SlideShare a Scribd company logo
LCU14-209: LLVMLinux 
Behan Webster, LCU14 
LCU14 BURLINGAME
Linaro Overview 
● About the LLVMLinux project 
● State of Linux kernel being compiled with clang 
● Details of specific known problems/patches still being upstreamed 
● How can you help? 
● Contact info
The LLVMLinux Project Goals 
● Fully build the Linux kernel for multiple 
architectures, using the Clang/LLVM toolchain 
● Discover LLVM/Kernel issues early and find fixes 
quickly across both communities 
● Upstream patches to the Linux Kernel and LLVM 
projects 
● Bring together like-minded developers 
● Enable the kernel community to do more in depth 
analysis of the kernel code
LLVMLinux Build/Test System 
● Fetches, patches, builds, tests: clang, kernel, qemu, etc 
○ git clone http://git.linuxfoundation.org/llvmlinux.git 
○ cd llvmlinux/target/vexpress (or vexpress64) 
○ make
Patched Mainline Kernel Tree 
● A mainline kernel tree with all LLVMLinux patches applied on top is 
now available: 
○ git://git.linuxfoundation.org/llvmlinux/kernel.git 
● Dated llvmlinux branches 
○ remotes/origin/llvmlinux-2014.09.16 
● The master branch is rebased regularly
LLVMLinux Project Status 
● LLVM/clang: 
○ All LLVMLinux patches for LLVM are Upstream 
○ Newer LLVM patches to support the Linux kernel are mostly 
being added by upstream maintainers 
● Linux Kernel: 
○ Roughly 49 kernel patches for various arches 
○ LLVMLinux branch in linux-next
Remaining LLVMLinux Kernel Patches 
● Patches still working their way upstream 
Architecture Number of 
Patches 
Patches 
Submitted 
Patches 
Accepted 
all 23 17 2 
arm 12 11 1 (+7) 
aarch64 11 8 7 
x86_64 3 1 1 
TOTAL 49 37 11
LLVMLinux Kernel Patches 
● Total patches currently required for a single architecture 
Architecture Number of 
Patches 
Patches 
Submitted 
Patches 
Accepted 
arm (+all) 35 (12+23) 28 (11+17) 3 (1+2) 
aarch64 (+all) 34 (11+23) 25 (8+17) 9 (7+2) 
x86_64 (+all) 26 (3+23) 18 (1+17) 3 (1+2)
Kbuild 
● Basic Kbuild support for clang and x86 has been upstreamed 
● Specific support for ARM and aarch64 ready to be upstreamed
Integrated Assembly Status 
● Renato Golin, Vicinius Tinti, Saleem Abdulrasool and Stepan 
Dyatkovskiy are working on fixing IA issues in clang to support the 
Linux ARM kernel code (and ultimately AARCH64) 
● David Woodhouse has added .code16 support for X86 ASM 
● For now we disable the IA and use GNU as instead
Different option passing 
● gcc passes -march to GNU as 
● clang doesn't... (Bug submitted PR) 
● Probably should be fixed in clang 
● Work around patch for now 
-CFLAGS_aes-ce-cipher.o += -march=armv8-a+crypto 
+CFLAGS_aes-ce-cipher.o += -march=armv8-a+crypto -Wa,-march=armv8-a+crypto
extern inline: Different for gnu89 and gnu99 
● GNU89/GNU90 (used by gcc) 
○ Function will be inlined where it is used 
○ No function definition is emitted 
○ A non-inlined function may also be provided 
● GNU99/C99 (used by clang) 
○ Function will be inlined where it is used 
○ An external function is emitted 
○ No other function of the same name may be provided. 
● Solution? Use “static inline” instead. 
● Only still an issue for ARM support for ftrace (submitted)
Attribute Order 
● gcc is less picky about placement of __attribute__(()) 
● clang requires it at the end of the type or variable 
-struct __read_mostly va_alignment va_align = { 
+struct va_alignment __read_mostly va_align = { 
● (This particular patch was just accepted)
Named Registers 
● (Named registers for X86 kernel have been removed from the 
mainline kernel by Andi Kleen) 
● ARM and AARCH64 still like using named registers 
● Clang now supports using a globally named register for the stack 
pointer (Thanks Renato!) 
