[AArch64] Refactor @plt, @gotpcrel, and @AUTH to use parseDataExpr #134202
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-mc @llvm/pr-subscribers-backend-aarch64 Author: Fangrui Song (MaskRay) ChangesFollowing PR #132569 (RISC-V), which added Note: AArch64 ELF lacks an official syntax for data directives AsmParser's In the test elf-reloc-ptrauth.s, many errors are now reported at parse Full diff: https://github.com/llvm/llvm-project/pull/134202.diff 7 Files Affected:
diff --git a/llvm/include/llvm/MC/MCParser/MCAsmParser.h b/llvm/include/llvm/MC/MCParser/MCAsmParser.h
index c65a38c944eea..bbe6d1f2a0082 100644
--- a/llvm/include/llvm/MC/MCParser/MCAsmParser.h
+++ b/llvm/include/llvm/MC/MCParser/MCAsmParser.h
@@ -333,6 +333,9 @@ class MCAsmParser {
/// Parse a .gnu_attribute.
bool parseGNUAttribute(SMLoc L, int64_t &Tag, int64_t &IntegerValue);
+
+ bool parseAtSpecifier(const MCExpr *&Res, SMLoc &EndLoc);
+ const MCExpr *applySpecifier(const MCExpr *E, uint32_t Variant);
};
/// Create an MCAsmParser instance for parsing assembly similar to gas syntax
diff --git a/llvm/lib/MC/MCParser/AsmParser.cpp b/llvm/lib/MC/MCParser/AsmParser.cpp
index 71f2bdbdf0b16..14faea7c48af3 100644
--- a/llvm/lib/MC/MCParser/AsmParser.cpp
+++ b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -670,8 +670,6 @@ class AsmParser : public MCAsmParser {
bool parseEscapedString(std::string &Data) override;
bool parseAngleBracketString(std::string &Data) override;
- const MCExpr *applySpecifier(const MCExpr *E, uint32_t Variant);
-
// Macro-like directives
MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
@@ -1194,7 +1192,7 @@ bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
Split = std::make_pair(Identifier, VName);
}
- } else {
+ } else if (Lexer.getAllowAtInIdentifier()) {
Split = Identifier.split('@');
}
} else if (MAI.useParensForSpecifier() &&
@@ -1342,7 +1340,7 @@ bool AsmParser::parseExpression(const MCExpr *&Res) {
return parseExpression(Res, EndLoc);
}
-const MCExpr *AsmParser::applySpecifier(const MCExpr *E, uint32_t Spec) {
+const MCExpr *MCAsmParser::applySpecifier(const MCExpr *E, uint32_t Spec) {
// Ask the target implementation about this expression first.
const MCExpr *NewE = getTargetParser().applySpecifier(E, Spec, Ctx);
if (NewE)
@@ -1433,6 +1431,23 @@ static std::string angleBracketString(StringRef AltMacroStr) {
return Res;
}
+bool MCAsmParser::parseAtSpecifier(const MCExpr *&Res, SMLoc &EndLoc) {
+ if (parseOptionalToken(AsmToken::At)) {
+ if (getLexer().isNot(AsmToken::Identifier))
+ return TokError("expected specifier following '@'");
+
+ auto Spec = MAI.getSpecifierForName(getTok().getIdentifier());
+ if (!Spec)
+ return TokError("invalid specifier '@" + getTok().getIdentifier() + "'");
+
+ const MCExpr *ModifiedRes = applySpecifier(Res, *Spec);
+ if (ModifiedRes)
+ Res = ModifiedRes;
+ Lex();
+ }
+ return false;
+}
+
/// Parse an expression and return it.
///
/// expr ::= expr &&,|| expr -> lowest.
@@ -1453,8 +1468,7 @@ bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
// As a special case, we support 'a op b @ modifier' by rewriting the
// expression to include the modifier. This is inefficient, but in general we
// expect users to use 'a@modifier op b'.
