error: invalid instruction arch/arm/mm/tlb-v7.S #1195

nickdesaulniers · 2020-11-09T22:56:34Z

building allyesconfig with LLVM_IAS=1:

$ ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- make LLVM=1 LLVM_IAS=1 -j71 arch/arm/mm/tlb-v7.o
...
clang-12: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument]
clang-12: warning: argument unused during compilation: '-Wa,-march=armv7-a' [-Wunused-command-line-argument]
arch/arm/mm/tlb-v7.S:35:2: error: invalid instruction
 dsb ish
 ^
arch/arm/mm/tlb-v7.S:56:2: error: invalid instruction
 dsb ish
 ^
arch/arm/mm/tlb-v7.S:69:2: error: invalid instruction
 dsb ish
 ^
arch/arm/mm/tlb-v7.S:84:2: error: invalid instruction
 dsb ish
 ^
arch/arm/mm/tlb-v7.S:85:2: error: instruction requires: data-barriers
 isb
 ^

similar to 7548bf8

diff --git a/arch/arm/mm/Makefile b/arch/arm/mm/Makefile
index 3510503bc5e6..6e9b54714500 100644
--- a/arch/arm/mm/Makefile
+++ b/arch/arm/mm/Makefile
@@ -73,7 +73,6 @@ obj-$(CONFIG_CPU_TLB_V7)      += tlb-v7.o
 obj-$(CONFIG_CPU_TLB_FA)       += tlb-fa.o
 
 AFLAGS_tlb-v6.o                :=-Wa,-march=armv6
-AFLAGS_tlb-v7.o                :=-Wa,-march=armv7-a
 
 obj-$(CONFIG_CPU_ARM7TDMI)     += proc-arm7tdmi.o
 obj-$(CONFIG_CPU_ARM720T)      += proc-arm720.o
diff --git a/arch/arm/mm/tlb-v7.S b/arch/arm/mm/tlb-v7.S
index 1bb28d7db567..2643e2f51b6a 100644
--- a/arch/arm/mm/tlb-v7.S
+++ b/arch/arm/mm/tlb-v7.S
@@ -16,6 +16,8 @@
 #include <asm/tlbflush.h>
 #include "proc-macros.S"
 
+.arch armv7-a
+
 /*
  *     v7wbi_flush_user_tlb_range(start, end, vma)
  *

fixes the error. cc @jcai19

The text was updated successfully, but these errors were encountered:

nickdesaulniers · 2020-11-09T22:59:17Z

eh, this is pretty common due to commit aff7b4f ("ARM: Ensure ARMv6/7 mm files are built using appropriate assembler options")

nickdesaulniers · 2020-11-09T23:26:50Z

tempted to mark this a dup of #957.

nickdesaulniers · 2020-11-10T00:38:20Z

Note that arch/arm64/Makefile has code like:

 91 ifeq ($(CONFIG_AS_HAS_PAC), y)                                                                                                                                                                                                         
 92 asm-arch := armv8.3-a                                                                                                                                                                                                                  
 93 endif                                                                                                                                                                                                                                  
 94 endif                                                                                                                                                                                                                                  
 95                                                                                                                                                                                                                                        
 96 KBUILD_CFLAGS += $(branch-prot-flags-y)                                                                                                                                                                                                
 97                                                                                                                                                                                                                                        
 98 ifeq ($(CONFIG_AS_HAS_ARMV8_4), y)                                                                                                                                                                                                     
 99 # make sure to pass the newest target architecture to -march.                                                                                                                                                                          
100 asm-arch := armv8.4-a                                                                                                                                                                                                                  
101 endif                                                                                                                                                                                                                                  
102                                                                                                                                                                                                                                        
103 ifdef asm-arch                                                                                                                                                                                                                         
104 KBUILD_CFLAGS += -Wa,-march=$(asm-arch) \                                                                                                                                                                                              
105        -DARM64_ASM_ARCH='"$(asm-arch)"'                                                                                                                                                                                                
106 endif

and it seems like the Kconfig checks are working, so is it the case that we support -Wa,-march= for aarch64 but not 32b arm?

nathanchance · 2020-11-10T01:26:45Z

Isn't the issue that -Wa,... is not used with the integrated assembler? That block was added due to the outcome of #1106.

nickdesaulniers · 2020-11-10T03:21:14Z

#1106 (comment)
#1106 (comment)

nickdesaulniers · 2021-01-27T02:23:51Z

https://llvm.org/pr48894

nickdesaulniers · 2021-01-27T23:06:24Z

ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- make LLVM=1 LLVM_IAS=1 -j72 defconfig enabled both

CONFIG_CPU_32v6K=y                                                                                                                                                         
CONFIG_CPU_32v7=y

cc @arndb should those be mutually exclusive?

nickdesaulniers · 2021-01-27T23:13:24Z

Modifying the kernel would involve touching at least 20 .S sources, and their Makefiles.

