[lld][AArch64][Build Attributes] Add support for converting AArch64 Build Attributes to GNU Properties #131990
There are no files selected for viewing
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -22,6 +22,7 @@ | |
| #include "llvm/ADT/STLExtras.h" | ||
| #include "llvm/LTO/LTO.h" | ||
| #include "llvm/Object/IRObjectFile.h" | ||
| #include "llvm/Support/AArch64AttributeParser.h" | ||
| #include "llvm/Support/ARMAttributeParser.h" | ||
| #include "llvm/Support/ARMBuildAttributes.h" | ||
| #include "llvm/Support/Endian.h" | ||
|
|
@@ -539,6 +540,65 @@ uint32_t ObjFile<ELFT>::getSectionIndex(const Elf_Sym &sym) const { | |
| this); | ||
| } | ||
|
|
||
| template <class ELFT> | ||
| static void | ||
| handleAArch64BAAndGnuProperties(const ELFT &tPointer, Ctx &ctx, bool isBE, | ||
| bool hasBA, bool hasGP, | ||
| const AArch64BuildAttrSubsections &baInfo, | ||
| const gnuPropertiesInfo &gpInfo) { | ||
|
|
||
| auto serializeUnsigned = [&](unsigned valueLow, unsigned valueHigh, | ||
| bool isBE) -> std::array<uint8_t, 16> { | ||
| std::array<uint8_t, 16> arr; | ||
| for (size_t i = 0; i < 8; ++i) { | ||
| arr[i] = static_cast<uint8_t>( | ||
| (static_cast<uint64_t>(valueLow) >> (8 * (isBE ? (7 - i) : i))) & | ||
| 0xFF); | ||
| arr[i + 8] = static_cast<uint8_t>( | ||
| (static_cast<uint64_t>(valueHigh) >> (8 * (isBE ? (7 - i) : i))) & | ||
| 0xFF); | ||
| }; | ||
| return arr; | ||
| }; | ||
|
|
||
| if (hasBA && hasGP) { | ||
| // Check for data mismatch | ||
| if (!gpInfo.aarch64PauthAbiCoreInfo.empty()) { | ||
| auto baPauth = serializeUnsigned(baInfo.Pauth.TagPlatform, | ||
| baInfo.Pauth.TagSchema, isBE); | ||
| if (gpInfo.aarch64PauthAbiCoreInfo != ArrayRef<uint8_t>(baPauth)) | ||
| Err(ctx) | ||
| << tPointer | ||
| << " Pauth Data mismatch: file contains both GNU properties and " | ||
| "AArch64 build attributes sections with different Pauth data"; | ||
| } | ||
| if (baInfo.AndFeatures != gpInfo.andFeatures) | ||
| Err(ctx) << tPointer | ||
| << " Features Data mismatch: file contains both GNU " | ||
| "properties and AArch64 build attributes sections with " | ||
| "different And Features data"; | ||
| } | ||
|
|
||
| if (hasBA && !hasGP) { | ||
| // Write missing data | ||
| // We can only know when Pauth is missing. | ||
| // Unlike AArch64 Build Attributes, GNU properties does not give a way to | ||
| // distinguish between no-value given to value of '0' given. | ||
| if (baInfo.Pauth.TagPlatform || baInfo.Pauth.TagSchema) { | ||
| tPointer->aarch64PauthAbiCoreInfoStorage = serializeUnsigned( | ||
| baInfo.Pauth.TagPlatform, baInfo.Pauth.TagSchema, isBE); | ||
| tPointer->aarch64PauthAbiCoreInfo = | ||
| tPointer->aarch64PauthAbiCoreInfoStorage; | ||
| } | ||
| tPointer->andFeatures = baInfo.AndFeatures; | ||
| } | ||
| } | ||
|
|
||
| // Forward declaration: | ||
| template <typename ELFT> | ||
| static gnuPropertiesInfo readGnuProperty(Ctx &, const InputSection &, | ||
| ObjFile<ELFT> &); | ||
|
|
||
| template <class ELFT> void ObjFile<ELFT>::parse(bool ignoreComdats) { | ||
| object::ELFFile<ELFT> obj = this->getObj(); | ||
| // Read a section table. justSymbols is usually false. | ||
|
|
@@ -554,8 +614,32 @@ template <class ELFT> void ObjFile<ELFT>::parse(bool ignoreComdats) { | |
| StringRef shstrtab = CHECK2(obj.getSectionStringTable(objSections), this); | ||