- if (Ctx.getAsmInfo()->useAtForSpecifier() &&
- parseOptionalToken(AsmToken::At)) {
+ if (Lexer.getAllowAtInIdentifier() && parseOptionalToken(AsmToken::At)) {
if (Lexer.isNot(AsmToken::Identifier))
return TokError("unexpected symbol modifier following '@'");
diff --git a/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp b/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
index 28b4cbb5efed8..894be7565fabe 100644
--- a/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
+++ b/llvm/lib/Target/AArch64/AsmParser/AArch64AsmParser.cpp
@@ -25,6 +25,7 @@
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/ADT/Twine.h"
+#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCExpr.h"
#include "llvm/MC/MCInst.h"
@@ -180,6 +181,7 @@ class AArch64AsmParser : public MCTargetAsmParser {
bool showMatchError(SMLoc Loc, unsigned ErrCode, uint64_t ErrorInfo,
OperandVector &Operands);
+ bool parseDataExpr(const MCExpr *&Res) override;
bool parseAuthExpr(const MCExpr *&Res, SMLoc &EndLoc);
bool parseDirectiveArch(SMLoc L);
@@ -335,8 +337,6 @@ class AArch64AsmParser : public MCTargetAsmParser {
unsigned validateTargetOperandClass(MCParsedAsmOperand &Op,
unsigned Kind) override;
- bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override;
-
static bool classifySymbolRef(const MCExpr *Expr,
AArch64MCExpr::Specifier &ELFSpec,
MCSymbolRefExpr::VariantKind &DarwinRefKind,
@@ -4478,6 +4478,19 @@ bool AArch64AsmParser::parseSymbolicImmVal(const MCExpr *&ImmVal) {
if (HasELFModifier)
ImmVal = AArch64MCExpr::create(ImmVal, RefKind, getContext());
+ SMLoc EndLoc;
+ if (getContext().getAsmInfo()->hasSubsectionsViaSymbols()) {
+ if (getParser().parseAtSpecifier(ImmVal, EndLoc))
+ return true;
+ const MCExpr *Term;
+ if (parseOptionalToken(AsmToken::Plus)) {
+ if (getParser().parseExpression(Term, EndLoc))
+ return true;
+ ImmVal =
+ MCBinaryExpr::create(MCBinaryExpr::Add, ImmVal, Term, getContext());
+ }
+ }
+
return false;
}
@@ -5007,11 +5020,18 @@ bool AArch64AsmParser::parseOperand(OperandVector &Operands, bool isCondCode,
// This was not a register so parse other operands that start with an
// identifier (like labels) as expressions and create them as immediates.
- const MCExpr *IdVal;
+ const MCExpr *IdVal, *Term;
S = getLoc();
if (getParser().parseExpression(IdVal))
return true;
- E = SMLoc::getFromPointer(getLoc().getPointer() - 1);
+ if (getParser().parseAtSpecifier(IdVal, E))
+ return true;
+ if (parseOptionalToken(AsmToken::Plus)) {
+ if (getParser().parseExpression(Term, E))
+ return true;
+ IdVal =
+ MCBinaryExpr::create(MCBinaryExpr::Add, IdVal, Term, getContext());
+ }
Operands.push_back(AArch64Operand::CreateImm(IdVal, S, E, getContext()));
// Parse an optional shift/extend modifier.
@@ -8086,11 +8106,56 @@ bool AArch64AsmParser::parseDirectiveAeabiAArch64Attr(SMLoc L) {
return false;
}
-bool AArch64AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
- // Try @AUTH expressions: they're more complex than the usual symbol variants.
- if (!parseAuthExpr(Res, EndLoc))
+bool AArch64AsmParser::parseDataExpr(const MCExpr *&Res) {
+ SMLoc EndLoc;
+
+ if (getParser().parseExpression(Res))
+ return true;
+ MCAsmParser &Parser = getParser();
+ if (!parseOptionalToken(AsmToken::At))
return false;
- return getParser().parsePrimaryExpr(Res, EndLoc, nullptr);
+ if (getLexer().getKind() != AsmToken::Identifier)
+ return Error(getLoc(), "expected relocation specifier");
+
+ std::string Identifier = Parser.getTok().getIdentifier().lower();
+ SMLoc Loc = getLoc();
+ Lex();
+ if (Identifier == "auth")
+ return parseAuthExpr(Res, EndLoc);
+
+ auto Spec = MCSymbolRefExpr::VK_None;
+ if (STI->getTargetTriple().isOSBinFormatMachO()) {
+ if (Identifier == "got")
+ Spec = MCSymbolRefExpr::VK_GOT;
+ } else {
+ // Unofficial, experimental syntax that will be changed.
