Linux: invalid output size for constraint '=q' in arch/x86/events/amd/core.c
In file included from arch/x86/events/amd/core.c:8:
arch/x86/events/amd/../perf_event.h:774:21: error: invalid output size for constraint '=q'
u64 disable_mask = __this_cpu_read(cpu_hw_events.perf_ctr_virt_mask);
^
./include/linux/percpu-defs.h:448:2: note: expanded from macro '__this_cpu_read'
raw_cpu_read(pcp); \
^
./include/linux/percpu-defs.h:422:28: note: expanded from macro 'raw_cpu_read'
#define raw_cpu_read(pcp) __pcpu_size_call_return(raw_cpu_read_, pcp)
^
./include/linux/percpu-defs.h:323:23: note: expanded from macro '__pcpu_size_call_return'
case 1: pscr_ret__ = stem##1(variable); break; \
^
<scratch space>:99:1: note: expanded from here
raw_cpu_read_1
^
./arch/x86/include/asm/percpu.h:394:30: note: expanded from macro 'raw_cpu_read_1'
#define raw_cpu_read_1(pcp) percpu_from_op("mov", pcp)
^
./arch/x86/include/asm/percpu.h:189:15: note: expanded from macro 'percpu_from_op'
: "=q" (pfo_ret__) \
^
In file included from arch/x86/events/amd/core.c:8:
arch/x86/events/amd/../perf_event.h:774:21: error: invalid output size for constraint '=q'
./include/linux/percpu-defs.h:448:2: note: expanded from macro '__this_cpu_read'
raw_cpu_read(pcp); \
^
./include/linux/percpu-defs.h:422:28: note: expanded from macro 'raw_cpu_read'
#define raw_cpu_read(pcp) __pcpu_size_call_return(raw_cpu_read_, pcp)
^
./include/linux/percpu-defs.h:324:23: note: expanded from macro '__pcpu_size_call_return'
case 2: pscr_ret__ = stem##2(variable); break; \
^
<scratch space>:108:1: note: expanded from here
raw_cpu_read_2
^
./arch/x86/include/asm/percpu.h:395:30: note: expanded from macro 'raw_cpu_read_2'
#define raw_cpu_read_2(pcp) percpu_from_op("mov", pcp)
^
./arch/x86/include/asm/percpu.h:189:15: note: expanded from macro 'percpu_from_op'
: "=q" (pfo_ret__) \
^
In file included from arch/x86/events/amd/core.c:8:
arch/x86/events/amd/../perf_event.h:774:21: error: invalid output size for constraint '=q'
./include/linux/percpu-defs.h:448:2: note: expanded from macro '__this_cpu_read'
raw_cpu_read(pcp); \
^
./include/linux/percpu-defs.h:422:28: note: expanded from macro 'raw_cpu_read'
#define raw_cpu_read(pcp) __pcpu_size_call_return(raw_cpu_read_, pcp)
^
./include/linux/percpu-defs.h:325:23: note: expanded from macro '__pcpu_size_call_return'
case 4: pscr_ret__ = stem##4(variable); break; \
^
<scratch space>:117:1: note: expanded from here
raw_cpu_read_4
^
./arch/x86/include/asm/percpu.h:396:30: note: expanded from macro 'raw_cpu_read_4'
#define raw_cpu_read_4(pcp) percpu_from_op("mov", pcp)
^
./arch/x86/include/asm/percpu.h:189:15: note: expanded from macro 'percpu_from_op'
: "=q" (pfo_ret__) \
^
3 errors generated.
