Re: [PATCH 2/2] selftests/fpu: Add an FPU selftest

From: Nick Desaulniers
Date: Mon Jun 22 2020 - 14:15:59 EST

Next message: Kees Cook: "[PATCH v2 0/8] selftests/harness: Switch to TAP output"
Previous message: Markus Elfring: "Re: [PATCH] i2c: sprd: Fix runtime PM imbalance on error"
In reply to: Borislav Petkov: "Re: [PATCH 2/2] selftests/fpu: Add an FPU selftest"
Next in thread: Borislav Petkov: "[PATCH 1/2] x86/fpu: Reset MXCSR to default in kernel_fpu_begin()"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Mon, Jun 22, 2020 at 10:04 AM Borislav Petkov <bp@xxxxxxxxx> wrote:
>
> On Sun, Jun 21, 2020 at 03:55:00AM +0800, kernel test robot wrote:
> > Hi Borislav,
> >
> > Thank you for the patch! Yet something to improve:
> >
> > [auto build test ERROR on kselftest/next]
> > [also build test ERROR on tip/auto-latest linus/master v5.8-rc1 next-20200618]
> > [cannot apply to tip/x86/core]
> > [If your patch is applied to the wrong git tree, kindly drop us a note.
> > And when submitting patch, we suggest to use as documented in
> > https://git-scm.com/docs/git-format-patch]
> >
> > url: https://github.com/0day-ci/linux/commits/Borislav-Petkov/x86-FPU-FPU-sanitization-for-in-kernel-use/20200620-014453
> > base: https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest.git next
> > config: x86_64-allmodconfig (attached as .config)
> > compiler: clang version 11.0.0 (https://github.com/llvm/llvm-project 487ca07fcc75d52755c9fe2ee05bcb3b6eeeec44)
> > reproduce (this is a W=1 build):
> > wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
> > chmod +x ~/bin/make.cross
> > # install x86_64 cross compiling tool for clang build
> > # apt-get install binutils-x86-64-linux-gnu
> > # save the attached .config to linux build tree
> > COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=x86_64
> >
> > If you fix the issue, kindly add following tag as appropriate
> > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> >
> > All errors (new ones prefixed by >>, old ones prefixed by <<):
> >
> > >> ERROR: modpost: "__gtdf2" [lib/test_fpu.ko] undefined!
> > >> ERROR: modpost: "__divdf3" [lib/test_fpu.ko] undefined!
> > >> ERROR: modpost: "__muldf3" [lib/test_fpu.ko] undefined!
> > >> ERROR: modpost: "__adddf3" [lib/test_fpu.ko] undefined!
>
> Hmm, the point here is to actually have SSE* code in a kernel module for
> testing. And gcc does the right thing, see below.
>
> LLVM folks, how can I make llvm not call those library functions and
> actually spit out SSE* code?
>
> Full build command is:
>
> /mnt/smr/share/src/llvm/tc-build/install/bin/clang-11 -Wp,-MMD,lib/.test_fpu.o.d -nostdinc -isystem /mnt/smr/share/src/llvm/tc-build/install/lib/clang/11.0.0/include -I./arch/x86/include -I./arch/x86/include/generated -I./include -I./arch/x86/include/uapi -I./arch/x86/include/generated/uapi -I./include/uapi -I./include/generated/uapi -include ./include/linux/kconfig.h -include ./include/linux/compiler_types.h -D__KERNEL__ -Qunused-arguments -Wall -Wundef -Werror=strict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -fshort-wchar -fno-PIE -Werror=implicit-function-declaration -Werror=implicit-int -Wno-format-security -std=gnu89 -no-integrated-as -Werror=unknown-warning-option -mno-sse -mno-mmx -mno-sse2 -mno-3dnow -mno-avx -m64 -mno-80387 -mstack-alignment=8 -march=k8 -mno-red-zone -mcmodel=kernel -Wno-sign-compare -fno-asynchronous-unwind-tables -mretpoline-external-thunk -fno-delete-null-pointer-checks -Wno-frame-address -Wno-address-of-packed-member -O2 -Wframe-larger-than=2048 -fstack-protector -Wno-format-invalid-specifier -Wno-gnu -mno-global-merge -Wno-unused-const-variable -g -DCC_USING_FENTRY -Wdeclaration-after-statement -Wvla -Wno-pointer-sign -Wno-array-bounds -fno-strict-overflow -fno-merge-all-constants -fno-stack-check -Werror=date-time -Werror=incompatible-pointer-types -fmacro-prefix-map=./= -fcf-protection=none -Wno-initializer-overrides -Wno-format -Wno-sign-compare -Wno-format-zero-length -Wno-pointer-to-enum-cast -Wno-tautological-constant-out-of-range-compare -mhard-float -msse -DMODULE -DKBUILD_BASENAME='"test_fpu"' -DKBUILD_MODNAME='"test_fpu"' -c -o lib/test_fpu.o lib/test_fpu.c

