[LLD][ELF] Allow memory region in OVERLAY #133540
Merged
Conversation
This file contains 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
|
@llvm/pr-subscribers-lld-elf @llvm/pr-subscribers-lld Author: Daniel Thornburgh (mysterymath) ChangesThis allows the contents of OVERLAYs to be attributed to memory regions. This is the only clean way to overlap VMAs in linker scripts that choose to primarily use memory regions to lay out addresses. This also simplifies OVERLAY expansion to better match GNU LD. Expressions for the first section's LMA and VMA are not generated if the user did not provide them. This allows the LMA/VMA offset to be preserved across multiple overlays in the same region, as with regular sections. Fixes #129816 Full diff: https://github.com/llvm/llvm-project/pull/133540.diff 5 Files Affected:
diff --git a/lld/ELF/LinkerScript.cpp b/lld/ELF/LinkerScript.cpp
index e19823f2ea752..8437049a209d9 100644
--- a/lld/ELF/LinkerScript.cpp
+++ b/lld/ELF/LinkerScript.cpp
@@ -182,7 +182,18 @@ void LinkerScript::expandMemoryRegions(uint64_t size) {
void LinkerScript::expandOutputSection(uint64_t size) {
state->outSec->size += size;
- expandMemoryRegions(size);
+ size_t regionSize = size;
+ if (state->outSec->inOverlay) {
+ // Expand the overlay if necessary, and expand the region by the
+ // corresponding amount.
+ if (state->outSec->size > state->overlaySize) {
+ regionSize = state->outSec->size - state->overlaySize;
+ state->overlaySize = state->outSec->size;
+ } else {
+ regionSize = 0;
+ }
+ }
+ expandMemoryRegions(regionSize);
}
void LinkerScript::setDot(Expr e, const Twine &loc, bool inSec) {
@@ -1273,6 +1284,8 @@ bool LinkerScript::assignOffsets(OutputSection *sec) {
// NOBITS TLS sections are similar. Additionally save the end address.
state->tbssAddr = dot;
dot = savedDot;
+ } else if (sec->lastInOverlay) {
+ state->overlaySize = 0;
}
return addressChanged;
}
diff --git a/lld/ELF/LinkerScript.h b/lld/ELF/LinkerScript.h
index 0a2dda13f4ef8..65c443ba74d27 100644
--- a/lld/ELF/LinkerScript.h
+++ b/lld/ELF/LinkerScript.h
@@ -311,6 +311,7 @@ class LinkerScript final {
MemoryRegion *lmaRegion = nullptr;
uint64_t lmaOffset = 0;
uint64_t tbssAddr = 0;
+ uint64_t overlaySize = 0;
};
Ctx &ctx;
diff --git a/lld/ELF/OutputSections.h b/lld/ELF/OutputSections.h
index 3ab36a21ce488..33516a62ecb43 100644
--- a/lld/ELF/OutputSections.h
+++ b/lld/ELF/OutputSections.h
@@ -102,6 +102,7 @@ class OutputSection final : public SectionBase {
bool expressionsUseSymbols = false;
bool usedInExpression = false;
bool inOverlay = false;
+ bool lastInOverlay = false;
// Tracks whether the section has ever had an input section added to it, even
// if the section was later removed (e.g. because it is a synthetic section
diff --git a/lld/ELF/ScriptParser.cpp b/lld/ELF/ScriptParser.cpp
index 4c52bfda7a70e..09786aebaf337 100644
--- a/lld/ELF/ScriptParser.cpp
+++ b/lld/ELF/ScriptParser.cpp
@@ -561,37 +561,40 @@ void ScriptParser::readSearchDir() {
// https://sourceware.org/binutils/docs/ld/Overlay-Description.html#Overlay-Description
SmallVector<SectionCommand *, 0> ScriptParser::readOverlay() {
Expr addrExpr;
- if (consume(":")) {
- addrExpr = [s = ctx.script] { return s->getDot(); };
- } else {
+ if (!consume(":")) {
addrExpr = readExpr();
expect(":");
}
- // When AT is omitted, LMA should equal VMA. script->getDot() when evaluating
- // lmaExpr will ensure this, even if the start address is specified.
- Expr lmaExpr = consume("AT") ? readParenExpr()
- : [s = ctx.script] { return s->getDot(); };
+ Expr lmaExpr = consume("AT") ? readParenExpr() : Expr{};
expect("{");
SmallVector<SectionCommand *, 0> v;
OutputSection *prev = nullptr;
while (!errCount(ctx) && !consume("}")) {
// VA is the same for all sections. The LMAs are consecutive in memory
- // starting from the base load address specified.
+ // starting from the base load address.
OutputDesc *osd = readOverlaySectionDescription();
osd->osec.addrExpr = addrExpr;
if (prev) {
osd->osec.lmaExpr = [=] { return prev->getLMA() + prev->size; };
} else {
osd->osec.lmaExpr = lmaExpr;
- // Use first section address for subsequent sections as initial addrExpr
- // can be DOT. Ensure the first section, even if empty, is not discarded.