tpimh
All 22 comments
arch/x86/kernel/cpu/intel.c:98:2: error: invalid input size for constraint 'qi'
this_cpu_or(msr_misc_features_shadow,
^
./include/linux/percpu-defs.h:513:32: note: expanded from macro 'this_cpu_or'
#define this_cpu_or(pcp, val) __pcpu_size_call(this_cpu_or_, pcp, val)
^
./include/linux/percpu-defs.h:379:11: note: expanded from macro '__pcpu_size_call'
case 1: stem##1(variable, __VA_ARGS__);break; \
^
<scratch space>:305:1: note: expanded from here
this_cpu_or_1
^
./arch/x86/include/asm/percpu.h:426:34: note: expanded from macro 'this_cpu_or_1'
#define this_cpu_or_1(pcp, val) percpu_to_op("or", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
arch/x86/kernel/cpu/intel.c:98:2: error: invalid input size for constraint 'qi'
./include/linux/percpu-defs.h:513:32: note: expanded from macro 'this_cpu_or'
#define this_cpu_or(pcp, val) __pcpu_size_call(this_cpu_or_, pcp, val)
^
./include/linux/percpu-defs.h:380:11: note: expanded from macro '__pcpu_size_call'
case 2: stem##2(variable, __VA_ARGS__);break; \
^
<scratch space>:314:1: note: expanded from here
this_cpu_or_2
^
./arch/x86/include/asm/percpu.h:427:34: note: expanded from macro 'this_cpu_or_2'
#define this_cpu_or_2(pcp, val) percpu_to_op("or", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
arch/x86/kernel/cpu/intel.c:98:2: error: invalid input size for constraint 'qi'
./include/linux/percpu-defs.h:513:32: note: expanded from macro 'this_cpu_or'
#define this_cpu_or(pcp, val) __pcpu_size_call(this_cpu_or_, pcp, val)
^
./include/linux/percpu-defs.h:381:11: note: expanded from macro '__pcpu_size_call'
case 4: stem##4(variable, __VA_ARGS__);break; \
^
<scratch space>:323:1: note: expanded from here
this_cpu_or_4
^
./arch/x86/include/asm/percpu.h:428:34: note: expanded from macro 'this_cpu_or_4'
#define this_cpu_or_4(pcp, val) percpu_to_op("or", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
arch/x86/kernel/cpu/intel.c:647:2: error: invalid input size for constraint 'qi'
this_cpu_write(msr_misc_features_shadow, 0);
^
./include/linux/percpu-defs.h:510:34: note: expanded from macro 'this_cpu_write'
#define this_cpu_write(pcp, val) __pcpu_size_call(this_cpu_write_, pcp, val)
^
./include/linux/percpu-defs.h:379:11: note: expanded from macro '__pcpu_size_call'
case 1: stem##1(variable, __VA_ARGS__);break; \
^
<scratch space>:187:1: note: expanded from here
this_cpu_write_1
^
./arch/x86/include/asm/percpu.h:417:36: note: expanded from macro 'this_cpu_write_1'
#define this_cpu_write_1(pcp, val) percpu_to_op("mov", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
arch/x86/kernel/cpu/intel.c:647:2: error: invalid input size for constraint 'qi'
./include/linux/percpu-defs.h:510:34: note: expanded from macro 'this_cpu_write'
#define this_cpu_write(pcp, val) __pcpu_size_call(this_cpu_write_, pcp, val)
^
./include/linux/percpu-defs.h:380:11: note: expanded from macro '__pcpu_size_call'
case 2: stem##2(variable, __VA_ARGS__);break; \
^
<scratch space>:196:1: note: expanded from here
this_cpu_write_2
^
./arch/x86/include/asm/percpu.h:418:36: note: expanded from macro 'this_cpu_write_2'
#define this_cpu_write_2(pcp, val) percpu_to_op("mov", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
arch/x86/kernel/cpu/intel.c:647:2: error: invalid input size for constraint 'qi'
./include/linux/percpu-defs.h:510:34: note: expanded from macro 'this_cpu_write'
#define this_cpu_write(pcp, val) __pcpu_size_call(this_cpu_write_, pcp, val)
^
./include/linux/percpu-defs.h:381:11: note: expanded from macro '__pcpu_size_call'
case 4: stem##4(variable, __VA_ARGS__);break; \
^
<scratch space>:205:1: note: expanded from here
this_cpu_write_4
^
./arch/x86/include/asm/percpu.h:419:36: note: expanded from macro 'this_cpu_write_4'
#define this_cpu_write_4(pcp, val) percpu_to_op("mov", (pcp), val)
^
./arch/x86/include/asm/percpu.h:102:15: note: expanded from macro 'percpu_to_op'
: "qi" ((pto_T__)(val))); \
^
6 errors generated.
