[llvm] [DWARFVerifier] Allow overlapping ranges for ICF-merged functions (PR #117952)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Nov 27 17:50:28 PST 2024
https://github.com/alx32 created https://github.com/llvm/llvm-project/pull/117952
This patch modifies the DWARF verifier to handle a valid case where two or more functions have identical address ranges due to being merged by ICF (Identical Code Folding). Previously, the verifier would incorrectly report these as errors, but functions merged via ICF (such as when using LLD's --keep-icf-stabs option) can legitimately share the same address range.
A new test case has been added to verify this behavior using YAML-based DWARF data that simulates two DW_TAG_subprogram entries with identical address ranges. The test ensures that the verifier correctly identifies this as a valid case and doesn't emit any errors, while still maintaining the existing verification for truly invalid overlapping ranges in other scenarios.
>From a5f73d2f55979a7bc612fa87e6aa5d3991a5f258 Mon Sep 17 00:00:00 2001
From: Alex B <alexborcan at meta.com>
Date: Wed, 27 Nov 2024 17:48:11 -0800
Subject: [PATCH] [DWARFVerifier] Allow overlapping ranges for ICF-merged
functions
This patch modifies the DWARF verifier to handle a valid case where two or more functions have identical address ranges due to being merged by ICF (Identical Code Folding). Previously, the verifier would incorrectly report these as errors, but functions merged via ICF (such as when using LLD's --keep-icf-stabs option) can legitimately share the same address range.
A new test case has been added to verify this behavior using YAML-based DWARF data that simulates two DW_TAG_subprogram entries with identical address ranges. The test ensures that the verifier correctly identifies this as a valid case and doesn't emit any errors, while still maintaining the existing verification for truly invalid overlapping ranges in other scenarios.
---
llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp | 21 +-
.../X86/verify_no_overlap_error_icf.yaml | 531 ++++++++++++++++++
2 files changed, 546 insertions(+), 6 deletions(-)
create mode 100644 llvm/test/tools/llvm-dwarfdump/X86/verify_no_overlap_error_icf.yaml
diff --git a/llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp b/llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp
index 1fe3eb1e90fe65..58b965892ecaf3 100644
--- a/llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp
+++ b/llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp
@@ -624,12 +624,21 @@ unsigned DWARFVerifier::verifyDieRanges(const DWARFDie &Die,
// Verify that children don't intersect.
const auto IntersectingChild = ParentRI.insert(RI);
if (IntersectingChild != ParentRI.Children.end()) {
- ++NumErrors;
- ErrorCategory.Report("DIEs have overlapping address ranges", [&]() {
- error() << "DIEs have overlapping address ranges:";
- dump(Die);
- dump(IntersectingChild->Die) << '\n';
- });
+ auto &IR = IntersectingChild->Ranges;
+ // Overlapping DW_TAG_subprogram can happen and are valid when multiple
+ // functions are merged via ICF. See --keep-icf-stabs in LLD.
+ bool isMergedFunc = (Die.getTag() == DW_TAG_subprogram) &&
+ (IR.size() == 1) && (IR == RI.Ranges);
+ if (!isMergedFunc) {
+ if (IntersectingChild->Ranges != ParentRI.Children.end()->Ranges) {
+ ++NumErrors;
+ ErrorCategory.Report("DIEs have overlapping address ranges", [&]() {
+ error() << "DIEs have overlapping address ranges:";
+ dump(Die);
+ dump(IntersectingChild->Die) << '\n';
+ });
+ }
+ }
}
// Verify that ranges are contained within their parent.
diff --git a/llvm/test/tools/llvm-dwarfdump/X86/verify_no_overlap_error_icf.yaml b/llvm/test/tools/llvm-dwarfdump/X86/verify_no_overlap_error_icf.yaml
new file mode 100644
index 00000000000000..6f3e6de28169dc
--- /dev/null
+++ b/llvm/test/tools/llvm-dwarfdump/X86/verify_no_overlap_error_icf.yaml
@@ -0,0 +1,531 @@
+# This test verifies that if two DW_TAG_subprogram's have a single identical
+# address range they are not flagged by dwarfdump as being invalid (having
+# overlapping address ranges).
