[ELF] Support NOCROSSREFS and NOCROSSERFS_TO #98773
Merged
Conversation
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
Created using spr 1.3.5-bogner
|
@llvm/pr-subscribers-lld-elf @llvm/pr-subscribers-lld Author: Fangrui Song (MaskRay) ChangesImplement the two commands described by After To support non SHF_ALLOC sections, Some parse code is adapted from #95714. Close #41825 Full diff: https://github.com/llvm/llvm-project/pull/98773.diff 6 Files Affected:
diff --git a/lld/ELF/LinkerScript.h b/lld/ELF/LinkerScript.h
index 43d0850eed718..b987026c00200 100644
--- a/lld/ELF/LinkerScript.h
+++ b/lld/ELF/LinkerScript.h
@@ -256,6 +256,16 @@ struct InsertCommand {
StringRef where;
};
+// A NOCROSSREFS/NOCROSSREFS_TO command that probits references among certain
+// output sections.
+struct NoCrossRefCommand {
+ SmallVector<StringRef, 0> outputSections;
+
+ // When true, this describes a NOCROSSREFS_TO command that probits references
+ // to the first output section from any of the other sections.
+ bool toFirst = false;
+};
+
struct PhdrsCommand {
StringRef name;
unsigned type = llvm::ELF::PT_NULL;
@@ -397,6 +407,9 @@ class LinkerScript final {
// OutputSections specified by OVERWRITE_SECTIONS.
SmallVector<OutputDesc *, 0> overwriteSections;
+ // NOCROSSREFS(_TO) commands.
+ SmallVector<NoCrossRefCommand, 0> noCrossRefs;
+
// Sections that will be warned/errored by --orphan-handling.
SmallVector<const InputSectionBase *, 0> orphanSections;
diff --git a/lld/ELF/Relocations.cpp b/lld/ELF/Relocations.cpp
index 9ad180306bcd8..6c1edaa70728a 100644
--- a/lld/ELF/Relocations.cpp
+++ b/lld/ELF/Relocations.cpp
@@ -2367,7 +2367,61 @@ void elf::hexagonTLSSymbolUpdate(ArrayRef<OutputSection *> outputSections) {
});
}
+static bool matchesRefTo(const NoCrossRefCommand &cmd, StringRef osec) {
+ if (cmd.toFirst)
+ return cmd.outputSections[0] == osec;
+ return llvm::is_contained(cmd.outputSections, osec);
+}
+
+template <class ELFT, class Rels>
+static void scanCrossRefs(const NoCrossRefCommand &cmd, OutputSection *osec,
+ InputSection *sec, Rels rels) {
+ for (const auto &r : rels) {
+ Symbol &sym = sec->file->getSymbol(r.getSymbol(config->isMips64EL));
+ // The destination output section can be nullptr, osec, or those described
+ // by the NOCROSSREFS/NOCROSSREFS_TO command.
