diff options
| author | Russell King <rmk+kernel@arm.linux.org.uk> | 2012-10-09 11:13:26 +0100 |
|---|---|---|
| committer | Russell King <rmk+kernel@arm.linux.org.uk> | 2012-10-09 12:16:46 +0100 |
| commit | 846a136881b8f73c1f74250bf6acfaa309cab1f2 (patch) | |
| tree | 56e17539b3b07ef5a17bcccb4174206f07610b0d | |
| parent | f5a246eab9a268f51ba8189ea5b098a1bfff200e (diff) | |
ARM: vfp: fix saving d16-d31 vfp registers on v6+ kernels
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.
Cc: <stable@vger.kernel.org>
Tested-By: Michael Olbrich <m.olbrich@pengutronix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
| -rw-r--r-- | arch/arm/include/asm/vfpmacros.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/arch/arm/include/asm/vfpmacros.h b/arch/arm/include/asm/vfpmacros.h index a7aadbd9a6d..6a6f1e485f4 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 |