● For ARM/AARCH64 move to using a global in asm/thread_info.h 
register unsigned long current_stack_pointer asm ("sp"); 
● Patches for AARCH64 now accepted, acked for ARM
ARM percpu patch 
● One of the uses of Named Registers in the ARM code is due to a 
deficiency in gcc 
● The new code which works with gcc fails in clang 
● Solution, provide routines for both, and choose at compile time 
● Gcc: 
asm("mrc p15, 0, %0, c13, c0, 4" : "=r" (off) : "Q" (*sp)); 
● Clang: 
asm("mrc p15, 0, %0, c13, c0, 4" : "=r" (off) : : "memory");
Missing “%a” for inline ASM 
● The following error is generated with clang for AArch64: 
error: invalid operand in inline asm: 'prfm pldl1keep, ${0:a}' 
● Per comments by Tim Northover on the LLVM Bug database: 
It's rather unclear how it's better than "prfm pstl1keep, [%0]" though. Not 
all instructions can make use of any offset, so wouldn't we have to be 
conservative and always map it to "[xN]"? 
● When %a0 is changed to [%x0] it uncovered a GCC bug: https://bugs. 
linaro.org/show_bug.cgi?id=635 
● Changing the "p" to "r" resolves the issue for both clang and GCC. 
- asm volatile("prfm pldl1keep, %a0n" : : "p" (ptr)); 
+ asm volatile("prfm pldl1keep, [%x0]n" : : "r" (ptr));
Section Mismatch Issues (MergedGlobals) 
● By default clang merges globals with internal linkage into one: 
MergedGlobals 
● Allows globals to be addressed using offsets from a base pointer 
● Can reduce the number of registers used 
● Modpost script in the Linux kernel uses symbol names to look for 
section mismatches (e.g. regular code calling init code) 
● MergedGlobals breaks modpost (false positive section 
mismatches) 
● Current solution: use -mno-global-merge to stop global merging 
● Updates to modpost may allow this optimization to be enabled 
again in the future
ARM eabi support 
● Clang emits code which uses the “aeabi” ARM calls which are 
implemented in compiler-rt (equivalient to libgcc) 
● Compiler-rt doesn't easily cross compile yet... 
void __aeabi_memcpy(void *dest, const void *src, size_t n) 
void __aeabi_memmove(void *dest, const void *src, size_t n) 
void __aeabi_memset(void *s, size_t n, int c) 
● Still needed for ARM 
● No longer required for AARCH64
Variable Length Arrays In Structs 
● VLAIS isn't supported by Clang (undocumented gcc extension) 
char vla[n]; /* Supported, C99/C11 */ 
struct { 
char flexible_member[]; /* Supported, C99/C11 */ 
} struct_with_flexible_member; 
struct { 
char vlais[n]; /* Explicitly not allowed by C99/C11 */ 
} variable_length_array_in_struct; 
● VLAIS is used in the Linux kernel in a number of places, spreading 
mostly through reusing patterns from data structures found in 
crypto
VLAIS Removal Example (from crypto/hmac.c) 
- struct { 
- struct shash_desc shash; 
- char ctx[crypto_shash_descsize(hash)]; 
- } desc; 
+ char desc[sizeof(struct shash_desc) 
+ 
+ crypto_shash_descsize(hash)] CRYPTO_MINALIGN_ATTR; 
+ struct shash_desc *shash = (struct shash_desc *)desc; 
unsigned int i; 
- desc.shash.tfm = hash; 
+ shash->tfm = hash;
VLAIS Removal Example (the missing pieces) 
#define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) 
#define CRYPTO_MINALIGN ARCH_KMALLOC_MINALIGN 
#define CRYPTO_MINALIGN_ATTR __attribute__ ((__aligned__(CRYPTO_MINALIGN))) 
struct shash_desc { 
struct crypto_shash *tfm; 
u32 flags; 
void *__ctx[] CRYPTO_MINALIGN_ATTR; 
};
Status of VLAIS in the Linux Kernel 
● USB Gadget patch is in mainline 
● Mac80211 patch is in mainline 
● Netfilter patch is in mainline 
● apparmor patch accepted 
● Bluetooth patch accepted 
● Only the VLAIS patches for crypto are left: 
(btrfs, dm-crypt, hmac, libcrc32c, testmgr, etc) 
● However, we recently found a few previously unknown instances of 
VLAIS in raid10 and exofs...