+ auto *dstOsec = sym.getOutputSection();
+ if (!dstOsec || dstOsec == osec || !matchesRefTo(cmd, dstOsec->name))
+ continue;
+
+ std::string toSymName;
+ if (!sym.isSection())
+ toSymName = toString(sym);
+ else if (auto *d = dyn_cast<Defined>(&sym))
+ toSymName = d->section->name;
+ errorOrWarn(sec->getLocation(r.r_offset) +
+ ": prohibited cross reference from '" + osec->name + "' to '" +
+ toSymName + "' in '" + dstOsec->name + "'");
+ }
+}
+
+template <class ELFT> void elf::checkNoCrossRefs() {
+ for (OutputSection *osec : outputSections) {
+ for (const NoCrossRefCommand &noxref : script->noCrossRefs) {
+ if (!llvm::is_contained(noxref.outputSections, osec->name))
+ continue;
+ for (SectionCommand *cmd : osec->commands) {
+ auto *isd = dyn_cast<InputSectionDescription>(cmd);
+ if (!isd)
+ continue;
+ parallelForEach(isd->sections, [&](InputSection *sec) {
+ const RelsOrRelas<ELFT> rels = sec->template relsOrRelas<ELFT>();
+ if (rels.areRelocsRel())
+ scanCrossRefs<ELFT>(noxref, osec, sec, rels.rels);
+ else
+ scanCrossRefs<ELFT>(noxref, osec, sec, rels.relas);
+ });
+ }
+ }
+ }
+}
+
template void elf::scanRelocations<ELF32LE>();
template void elf::scanRelocations<ELF32BE>();
template void elf::scanRelocations<ELF64LE>();
template void elf::scanRelocations<ELF64BE>();
+
+template void elf::checkNoCrossRefs<ELF32LE>();
+template void elf::checkNoCrossRefs<ELF32BE>();
+template void elf::checkNoCrossRefs<ELF64LE>();
+template void elf::checkNoCrossRefs<ELF64BE>();
diff --git a/lld/ELF/Relocations.h b/lld/ELF/Relocations.h
index e299d23dd7db3..1bee0dedf8587 100644
--- a/lld/ELF/Relocations.h
+++ b/lld/ELF/Relocations.h
@@ -141,6 +141,7 @@ struct JumpInstrMod {
// Call reportUndefinedSymbols() after calling scanRelocations() to emit
// the diagnostics.
template <class ELFT> void scanRelocations();
+template <class ELFT> void checkNoCrossRefs();
void reportUndefinedSymbols();
void postScanRelocations();
void addGotEntry(Symbol &sym);
diff --git a/lld/ELF/ScriptParser.cpp b/lld/ELF/ScriptParser.cpp
index 41bd9a95053f7..65fca3d0d090e 100644
--- a/lld/ELF/ScriptParser.cpp
+++ b/lld/ELF/ScriptParser.cpp
@@ -87,6 +87,7 @@ class ScriptParser final : ScriptLexer {
void readTarget();
void readVersion();
void readVersionScriptCommand();
+ void readNoCrossRefs(bool to);
SymbolAssignment *readSymbolAssignment(StringRef name);
ByteCommand *readByteCommand(StringRef tok);
@@ -280,6 +281,10 @@ void ScriptParser::readLinkerScript() {
readTarget();
} else if (tok == "VERSION") {
readVersion();
+ } else if (tok == "NOCROSSREFS") {
+ readNoCrossRefs(/*to=*/false);
+ } else if (tok == "NOCROSSREFS_TO") {
+ readNoCrossRefs(/*to=*/true);
} else if (SymbolAssignment *cmd = readAssignment(tok)) {
script->sectionCommands.push_back(cmd);
} else {
@@ -299,6 +304,16 @@ void ScriptParser::readDefsym(StringRef name) {
script->sectionCommands.push_back(cmd);
}
+void ScriptParser::readNoCrossRefs(bool to) {
+ expect("(");
+ NoCrossRefCommand cmd{{}, to};
+ while (!atEOF() && !errorCount() && peek() != ")")
+ cmd.outputSections.push_back(next());
+ if (cmd.outputSections.size() >= 2)
+ script->noCrossRefs.push_back(std::move(cmd));
+ expect(")");
+}
+
void ScriptParser::addFile(StringRef s) {
if (isUnderSysroot && s.starts_with("/")) {
SmallString<128> pathData;
diff --git a/lld/ELF/Writer.cpp b/lld/ELF/Writer.cpp
index 8940a1c5d5113..5cffdb771a738 100644
--- a/lld/ELF/Writer.cpp
+++ b/lld/ELF/Writer.cpp
@@ -1943,6 +1943,11 @@ template <class ELFT> void Writer<ELFT>::finalizeSections() {
// have the headers, we can find out which sections they point to.
