<InputSectionBase>(rel))
+          add(relIS);
         if (auto *relIS = dyn_cast_or_null<InputSectionBase>(rel->parent))
           add(relIS);
+      }
+    }
     add(isec);
     if (ctx.arg.relocatable)
       for (InputSectionBase *depSec : isec->dependentSections)
diff --git a/lld/test/ELF/linkerscript/orphan-sections-init.s b/lld/test/ELF/linkerscript/orphan-sections-init.s
new file mode 100644
index 0000000000000..1701336f098e2
--- /dev/null
+++ b/lld/test/ELF/linkerscript/orphan-sections-init.s
@@ -0,0 +1,30 @@
+# REQUIRES: x86
+# RUN: rm -rf %t && mkdir -p %t
+# RUN: split-file %s %t && cd %t
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64 foo.s -o foo.o
+
+# RUN: ld.lld -r  foo.o -T script.ld -o foo_mc.o
+
+# RUN: llvm-objcopy --rename-section .text=.com.text foo_mc.o foo_mc.o
+# RUN: llvm-objcopy --rename-section .rela.text=.rela.com.text foo_mc.o foo_mc.o
+
+# RUN: ld.lld -r foo_mc.o  -T script.ld -o foo_mc_after.o
+
+#--- foo.s
+  .text
+  .globl	foo
+  .p2align	4
+  .type	foo,@function
+foo:
+  mov $bar, %rax
+
+
+
+#--- script.ld
+SECTIONS
+{
+  .rela.text    0 : { *(.rela.text) }
+  .text         0 : { *(.text) }
+}
+

@llvmbot
Copy link
Member

llvmbot commented Sep 1, 2025

@llvm/pr-subscribers-lld-elf

Author: None (mykouHW)

Changes

Fix the error generated during the linking process when the relocation section is placed before the relocated section and the relocated section is not defined in the linker script.

Issue Cause:
In the judgment logic, addOrphanSections assumes that the RelocatedSection must be processed before the RelocationSection. Under this assumption, the OutputSection for the RelocatedSection has already been constructed, and the parent relationship associated with the InputSectionBase has been established.

If the RelocationSection is processed before the RelocatedSection, this assumption is violated. As a result, the condition rel-&gt;parent evaluates to null, causing add(relIS) to not execute. This skips the registration and construction process of the RelocatedSection, since its createOutputSection and recordSection methods have not yet been called at this point.

However, during the construction and registration of the RelocationSection in the addInputSec function, the RelocatedSection is accessed. Since the RelocatedSection has not been constructed yet, attempting to access it results in a null pointer error.

Solution:
Before processing the RelocationSection, ensure that the OutputSection for the RelocatedSection is created and registered. The creation and registration logic is protected by the add function, which prevents duplicate creation. However, it may result in duplicate establishment of the parent relationship, which does not affect correctness.


Full diff: https://github.com/llvm/llvm-project/pull/156354.diff

2 Files Affected:

  • (modified) lld/ELF/LinkerScript.cpp (+6-2)
  • (added) lld/test/ELF/linkerscript/orphan-sections-init.s (+30)
diff --git a/lld/ELF/LinkerScript.cpp b/lld/ELF/LinkerScript.cpp
index 921128dae2bdb..067abbc42a13d 100644
--- a/lld/ELF/LinkerScript.cpp
+++ b/lld/ELF/LinkerScript.cpp
@@ -1037,10 +1037,14 @@ void LinkerScript::addOrphanSections() {
     if (ctx.arg.relocatable && (isec->flags & SHF_LINK_ORDER))
       continue;
 
-    if (auto *sec = dyn_cast<InputSection>(isec))
-      if (InputSectionBase *rel = sec->getRelocatedSection())
+    if (auto *sec = dyn_cast<InputSection>(isec)){
+      if (InputSectionBase *rel = sec->getRelocatedSection()){
+        if (auto *relIS = dyn_cast_or_null<InputSectionBase>(rel))
+          add(relIS);
         if (auto *relIS = dyn_cast_or_null<InputSectionBase>(rel->parent))
           add(relIS);
+      }
+    }
     add(isec);
     if (ctx.arg.relocatable)
       for (InputSectionBase *depSec : isec->dependentSections)
diff --git a/lld/test/ELF/linkerscript/orphan-sections-init.s b/lld/test/ELF/linkerscript/orphan-sections-init.s
new file mode 100644
index 0000000000000..1701336f098e2
--- /dev/null
+++ b/lld/test/ELF/linkerscript/orphan-sections-init.s
@@ -0,0 +1,30 @@
+# REQUIRES: x86
+# RUN: rm -rf %t && mkdir -p %t
+# RUN: split-file %s %t && cd %t
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64 foo.s -o foo.o
+
+# RUN: ld.lld -r  foo.o -T script.ld -o foo_mc.o
+
+# RUN: llvm-objcopy --rename-section .text=.com.text foo_mc.o foo_mc.o
+# RUN: llvm-objcopy --rename-section .rela.text=.rela.com.text foo_mc.o foo_mc.o
+
+# RUN: ld.lld -r foo_mc.o  -T script.ld -o foo_mc_after.o
+
+#--- foo.s
+  .text
+  .globl	foo
+  .p2align	4
+  .type	foo,@function
+foo:
+  mov $bar, %rax
+
+
+
+#--- script.ld
+SECTIONS
+{
+  .rela.text    0 : { *(.rela.text) }
+  .text         0 : { *(.text) }
+}
+