Add -msse2. See also:
commit e8a170ff9a35 ("drm/amdgpu: enable -msse2 for GCC 7.1+ users")
potentially relevant, though less so:
commit 00db297106e8 ("drm/amdgpu: fix stack alignment ABI mismatch for
GCC 7.1+")
commit c868868f6b6a ("drm/amdgpu: fix stack alignment ABI mismatch for Clang")
and may your stacks always share the same alignment.

>
> Thx.
>
> --- /tmp/test_fpu.s.gcc 2020-06-22 18:51:25.828615337 +0200
> +++ /tmp/test_fpu.s.llvm-11 2020-06-22 18:50:35.080614534 +0200
> @@ -1,74 +1,49 @@
> -test_fpu_get:
> - pushq %r12 #
> - pushq %rbx #
> - movq %rsi, %rbx # tmp119, val
> - subq $56, %rsp #,
> -# lib/test_fpu.c:58: kernel_fpu_begin();
> - call kernel_fpu_begin #
> -# lib/test_fpu.c:32: a = 4.0;
> - movlpd .LC1(%rip), %xmm0 #, tmp106
> - movsd %xmm0, (%rsp) # tmp106, a
> -# lib/test_fpu.c:33: b = 1e-15;
> - movlpd .LC2(%rip), %xmm0 #, tmp107
> - movsd %xmm0, 8(%rsp) # tmp107, b
> -# lib/test_fpu.c:34: c = 1e-310;
> - movlpd .LC3(%rip), %xmm0 #, tmp108
> - movsd %xmm0, 16(%rsp) # tmp108, c
> -# lib/test_fpu.c:37: d = a + b;
> - movlpd (%rsp), %xmm0 # a, a.2_8
> - movlpd 8(%rsp), %xmm1 # b, b.3_9
> - addsd %xmm1, %xmm0 # b.3_9, _10
> -# lib/test_fpu.c:37: d = a + b;
> - movsd %xmm0, 24(%rsp) # _10, d
> -# lib/test_fpu.c:40: e = a + b / 2;
> - movlpd 8(%rsp), %xmm0 # b, b.4_11
> -# lib/test_fpu.c:40: e = a + b / 2;
> - movlpd (%rsp), %xmm1 # a, a.5_13
> -# lib/test_fpu.c:40: e = a + b / 2;
> - mulsd .LC4(%rip), %xmm0 #, tmp109
> -# lib/test_fpu.c:40: e = a + b / 2;
> - addsd %xmm1, %xmm0 # a.5_13, _14
> -# lib/test_fpu.c:40: e = a + b / 2;
> - movsd %xmm0, 32(%rsp) # _14, e
> -# lib/test_fpu.c:43: f = b / c;
> - movlpd 8(%rsp), %xmm0 # b, b.6_15
> - movlpd 16(%rsp), %xmm1 # c, c.7_16
> - divsd %xmm1, %xmm0 # c.7_16, _17
> -# lib/test_fpu.c:43: f = b / c;
> - movsd %xmm0, 40(%rsp) # _17, f
> -# lib/test_fpu.c:46: g = a + c * f;
> - movlpd 16(%rsp), %xmm0 # c, c.8_18
> - movlpd 40(%rsp), %xmm2 # f, f.9_19
> -# lib/test_fpu.c:46: g = a + c * f;
> - movlpd (%rsp), %xmm1 # a, a.10_21
> -# lib/test_fpu.c:46: g = a + c * f;
> - mulsd %xmm2, %xmm0 # f.9_19, tmp111
> -# lib/test_fpu.c:46: g = a + c * f;
> - addsd %xmm1, %xmm0 # a.10_21, _22
> -# lib/test_fpu.c:46: g = a + c * f;
> - movsd %xmm0, 48(%rsp) # _22, g
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - movlpd 24(%rsp), %xmm1 # d, d.11_23
> - movlpd (%rsp), %xmm0 # a, a.12_24
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - comisd %xmm0, %xmm1 # a.12_24, d.11_23
> - jbe .L15 #,
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - movlpd 32(%rsp), %xmm1 # e, e.13_25
> - movlpd (%rsp), %xmm0 # a, a.14_26