+ if (Identifier == "gotpcrel")
+ Spec = MCSymbolRefExpr::VK_GOTPCREL;
+ else if (Identifier == "plt")
+ Spec = MCSymbolRefExpr::VK_PLT;
+ }
+ if (Spec == MCSymbolRefExpr::VK_None)
+ return Error(Loc, "invalid relocation specifier");
+ if (auto *SRE = dyn_cast<MCSymbolRefExpr>(Res))
+ Res = MCSymbolRefExpr::create(&SRE->getSymbol(), Spec, getContext(),
+ SRE->getLoc());
+ else
+ return Error(Loc, "@ specifier only allowed after a symbol");
+
+ for (;;) {
+ std::optional<MCBinaryExpr::Opcode> Opcode;
+ if (parseOptionalToken(AsmToken::Plus))
+ Opcode = MCBinaryExpr::Add;
+ else if (parseOptionalToken(AsmToken::Minus))
+ Opcode = MCBinaryExpr::Sub;
+ else
+ break;
+ const MCExpr *Term;
+ if (getParser().parsePrimaryExpr(Term, EndLoc, nullptr))
+ return true;
+ Res = MCBinaryExpr::create(*Opcode, Res, Term, getContext());
+ }
+ return false;
}
/// parseAuthExpr
@@ -8100,54 +8165,8 @@ bool AArch64AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
bool AArch64AsmParser::parseAuthExpr(const MCExpr *&Res, SMLoc &EndLoc) {
MCAsmParser &Parser = getParser();
MCContext &Ctx = getContext();
-
AsmToken Tok = Parser.getTok();
- // Look for '_sym@AUTH' ...
- if (Tok.is(AsmToken::Identifier) && Tok.getIdentifier().ends_with("@AUTH")) {
- StringRef SymName = Tok.getIdentifier().drop_back(strlen("@AUTH"));
- if (SymName.contains('@'))
- return TokError(
- "combination of @AUTH with other modifiers not supported");
- Res = MCSymbolRefExpr::create(Ctx.getOrCreateSymbol(SymName), Ctx);
-
- Parser.Lex(); // Eat the identifier.
- } else {
- // ... or look for a more complex symbol reference, such as ...
- SmallVector<AsmToken, 6> Tokens;
-
- // ... '"_long sym"@AUTH' ...
- if (Tok.is(AsmToken::String))
- Tokens.resize(2);
- // ... or '(_sym + 5)@AUTH'.
- else if (Tok.is(AsmToken::LParen))
- Tokens.resize(6);
- else
- return true;
-
- if (Parser.getLexer().peekTokens(Tokens) != Tokens.size())
- return true;
-
- // In either case, the expression ends with '@' 'AUTH'.
- if (Tokens[Tokens.size() - 2].isNot(AsmToken::At) ||
- Tokens[Tokens.size() - 1].isNot(AsmToken::Identifier) ||
- Tokens[Tokens.size() - 1].getIdentifier() != "AUTH")
- return true;
-
- if (Tok.is(AsmToken::String)) {
- StringRef SymName;
- if (Parser.parseIdentifier(SymName))
- return true;
- Res = MCSymbolRefExpr::create(Ctx.getOrCreateSymbol(SymName), Ctx);
- } else {
- if (Parser.parsePrimaryExpr(Res, EndLoc, nullptr))
- return true;
- }
-
- Parser.Lex(); // '@'
- Parser.Lex(); // 'AUTH'
- }
-
// At this point, we encountered "<id>@AUTH". There is no fallback anymore.