$ ag AFLAGS | grep march
mm/Makefile:36:AFLAGS_abort-ev6.o       :=-Wa,-march=armv6k
mm/Makefile:37:AFLAGS_abort-ev7.o       :=-Wa,-march=armv7-a
mm/Makefile:52:AFLAGS_cache-v6.o        :=-Wa,-march=armv6
mm/Makefile:53:AFLAGS_cache-v7.o        :=-Wa,-march=armv7-a
mm/Makefile:54:AFLAGS_cache-v7m.o       :=-Wa,-march=armv7-m
mm/Makefile:75:AFLAGS_tlb-v6.o          :=-Wa,-march=armv6
mm/Makefile:76:AFLAGS_tlb-v7.o          :=-Wa,-march=armv7-a
mm/Makefile:104:AFLAGS_proc-v6.o        :=-Wa,-march=armv6
mm/Makefile:105:AFLAGS_proc-v7.o        :=-Wa,-march=armv7-a
mach-npcm/Makefile:2:AFLAGS_headsmp.o           += -march=armv7-a
kernel/Makefile:102:AFLAGS_hyp-stub.o           :=-Wa,-march=armv7-a
mach-imx/Makefile:39:AFLAGS_headsmp.o :=-Wa,-march=armv7-a
mach-imx/Makefile:53:AFLAGS_suspend-imx6.o :=-Wa,-march=armv7-a
mach-imx/Makefile:58:AFLAGS_resume-imx6.o :=-Wa,-march=armv7-a
boot/compressed/Makefile:196:AFLAGS_hyp-stub.o := -Wa,-march=armv7-a
mach-at91/Makefile:17:AFLAGS_pm_suspend.o := -march=armv7-a
lib/Makefile:40:  AFLAGS_delay-loop.o           += -march=armv4
mach-mvebu/Makefile:4:AFLAGS_coherency_ll.o             := -Wa,-march=armv7-a
common/Makefile:17:AFLAGS_mcpm_head.o           := -march=armv7-a
common/Makefile:18:AFLAGS_vlock.o                       := -march=armv7-a
$ ag AFLAGS | grep march | wc -l
20

arndb · 2021-01-27T23:23:14Z

cc @arndb should those be mutually exclusive?

CONFIG_CPU_32v7 and CONFIG_CPU_32v6 are mutually exclusive, however CONFIG_CPU_32v6K is a special symbol that tells us that the 'K' extension of v6 is present. As 'K' is implied by ARMv7, CONFIG_CPU_32v6K is always enabled on any ARMv7 configuration.

I agree this is really confusing, but it works as designed.

nickdesaulniers · 2021-02-02T17:10:25Z

@DavidSpickett at Linaro also has a patch for LLVM: https://reviews.llvm.org/D95872. I'll try to get it reviewed and tested asap.

nickdesaulniers · 2021-02-12T22:57:11Z

Fixed: https://reviews.llvm.org/rG1d51c699b9e2ebc5bcfdbe85c74cc871426333d4

arndb · 2021-02-27T21:26:11Z

I still get the "argument unused during compilation" warning, though I no longer get the "invalid instruction" errors.

I think something is still wrong here. For instance when building the kernel without my earlier patch:

clang-13 -Wp,-MMD,drivers/memory/.ti-emif-sram-pm.o.d -nostdinc -isystem /usr/lib/llvm-13/lib/clang/13.0.0/include -I/git/arm-soc/arch/arm/include -I./arch/arm/include/generated -I/git/arm-soc/include -I./include -I/git/arm-soc/arch/arm/include/uapi -I./arch/arm/include/generated/uapi -I/git/arm-soc/include/uapi -I./include/generated/uapi -include /git/arm-soc/include/linux/kconfig.h -D__KERNEL__ -mlittle-endian -fmacro-prefix-map=/git/arm-soc/= -D__ASSEMBLY__ -fno-PIE --target=arm-linux-gnueabi --prefix=/home/arnd/cross/x86_64/gcc-10.1.0-nolibc/arm-linux-gnueabi/bin/arm-linux-gnueabi- --gcc-toolchain=/home/arnd/cross/x86_64/gcc-10.1.0-nolibc/arm-linux-gnueabi -Werror=unknown-warning-option -mabi=aapcs-linux -mfpu=vfp -meabi gnu -marm -Wa,-W -D__LINUX_ARM_ARCH__=6 -march=armv6k -mtune=arm1136j-s -include asm/unified.h -msoft-float -Wa,--fatal-warnings -Wa,-march=armv7-a -I /git/arm-soc/drivers/memory -I ./drivers/memory -DMODULE -c -o drivers/memory/ti-emif-sram-pm.o /git/arm-soc/drivers/memory/ti-emif-sram-pm.S
clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument]

It seems that clang now accepts the "-Wa,-march=armv7" bit and correctly builds the file with that, but when we pass both "-march=" and "-Wa,-march=" options, it complains about one of them being unused.

nickdesaulniers · 2021-03-01T01:12:53Z

That looks like a separate issue. Let's track it in #1315.

ClangBuiltLinux/linux#1195 was not resolved in LLVM 12, resulting in errors of that nature when using the integrated assembler: https://github.com/ClangBuiltLinux/continuous-integration2/runs/2434374437?check_suite_focus=true Make these builds just LLVM=1 to keep the arm coverage on Android. Signed-off-by: Nathan Chancellor <nathan@kernel.org>

Similar to commit a6c3087 ("ARM: 8989/1: use .fpu assembler directives instead of assembler arguments"). GCC and GNU binutils support setting the "sub arch" via -march=, -Wa,-march, target function attribute, and .arch assembler directive. Clang's integrated assembler does not support -Wa,-march (and the logic to overrule one when multiple of the above are used), and this can cause annoying warnings such as: clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] Since most assembler is non-conditionally assembled with one sub arch (modulo arch/arm/delay-loop.S which conditionally is assembled as armv4 based on CONFIG_ARCH_RPC, and arch/arm/mach-at91/pm-suspend.S which is conditionally assembled as armv7-a based on CONFIG_CPU_V7), prefer the .arch assembler directive. Link: https://bugs.llvm.org/show_bug.cgi?id=48894 Link: ClangBuiltLinux#1195 Suggested-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> [arnd] add a few more instances found in compile testing Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Similar to commit a6c3087 ("ARM: 8989/1: use .fpu assembler directives instead of assembler arguments"). GCC and GNU binutils support setting the "sub arch" via -march=, -Wa,-march, target function attribute, and .arch assembler directive. Clang was missing support for -Wa,-march=, but this was implemented in clang-13. The behavior of both GCC and Clang is to prefer -Wa,-march= over -march= for assembler and assembler-with-cpp sources, but Clang will warn about the -march= being unused. clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] Since most assembler is non-conditionally assembled with one sub arch (modulo arch/arm/delay-loop.S which conditionally is assembled as armv4 based on CONFIG_ARCH_RPC, and arch/arm/mach-at91/pm-suspend.S which is conditionally assembled as armv7-a based on CONFIG_CPU_V7), prefer the .arch assembler directive. Link: llvm/llvm-project@1d51c69 Link: https://bugs.llvm.org/show_bug.cgi?id=48894 Link: ClangBuiltLinux#1195 Link: ClangBuiltLinux#1315 Suggested-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> [arnd] add a few more instances found in compile testing Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Similar to commit a6c3087 ("ARM: 8989/1: use .fpu assembler directives instead of assembler arguments"). GCC and GNU binutils support setting the "sub arch" via -march=, -Wa,-march, target function attribute, and .arch assembler directive. Clang was missing support for -Wa,-march=, but this was implemented in clang-13. The behavior of both GCC and Clang is to prefer -Wa,-march= over -march= for assembler and assembler-with-cpp sources, but Clang will warn about the -march= being unused. clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] Since most assembler is non-conditionally assembled with one sub arch (modulo arch/arm/delay-loop.S which conditionally is assembled as armv4 based on CONFIG_ARCH_RPC, and arch/arm/mach-at91/pm-suspend.S which is conditionally assembled as armv7-a based on CONFIG_CPU_V7), prefer the .arch assembler directive. Add a few more instances found in compile testing as found by Arnd and Nathan. Link: llvm/llvm-project@1d51c69 Link: https://bugs.llvm.org/show_bug.cgi?id=48894 Link: ClangBuiltLinux#1195 Link: ClangBuiltLinux#1315 Suggested-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>

