+/*
+ * Copyright (c) 2017-2019 Apple Inc. All rights reserved.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
+ *
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ *
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ *
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
+ */
/*
* ucode.c
*
#include <i386/proc_reg.h>
#include <i386/cpuid.h>
#include <vm/vm_kern.h>
-#include <i386/mp.h> // mp_broadcast
+#include <i386/mp.h> // mp_cpus_call
+#include <i386/commpage/commpage.h>
+#include <i386/fpu.h>
#include <machine/cpu_number.h> // cpu_number
+#include <pexpert/pexpert.h> // boot-args
#define IA32_BIOS_UPDT_TRIG (0x79) /* microcode update trigger MSR */
register_locks(void)
{
/* already allocated? */
- if (ucode_slock_grp_attr && ucode_slock_grp && ucode_slock_attr && ucode_slock)
+ if (ucode_slock_grp_attr && ucode_slock_grp && ucode_slock_attr && ucode_slock) {
return KERN_SUCCESS;
+ }
/* allocate lock group attribute and group */
- if (!(ucode_slock_grp_attr = lck_grp_attr_alloc_init()))
+ if (!(ucode_slock_grp_attr = lck_grp_attr_alloc_init())) {
goto nomem_out;
+ }
- lck_grp_attr_setstat(ucode_slock_grp_attr);
-
- if (!(ucode_slock_grp = lck_grp_alloc_init("uccode_lock", ucode_slock_grp_attr)))
+ if (!(ucode_slock_grp = lck_grp_alloc_init("uccode_lock", ucode_slock_grp_attr))) {
goto nomem_out;
+ }
/* Allocate lock attribute */
- if (!(ucode_slock_attr = lck_attr_alloc_init()))
+ if (!(ucode_slock_attr = lck_attr_alloc_init())) {
goto nomem_out;
+ }
/* Allocate the spin lock */
/* We keep one global spin-lock. We could have one per update
* request... but srsly, why would you update microcode like that?
*/
- if (!(ucode_slock = lck_spin_alloc_init(ucode_slock_grp, ucode_slock_attr)))
+ if (!(ucode_slock = lck_spin_alloc_init(ucode_slock_grp, ucode_slock_attr))) {
goto nomem_out;
+ }
return KERN_SUCCESS;
nomem_out:
/* clean up */
- if (ucode_slock)
+ if (ucode_slock) {
lck_spin_free(ucode_slock, ucode_slock_grp);
- if (ucode_slock_attr)
+ }
+ if (ucode_slock_attr) {
lck_attr_free(ucode_slock_attr);
- if (ucode_slock_grp)
+ }
+ if (ucode_slock_grp) {
lck_grp_free(ucode_slock_grp);
- if (ucode_slock_grp_attr)
+ }
+ if (ucode_slock_grp_attr) {
lck_grp_attr_free(ucode_slock_grp_attr);
+ }
return KERN_NO_SPACE;
}
/* Copy in enough header to peek at the size */
error = copyin((user_addr_t)inaddr, (void *)&update_header, sizeof(update_header));
- if (error)
+ if (error) {
return error;
+ }
/* Get the actual, alleged size */
size = update_header.total_size;
/* huge bogus piece of data that somehow made it through? */
- if (size >= 1024 * 1024)
+ if (size >= 1024 * 1024) {
return ENOMEM;
+ }
/* Old microcodes? */
- if (size == 0)
+ if (size == 0) {
size = 2048; /* default update size; see SDM */
-
+ }
/*
* create the buffer for the update
* It need only be aligned to 16-bytes, according to the SDM.