+# This is a valid case that can happen when ICF merges multiple functions
+# together. See --keep-icf-stabs in LLD.
+#
+# The DWARF looks like:
+# 0x0000002e: DW_TAG_subprogram
+# DW_AT_low_pc (0x00000001000002f0)
+# DW_AT_high_pc (0x00000001000002fc)
+# DW_AT_name ("function1")
+# DW_AT_decl_file ("/tmp/out/my_code.cpp")
+# [...]
+#
+# 0x00000071: DW_TAG_subprogram
+# DW_AT_low_pc (0x00000001000002f0)
+# DW_AT_high_pc (0x00000001000002fc)
+# DW_AT_name ("function2")
+# DW_AT_decl_file ("/tmp/out/my_code.cpp")
+# [...]
+#
+
+# RUN: yaml2obj %s | llvm-dwarfdump --error-display=details --verify - | FileCheck %s
+# CHECK: No errors.
+
+
+## YAML below was generating using this shell script:
+# #!/bin/bash
+# set -ex
+#
+# TOOLCHAIN_DIR="Debug/bin"
+# SCRIPT_DIR="$(cd "$(dirname "$0")"; pwd)"
+# OUT_DIR="$SCRIPT_DIR/out"
+# mkdir -p $OUT_DIR
+#
+# cd $SCRIPT_DIR
+#
+# cat > "$OUT_DIR/my_code.cpp" << EOF
+# #define ATTRIB extern "C" __attribute__((noinline))
+#
+# ATTRIB int function1(int a) {
+# int b = a * 4;
+# int result = b + 4;
+# return result;
+# }
+#
+# ATTRIB int function2(int a) {
+# int b = a * 4;
+# int result = b + 4;
+# return result;
+# }
+#
+# int main() {
+# return function1(1) + function2(2);
+# }
+# EOF
+#
+# "$TOOLCHAIN_DIR/clang++" --target=arm64-apple-macos11 -c -g -O3 "$OUT_DIR/my_code.cpp" -o "$OUT_DIR/my_code.o" \
+# -gmlt -fno-unwind-tables -fno-exceptions
+# "$TOOLCHAIN_DIR/ld64.lld" -arch arm64 -platform_version macos 11.0.0 11.0.0 -o "$OUT_DIR/my_bin_icf.exe" \
+# "$OUT_DIR/my_code.o" -dead_strip --icf=all --keep-icf-stabs
+#
+# $TOOLCHAIN_DIR/dsymutil --flat "$OUT_DIR/my_bin_icf.exe" -o "$OUT_DIR/my_bin_icf.dSYM" --verify-dwarf=none
+# $TOOLCHAIN_DIR/obj2yaml "$OUT_DIR/my_bin_icf.dSYM" > $OUT_DIR/my_bin_icf.dSYM.yaml
+
+
+
+--- !mach-o
+FileHeader:
+ magic: 0xFEEDFACF
+ cputype: 0x100000C
+ cpusubtype: 0x0
+ filetype: 0xA
+ ncmds: 7
+ sizeofcmds: 1240
+ flags: 0x0
+ reserved: 0x0
+LoadCommands:
+ - cmd: LC_UUID
+ cmdsize: 24
+ uuid: 4C4C449C-5555-3144-A108-0F6123B3CB8A
+ - cmd: LC_BUILD_VERSION
+ cmdsize: 24
+ platform: 1
+ minos: 720896
+ sdk: 720896
+ ntools: 0
+ - cmd: LC_SYMTAB
+ cmdsize: 24
+ symoff: 4096
+ nsyms: 4
+ stroff: 4160
+ strsize: 50
+ - cmd: LC_SEGMENT_64
+ cmdsize: 72
+ segname: __PAGEZERO
+ vmaddr: 0
+ vmsize: 4294967296
+ fileoff: 0
+ filesize: 0
+ maxprot: 0
+ initprot: 0
+ nsects: 0
+ flags: 0
+ - cmd: LC_SEGMENT_64
+ cmdsize: 152
+ segname: __TEXT
+ vmaddr: 4294967296
+ vmsize: 16384
+ fileoff: 0
+ filesize: 0
+ maxprot: 5
+ initprot: 5
+ nsects: 1
+ flags: 0
+ Sections:
+ - sectname: __text
+ segname: __TEXT
+ addr: 0x1000002A0