if (parseToken(AsmToken::LParen, "expected '('"))
return true;
diff --git a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
index 9ff53631a995e..4bc84ce9b8e80 100644
--- a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
+++ b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MCAsmInfo.cpp
@@ -61,6 +61,7 @@ AArch64MCAsmInfoDarwin::AArch64MCAsmInfoDarwin(bool IsILP32) {
UsesELFSectionDirectiveForBSS = true;
SupportsDebugInformation = true;
UseDataRegionDirectives = true;
+ UseAtForSpecifier = false;
ExceptionsType = ExceptionHandling::DwarfCFI;
@@ -105,6 +106,7 @@ AArch64MCAsmInfoELF::AArch64MCAsmInfoELF(const Triple &T) {
Data64bitsDirective = "\t.xword\t";
UseDataRegionDirectives = false;
+ UseAtForSpecifier = false;
WeakRefDirective = "\t.weak\t";
diff --git a/llvm/test/MC/AArch64/data-directive-specifier.s b/llvm/test/MC/AArch64/data-directive-specifier.s
index 3a8665126097a..5410bcb4a4211 100644
--- a/llvm/test/MC/AArch64/data-directive-specifier.s
+++ b/llvm/test/MC/AArch64/data-directive-specifier.s
@@ -1,5 +1,6 @@
# RUN: llvm-mc -triple=aarch64 -filetype=obj %s | llvm-readobj -r - | FileCheck %s
-# RUN: not llvm-mc -triple=aarch64 -filetype=obj %s --defsym ERR=1 -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR --implicit-check-not=error:
+# RUN: not llvm-mc -triple=aarch64 %s --defsym ERR=1 -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR --implicit-check-not=error:
+# RUN: not llvm-mc -triple=aarch64 -filetype=obj %s --defsym OBJERR=1 -o /dev/null 2>&1 | FileCheck %s --check-prefix=OBJERR --implicit-check-not=error:
.globl g
g:
@@ -32,13 +33,21 @@ data1:
.word extern@GOTPCREL-5
.ifdef ERR
-# ERR: [[#@LINE+1]]:7: error: symbol 'und' can not be undefined in a subtraction expression
-.word extern@plt - und
+# ERR: [[#@LINE+1]]:9: error: @ specifier only allowed after a symbol
+.quad 3@plt - .
+
+# ERR: [[#@LINE+1]]:9: error: expected ')'
+.quad (l@plt - .)
+.endif
+.ifdef OBJERR
.quad g@plt - .
.word extern@gotpcrel - .
-# ERR: [[#@LINE+1]]:7: error: symbol 'und' can not be undefined in a subtraction expression
+# OBJERR: [[#@LINE+1]]:7: error: symbol 'und' can not be undefined in a subtraction expression
+.word extern@plt - und
+
+# OBJERR: [[#@LINE+1]]:7: error: symbol 'und' can not be undefined in a subtraction expression
.word extern@gotpcrel - und
.endif
diff --git a/llvm/test/MC/AArch64/elf-reloc-ptrauth.s b/llvm/test/MC/AArch64/elf-reloc-ptrauth.s
index bed85bcc5798b..263ed91ec8e99 100644
--- a/llvm/test/MC/AArch64/elf-reloc-ptrauth.s
+++ b/llvm/test/MC/AArch64/elf-reloc-ptrauth.s
@@ -1,4 +1,4 @@
-// RUN: llvm-mc -triple=aarch64 %s --defsym=ASMONLY=1 | FileCheck %s --check-prefix=ASM
+// RUN: llvm-mc -triple=aarch64 %s | FileCheck %s --check-prefix=ASM
// RUN: llvm-mc -triple=aarch64 -filetype=obj %s | \
// RUN: llvm-readelf -S -r -x .test - | FileCheck %s --check-prefix=RELOC
@@ -41,8 +41,6 @@
// RELOC-NEXT: 70 00000000 10000000
// ^^^^ discriminator
// ^^ 0 no addr diversity 0 reserved 00 ia key 0000 reserved
-// RELOC-NEXT: 80 04000000 00000000
-// Folded to constant 4 bytes difference between _g9 and _g8
.section .helper
.local "_g 6"
@@ -63,12 +61,12 @@ _g9:
.quad _g0@AUTH(ia,42)
.quad 0
-// ASM: .xword _g1@AUTH(ib,0)