Huge use of VLAIS in fs/exofs/ore_raid.c 
static int _sp2d_alloc(unsigned pages_in_unit, unsigned group_width, 
unsigned parity, struct __stripe_pages_2d **psp2d) 
[...] 
unsigned data_devs = group_width - parity; 
struct _alloc_all_bytes { 
struct __alloc_stripe_pages_2d { 
struct __stripe_pages_2d sp2d; 
struct __1_page_stripe _1p_stripes[pages_in_unit]; 
} __asp2d; 
struct __alloc_1p_arrays { 
struct page *pages[group_width]; 
struct page *scribble[group_width]; 
char page_is_read[data_devs]; 
} __a1pa[pages_in_unit]; 
} *_aab;
How Can You Help? 
● Don’t use the non-C99 practices we showed in previous slides 
● Make it known you want to be able to use Clang to compile the 
kernel (tell your Linaro representative!) 
● Test LLVMLinux patches 
● Report bugs to the LLVMLinux mailing list 
● Help get LLVMLinux patches upstream 
● Work on unsupported features and Bugs 
○ http://llvm.linuxfoundation.org/index.php/Broken_kernel_options 
● Submit new targets and arch support 
● Patches welcome
Embrace the 
Dragon. 
He's cuddly. 
Thank you 
http://llvm.linuxfoundation.org
Contribute to the LLVMLinux Project 
● Project wiki page 
○ http://llvm.linuxfoundation.org 
● Project Mailing List 
○ http://lists.linuxfoundation.org/mailman/listinfo/llvmlinux 
○ http://lists.linuxfoundation.org/pipermail/llvmlinux/ 
● IRC Channel 
○ #llvmlinux on OFTC 
○ http://buildbot.llvm.linuxfoundation.org/irclogs/OFTC/%23llvmlinux/ 
● LLVMLinux Community on Google Plus
More about Linaro Connect: connect.linaro.org 
Linaro members: www.linaro.org/members 
More about Linaro: www.linaro.org/about/

More Related Content

PDF
LCA14: LCA14-412: GPGPU on ARM SoC session
PPTX
Onnc intro
PDF
[COSCUP 2021] A trip about how I contribute to LLVM
PDF
netfilter and iptables
PDF
LAS16-501: Introduction to LLVM - Projects, Components, Integration, Internals
PDF
Linux Kernel Cryptographic API and Use Cases
PDF
NIR on the Mesa i965 backend (FOSDEM 2016)
PDF
Kernel Recipes 2016 - Landlock LSM: Unprivileged sandboxing
LCA14: LCA14-412: GPGPU on ARM SoC session
Onnc intro
[COSCUP 2021] A trip about how I contribute to LLVM
netfilter and iptables
LAS16-501: Introduction to LLVM - Projects, Components, Integration, Internals
Linux Kernel Cryptographic API and Use Cases
NIR on the Mesa i965 backend (FOSDEM 2016)
Kernel Recipes 2016 - Landlock LSM: Unprivileged sandboxing

What's hot (20)

PDF
Cilium - Fast IPv6 Container Networking with BPF and XDP
PDF
Q4.11: NEON Intrinsics
PDF
Multithreading done right
PPTX
The Silence of the Canaries
PPTX
Kernel Proc Connector and Containers
PDF
CSW2017 Amanda rousseau cansecwest2017_net_hijacking_powershell
PDF
LAS16-504: Secure Storage updates in OP-TEE
PDF
Kernel Recipes 2017 - EBPF and XDP - Eric Leblond
PDF
Kernel Recipes 2018 - A year of fixing Coverity issues all over the Linux ker...
PPTX
Low-level Shader Optimization for Next-Gen and DX11 by Emil Persson
PDF
Specializing the Data Path - Hooking into the Linux Network Stack
PDF
Make Your Own Developement Board @ 2014.4.21 JuluOSDev
PDF
Kernel Recipes 2016 - Upstream Kernel Graphics is (Finally) Winning
PPTX
An Open Discussion of RISC-V BitManip, trends, and comparisons _ Claire
PDF
BPF - All your packets belong to me
PDF
#Include os - From bootloader to REST API with the new C++
PDF
BKK16-103 OpenCSD - Open for Business!
PDF
BKK16-409 VOSY Switch Port to ARMv8 Platforms and ODP Integration
PPTX
[Sitcon2018] Analysis and Improvement of IOTA PoW Implementation
PDF
Comprehensive XDP Off‌load-handling the Edge Cases
Cilium - Fast IPv6 Container Networking with BPF and XDP
Q4.11: NEON Intrinsics
Multithreading done right
The Silence of the Canaries
Kernel Proc Connector and Containers
CSW2017 Amanda rousseau cansecwest2017_net_hijacking_powershell
LAS16-504: Secure Storage updates in OP-TEE
Kernel Recipes 2017 - EBPF and XDP - Eric Leblond
Kernel Recipes 2018 - A year of fixing Coverity issues all over the Linux ker...