Guessing this error is what this patch is addressing: https://github.com/nathanchance/patches/blob/19158418f43c6334a6c0ed47b11147b70853f7e3/linux/other-hax/0001-Make-clang-hate-percpu.h-less-in-32-bit-mode.patch
nathanchance
on 8 Oct 2018
I submitted it to LLVM bugzilla: https://bugs.llvm.org/show_bug.cgi?id=33587
tpimh
on 8 Oct 2018
It's already in this issue tracker as #3, should this now be closed as duplicate?
tpimh
on 9 Oct 2018
Reopening this to track linux patches fixing this problem. Currently, two independent solutions exist:
- By @arndb: https://lkml.org/lkml/2019/7/11/263, modified to be applied to newer kernels:
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -99,7 +99,7 @@ do { \
case 1: \
asm qual (op "b %1,"__percpu_arg(0) \
: "+m" (var) \
- : "qi" ((pto_T__)(val))); \
+ : "ri" ((pto_T__)(val))); \
break; \
case 2: \
asm qual (op "w %1,"__percpu_arg(0) \
@@ -144,7 +144,7 @@ do { \
else \
asm qual ("addb %1, "__percpu_arg(0) \
: "+m" (var) \
- : "qi" ((pao_T__)(val))); \
+ : "ri" ((pao_T__)(val))); \
break; \
case 2: \
if (pao_ID__ == 1) \
@@ -186,7 +186,7 @@ do { \
switch (sizeof(var)) { \
case 1: \
asm qual (op "b "__percpu_arg(1)",%0" \
- : "=q" (pfo_ret__) \
+ : "=r" (pfo_ret__) \
: "m" (var)); \
break; \
case 2: \
@@ -215,7 +215,7 @@ do { \
switch (sizeof(var)) { \
case 1: \
asm(op "b "__percpu_arg(P1)",%0" \
- : "=q" (pfo_ret__) \
+ : "=r" (pfo_ret__) \
: "p" (&(var))); \
break; \
case 2: \
- By @dwmw2: https://lkml.org/lkml/2018/2/9/529 (lore), modified to be applied to newer kernels:
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -99,7 +99,7 @@ do { \
case 1: \
asm qual (op "b %1,"__percpu_arg(0) \
: "+m" (var) \
- : "qi" ((pto_T__)(val))); \
+ : "qi" ((unsigned char)(unsigned long)(val))); \
break; \
case 2: \
asm qual (op "w %1,"__percpu_arg(0) \
@@ -144,7 +144,7 @@ do { \
else \
asm qual ("addb %1, "__percpu_arg(0) \
: "+m" (var) \
- : "qi" ((pao_T__)(val))); \
+ : "qi" ((unsigned char)(unsigned long)(val))); \
break; \
case 2: \
if (pao_ID__ == 1) \
@@ -182,12 +182,14 @@ do { \
#define percpu_from_op(qual, op, var) \
({ \
+ unsigned char pfo_u8__; \
typeof(var) pfo_ret__; \
switch (sizeof(var)) { \
case 1: \
asm qual (op "b "__percpu_arg(1)",%0" \
- : "=q" (pfo_ret__) \
+ : "=q" (pfo_u8__) \
: "m" (var)); \
+ pfo_ret__ = (typeof(var))(unsigned long)pfo_u8__; \
break; \
case 2: \
asm qual (op "w "__percpu_arg(1)",%0" \
@@ -211,12 +213,14 @@ do { \
#define percpu_stable_op(op, var) \
({ \
+ unsigned char pfo_u8__; \
typeof(var) pfo_ret__; \
switch (sizeof(var)) { \
case 1: \
asm(op "b "__percpu_arg(P1)",%0" \
- : "=q" (pfo_ret__) \
+ : "=q" (pfo_u8__) \
: "p" (&(var))); \
+ pfo_ret__ = (typeof(var))(unsigned long)pfo_u8__; \
break; \
case 2: \
asm(op "w "__percpu_arg(P1)",%0" \
They both work equally good, but I can't tell which solution is better. Anyway I believe one of these patches should be resent to LKML to be included in the kernel.