setReservedSymbolSections();
+ if (script->noCrossRefs.size()) {
+ llvm::TimeTraceScope timeScope("Check NOCROSSREFS");
+ checkNoCrossRefs<ELFT>();
+ }
+
{
llvm::TimeTraceScope timeScope("Finalize synthetic sections");
diff --git a/lld/test/ELF/linkerscript/nocrossrefs.test b/lld/test/ELF/linkerscript/nocrossrefs.test
new file mode 100644
index 0000000000000..5ec6607959494
--- /dev/null
+++ b/lld/test/ELF/linkerscript/nocrossrefs.test
@@ -0,0 +1,84 @@
+# REQUIRES: x86
+# RUN: rm -rf %t && split-file %s %t && cd %t
+
+# RUN: llvm-mc --triple=x86_64 -filetype=obj a.s -o a.o
+# RUN: llvm-mc --triple=x86_64 -filetype=obj data.s -o data.o
+# RUN: ld.lld a.o data.o -T 0.t
+
+# RUN: not ld.lld a.o data.o -T 1.t 2>&1 | FileCheck %s --check-prefix=ERR1 --implicit-check-not=error:
+# ERR1: error: a.o:(.text.start+0x11): prohibited cross reference from '.text' to 'data' in '.data'
+
+## .text and .text1 are in two NOCROSSREFS commands. Violations are reported twice.
+# RUN: not ld.lld --threads=1 a.o data.o -T 2.t 2>&1 | FileCheck %s --check-prefix=ERR2 --implicit-check-not=error:
+# ERR2: error: a.o:(.text.start+0x6): prohibited cross reference from '.text' to '.text1' in '.text1'
+# ERR2-NEXT: error: a.o:(.text.start+0x6): prohibited cross reference from '.text' to '.text1' in '.text1'
+# ERR2-NEXT: error: a.o:(.text.start+0xb): prohibited cross reference from '.text' to 'foo2' in '.text2'
+# ERR2-NEXT: error: a.o:(.text.start+0x11): prohibited cross reference from '.text' to 'data' in '.data'
+# ERR2-NEXT: error: a.o:(.text.start+0x17): prohibited cross reference from '.text' to 'str1' in '.rodata'
+## .data occurs twice in the command, but the violation is only reported once.
+# ERR2-NEXT: error: a.o:(.text1+0x1): prohibited cross reference from '.text1' to '_edata' in '.data'
+# ERR2-NEXT: error: a.o:(.nonalloc+0x0): prohibited cross reference from '.nonalloc' to '.text' in '.text'
+# ERR2-NEXT: error: a.o:(.nonalloc+0x10): prohibited cross reference from '.nonalloc' to 'data' in '.data'
+
+# RUN: not ld.lld a.o data.o -T 3.t 2>&1 | FileCheck %s --check-prefix=ERR3 --implicit-check-not=error:
+# ERR3: error: a.o:(.nonalloc+0x0): prohibited cross reference from '.nonalloc' to '.text' in '.text'
+
+#--- 0.t
+## Some cases that do not cause errors.
+abs = 42;
+NOCROSSREFS();
+NOCROSSREFS(.text);
+NOCROSSREFS( .text .text3 );
+NOCROSSREFS_TO(.text .text2 .text3 .data );
+NOCROSSREFS_TO(.data .text2 .text3);
+
+#--- 1.t
+abs = 42;
+NOCROSSREFS(.text .data);
+
+#--- 2.t
+abs = 42;
+NOCROSSREFS(.text .text1 .text .text1);
+NOCROSSREFS(.text .text1 .text2 .data .rodata .data .nonalloc);
+
+#--- 3.t
+abs = 42;
+NOCROSSREFS_TO(.text .text1 .text2 .data .nonalloc);
+
+#--- a.s
+.global _start, foo1, foo2, foo3
+.section .text.start,"ax"
+_start:
+ call _start
+ call .text1
+ call foo2
+ movl data(%rip), %eax
+ movl str1(%rip), %eax
+
+.section .text1,"ax"
+foo1:
+ call _edata
+
+.section .text2,"ax"
+foo2:
+ call foo3
+
+.section .text3,"ax"
+foo3:
+ call foo2
+
+.section .rodata.str1.1,"aMS",@progbits,1
+str1:
+ .asciz "abc"
+
+.section .nonalloc,""
+ .quad .text
+ .quad .text3
+ .quad data
+
+#--- data.s
+.section .data,"aw"
+.globl data
+data:
+ .byte 0
+ .quad abs
|
Created using spr 1.3.5-bogner
smithp35
reviewed
Jul 16, 2024
lld/ELF/LinkerScript.h
Outdated
| @@ -256,6 +256,16 @@ struct InsertCommand { | |||
| StringRef where; | |||
| }; | |||
|
|
|||
| // A NOCROSSREFS/NOCROSSREFS_TO command that probits references among certain | |||
There was a problem hiding this comment.