Low-level Shader Optimization for Next-Gen and DX11 by Emil Persson
Specializing the Data Path - Hooking into the Linux Network Stack
Make Your Own Developement Board @ 2014.4.21 JuluOSDev
Kernel Recipes 2016 - Upstream Kernel Graphics is (Finally) Winning
An Open Discussion of RISC-V BitManip, trends, and comparisons _ Claire
BPF - All your packets belong to me
#Include os - From bootloader to REST API with the new C++
BKK16-103 OpenCSD - Open for Business!
BKK16-409 VOSY Switch Port to ARMv8 Platforms and ODP Integration
[Sitcon2018] Analysis and Improvement of IOTA PoW Implementation
Comprehensive XDP Off‌load-handling the Edge Cases
Ad

Viewers also liked (18)

PDF
LCU14 201- Binary Analysis Tools
PDF
Lca14 14-501- glibc-eglibc
PDF
LCU14-410: How to build an Energy Model for your SoC
PDF
LCA14: LCA14-415: ACPI Power Management
PDF
LCU14 114- Upstreaming 201
PDF
Lcu14 306 - OP-TEE Future Enhancements
PDF
Lcu14 101- coresight overview
PDF
Lcu14 107- op-tee on ar mv8
PDF
HKG15-104: What is Linaro working on - core development lightning talks
PDF
LCU14 500 ARM Trusted Firmware
PDF
LCU14-103: How to create and run Trusted Applications on OP-TEE
PDF
LCU14 302- How to port OP-TEE to another platform
PDF
LCU14 208- Chromium-Blink Migration for RDK
PDF
LCU14 206- Tools to Analyse Scheduling Behaviour and Its Impact on Power Mana...
PDF
LCA14: LCA14-506: Comparative analysis of preemption vs preempt-rt
PDF
LCA14: LCA14-502: The way to a generic TrustZone® solution
PDF
LCU14 303- Toolchain Collaboration
PDF
LCA14: LCA14-418: Testing a secure framework
LCU14 201- Binary Analysis Tools
Lca14 14-501- glibc-eglibc
LCU14-410: How to build an Energy Model for your SoC
LCA14: LCA14-415: ACPI Power Management
LCU14 114- Upstreaming 201
Lcu14 306 - OP-TEE Future Enhancements
Lcu14 101- coresight overview
Lcu14 107- op-tee on ar mv8
HKG15-104: What is Linaro working on - core development lightning talks
LCU14 500 ARM Trusted Firmware
LCU14-103: How to create and run Trusted Applications on OP-TEE
LCU14 302- How to port OP-TEE to another platform
LCU14 208- Chromium-Blink Migration for RDK
LCU14 206- Tools to Analyse Scheduling Behaviour and Its Impact on Power Mana...
LCA14: LCA14-506: Comparative analysis of preemption vs preempt-rt
LCA14: LCA14-502: The way to a generic TrustZone® solution
LCU14 303- Toolchain Collaboration
LCA14: LCA14-418: Testing a secure framework
Ad

Similar to LCU14 209- LLVM Linux (20)

PDF
Clang: More than just a C/C++ Compiler
PPTX
Linux Initialization Process (1)
PPTX
Qt5 on ti processors
PDF
LMG Lightning Talks - SFO17-205
PDF
Scylla Summit 2022: ScyllaDB Rust Driver: One Driver to Rule Them All
PDF
Valgrind
PDF
Haskell Symposium 2010: An LLVM backend for GHC
PDF
Systemd: the modern Linux init system you will learn to love
PPTX
Adventures in Thread-per-Core Async with Redpanda and Seastar
PPTX
Exploring Compiler Optimization Opportunities for the OpenMP 4.x Accelerator...
PPT
Linux Device Driver for Writing a real world driver for embedded Linux
PDF
Share the Experience of Using Embedded Development Board
PDF
CONFidence 2017: Escaping the (sand)box: The promises and pitfalls of modern ...