-.quad _g1@AUTH(ib,0)
+// ASM: .xword (+_g1)@AUTH(ib,0)
+.quad +_g1@AUTH(ib,0)
.quad 0
// ASM: .xword _g2@AUTH(da,5,addr)
-.quad _g2@AUTH(da,5,addr)
+.quad _g2 @ AUTH(da,5,addr)
.quad 0
// ASM: .xword _g3@AUTH(db,65535,addr)
@@ -91,33 +89,20 @@ _g9:
.quad ("_g 7" + 7)@AUTH(ia,16)
.quad 0
-// ASM: .xword _g9@AUTH(ia,42)-_g8@AUTH(ia,42)
-.quad _g9@AUTH(ia,42) - _g8@AUTH(ia,42)
-.quad 0
+// RUN: not llvm-mc -triple=aarch64 --defsym=ERR=1 %s 2>&1 | \
+// RUN: FileCheck %s --check-prefix=ERR
-.ifdef ASMONLY
+.ifdef ERR
-// ASM: .xword _g10@AUTH(ia,42)+1
.quad _g10@AUTH(ia,42) + 1
-// ASM: .xword 1+_g11@AUTH(ia,42)
.quad 1 + _g11@AUTH(ia,42)
-// ASM: .xword 1+_g12@AUTH(ia,42)+1
.quad 1 + _g12@AUTH(ia,42) + 1
-// ASM: .xword _g13@AUTH(ia,42)+_g14@AUTH(ia,42)
.quad _g13@AUTH(ia,42) + _g14@AUTH(ia,42)
-// ASM: .xword _g9@AUTH(ia,42)-_g8
.quad _g9@AUTH(ia,42) - _g8
-.quad 0
-
-.endif // ASMONLY
-
-.ifdef ERR
-// RUN: not llvm-mc -triple=aarch64 --defsym=ERR=1 %s 2>&1 | \
-// RUN: FileCheck %s --check-prefix=ERR
// ERR: :[[#@LINE+1]]:15: error: expected '('
.quad sym@AUTH)ia,42)
@@ -143,51 +128,28 @@ _g9:
// ERR: :[[#@LINE+1]]:21: error: expected ')'
.quad sym@AUTH(ia,42(
-// ERR: :[[#@LINE+1]]:7: error: combination of @AUTH with other modifiers not supported
+// ERR: :[[#@LINE+1]]:14: error: unexpected token
.quad sym@PLT@AUTH(ia,42)
-// ERR: :[[#@LINE+1]]:11: error: invalid variant 'AUTH@GOT'
+// ERR: :[[#@LINE+1]]:15: error: expected '('
.quad sym@AUTH@GOT(ia,42)
-// ERR: :[[#@LINE+1]]:18: error: invalid variant 'TLSDESC@AUTH'
-.quad "long sym"@TLSDESC@AUTH(ia,42)
-
-// ERR: :[[#@LINE+1]]:18: error: invalid variant 'AUTH@PLT'
+// ERR: :[[#@LINE+1]]:22: error: expected '('
.quad "long sym"@AUTH@PLT(ia,42)
-// ERR: :[[#@LINE+1]]:17: error: invalid variant 'GOT@AUTH'
+// ERR: :[[#@LINE+1]]:17: error: invalid relocation specifier
.quad (sym - 5)@GOT@AUTH(ia,42)
-// ERR: :[[#@LINE+1]]:17: error: invalid variant 'AUTH@TLSDESC'
-.quad (sym + 5)@AUTH@TLSDESC(ia,42)
-
-// ERR: :[[#@LINE+1]]:12: error: invalid variant 'AUTH'
-.quad +sym@AUTH(ia,42)
-
-.endif // ERR
-
-.ifdef ERROBJ
-// RUN: not llvm-mc -triple=aarch64 -filetype=obj --defsym=ERROBJ=1 %s -o /dev/null 2>&1 | \
-// RUN: FileCheck %s --check-prefix=ERROBJ
-
-// ERROBJ: :[[#@LINE+1]]:7: error: expected relocatable expression
+// ERR: :[[#@LINE+1]]:23: error: unexpected token
.quad sym@AUTH(ia,42) + 1
-// ERROBJ: :[[#@LINE+1]]:7: error: expected relocatable expression
-.quad 1 + sym@AUTH(ia,42)
-
-// ERROBJ: :[[#@LINE+1]]:7: error: expected relocatable expression
+// ERR: :[[#@LINE+1]]:27: error: unexpected token
.quad 1 + sym@AUTH(ia,42) + 1
-// ERROBJ: :[[#@LINE+1]]:7: error: expected relocatable expression
-.quad sym@AUTH(ia,42) + sym@AUTH(ia,42)
-
-// TODO: do we really want to emit an error here? It might not be important
-// whether a symbol has an AUTH modifier or not since the compile-time computed
-// distance remains the same. Leave it in such state as for now since it
-// makes code simpler: subtraction of a non-AUTH symbol and of a constant
-// are handled identically.
-// ERROBJ: :[[#@LINE+1]]:7: error: expected relocatable expression
+/// @AUTH applies to the whole operand instead of an individual term.
