[ARM] Fix VFP initialisation issue for SMP systems

When we install the handlers for context switching, we must enable
VFP on all CPU cores, otherwise undefined (and random) effects
occur.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This commit is contained in:
Russell King 2007-01-02 23:40:30 +00:00 committed by Russell King
parent 1c9d3df5e8
commit 8e140362f7
1 changed files with 15 additions and 0 deletions

View File

@ -264,6 +264,18 @@ void VFP9_bounce(u32 trigger, u32 fpexc, struct pt_regs *regs)
vfp_raise_exceptions(exceptions, trigger, orig_fpscr, regs);
}
static void vfp_enable(void *unused)
{
u32 access = get_copro_access();
/*
* Enable full access to VFP (cp10 and cp11)
*/
set_copro_access(access | CPACC_FULL(10) | CPACC_FULL(11));
}
#include <linux/smp.h>
/*
* VFP support code initialisation.
*/
@ -288,6 +300,7 @@ static int __init vfp_init(void)
* we just need to read the VFPSID register.
*/
vfpsid = fmrx(FPSID);
barrier();
printk(KERN_INFO "VFP support v0.3: ");
if (VFP_arch) {
@ -301,6 +314,8 @@ static int __init vfp_init(void)
} else if (vfpsid & FPSID_NODOUBLE) {
printk("no double precision support\n");
} else {
smp_call_function(vfp_enable, NULL, 1, 1);
VFP_arch = (vfpsid & FPSID_ARCH_MASK) >> FPSID_ARCH_BIT; /* Extract the architecture version */
printk("implementor %02x architecture %d part %02x variant %x rev %x\n",
(vfpsid & FPSID_IMPLEMENTER_MASK) >> FPSID_IMPLEMENTER_BIT,