PDF
Lecture 6 Kernel Debugging + Ports Development
PDF
GDG Cloud Iasi - Docker For The Busy Developer.pdf
PDF
不深不淺,帶你認識 LLVM (Found LLVM in your life)
PDF
LXC on Ganeti
PDF
Lightweight Virtualization with Linux Containers and Docker I YaC 2013
PDF
Lightweight Virtualization with Linux Containers and Docker | YaC 2013
PPT
Linux Kernel Debugging
Clang: More than just a C/C++ Compiler
Linux Initialization Process (1)
Qt5 on ti processors
LMG Lightning Talks - SFO17-205
Scylla Summit 2022: ScyllaDB Rust Driver: One Driver to Rule Them All
Valgrind
Haskell Symposium 2010: An LLVM backend for GHC
Systemd: the modern Linux init system you will learn to love
Adventures in Thread-per-Core Async with Redpanda and Seastar
Exploring Compiler Optimization Opportunities for the OpenMP 4.x Accelerator...
Linux Device Driver for Writing a real world driver for embedded Linux
Share the Experience of Using Embedded Development Board
CONFidence 2017: Escaping the (sand)box: The promises and pitfalls of modern ...
Lecture 6 Kernel Debugging + Ports Development
GDG Cloud Iasi - Docker For The Busy Developer.pdf
不深不淺,帶你認識 LLVM (Found LLVM in your life)
LXC on Ganeti
Lightweight Virtualization with Linux Containers and Docker I YaC 2013
Lightweight Virtualization with Linux Containers and Docker | YaC 2013
Linux Kernel Debugging

More from Linaro (20)

PDF
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
PDF
Arm Architecture HPC Workshop Santa Clara 2018 - Kanta Vekaria
PDF
Huawei’s requirements for the ARM based HPC solution readiness - Joshua Mora
PDF
Bud17 113: distribution ci using qemu and open qa
PDF
OpenHPC Automation with Ansible - Renato Golin - Linaro Arm HPC Workshop 2018
PDF
HPC network stack on ARM - Linaro HPC Workshop 2018
PDF
It just keeps getting better - SUSE enablement for Arm - Linaro HPC Workshop ...
PDF
Intelligent Interconnect Architecture to Enable Next Generation HPC - Linaro ...
PDF
Yutaka Ishikawa - Post-K and Arm HPC Ecosystem - Linaro Arm HPC Workshop Sant...
PDF
Andrew J Younge - Vanguard Astra - Petascale Arm Platform for U.S. DOE/ASC Su...
PDF
HKG18-501 - EAS on Common Kernel 4.14 and getting (much) closer to mainline
PDF
HKG18-100K1 - George Grey: Opening Keynote
PDF
HKG18-318 - OpenAMP Workshop
PDF
HKG18-501 - EAS on Common Kernel 4.14 and getting (much) closer to mainline
PDF
HKG18-315 - Why the ecosystem is a wonderful thing, warts and all
PDF
HKG18- 115 - Partitioning ARM Systems with the Jailhouse Hypervisor
PDF
HKG18-TR08 - Upstreaming SVE in QEMU
PDF
HKG18-113- Secure Data Path work with i.MX8M
PPTX
HKG18-120 - Devicetree Schema Documentation and Validation
PPTX
HKG18-223 - Trusted FirmwareM: Trusted boot
Deep Learning Neural Network Acceleration at the Edge - Andrea Gallo
Arm Architecture HPC Workshop Santa Clara 2018 - Kanta Vekaria
Huawei’s requirements for the ARM based HPC solution readiness - Joshua Mora
Bud17 113: distribution ci using qemu and open qa
OpenHPC Automation with Ansible - Renato Golin - Linaro Arm HPC Workshop 2018
HPC network stack on ARM - Linaro HPC Workshop 2018
It just keeps getting better - SUSE enablement for Arm - Linaro HPC Workshop ...
Intelligent Interconnect Architecture to Enable Next Generation HPC - Linaro ...
Yutaka Ishikawa - Post-K and Arm HPC Ecosystem - Linaro Arm HPC Workshop Sant...
Andrew J Younge - Vanguard Astra - Petascale Arm Platform for U.S. DOE/ASC Su...