+ size: 60
+ offset: 0x0
+ align: 2
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x80000400
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: CFFAEDFE0C000001000000000A00000007000000D804000000000000000000001B000000180000004C4C449C55553144A1080F6123B3CB8A32000000
+ - cmd: LC_SEGMENT_64
+ cmdsize: 72
+ segname: __LINKEDIT
+ vmaddr: 4294983680
+ vmsize: 4096
+ fileoff: 4096
+ filesize: 114
+ maxprot: 1
+ initprot: 1
+ nsects: 0
+ flags: 0
+ - cmd: LC_SEGMENT_64
+ cmdsize: 872
+ segname: __DWARF
+ vmaddr: 4294987776
+ vmsize: 4096
+ fileoff: 8192
+ filesize: 871
+ maxprot: 7
+ initprot: 3
+ nsects: 10
+ flags: 0
+ Sections:
+ - sectname: __debug_line
+ segname: __DWARF
+ addr: 0x100005000
+ size: 123
+ offset: 0x2000
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ - sectname: __debug_aranges
+ segname: __DWARF
+ addr: 0x10000507B
+ size: 48
+ offset: 0x207B
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ - sectname: __debug_info
+ segname: __DWARF
+ addr: 0x1000050AB
+ size: 125
+ offset: 0x20AB
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ - sectname: __debug_frame
+ segname: __DWARF
+ addr: 0x100005128
+ size: 112
+ offset: 0x2128
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: 14000000FFFFFFFF0400080001781E0C1F000000000000001400000000000000A0020000010000000C000000000000001400000000000000A0020000010000000C000000000000002400000000000000AC0200000100000030000000000000004C0C1D109E019D029303940400000000
+ - sectname: __debug_abbrev
+ segname: __DWARF
+ addr: 0x100005198
+ size: 64
+ offset: 0x2198
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ - sectname: __debug_str
+ segname: __DWARF
+ addr: 0x1000051D8
+ size: 163
+ offset: 0x21D8
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ - sectname: __apple_namespac
+ segname: __DWARF
+ addr: 0x10000527B
+ size: 36
+ offset: 0x227B
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: 485341480100000001000000000000000C000000000000000100000001000600FFFFFFFF
+ - sectname: __apple_names
+ segname: __DWARF
+ addr: 0x10000529F
+ size: 116
+ offset: 0x229F
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: 485341480100000003000000030000000C000000000000000100000001000600FFFFFFFF0000000002000000DC41AB586A7F9A7CDD41AB584400000054000000640000008A000000010000002E000000000000009E00000001000000500000000000000094000000010000003F00000000000000
+ - sectname: __apple_types
+ segname: __DWARF
+ addr: 0x100005313
+ size: 48
+ offset: 0x2313
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: 48534148010000000100000000000000180000000000000004000000010006000300050005000B0006000600FFFFFFFF
+ - sectname: __apple_objc