+/// Trailing expression parts are not allowed even if the logical subtraction
+/// result might make sense.
+// ERR: :[[#@LINE+1]]:23: error: unexpected token
.quad _g9@AUTH(ia,42) - _g8
-.endif // ERROBJ
+.endif // ERR
diff --git a/llvm/test/MC/AArch64/label-arithmetic-diags-darwin.s b/llvm/test/MC/AArch64/label-arithmetic-diags-darwin.s
index e32db7c125bb4..357e04a828f8e 100644
--- a/llvm/test/MC/AArch64/label-arithmetic-diags-darwin.s
+++ b/llvm/test/MC/AArch64/label-arithmetic-diags-darwin.s
@@ -1,9 +1,17 @@
+// RUN: not llvm-mc -triple aarch64-darwin -filetype=obj --defsym PARSE=1 %s -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR
// RUN: not llvm-mc -triple aarch64-darwin -filetype=obj %s -o /dev/null 2>&1 | FileCheck %s
// RUN: not llvm-mc -triple aarch64-ios -filetype=obj %s -o /dev/null 2>&1 | FileCheck %s
Lstart:
.space 8
Lend:
+.ifdef PARSE
+ add w0, w1, #(Lend - var@TLVPPAGEOFF)
+ // ERR: [[#@LINE-1]]:27: error: expected ')'
+ cmp w0, #(Lend - var@TLVPPAGEOFF)
+ // ERR: [[#@LINE-1]]:23: error: expected ')'
+
+.else
add w0, w1, #(Lend - external)
cmp w0, #(Lend - external)
// CHECK: error: unknown AArch64 fixup kind!
@@ -13,15 +21,6 @@ Lend:
// CHECK-NEXT: cmp w0, #(Lend - external)
// CHECK-NEXT: ^
- add w0, w1, #(Lend - var@TLVPPAGEOFF)
- cmp w0, #(Lend - var@TLVPPAGEOFF)
- // CHECK: error: unsupported subtraction of qualified symbol
- // CHECK-NEXT: add w0, w1, #(Lend - var@TLVPPAGEOFF)
- // CHECK-NEXT: ^
- // CHECK: error: unsupported subtraction of qualified symbol
- // CHECK-NEXT: cmp w0, #(Lend - var@TLVPPAGEOFF)
- // CHECK-NEXT: ^
-
add w0, w1, #(Lstart - Lend)
cmp w0, #(Lstart - Lend)
// CHECK: error: fixup value out of range
@@ -66,3 +65,4 @@ Lend_across_sec:
// CHECK: error: unknown AArch64 fixup kind!
// CHECK-NEXT: cmp w0, #(Lend_across_sec - Lprivate2)
// CHECK-NEXT: ^
+.endif
|
|
I submitted issue #132569 and created this PR to address the syntax issues in I propose introducing a breaking change to eliminate future dependencies on While I’d prefer for Mach-O to phase out
|
smithp35
reviewed
Apr 4, 2025
There was a problem hiding this comment.
Overall looks good to me.
To the best that I can infer from reading the code this doesn't limit affect any data directive that is legal Today.
It is a bit of shame that we lose some of the context from the error messages, although I can't easily see how to put that back at the time the error is detected.