| uint64_t size = objSections.size(); | ||
| sections.resize(size); | ||
|
|
||
| // For handling AArch64 Build attributes and GNU properties | ||
| AArch64BuildAttrSubsections aarch64BAsubSections; | ||
| gnuPropertiesInfo gnuPropertiesInformation; | ||
| bool hasAArch64BuildAttributes = false; | ||
| bool hasGNUProperties = false; | ||
|
|
||
| for (size_t i = 0; i != size; ++i) { | ||
| const Elf_Shdr &sec = objSections[i]; | ||
|
|
||
|
There was a problem hiding this comment. It looks like BuildAttributes are read in parse() but gnu properties are read in as part of initializeSections. As neither GNU properties or BuildAttributes rely on symbol information then it should be possible to read them here, move the BuildAttributes reading to initializeSections or delay the merging of BuildAttributes and GNU properties to after initializeSections. If reading GNU properties is moved here, remember to remove them from createInputSection or they will get processed twice. |
||
| // Object files that use processor features such as Intel Control-Flow | ||
| // Enforcement (CET) or AArch64 Branch Target Identification BTI, use a | ||
| // .note.gnu.property section containing a bitfield of feature bits like the | ||
| // GNU_PROPERTY_X86_FEATURE_1_IBT flag. Read a bitmap containing the flag. | ||
| if (check(obj.getSectionName(sec, shstrtab)) == ".note.gnu.property") { | ||
| gnuPropertiesInformation = readGnuProperty( | ||
| ctx, | ||
| InputSection(*this, sec, check(obj.getSectionName(sec, shstrtab))), | ||
| *this); | ||
| hasGNUProperties = true; | ||
| // Since we merge bitmaps from multiple object files to create a new | ||
| // .note.gnu.property containing a single AND'ed bitmap, we discard an | ||
| // input file's .note.gnu.property section. | ||
| sections[i] = &InputSection::discarded; | ||
| } | ||
|
|
||
| if (LLVM_LIKELY(sec.sh_type == SHT_PROGBITS)) | ||
| continue; | ||
| if (LLVM_LIKELY(sec.sh_type == SHT_GROUP)) { | ||
|
|
@@ -639,16 +723,30 @@ template <class ELFT> void ObjFile<ELFT>::parse(bool ignoreComdats) { | |
| } | ||
| break; | ||
| case EM_AARCH64: | ||
| // At this stage AArch64 Build Attributes does not replace GNU Properties. | ||
| // When both exists, their values must match. | ||
| // When both exists and contain different attributes, they complement each | ||
| // other. Currently attributes are represented in the linked object file | ||
| // as GNU properties, which are already supported by the Linux kernel and | ||
| // the dynamic loader. In the future, when relocatable linking (`-r` flag) | ||
| // is performed, a single merged AArch64 Build Attributes section will be | ||
| // emitted. | ||
| if (sec.sh_type == SHT_AARCH64_ATTRIBUTES) { | ||
| ArrayRef<uint8_t> contents = check(obj.getSectionContents(sec)); | ||
| AArch64AttributeParser attributes; | ||
| StringRef name = check(obj.getSectionName(sec, shstrtab)); | ||
| InputSection isec(*this, sec, name); | ||
| if (Error e = attributes.parse(contents, ELFT::Endianness)) { | ||
| Warn(ctx) << &isec << ": " << std::move(e); | ||
| } else { | ||
| aarch64BAsubSections = extractBuildAttributesSubsections(attributes); | ||
| hasAArch64BuildAttributes = true; | ||
| } | ||
| sections[i] = &InputSection::discarded; | ||
| } | ||
sivan-shani marked this conversation as resolved.