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - comisd %xmm0, %xmm1 # a.14_26, e.13_25
> - jbe .L15 #,
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - movlpd 48(%rsp), %xmm1 # g, g.15_27
> -# lib/test_fpu.c:51: return -EINVAL;
> - movl $0, %r12d #, tmp117
> - movl $-22, %eax #, tmp118
> -# lib/test_fpu.c:48: if (d > a && e > a && g > a)
> - movlpd (%rsp), %xmm0 # a, a.16_28
> -# lib/test_fpu.c:51: return -EINVAL;
> - comisd %xmm0, %xmm1 # a.16_28, g.15_27
> - cmovbe %eax, %r12d # tmp117,, tmp118, <retval>
> -.L4:
> -# lib/test_fpu.c:60: kernel_fpu_end();
> - call kernel_fpu_end #
> +test_fpu_get: # @test_fpu_get
> +# %bb.0:
> + pushq %rbp
> + pushq %rbx
> + subq $56, %rsp
> + movq %rsi, %rbx
> + callq kernel_fpu_begin
> + movabsq $4616189618054758400, %rax # imm = 0x4010000000000000
> + movq %rax, (%rsp)
> + movabsq $4382569440205035030, %rax # imm = 0x3CD203AF9EE75616
> + movq %rax, 8(%rsp)
> + movabsq $20240225330731, %rax # imm = 0x12688B70E62B
> + movq %rax, 16(%rsp)
> + movq (%rsp), %rdi
> + movq 8(%rsp), %rsi
> + callq __adddf3
> + movq %rax, 48(%rsp)
> + movq (%rsp), %rbp
> + movq 8(%rsp), %rdi
> + movabsq $4602678819172646912, %rsi # imm = 0x3FE0000000000000
> + callq __muldf3
> + movq %rbp, %rdi
> + movq %rax, %rsi
> + callq __adddf3
> + movq %rax, 40(%rsp)
> + movq 8(%rsp), %rdi
> + movq 16(%rsp), %rsi
> + callq __divdf3
> + movq %rax, 32(%rsp)
> + movq (%rsp), %rbp
> + movq 16(%rsp), %rdi
> + movq 32(%rsp), %rsi
> + callq __muldf3
> + movq %rbp, %rdi
> + movq %rax, %rsi
> + callq __adddf3
> + movq %rax, 24(%rsp)
> + movq 48(%rsp), %rdi
> + movq (%rsp), %rsi
> + callq __gtdf2
> + testl %eax, %eax
> + jle .LBB3_3
> +# %bb.1:
> + movq 40(%rsp), %rdi
> + movq (%rsp), %rsi
> + callq __gtdf2
> + testl %eax, %eax
> + jle .LBB3_3
> +# %bb.2:
>
> --
> Regards/Gruss,
> Boris.
>
> https://people.kernel.org/tglx/notes-about-netiquette
>
> --
> You received this message because you are subscribed to the Google Groups "Clang Built Linux" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to clang-built-linux+unsubscribe@xxxxxxxxxxxxxxxxx
> To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/20200622170446.GG32200%40zn.tnic.

--
Thanks,
~Nick Desaulniers

Next message: Kees Cook: "[PATCH v2 0/8] selftests/harness: Switch to TAP output"
Previous message: Markus Elfring: "Re: [PATCH] i2c: sprd: Fix runtime PM imbalance on error"
In reply to: Borislav Petkov: "Re: [PATCH 2/2] selftests/fpu: Add an FPU selftest"
Next in thread: Borislav Petkov: "[PATCH 1/2] x86/fpu: Reset MXCSR to default in kernel_fpu_begin()"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]