HKG18-501 - EAS on Common Kernel 4.14 and getting (much) closer to mainline
HKG18-100K1 - George Grey: Opening Keynote
HKG18-318 - OpenAMP Workshop
HKG18-501 - EAS on Common Kernel 4.14 and getting (much) closer to mainline
HKG18-315 - Why the ecosystem is a wonderful thing, warts and all
HKG18- 115 - Partitioning ARM Systems with the Jailhouse Hypervisor
HKG18-TR08 - Upstreaming SVE in QEMU
HKG18-113- Secure Data Path work with i.MX8M
HKG18-120 - Devicetree Schema Documentation and Validation
HKG18-223 - Trusted FirmwareM: Trusted boot

Recently uploaded (20)

PDF
Softaken Excel to vCard Converter Software.pdf
PDF
How Creative Agencies Leverage Project Management Software.pdf
PPTX
Agentic AI : A Practical Guide. Undersating, Implementing and Scaling Autono...
PDF
2025 Textile ERP Trends: SAP, Odoo & Oracle
PDF
T3DD25 TYPO3 Content Blocks - Deep Dive by André Kraus
PPTX
Transform Your Business with a Software ERP System
PPTX
VVF-Customer-Presentation2025-Ver1.9.pptx
PDF
System and Network Administration Chapter 2
PDF
Adobe Illustrator 28.6 Crack My Vision of Vector Design
PDF
How to Choose the Right IT Partner for Your Business in Malaysia
PDF
Audit Checklist Design Aligning with ISO, IATF, and Industry Standards — Omne...
PDF
Wondershare Filmora 15 Crack With Activation Key [2025
PPTX
CHAPTER 2 - PM Management and IT Context
PDF
Why TechBuilder is the Future of Pickup and Delivery App Development (1).pdf
PDF
Digital Strategies for Manufacturing Companies
PPTX
L1 - Introduction to python Backend.pptx
PDF
Design an Analysis of Algorithms I-SECS-1021-03
PDF
top salesforce developer skills in 2025.pdf
PDF
AI in Product Development-omnex systems
PDF
Addressing The Cult of Project Management Tools-Why Disconnected Work is Hold...
Softaken Excel to vCard Converter Software.pdf
How Creative Agencies Leverage Project Management Software.pdf
Agentic AI : A Practical Guide. Undersating, Implementing and Scaling Autono...
2025 Textile ERP Trends: SAP, Odoo & Oracle
T3DD25 TYPO3 Content Blocks - Deep Dive by André Kraus
Transform Your Business with a Software ERP System
VVF-Customer-Presentation2025-Ver1.9.pptx
System and Network Administration Chapter 2
Adobe Illustrator 28.6 Crack My Vision of Vector Design
How to Choose the Right IT Partner for Your Business in Malaysia
Audit Checklist Design Aligning with ISO, IATF, and Industry Standards — Omne...
Wondershare Filmora 15 Crack With Activation Key [2025
CHAPTER 2 - PM Management and IT Context
Why TechBuilder is the Future of Pickup and Delivery App Development (1).pdf
Digital Strategies for Manufacturing Companies
L1 - Introduction to python Backend.pptx
Design an Analysis of Algorithms I-SECS-1021-03
top salesforce developer skills in 2025.pdf
AI in Product Development-omnex systems
Addressing The Cult of Project Management Tools-Why Disconnected Work is Hold...

LCU14 209- LLVM Linux

  • 1. LCU14-209: LLVMLinux Behan Webster, LCU14 LCU14 BURLINGAME
  • 2. Linaro Overview ● About the LLVMLinux project ● State of Linux kernel being compiled with clang ● Details of specific known problems/patches still being upstreamed ● How can you help? ● Contact info
  • 3. The LLVMLinux Project Goals ● Fully build the Linux kernel for multiple architectures, using the Clang/LLVM toolchain ● Discover LLVM/Kernel issues early and find fixes quickly across both communities ● Upstream patches to the Linux Kernel and LLVM projects ● Bring together like-minded developers ● Enable the kernel community to do more in depth analysis of the kernel code
  • 4. LLVMLinux Build/Test System ● Fetches, patches, builds, tests: clang, kernel, qemu, etc ○ git clone http://git.linuxfoundation.org/llvmlinux.git ○ cd llvmlinux/target/vexpress (or vexpress64) ○ make
  • 5. Patched Mainline Kernel Tree ● A mainline kernel tree with all LLVMLinux patches applied on top is now available: ○ git://git.linuxfoundation.org/llvmlinux/kernel.git ● Dated llvmlinux branches ○ remotes/origin/llvmlinux-2014.09.16 ● The master branch is rebased regularly