Show resolved
Hide resolved
|
||
| // Producing a static binary with MTE globals is not currently supported, | ||
| // remove all SHT_AARCH64_MEMTAG_GLOBALS_STATIC sections as they're unused | ||
| // metadata, and we don't want them to end up in the output file for | ||
| // static executables. | ||
| if (sec.sh_type == SHT_AARCH64_MEMTAG_GLOBALS_STATIC && | ||
| !canHaveMemtagGlobals(ctx)) | ||
|
|
@@ -657,6 +755,16 @@ template <class ELFT> void ObjFile<ELFT>::parse(bool ignoreComdats) { | |
| } | ||
| } | ||
|
|
||
Stylie777 marked this conversation as resolved.
Show resolved
Hide resolved
|
||
| if (hasAArch64BuildAttributes) { | ||
| bool isBE = ELFT::Endianness == llvm::endianness::big; | ||
| // Handle AArch64 Build Attributes and GNU properties: | ||
| // - Err on mismatched values. | ||
| // - Store missing values as GNU properties. | ||
| handleAArch64BAAndGnuProperties(this, ctx, isBE, hasAArch64BuildAttributes, | ||
| hasGNUProperties, aarch64BAsubSections, | ||
| gnuPropertiesInformation); | ||
| } | ||
|
|
||
| // Read a symbol table. | ||
| initializeSymbols(obj); | ||
| } | ||
|
|
@@ -968,14 +1076,15 @@ static void parseGnuPropertyNote(Ctx &ctx, ELFFileBase &f, | |
| desc = desc.slice(alignTo<(ELFT::Is64Bits ? 8 : 4)>(size)); | ||
| } | ||
| } | ||
|
|
||
| // Read the following info from the .note.gnu.property section and write it to | ||
| // the corresponding fields in `ObjFile`: | ||
| // - Feature flags (32 bits) representing x86 or AArch64 features for | ||
| // hardware-assisted call flow control; | ||
| // - AArch64 PAuth ABI core info (16 bytes). | ||
| template <class ELFT> | ||
| static gnuPropertiesInfo readGnuProperty(Ctx &ctx, const InputSection &sec, | ||
| ObjFile<ELFT> &f) { | ||
| using Elf_Nhdr = typename ELFT::Nhdr; | ||
| using Elf_Note = typename ELFT::Note; | ||
|
|
||
|
|
@@ -991,7 +1100,7 @@ static void readGnuProperty(Ctx &ctx, const InputSection &sec, | |
| auto *nhdr = reinterpret_cast<const Elf_Nhdr *>(data.data()); | ||
| if (data.size() < sizeof(Elf_Nhdr) || | ||
| data.size() < nhdr->getSize(sec.addralign)) | ||
| return (err(data.data()) << "data is too short", gnuPropertiesInfo{}); | ||
|
|
||
| Elf_Note note(*nhdr); | ||
| if (nhdr->n_type != NT_GNU_PROPERTY_TYPE_0 || note.getName() != "GNU") { | ||
|
|
@@ -1011,6 +1120,7 @@ static void readGnuProperty(Ctx &ctx, const InputSection &sec, | |
| // Go to next NOTE record to look for more FEATURE_1_AND descriptions. | ||
| data = data.slice(nhdr->getSize(sec.addralign)); | ||
| } | ||
| return gnuPropertiesInfo{f.andFeatures, f.aarch64PauthAbiCoreInfo}; | ||
| } | ||
|
|
||
| template <class ELFT> | ||
|
|
@@ -1065,19 +1175,6 @@ InputSectionBase *ObjFile<ELFT>::createInputSection(uint32_t idx, | |
| return &InputSection::discarded; | ||
| } | ||
|
|
||
| // Split stacks is a feature to support a discontiguous stack, | ||
| // commonly used in the programming language Go. For the details, | ||
| // see https://gcc.gnu.org/wiki/SplitStacks. An object file compiled | ||
|
|
||
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,10 @@ | ||
| // Declare file properties exclusively with aarch64 build attributes. | ||
|
|
||
| .aeabi_subsection aeabi_feature_and_bits, optional, uleb128 | ||
| .aeabi_attribute Tag_Feature_PAC, 1 | ||
|
|
||
| .text | ||
| .globl func3 | ||