…lags Similar to commit a6c3087 ("ARM: 8989/1: use .fpu assembler directives instead of assembler arguments"). GCC and GNU binutils support setting the "sub arch" via -march=, -Wa,-march, target function attribute, and .arch assembler directive. Clang was missing support for -Wa,-march=, but this was implemented in clang-13. The behavior of both GCC and Clang is to prefer -Wa,-march= over -march= for assembler and assembler-with-cpp sources, but Clang will warn about the -march= being unused. clang: warning: argument unused during compilation: '-march=armv6k' [-Wunused-command-line-argument] Since most assembler is non-conditionally assembled with one sub arch (modulo arch/arm/delay-loop.S which conditionally is assembled as armv4 based on CONFIG_ARCH_RPC, and arch/arm/mach-at91/pm-suspend.S which is conditionally assembled as armv7-a based on CONFIG_CPU_V7), prefer the .arch assembler directive. Add a few more instances found in compile testing as found by Arnd and Nathan. Link: llvm/llvm-project@1d51c69 Link: https://bugs.llvm.org/show_bug.cgi?id=48894 Link: ClangBuiltLinux#1195 Link: ClangBuiltLinux#1315 Suggested-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Tested-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>

nickdesaulniers mentioned this issue Nov 9, 2020

warning: argument unused during compilation: '-march=armv7-a' #957

Closed

nickdesaulniers added the Reported upstream This bug was filed on LLVM’s issue tracker, Phabricator, or the kernel mailing list. label Jan 27, 2021

nickdesaulniers mentioned this issue Jan 27, 2021

clang ias: warning: argument unused during compilation: '-march=armv7-a' [-Wunused-command-line-argument] #1275

Closed

nickdesaulniers mentioned this issue Jan 28, 2021

Inconsistent kallsyms data #1280

Closed

nickdesaulniers closed this as completed Feb 12, 2021

nickdesaulniers added [FIXED][LLVM] 13 This bug was fixed in LLVM 13.x and removed [PATCH] Exists There is a patch that fixes this issue labels Feb 12, 2021

This was referenced Feb 22, 2021

invalid instruction in arch/arm/kernel/hyp-stub.S #1155

Closed

deprecated since v7, use 'dsb' #956

Closed

clang ias: invalid instruction in arch/arm/mach-omap1/ams-delta-fiq-handler.S: #1281

Closed

arndb reopened this Feb 27, 2021

nickdesaulniers mentioned this issue Mar 1, 2021

argument unused during compilation: '-march=armv6k' #1315

Closed

nickdesaulniers closed this as completed Mar 1, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

error: invalid instruction arch/arm/mm/tlb-v7.S #1195

error: invalid instruction arch/arm/mm/tlb-v7.S #1195

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 10, 2020

nathanchance commented Nov 10, 2020

nickdesaulniers commented Nov 10, 2020

nickdesaulniers commented Jan 27, 2021 •

edited by nathanchance

Loading

nickdesaulniers commented Jan 27, 2021

nickdesaulniers commented Jan 27, 2021

arndb commented Jan 27, 2021

nickdesaulniers commented Feb 2, 2021 •

edited

Loading

nickdesaulniers commented Feb 12, 2021

arndb commented Feb 27, 2021

nickdesaulniers commented Mar 1, 2021

error: invalid instruction arch/arm/mm/tlb-v7.S #1195

error: invalid instruction arch/arm/mm/tlb-v7.S #1195

Comments

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 9, 2020

nickdesaulniers commented Nov 10, 2020

nathanchance commented Nov 10, 2020

nickdesaulniers commented Nov 10, 2020

nickdesaulniers commented Jan 27, 2021 • edited by nathanchance Loading

nickdesaulniers commented Jan 27, 2021

nickdesaulniers commented Jan 27, 2021

arndb commented Jan 27, 2021

nickdesaulniers commented Feb 2, 2021 • edited Loading

nickdesaulniers commented Feb 12, 2021

arndb commented Feb 27, 2021

nickdesaulniers commented Mar 1, 2021

nickdesaulniers commented Jan 27, 2021 •

edited by nathanchance

Loading

nickdesaulniers commented Feb 2, 2021 •

edited

Loading