tpimh
on 1 Nov 2019
Cool, thanks for finding and documenting both of these; I suspect we can start a discussion with both @arndb and dwm on LKML+x86 mailing list for thoughts.
nickdesaulniers
on 1 Nov 2019
technically a case of #3
nickdesaulniers
on 8 Feb 2020
The problem with my patch is that it does not work when the compiler picks register ESI/EDI/EBP on i386, as using the lower 8 bits only works on registers EAX/EBX/ECX/EDX. This causes occasional build failures in random files. @dwmw2's patch looks correct to me, I'll give it some more randconfig testing.
arndb
on 28 Apr 2020
nickdesaulniers
on 5 May 2020
The whole series for this has landed in the percpu tree, scheduled for 5.9 (see 11 patches from Brian Gerst and myself): https://git.kernel.org/pub/scm/linux/kernel/git/dennis/percpu.git/log/?h=for-5.9
nickdesaulniers
on 22 Jul 2020
It's cool man.
How do I pass i386_defconfig to make in an x86-64 build-environment?
And which prerequirements do I need?
Here I have an x86 multiarch environment.
I tried this but did not a full i386_defconfig build:
MAKE_OPTS="HOSTCC=clang-11 HOSTCXX=clang++-11 HOSTLD=ld.lld-11 HOSTAR=llvm-ar-11 CC=clang-11 LD=ld.lld-11 AR=llvm-ar-11 NM=llvm-nm-11 OBJCOPY=llvm-objcopy-11 OBJDUMP=llvm-objdump-11 OBJSIZE=llvm-size-11 READELF=llvm-readelf-11 STRIP=llvm-strip-11 LLVM_IAS=1"
$ ARCH=x86 make V=1 -j3 $MAKE_OPTS i386_defconfig
$ ARCH=x86 make V=1 -j3 $MAKE_OPTS
i386_defconfig_config-generated.txt has:
# Linux/x86 5.8.0-rc6 Kernel Configuration
CONFIG_X86_32=y
CONFIG_X86=y
CONFIG_X86_32_SMP=y
Generated linux-config is attached.
dileks
on 22 Jul 2020
IIUC, I think:
$ make ... i386_defconfig
and
$ ARCH=i386 make defconfig
are equivalent, though I haven't compared .configs from the two.
$ make allnoconfig
happens to be i386, too.
I usually test with make ... i386_defconfig.
nickdesaulniers
on 22 Jul 2020
Doing:
$ ARCH=i386 make V=1 -j3 $MAKE_OPTS i386_defconfig
...and looking at the diff:
$ diff -uprN ../i386_defconfig.config ../i386_defconfig.config-v2
--- ../i386_defconfig.config 2020-07-22 14:58:36.346842915 +0200
+++ ../i386_defconfig.config-v2 2020-07-22 20:37:00.265095068 +0200
@@ -1,6 +1,6 @@
#
# Automatically generated file; DO NOT EDIT.
-# Linux/x86 5.8.0-rc6 Kernel Configuration
+# Linux/i386 5.8.0-rc6 Kernel Configuration
#
CONFIG_CC_VERSION_TEXT="clang version 11.0.0 (https://github.com/llvm/llvm-project cebd637c88624dfd44520848cb1f43dc8a02ba80)"
CONFIG_GCC_VERSION=0
@@ -241,7 +241,6 @@ CONFIG_PROFILING=y
CONFIG_TRACEPOINTS=y
# end of General setup
-# CONFIG_64BIT is not set
CONFIG_X86_32=y
CONFIG_X86=y
CONFIG_INSTRUCTION_DECODER=y
Hmm, there is no CONFIG_64BIT - isn't that checked for in some places to distinguish with x86 (32bit) and/or x86_32?
dileks
on 22 Jul 2020
Produces the same .config:
$ ARCH=x86 make V=1 -j3 $MAKE_OPTS i386_defconfig
$ make V=1 -j3 $MAKE_OPTS i386_defconfig
Both ARCH=i386 and ARCH=x86 let clang-11 compile with -march=i686 -mtune=generic.
So hat config should be named x86_defconfig not i386_defconfig (might be historical reasons).
AFAICS i386code was thrown out and is no more supported.
x86_defconfig would be consistent with x86_64_defconfig:
$ find arch/x86 -name '*defconfig'
arch/x86/configs/i386_defconfig
arch/x86/configs/x86_64_defconfig
This has nothing to do with i386 see CONFIG_CRYPTO_AES_586:
$ grep 86 arch/x86/configs/i386_defconfig
CONFIG_X86_GENERIC=y
CONFIG_X86_REROUTE_FOR_BROKEN_BOOT_IRQS=y
CONFIG_X86_MCE=y
CONFIG_X86_REBOOTFIXUPS=y
CONFIG_X86_MSR=y
CONFIG_X86_CPUID=y
CONFIG_X86_CHECK_BIOS_CORRUPTION=y
CONFIG_X86_ACPI_CPUFREQ=y
CONFIG_CRYPTO_AES_586=y
That seems to be some old relicts?
$ git grep "ARCH=i386"
Documentation/kbuild/headers_install.rst: make headers_install ARCH=i386 INSTALL_HDR_PATH=/usr
tools/testing/ktest/examples/crosstests.conf:MAKE_CMD = make ARCH=i386
tools/testing/ktest/sample.conf:#MAKE_CMD = CC=i386-gcc AS=i386-as make ARCH=i386
So minimum supported CPU is CONFIG_M486 - there is no CONFIG_M386.
dileks
on 22 Jul 2020
Checked with gcc-9 - all generated .config file set CONFIG_M686=y.
dileks
on 22 Jul 2020
Pah, that's also some old cruft:
$ git grep CRYPTO_AES_586
arch/x86/configs/i386_defconfig:CONFIG_CRYPTO_AES_586=y
Patchset now in <tip.git#x86/asm>.
[1] https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/log/?h=x86/asm
dileks
on 23 Jul 2020
👍2
As a side-note:
"x86/defconfigs: Remove CONFIG_CRYPTO_AES_586 from i386_defconfig" landed in <tip.git#x86/build>.
[1] https://git.kernel.org/tip/tip/c/6526b12de07588253a52577f42ec99fc7ca26a1f
dileks
on 24 Jul 2020
Now in Linus Git tree.
dileks
on 4 Aug 2020
🎉1
A nice i386_defconfig cleanup:
"x86/defconfigs/32: Unset 64BIT"
dileks
on 10 Aug 2020
Was this page helpful?
0 / 5 - 0 ratings
Related issues
nickdesaulniers
路
4Comments
nathanchance
路
4Comments
warning: optimization flag '-fno-strength-reduce' is not supported [-Wignored-optimization-argument]
nickdesaulniers
路
3Comments
nickdesaulniers
路
4Comments
tpgxyz
路
4Comments
Most helpful comment
Patchset now in
<tip.git#x86/asm>.[1] https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/log/?h=x86/asm