  • 6. LLVMLinux Project Status ● LLVM/clang: ○ All LLVMLinux patches for LLVM are Upstream ○ Newer LLVM patches to support the Linux kernel are mostly being added by upstream maintainers ● Linux Kernel: ○ Roughly 49 kernel patches for various arches ○ LLVMLinux branch in linux-next
  • 7. Remaining LLVMLinux Kernel Patches ● Patches still working their way upstream Architecture Number of Patches Patches Submitted Patches Accepted all 23 17 2 arm 12 11 1 (+7) aarch64 11 8 7 x86_64 3 1 1 TOTAL 49 37 11
  • 8. LLVMLinux Kernel Patches ● Total patches currently required for a single architecture Architecture Number of Patches Patches Submitted Patches Accepted arm (+all) 35 (12+23) 28 (11+17) 3 (1+2) aarch64 (+all) 34 (11+23) 25 (8+17) 9 (7+2) x86_64 (+all) 26 (3+23) 18 (1+17) 3 (1+2)
  • 9. Kbuild ● Basic Kbuild support for clang and x86 has been upstreamed ● Specific support for ARM and aarch64 ready to be upstreamed
  • 10. Integrated Assembly Status ● Renato Golin, Vicinius Tinti, Saleem Abdulrasool and Stepan Dyatkovskiy are working on fixing IA issues in clang to support the Linux ARM kernel code (and ultimately AARCH64) ● David Woodhouse has added .code16 support for X86 ASM ● For now we disable the IA and use GNU as instead
  • 11. Different option passing ● gcc passes -march to GNU as ● clang doesn't... (Bug submitted PR) ● Probably should be fixed in clang ● Work around patch for now -CFLAGS_aes-ce-cipher.o += -march=armv8-a+crypto +CFLAGS_aes-ce-cipher.o += -march=armv8-a+crypto -Wa,-march=armv8-a+crypto
  • 12. extern inline: Different for gnu89 and gnu99 ● GNU89/GNU90 (used by gcc) ○ Function will be inlined where it is used ○ No function definition is emitted ○ A non-inlined function may also be provided ● GNU99/C99 (used by clang) ○ Function will be inlined where it is used ○ An external function is emitted ○ No other function of the same name may be provided. ● Solution? Use “static inline” instead. ● Only still an issue for ARM support for ftrace (submitted)
  • 13. Attribute Order ● gcc is less picky about placement of __attribute__(()) ● clang requires it at the end of the type or variable -struct __read_mostly va_alignment va_align = { +struct va_alignment __read_mostly va_align = { ● (This particular patch was just accepted)
  • 14. Named Registers ● (Named registers for X86 kernel have been removed from the mainline kernel by Andi Kleen) ● ARM and AARCH64 still like using named registers ● Clang now supports using a globally named register for the stack pointer (Thanks Renato!) ● For ARM/AARCH64 move to using a global in asm/thread_info.h register unsigned long current_stack_pointer asm ("sp"); ● Patches for AARCH64 now accepted, acked for ARM
  • 15. ARM percpu patch ● One of the uses of Named Registers in the ARM code is due to a deficiency in gcc ● The new code which works with gcc fails in clang ● Solution, provide routines for both, and choose at compile time ● Gcc: asm("mrc p15, 0, %0, c13, c0, 4" : "=r" (off) : "Q" (*sp)); ● Clang: asm("mrc p15, 0, %0, c13, c0, 4" : "=r" (off) : : "memory");
  • 16. Missing “%a” for inline ASM ● The following error is generated with clang for AArch64: error: invalid operand in inline asm: 'prfm pldl1keep, ${0:a}' ● Per comments by Tim Northover on the LLVM Bug database: It's rather unclear how it's better than "prfm pstl1keep, [%0]" though. Not all instructions can make use of any offset, so wouldn't we have to be conservative and always map it to "[xN]"? ● When %a0 is changed to [%x0] it uncovered a GCC bug: https://bugs. linaro.org/show_bug.cgi?id=635 ● Changing the "p" to "r" resolves the issue for both clang and GCC. - asm volatile("prfm pldl1keep, %a0n" : : "p" (ptr)); + asm volatile("prfm pldl1keep, [%x0]n" : : "r" (ptr));
