x86/fpu: Fix FPU related boot regression when CPUID masking BIOS feat…

…ure is enabled

Mike Galbraith reported:

  " My i7-4790 box is having one hell of a time with this merge
    window, dead in the water.

    BIOS setting "Limit CPUID Maximum" upsets new fpu code
    mightily. "

It turns out that Linux does a double workaround here, as per:

  066941b ("x86: unmask CPUID levels on Intel CPUs")

it undoes the BIOS workaround - but as a side effect the CPUID
state is not completely constant during early init anymore,
and the new FPU init code did not take this into account.

So what happened is that the xstate init code did not have full
CPUID available, which broke subsequent attempts to use xstate
features.

Fix this by ordering the early FPU init code to after we've
stabilized the CPUID state.

Reported-bisected-and-tested-by: Mike Galbraith <umgwanakikbuti@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <umgwanakikbuti@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20150627082514.GA10894@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/kernel/cpu/common.c

@@ -742,7 +742,6 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 	cpu_detect(c);
 	get_cpu_vendor(c);
 	get_cpu_cap(c);
-	fpu__init_system(c);
 
 	if (this_cpu->c_early_init)
 		this_cpu->c_early_init(c);
@@ -754,6 +753,7 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 		this_cpu->c_bsp_init(c);
 
 	setup_force_cpu_cap(X86_FEATURE_ALWAYS);
+	fpu__init_system(c);
 }
 
 void __init early_cpu_init(void)