From a92fc36ebe032a04313cdf16b2152e36ca7c20c3 Mon Sep 17 00:00:00 2001 From: Russell King Date: Tue, 9 Oct 2012 11:13:26 +0100 Subject: [PATCH] ARM: vfp: fix saving d16-d31 vfp registers on v6+ kernels commit 846a136881b8f73c1f74250bf6acfaa309cab1f2 upstream. Michael Olbrich reported that his test program fails when built with -O2 -mcpu=cortex-a8 -mfpu=neon, and a kernel which supports v6 and v7 CPUs: volatile int x = 2; volatile int64_t y = 2; int main() { volatile int a = 0; volatile int64_t b = 0; while (1) { a = (a + x) % (1 << 30); b = (b + y) % (1 << 30); assert(a == b); } } and two instances are run. When built for just v7 CPUs, this program works fine. It uses the "vadd.i64 d19, d18, d16" VFP instruction. It appears that we do not save the high-16 double VFP registers across context switches when the kernel is built for v6 CPUs. Fix that. Tested-By: Michael Olbrich Signed-off-by: Russell King Signed-off-by: Ben Hutchings --- arch/arm/include/asm/vfpmacros.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/arch/arm/include/asm/vfpmacros.h b/arch/arm/include/asm/vfpmacros.h index 3d5fc41..bf53047 100644 --- a/arch/arm/include/asm/vfpmacros.h +++ b/arch/arm/include/asm/vfpmacros.h @@ -28,7 +28,7 @@ ldr \tmp, =elf_hwcap @ may not have MVFR regs ldr \tmp, [\tmp, #0] tst \tmp, #HWCAP_VFPv3D16 - ldceq p11, cr0, [\base],#32*4 @ FLDMIAD \base!, {d16-d31} + ldceql p11, cr0, [\base],#32*4 @ FLDMIAD \base!, {d16-d31} addne \base, \base, #32*4 @ step over unused register space #else VFPFMRX \tmp, MVFR0 @ Media and VFP Feature Register 0 @@ -52,7 +52,7 @@ ldr \tmp, =elf_hwcap @ may not have MVFR regs ldr \tmp, [\tmp, #0] tst \tmp, #HWCAP_VFPv3D16 - stceq p11, cr0, [\base],#32*4 @ FSTMIAD \base!, {d16-d31} + stceql p11, cr0, [\base],#32*4 @ FSTMIAD \base!, {d16-d31} addne \base, \base, #32*4 @ step over unused register space #else VFPFMRX \tmp, MVFR0 @ Media and VFP Feature Register 0 -- 2.7.4