Created using spr 1.3.5-bogner
Created using spr 1.3.5-bogner
smithp35
approved these changes
Apr 7, 2025
|
LGTM as long as there are no other objections, thanks! |
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Apr 8, 2025
…DataExpr Following PR #132569 (RISC-V), which added `parseDataExpr` for parsing expressions in data directives (e.g., `.word`), this PR migrates AArch64 `@plt`, `@gotpcrel`, and `@AUTH` from the `parsePrimaryExpr` workaround to `parseDataExpr`. The goal is to align with the GNU assembler model, where relocation specifiers apply to the entire operand rather than individual terms, reducing complexity-especially evident in `@AUTH` parsing. Note: AArch64 ELF lacks an official syntax for data directives (#132570). A prefix notation might be a preferable future direction. I recommend `%specifier(expr)`. AsmParser's `@specifier` parsing is suboptimal, necessitating lexer workarounds. `@` might appear multiple times in an operand. We should not use `@` beyond the existing AArch64 Mach-O instruction operands. In the test elf-reloc-ptrauth.s, many errors are now reported at parse time. Pull Request: llvm/llvm-project#134202
|
I'm not sure because I haven't been able to reproduce locally (cross-compiling), but I suspect this might be causing failures building chromium. Apologies if this is ends up being the wrong change. We get the following output (example failed build).: The source file in question can be viewed here. I'm not familiar with this code, but it seems like it's complaining about the |
|
The relevant source file is actually synthesized from a perl script (don't ask; it's OpenSSL's fault) that transforms ELF-style I take it That said, it looks like OpenSSL (which has a similar file) does the same thing, so I suspect LLVM will need to support |
|
I'm now not so sure this syntax is invalid. Here's what Xcode's clang outputs: That is definitely not the syntax I would have expected, given that |
MaskRay
added a commit
that referenced
this pull request
Apr 10, 2025
#134202 removed support for `sym@page-offset` in instruction operands. This change is generally reasonable since subtracting an offset from a symbol typically doesn’t make sense for Mach-O due to its .subsections_via_symbols mechanism, which treats them as separate atoms. However, BoringSSL relies on a temporary symbol with a negative offset, which can be meaningful when the symbol and the referenced location are within the same atom. ``` ../../third_party/boringssl/src/gen/bcm/p256-armv8-asm-apple.S:1160:25: error: unexpected token in argument list adrp x23,Lone_mont@PAGE-64 ``` It's worth noting that expressions involving @ can be complex and brittle in MCParser, and much of the Mach-O @ offsets remains under-tested. * Allow default argument for parsePrimaryExpr. The argument, used by the niche llvm-ml, should not require other targets to adapt.
While However, Restored the support in 3fd0d22. Sorry for the trouble! |
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Apr 10, 2025
llvm/llvm-project#134202 removed support for `sym@page-offset` in instruction operands. This change is generally reasonable since subtracting an offset from a symbol typically doesn’t make sense for Mach-O due to its .subsections_via_symbols mechanism, which treats them as separate atoms. However, BoringSSL relies on a temporary symbol with a negative offset, which can be meaningful when the symbol and the referenced location are within the same atom. ``` ../../third_party/boringssl/src/gen/bcm/p256-armv8-asm-apple.S:1160:25: error: unexpected token in argument list adrp x23,Lone_mont@PAGE-64 ``` It's worth noting that expressions involving @ can be complex and brittle in MCParser, and much of the Mach-O @ offsets remains under-tested. * Allow default argument for parsePrimaryExpr. The argument, used by the niche llvm-ml, should not require other targets to adapt.
|
Ah yeah, our asm files also don't yet set |
var-const
pushed a commit
to ldionne/llvm-project
that referenced
this pull request
Apr 17, 2025
llvm#134202 removed support for `sym@page-offset` in instruction operands. This change is generally reasonable since subtracting an offset from a symbol typically doesn’t make sense for Mach-O due to its .subsections_via_symbols mechanism, which treats them as separate atoms. However, BoringSSL relies on a temporary symbol with a negative offset, which can be meaningful when the symbol and the referenced location are within the same atom. ``` ../../third_party/boringssl/src/gen/bcm/p256-armv8-asm-apple.S:1160:25: error: unexpected token in argument list adrp x23,Lone_mont@PAGE-64 ``` It's worth noting that expressions involving @ can be complex and brittle in MCParser, and much of the Mach-O @ offsets remains under-tested. * Allow default argument for parsePrimaryExpr. The argument, used by the niche llvm-ml, should not require other targets to adapt.
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.
Following PR #132569 (RISC-V), which added
parseDataExprfor parsingexpressions in data directives (e.g.,
.word), this PR migrates AArch64@plt,@gotpcrel, and@AUTHfrom theparsePrimaryExprworkaroundto
parseDataExpr. The goal is to align with the GNU assembler model,where relocation specifiers apply to the entire operand rather than
individual terms, reducing complexity-especially evident in
@AUTHparsing.
Note: AArch64 ELF lacks an official syntax for data directives
(#132570). A prefix notation might be a preferable future direction.
I recommend
%specifier(expr).AsmParser's
@specifierparsing is suboptimal, necessitating lexerworkarounds.
@might appear multiple times in an operand.We should not use
@beyond the existing AArch64 Mach-O instructionoperands.
In the test elf-reloc-ptrauth.s, many errors are now reported at parse
time.