| .type func3,@function | ||
| func3: | ||
| ret |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,13 @@ | ||
| // This file replace gnu properties with aarch64 build attributes. | ||
|
|
||
| .aeabi_subsection aeabi_feature_and_bits, optional, uleb128 | ||
| .aeabi_attribute Tag_Feature_PAC, 1 | ||
|
|
||
| .text | ||
| .globl func2 | ||
| .type func2,@function | ||
| func2: | ||
| .globl func3 | ||
| .type func3, @function | ||
| bl func3 | ||
| ret |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,50 @@ | ||
| // REQUIRES: aarch64 | ||
| // RUN: llvm-mc -triple=aarch64_be %s -filetype=obj -o %t.o | ||
| // RUN: ld.lld %t.o --shared -o %t.so | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
|
|
||
| /// The Build attributes section appearing in the output of | ||
| /// llvm-mc should not appear in the output of lld, because | ||
| /// AArch64 build attributes are being transformed into .gnu.properties. | ||
|
|
||
| /// Test mc -> big endian, lld -> little endian | ||
| // RUN: llvm-mc -triple=aarch64_be %s -filetype=obj -o %t.o | ||
| // RUN: ld.lld %t.o --shared -o %t.so | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld %t.o -o %t | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld -r %t.o -o %t2.o | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
|
|
||
| /// Test mc -> little endian, lld -> big endian | ||
| // RUN: llvm-mc -triple=aarch64 %s -filetype=obj -o %t.o | ||
| // RUN: ld.lld --EB %t.o --shared -o %t.so | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld --EB %t.o -o %t | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld --EB -r %t.o -o %t2.o | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
|
|
||
| /// Test mc -> big endian, lld -> big endian | ||
| // RUN: llvm-mc -triple=aarch64_be %s -filetype=obj -o %t.o | ||
| // RUN: ld.lld --EB %t.o --shared -o %t.so | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld --EB %t.o -o %t | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
| // RUN: ld.lld --EB -r %t.o -o %t2.o | ||
| // RUN: llvm-readelf -n %t.so | FileCheck %s --check-prefix=NOTE | ||
|
|
||
| // NOTE: Displaying notes found in: .note.gnu.property | ||
| // NOTE-NEXT: Owner Data size Description | ||
| // NOTE-NEXT: GNU 0x00000028 NT_GNU_PROPERTY_TYPE_0 (property note) | ||
| // NOTE-NEXT: Properties: aarch64 feature: BTI, PAC, GCS | ||
| // NOTE-NEXT: AArch64 PAuth ABI core info: platform 0x89abcdef (unknown), version 0x89abcdef | ||
|
|
||
|
|
||
| .aeabi_subsection aeabi_pauthabi, required, uleb128 | ||
| .aeabi_attribute Tag_PAuth_Platform, 81985529216486895 | ||
| .aeabi_attribute Tag_PAuth_Schema, 81985529216486895 | ||
| .aeabi_subsection aeabi_feature_and_bits, optional, uleb128 | ||
| .aeabi_attribute Tag_Feature_BTI, 1 | ||
| .aeabi_attribute Tag_Feature_PAC, 1 | ||
| .aeabi_attribute Tag_Feature_GCS, 1 |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,35 @@ | ||
| // REQUIRES: aarch64 | ||
|
|
||
| // RUN: llvm-mc -triple=aarch64 %s -filetype=obj -o %t.o | ||
| // RUN: not ld.lld %t.o -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR | ||
|
|
||
| // ERR: Pauth Data mismatch: file contains both GNU properties and AArch64 build attributes sections with different Pauth data | ||
| // ERR-NEXT: Features Data mismatch: file contains both GNU properties and AArch64 build attributes sections with different And Features data | ||