+ segname: __DWARF
+ addr: 0x100005343
+ size: 36
+ offset: 0x2343
+ align: 0
+ reloff: 0x0
+ nreloc: 0
+ flags: 0x0
+ reserved1: 0x0
+ reserved2: 0x0
+ reserved3: 0x0
+ content: 485341480100000001000000000000000C000000000000000100000001000600FFFFFFFF
+LinkEditData:
+ NameList:
+ - n_strx: 2
+ n_type: 0xF
+ n_sect: 1
+ n_desc: 0
+ n_value: 4294967980
+ - n_strx: 8
+ n_type: 0xF
+ n_sect: 1
+ n_desc: 0
+ n_value: 4294967968
+ - n_strx: 19
+ n_type: 0xF
+ n_sect: 1
+ n_desc: 0
+ n_value: 4294967968
+ - n_strx: 30
+ n_type: 0xF
+ n_sect: 1
+ n_desc: 16
+ n_value: 4294967296
+ StringTable:
+ - ''
+ - ''
+ - _main
+ - _function1
+ - _function2
+ - __mh_execute_header
+DWARF:
+ debug_str:
+ - ''
+ - 'clang version 20.0.0git (https://github.com/alx32/llvm-project.git 92a15dd7482ff4e1fae7a07f888564e5b1d53eee)'
+ - '/tmp/out/my_code.cpp'
+ - '/'
+ - '/tmp'
+ - function1
+ - function2
+ - main
+ debug_abbrev:
+ - ID: 0
+ Table:
+ - Code: 0x1
+ Tag: DW_TAG_compile_unit
+ Children: DW_CHILDREN_yes
+ Attributes:
+ - Attribute: DW_AT_producer
+ Form: DW_FORM_strp
+ - Attribute: DW_AT_language
+ Form: DW_FORM_data2
+ - Attribute: DW_AT_name
+ Form: DW_FORM_strp
+ - Attribute: DW_AT_LLVM_sysroot
+ Form: DW_FORM_strp
+ - Attribute: DW_AT_stmt_list
+ Form: DW_FORM_sec_offset
+ - Attribute: DW_AT_comp_dir
+ Form: DW_FORM_strp
+ - Attribute: DW_AT_APPLE_optimized
+ Form: DW_FORM_flag_present
+ - Attribute: DW_AT_low_pc
+ Form: DW_FORM_addr
+ - Attribute: DW_AT_high_pc
+ Form: DW_FORM_data4
+ - Code: 0x2
+ Tag: DW_TAG_subprogram
+ Children: DW_CHILDREN_no
+ Attributes:
+ - Attribute: DW_AT_low_pc
+ Form: DW_FORM_addr
+ - Attribute: DW_AT_high_pc
+ Form: DW_FORM_data4
+ - Attribute: DW_AT_APPLE_omit_frame_ptr
+ Form: DW_FORM_flag_present
+ - Attribute: DW_AT_call_all_calls
+ Form: DW_FORM_flag_present
+ - Attribute: DW_AT_name
+ Form: DW_FORM_strp
+ - Code: 0x3
+ Tag: DW_TAG_subprogram
+ Children: DW_CHILDREN_yes
+ Attributes:
+ - Attribute: DW_AT_low_pc
+ Form: DW_FORM_addr
+ - Attribute: DW_AT_high_pc
+ Form: DW_FORM_data4
+ - Attribute: DW_AT_call_all_calls
+ Form: DW_FORM_flag_present
+ - Attribute: DW_AT_name
+ Form: DW_FORM_strp
+ - Code: 0x4
+ Tag: DW_TAG_call_site
+ Children: DW_CHILDREN_no
+ Attributes:
+ - Attribute: DW_AT_call_origin
+ Form: DW_FORM_ref4
+ - Attribute: DW_AT_call_return_pc
+ Form: DW_FORM_addr
+ debug_aranges:
+ - Length: 0x2C
+ Version: 2
+ CuOffset: 0x0
+ AddressSize: 0x8
+ Descriptors:
+ - Address: 0x1000002A0
+ Length: 0x3C
+ debug_info:
+ - Length: 0x79
+ Version: 4
+ AbbrevTableID: 0
+ AbbrOffset: 0x0
+ AddrSize: 8
+ Entries:
+ - AbbrCode: 0x1
+ Values:
+ - Value: 0x1
+ - Value: 0x21
+ - Value: 0x6E
+ - Value: 0x83
+ - Value: 0x0
+ - Value: 0x85
+ - Value: 0x1
+ - Value: 0x1000002A0
+ - Value: 0x3C
+ - AbbrCode: 0x2
+ Values:
+ - Value: 0x1000002A0
+ - Value: 0xC
+ - Value: 0x1
+ - Value: 0x1
+ - Value: 0x8A
+ - AbbrCode: 0x2
+ Values:
+ - Value: 0x1000002A0
+ - Value: 0xC
+ - Value: 0x1
+ - Value: 0x1
+ - Value: 0x94
+ - AbbrCode: 0x3
+ Values:
+ - Value: 0x1000002AC
+ - Value: 0x30
+ - Value: 0x1
+ - Value: 0x9E
+ - AbbrCode: 0x4
+ Values:
+ - Value: 0x2E
+ - Value: 0x1000002C0
+ - AbbrCode: 0x4
+ Values:
+ - Value: 0x3F
+ - Value: 0x1000002CC
+ - AbbrCode: 0x0
+ - AbbrCode: 0x0
+ debug_line:
+ - Length: 119
+ Version: 4
+ PrologueLength: 39
+ MinInstLength: 1
+ MaxOpsPerInst: 1
+ DefaultIsStmt: 1
+ LineBase: 251
+ LineRange: 14
+ OpcodeBase: 13
+ StandardOpcodeLengths: [ 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 1 ]
+ IncludeDirs:
+ - out
+ Files:
+ - Name: my_code.cpp
+ DirIdx: 1
+ ModTime: 0
+ Length: 0
+ Opcodes:
+ - Opcode: DW_LNS_extended_op
+ ExtLen: 9
+ SubOpcode: DW_LNE_set_address
+ Data: 4294967968
+ - Opcode: DW_LNS_set_column
+ Data: 15
+ - Opcode: DW_LNS_set_prologue_end
+ Data: 0
+ - Opcode: DW_LNS_advance_line
+ SData: 9
+ Data: 0
+ - Opcode: DW_LNS_copy
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 20
+ - Opcode: 0x4B
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 5
+ - Opcode: 0x4B
+ Data: 0
+ - Opcode: DW_LNS_advance_pc
+ Data: 4
+ - Opcode: DW_LNS_extended_op
+ ExtLen: 1
+ SubOpcode: DW_LNE_end_sequence
+ Data: 0
+ - Opcode: DW_LNS_extended_op
+ ExtLen: 9
+ SubOpcode: DW_LNE_set_address
+ Data: 4294967968
+ - Opcode: DW_LNS_set_column
+ Data: 15
+ - Opcode: DW_LNS_set_prologue_end
+ Data: 0
+ - Opcode: 0x15
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 20
+ - Opcode: 0x4B
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 5
+ - Opcode: 0x4B
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 0
+ - Opcode: DW_LNS_advance_line
+ SData: 9
+ Data: 0
+ - Opcode: 0x4A
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 12
+ - Opcode: DW_LNS_set_prologue_end
+ Data: 0
+ - Opcode: 0xBB
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 27
+ - Opcode: DW_LNS_negate_stmt
+ Data: 0
+ - Opcode: 0xBA
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 25
+ - Opcode: 0x82
+ Data: 0
+ - Opcode: DW_LNS_set_column
+ Data: 5
+ - Opcode: DW_LNS_set_epilogue_begin
+ Data: 0
+ - Opcode: 0x4A
+ Data: 0
+ - Opcode: DW_LNS_advance_pc
+ Data: 12
+ - Opcode: DW_LNS_extended_op
+ ExtLen: 1
+ SubOpcode: DW_LNE_end_sequence
+ Data: 0
+...
More information about the llvm-commits
mailing list