+ // Use first section address for subsequent sections. Ensure the first
+ // section, even if empty, is not discarded.
osd->osec.usedInExpression = true;
addrExpr = [=]() -> ExprValue { return {&osd->osec, false, 0, ""}; };
}
v.push_back(osd);
prev = &osd->osec;
}
+ if (!v.empty())
+ static_cast<OutputDesc *>(v.back())->osec.lastInOverlay = true;
+ if (consume(">")) {
+ StringRef regionName = readName();
+ for (SectionCommand *od : v)
+ static_cast<OutputDesc *>(od)->osec.memoryRegionName =
+ std::string(regionName);
+ }
// According to the specification, at the end of the overlay, the location
// counter should be equal to the overlay base address plus size of the
diff --git a/lld/test/ELF/linkerscript/overlay.test b/lld/test/ELF/linkerscript/overlay.test
index 7c64303b45659..8324c81a57092 100644
--- a/lld/test/ELF/linkerscript/overlay.test
+++ b/lld/test/ELF/linkerscript/overlay.test
@@ -9,6 +9,7 @@
## .text does not cause overlapping error and that
## .text's VA is 0x1000 + max(sizeof(.out.big), sizeof(.out.small)).
+# RUN: ld.lld a.o -T a.t -o a
# RUN: llvm-readelf --sections -l a | FileCheck %s
# CHECK: Name Type Address Off Size
@@ -41,6 +42,23 @@
# ERR2-NEXT:>>> .out.aaa { *(.aaa) } > AX AT>FLASH
# ERR2-NEXT:>>> ^
+# RUN: ld.lld a.o -T region.t -o region
+# RUN: llvm-readelf --sections -l region | FileCheck --check-prefix=REGION %s
+
+# REGION: Name Type Address Off Size
+# REGION: .big1 PROGBITS 0000000000001000 001000 000008
+# REGION-NEXT: .small1 PROGBITS 0000000000001000 002000 000004
+# REGION: .big2 PROGBITS 0000000000001008 002008 000008
+# REGION-NEXT: .small2 PROGBITS 0000000000001008 003008 000004
+# REGION-NEXT: .text PROGBITS 0000000000001010 003010 000001
+
+# REGION: Program Headers:
+# REGION: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
+# REGION-NEXT: LOAD 0x001000 0x0000000000001000 0x0000000000001000 0x000008 0x000008 R 0x1000
+# REGION-NEXT: LOAD 0x002000 0x0000000000001000 0x0000000000001008 0x000010 0x000010 R 0x1000
+# REGION-NEXT: LOAD 0x003008 0x0000000000001008 0x0000000000001018 0x000004 0x000004 R 0x1000
+# REGION-NEXT: LOAD 0x003010 0x0000000000001010 0x0000000000001020 0x000001 0x000001 R E 0x1000
+
#--- a.s
.globl _start
_start:
@@ -76,6 +94,22 @@ SECTIONS {
.text : { *(.text) }
}
+#--- region.t
+MEMORY { region : ORIGIN = 0x1000, LENGTH = 0x1000 }
+SECTIONS {
+## Memory region instead of explicit address.
+ OVERLAY : {
+ .big1 { *(.big1) }
+ .small1 { *(.small1) }
+ } >region
+ OVERLAY : {
+ .big2 { *(.big2) }
+ .small2 { *(.small2) }
+ } >region
+ .text : { *(.text) } >region
+ /DISCARD/ : { *(.big* .small*) }
+}
+
#--- err1.t
SECTIONS {
OVERLAY 0x1000 : AT ( 0x2000 ) {
|
MaskRay
approved these changes
Mar 29, 2025
There was a problem hiding this comment.
Thanks! Code looks great. Consider renaming lastInOverlay to firstInOverlay and moving the code to expand* to place overlay code together.
diff --git i/lld/ELF/LinkerScript.cpp w/lld/ELF/LinkerScript.cpp
index 8437049a209d..58bea55e4e97 100644
--- i/lld/ELF/LinkerScript.cpp
+++ w/lld/ELF/LinkerScript.cpp
@@ -186,2 +186,4 @@ void LinkerScript::expandOutputSection(uint64_t size) {
if (state->outSec->inOverlay) {
+ if (state->outSec->firstInOverlay)
+ state->overlaySize = 0;
// Expand the overlay if necessary, and expand the region by the
@@ -1286,4 +1288,2 @@ bool LinkerScript::assignOffsets(OutputSection *sec) {
dot = savedDot;
- } else if (sec->lastInOverlay) {
- state->overlaySize = 0;
}
diff --git i/lld/ELF/ScriptParser.cpp w/lld/ELF/ScriptParser.cpp
index 09786aebaf33..555fceccc4a7 100644
--- i/lld/ELF/ScriptParser.cpp
+++ w/lld/ELF/ScriptParser.cpp
@@ -590,3 +590,3 @@ SmallVector<SectionCommand *, 0> ScriptParser::readOverlay() {
if (!v.empty())
- static_cast<OutputDesc *>(v.back())->osec.lastInOverlay = true;
+ static_cast<OutputDesc *>(v[0])->osec.firstInOverlay = true;
if (consume(">")) {
As this is a feature request we prefer |
As written, this wouldn't be correct when the first overlay section is empty, as |
This allows the contents of OVERLAYs to be attributed to memory regions. This is the only clean way to overlap VMAs in linker scripts that choose to primarily use memory regions to lay out addresses. This also simplifies OVERLAY expansion to better match GNU LD. Expressions for the first section's LMA and VMA are not generated if the user did not provide them. This allows the LMA/VMA offset to be preserved across multiple overlays in the same region, as with regular sections. Fixes llvm#129816
mysterymath
force-pushed
the
lld/overlay
branch
from
e7994b7 to
fc4c2c7
Compare
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.
This allows the contents of OVERLAYs to be attributed to memory regions. This is the only clean way to overlap VMAs in linker scripts that choose to primarily use memory regions to lay out addresses.
This also simplifies OVERLAY expansion to better match GNU LD. Expressions for the first section's LMA and VMA are not generated if the user did not provide them. This allows the LMA/VMA offset to be preserved across multiple overlays in the same region, as with regular sections.
Closes #129816