|
|
||
| .aeabi_subsection aeabi_pauthabi, required, uleb128 | ||
| .aeabi_attribute Tag_PAuth_Platform, 5 | ||
| .aeabi_attribute Tag_PAuth_Schema, 5 | ||
| .aeabi_subsection aeabi_feature_and_bits, optional, uleb128 | ||
| .aeabi_attribute Tag_Feature_BTI, 1 | ||
| .aeabi_attribute Tag_Feature_PAC, 1 | ||
| .aeabi_attribute Tag_Feature_GCS, 1 | ||
|
|
||
| .section ".note.gnu.property", "a" | ||
| .long 4 | ||
| .long 0x10 | ||
| .long 0x5 | ||
| .asciz "GNU" | ||
| .long 0xc0000000 // GNU_PROPERTY_AARCH64_FEATURE_1_AND | ||
| .long 4 | ||
| .long 2 // GNU_PROPERTY_AARCH64_FEATURE_1_PAC | ||
| .long 0 | ||
|
|
||
| .section ".note.gnu.property", "a" | ||
| .long 4 | ||
| .long 24 | ||
| .long 5 | ||
| .asciz "GNU" | ||
| .long 0xc0000001 | ||
| .long 16 | ||
| .quad 305419896 // platform | ||
| .quad 2271560481 // version |
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,16 @@ | ||
| # RUN: llvm-mc -triple=aarch64 -filetype=obj %s -o %t.o | ||
| # RUN: ld.lld %t.o /dev/null 2>&1 | FileCheck %s | ||
|
|
||
| # CHECK: (.ARM.attributes): invalid Extended Build Attributes subsection size at offset: 39 | ||
|
|
||
| .section .ARM.attributes,"",%0x70000003 | ||
| .byte 0x41 // Tag 'A' (format version) | ||
| .long 0x00000019 // Subsection length | ||
| .asciz "aeabi_pauthabi" // Subsection name | ||
| .byte 0x00, 0x00 // Optionality and Type | ||
| .byte 0x01, 0x01, 0x02, 0x01 // PAuth_Platform and PAuth_Schema | ||
| .long 0x00000023 // Subsection length | ||
| .asciz "aeabi_feature_and_bits" // Subsection name | ||
| .byte 0x01, 0x00 // Optionality and Type | ||
| .byte 0x00, 0x01, 0x01, 0x01, 0x02, 0x01 // BTI, PAC, GCS | ||
| .byte 0x00, 0x00 // This is the malformation, data is too long. |
Oops, something went wrong.
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've simplified this in my pr131990 branch, but why do we need this serialization at all?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There are 2 reasons for the serialization of
baInfo.Pauth.TagPlatformandbaInfo.Pauth.TagSchemainto an array:gpInfo.aarch64PauthAbiCoreInfo)file->aarch64PauthAbiCoreInfoStorageWhile the second seems unavoidable, the first could be replaced with reading the values from the
gpInfo.aarch64PauthAbiCoreInfoarray.Reading the values from the array has the advantage of allowing more granular error messages.
On the other hand, it require some more operations.
Will commit a version that reads the values from the array separately so it can be decided which one is preferable.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In some conversations I've had with the PAuthABI team, they've wanted a property/build-attribute that would be the equivalent of the command-line option
-z pac-pltostensibly to prevent user-error if the linker is called directly (the clang driver could add this automatically).In that case it would be inconvenient to store the PAuthABICoreInfo as an array_ref only used for comparison.
Would definitely like to see an implementation where both BuildAttributes and GNU properties decode to the same individual properties. Then in https://github.com/llvm/llvm-project/blob/main/lld/ELF/SyntheticSections.cpp#L347 was can use a
write32()call for each property rather than memcpy.