Copy link
Collaborator

@smithp35 smithp35 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I recommend not using the word vulnerability in titles and description as this is usually reserved for security bugs.

I suggest something like `[lld][ELF] Fix crash when relocations proceed relocated section.

I also recommend putting the analysis in the linked issue into the description as this will form the commit message. It is much easier to find and search for changes with git log when the commit message has all the details.

if (auto *sec = dyn_cast<InputSection>(isec))
if (InputSectionBase *rel = sec->getRelocatedSection())
if (auto *sec = dyn_cast<InputSection>(isec)){
if (InputSectionBase *rel = sec->getRelocatedSection()){
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

could rel be renamed to relocated here?

Not part of your change, but I think it would be an improvement. When reviewing just the diff, it was difficult to tell from the name rel whether we are referring to the relocation section, or the section being relocated.

@@ -1037,10 +1037,14 @@ void LinkerScript::addOrphanSections() {
if (ctx.arg.relocatable && (isec->flags & SHF_LINK_ORDER))
continue;

if (auto *sec = dyn_cast<InputSection>(isec))
if (InputSectionBase *rel = sec->getRelocatedSection())
if (auto *sec = dyn_cast<InputSection>(isec)){
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you git clang-format the patch? I would expect a space between ){ i.e. (isec)) {

if (InputSectionBase *rel = sec->getRelocatedSection())
if (auto *sec = dyn_cast<InputSection>(isec)){
if (InputSectionBase *rel = sec->getRelocatedSection()){
if (auto *relIS = dyn_cast_or_null<InputSectionBase>(rel))
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could we add a comment before the if statement like
// Ensure creation of OutputSection for relocated section before relocation section.


# RUN: ld.lld -r foo.o -T script.ld -o foo_mc.o

# RUN: llvm-objcopy --rename-section .text=.com.text foo_mc.o foo_mc.o
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can merge do both -rename-section commands with a single instance of llvm-objcopy.

foo:
mov $bar, %rax


Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I recommend a CHECK of the llvm-readelf --sections output to make sure lld has created the link from the relocations correctly.

I've seen some scripts been run to find tests with no FileCheck output before so it is worth putting at least one in.

# REQUIRES: x86
# RUN: rm -rf %t && mkdir -p %t
# RUN: split-file %s %t && cd %t

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you add a comment to explain the test. Without context the steps look strange. LLD uses an additional comment character to distinguish comments from lit commands, so comment lines would start ##

For example:

Test that lld's orphan section placement can handle a relocatable link where the relocation section is seen before the relocated section. To create a test case we need to first need to create a relocatable object with
the relocations before the relocated section. Then we rename these to make these orphans.

I tried to see if I could create the test case without llvm-objcopy but it seemed like I needed the linker script to have .rela.text as the pattern to get it to place first.

@mykouHW mykouHW changed the title [lld][ELF] Improve the vulnerability in Orphan Sections initialization [lld][ELF] Fix crash when relocations proceed relocated section Sep 3, 2025
Fix the error generated during the linking process when the relocation section is placed before the relocated section and the relocated section is not defined in the linker script.
@mykouHW mykouHW force-pushed the fix-orphan-sections-init branch from fb38a50 to 3cc58a3 Compare September 4, 2025 08:43
@smithp35 smithp35 requested a review from MaskRay September 4, 2025 12:40
Copy link
Collaborator

@smithp35 smithp35 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the update. I've enabled the CI. Assuming that goes well then this looks good to me.

I've added the maintainer @MaskRay to see if he has any opinions.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[LLD] Crash when relocations proceed relocated section in Orphan Sections Initialization
3 participants