  • 17. Section Mismatch Issues (MergedGlobals) ● By default clang merges globals with internal linkage into one: MergedGlobals ● Allows globals to be addressed using offsets from a base pointer ● Can reduce the number of registers used ● Modpost script in the Linux kernel uses symbol names to look for section mismatches (e.g. regular code calling init code) ● MergedGlobals breaks modpost (false positive section mismatches) ● Current solution: use -mno-global-merge to stop global merging ● Updates to modpost may allow this optimization to be enabled again in the future
  • 18. ARM eabi support ● Clang emits code which uses the “aeabi” ARM calls which are implemented in compiler-rt (equivalient to libgcc) ● Compiler-rt doesn't easily cross compile yet... void __aeabi_memcpy(void *dest, const void *src, size_t n) void __aeabi_memmove(void *dest, const void *src, size_t n) void __aeabi_memset(void *s, size_t n, int c) ● Still needed for ARM ● No longer required for AARCH64
  • 19. Variable Length Arrays In Structs ● VLAIS isn't supported by Clang (undocumented gcc extension) char vla[n]; /* Supported, C99/C11 */ struct { char flexible_member[]; /* Supported, C99/C11 */ } struct_with_flexible_member; struct { char vlais[n]; /* Explicitly not allowed by C99/C11 */ } variable_length_array_in_struct; ● VLAIS is used in the Linux kernel in a number of places, spreading mostly through reusing patterns from data structures found in crypto
  • 20. VLAIS Removal Example (from crypto/hmac.c) - struct { - struct shash_desc shash; - char ctx[crypto_shash_descsize(hash)]; - } desc; + char desc[sizeof(struct shash_desc) + + crypto_shash_descsize(hash)] CRYPTO_MINALIGN_ATTR; + struct shash_desc *shash = (struct shash_desc *)desc; unsigned int i; - desc.shash.tfm = hash; + shash->tfm = hash;
  • 21. VLAIS Removal Example (the missing pieces) #define ARCH_KMALLOC_MINALIGN __alignof__(unsigned long long) #define CRYPTO_MINALIGN ARCH_KMALLOC_MINALIGN #define CRYPTO_MINALIGN_ATTR __attribute__ ((__aligned__(CRYPTO_MINALIGN))) struct shash_desc { struct crypto_shash *tfm; u32 flags; void *__ctx[] CRYPTO_MINALIGN_ATTR; };
  • 22. Status of VLAIS in the Linux Kernel ● USB Gadget patch is in mainline ● Mac80211 patch is in mainline ● Netfilter patch is in mainline ● apparmor patch accepted ● Bluetooth patch accepted ● Only the VLAIS patches for crypto are left: (btrfs, dm-crypt, hmac, libcrc32c, testmgr, etc) ● However, we recently found a few previously unknown instances of VLAIS in raid10 and exofs...
  • 23. Huge use of VLAIS in fs/exofs/ore_raid.c static int _sp2d_alloc(unsigned pages_in_unit, unsigned group_width, unsigned parity, struct __stripe_pages_2d **psp2d) [...] unsigned data_devs = group_width - parity; struct _alloc_all_bytes { struct __alloc_stripe_pages_2d { struct __stripe_pages_2d sp2d; struct __1_page_stripe _1p_stripes[pages_in_unit]; } __asp2d; struct __alloc_1p_arrays { struct page *pages[group_width]; struct page *scribble[group_width]; char page_is_read[data_devs]; } __a1pa[pages_in_unit]; } *_aab;
  • 24. How Can You Help? ● Don’t use the non-C99 practices we showed in previous slides ● Make it known you want to be able to use Clang to compile the kernel (tell your Linaro representative!) ● Test LLVMLinux patches ● Report bugs to the LLVMLinux mailing list ● Help get LLVMLinux patches upstream ● Work on unsupported features and Bugs ○ http://llvm.linuxfoundation.org/index.php/Broken_kernel_options ● Submit new targets and arch support ● Patches welcome
  • 25. Embrace the Dragon. He's cuddly. Thank you http://llvm.linuxfoundation.org
  • 26. Contribute to the LLVMLinux Project ● Project wiki page ○ http://llvm.linuxfoundation.org ● Project Mailing List ○ http://lists.linuxfoundation.org/mailman/listinfo/llvmlinux ○ http://lists.linuxfoundation.org/pipermail/llvmlinux/ ● IRC Channel ○ #llvmlinux on OFTC ○ http://buildbot.llvm.linuxfoundation.org/irclogs/OFTC/%23llvmlinux/ ● LLVMLinux Community on Google Plus
  • 27. More about Linaro Connect: connect.linaro.org Linaro members: www.linaro.org/members More about Linaro: www.linaro.org/about/