typo probits -> prohibits.
I'd also say references between certain output sections but that just reads a bit more naturally to me.
| cmd.outputSections.push_back(unquote(next())); | ||
| if (cmd.outputSections.size() >= 2) | ||
| script->noCrossRefs.push_back(std::move(cmd)); | ||
| } |
There was a problem hiding this comment.
Would it be worth a warning if the outputSections.size() < 2 I don't think there could be a legal use case for NOCROSSREFS() or NOCROSSREFS(<section>).
lld/ELF/Relocations.cpp
Outdated
| InputSection *sec, Rels rels) { | ||
| for (const auto &r : rels) { | ||
| Symbol &sym = sec->file->getSymbol(r.getSymbol(config->isMips64EL)); | ||
| // The destination output section can be nullptr, osec, or those described |
There was a problem hiding this comment.
I'm don't fully understand what
or those described by the NOCROSSREFS/NOCROSSREFS_TO command. means in this context.
If the comment is describing permitted cross-references then I'd expect something like:
A legal cross-reference is when the destination output section is nullptr, osec for a self-reference, or a section that is not in the set described by the NOCROSSREFS/NOCROSSREFS_TO command.
There was a problem hiding this comment.
Thanks for the suggestion. Improved the comment.
|
|
||
| template <class ELFT, class Rels> | ||
| static void scanCrossRefs(const NoCrossRefCommand &cmd, OutputSection *osec, | ||
| InputSection *sec, Rels rels) { |
There was a problem hiding this comment.
For a NoCrossRefCommand cmd with toFirst == true can we early exit if osec == cmd.outputSections[0]?
This would be the case where we search the relocations from tosection.
There was a problem hiding this comment.
Added an early return to elf::checkNoCrossRefs: (noxref.toFirst && noxref.outputSections[0] == osec->name)
Created using spr 1.3.5-bogner
Created using spr 1.3.5-bogner
Created using spr 1.3.5-bogner
smithp35
approved these changes
Jul 17, 2024
MaskRay
deleted the
users/MaskRay/spr/elf-support-nocrossrefs-and-nocrosserfs_to
branch
yuxuanchen1997
pushed a commit
that referenced
this pull request
Jul 25, 2024
Implement the two commands described by https://sourceware.org/binutils/docs/ld/Miscellaneous-Commands.html After `outputSections` is available, check each output section described by at least one `NOCROSSREFS`/`NOCROSSERFS_TO` command. For each checked output section, scan relocations from its input sections. This step is slow, therefore utilize `parallelForEach(isd->sections, ...)`. To support non SHF_ALLOC sections, `InputSectionBase::relocations` (empty) cannot be used. In addition, we may explore eliminating this member to speed up relocation scanning. Some parse code is adapted from #95714. Close #41825 Pull Request: #98773
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Implement the two commands described by
https://sourceware.org/binutils/docs/ld/Miscellaneous-Commands.html
After
outputSectionsis available, check each output section describedby at least one
NOCROSSREFS/NOCROSSERFS_TOcommand. For each checkedoutput section, scan relocations from its input sections.
This step is slow, therefore utilize
parallelForEach(isd->sections, ...).To support non SHF_ALLOC sections,
InputSectionBase::relocations(empty) cannot be used. In addition, we may explore eliminating this
member to speed up relocation scanning.
Some parse code is adapted from #95714.
Close #41825