* This also wires it down
*/
- ret = kmem_alloc_kobject(kernel_map, (vm_offset_t *)&update, size);
- if (ret != KERN_SUCCESS)
+ ret = kmem_alloc_kobject(kernel_map, (vm_offset_t *)&update, size, VM_KERN_MEMORY_OSFMK);
+ if (ret != KERN_SUCCESS) {
return ENOMEM;
+ }
/* Copy it in */
error = copyin((user_addr_t)inaddr, (void*)update, size);
return 0;
}
+static void
+cpu_apply_microcode(void)
+{
+ /* grab the lock */
+ lck_spin_lock(ucode_slock);
+
+ /* execute the update */
+ update_microcode();
+
+ /* release the lock */
+ lck_spin_unlock(ucode_slock);
+}
+
+static void
+cpu_update(__unused void *arg)
+{
+ cpu_apply_microcode();
+
+ cpuid_do_was();
+}
+
/*
* This is called once by every CPU on a wake from sleep/hibernate
* and is meant to re-apply a microcode update that got lost
{
if (global_update) {
kprintf("ucode: Re-applying update after wake (CPU #%d)\n", cpu_number());
- update_microcode();
-#ifdef DEBUG
+ cpu_update(NULL);
+#if DEBUG
} else {
kprintf("ucode: No update to apply (CPU #%d)\n", cpu_number());
#endif
}
static void
-cpu_update(__unused void *arg)
+ucode_cpuid_set_info(void)
{
- /* grab the lock */
- lck_spin_lock(ucode_slock);
+ uint64_t saved_xcr0, dest_xcr0;
+ int need_xcr0_restore = 0;
+ boolean_t intrs_enabled = ml_set_interrupts_enabled(FALSE);
- /* execute the update */
- update_microcode();
+ /*
+ * Before we cache the CPUID information, we must configure XCR0 with the maximal set of
+ * features to ensure the save area returned in the xsave leaf is correctly-sized.
+ *
+ * Since we are guaranteed that init_fpu() has already happened, we can use state
+ * variables set there that were already predicated on the presence of explicit
+ * boot-args enables/disables.
+ */
- /* if CPU #0, update global CPU information */
- if (!cpu_number())
- cpuid_set_info();
+ if (fpu_capability == AVX512 || fpu_capability == AVX) {
+ saved_xcr0 = xgetbv(XCR0);
+ dest_xcr0 = (fpu_capability == AVX512) ? AVX512_XMASK : AVX_XMASK;
+ assert((get_cr4() & CR4_OSXSAVE) != 0);
+ if (saved_xcr0 != dest_xcr0) {
+ need_xcr0_restore = 1;
+ xsetbv(dest_xcr0 >> 32, dest_xcr0 & 0xFFFFFFFFUL);
+ }
+ }
- /* release the lock */
- lck_spin_unlock(ucode_slock);
+ cpuid_set_info();
+
+ if (need_xcr0_restore) {
+ xsetbv(saved_xcr0 >> 32, saved_xcr0 & 0xFFFFFFFFUL);
+ }
+
+ ml_set_interrupts_enabled(intrs_enabled);
}
/* Farm an update out to all CPUs */
static void
xcpu_update(void)
{
- if (register_locks() != KERN_SUCCESS)
+ cpumask_t dest_cpumask;
+
+ if (register_locks() != KERN_SUCCESS) {
return;
+ }
+
+ mp_disable_preemption();
+ dest_cpumask = CPUMASK_OTHERS;
+ cpu_apply_microcode();
+ /* Update the cpuid info */
+ ucode_cpuid_set_info();
+ mp_enable_preemption();
+
+ /* Get all other CPUs to perform the update */
+ /*
+ * Calling mp_cpus_call with the ASYNC flag ensures that the
+ * IPI dispatch occurs in parallel, but that we will not
+ * proceed until all targeted CPUs complete the microcode
+ * update.
+ */
+ mp_cpus_call(dest_cpumask, ASYNC, cpu_update, NULL);
- /* Get all CPUs to perform the update */
- mp_broadcast(cpu_update, NULL);
+ /* Update the commpage only after we update all CPUs' microcode */
+ commpage_post_ucode_update();
}
/*
ucode_interface(uint64_t addr)
{
int error;
+ char arg[16];
+
+ if (PE_parse_boot_argn("-x", arg, sizeof(arg))) {
+ printf("ucode: no updates in safe mode\n");
+ return EPERM;
+ }
#if !DEBUG
/*
* would not make sense (all updates are cumulative), and also
* leak memory, because we don't free previous updates.
*/
- if (global_update)
+ if (global_update) {
return EPERM;
+ }
#endif
/* Get the whole microcode */
error = copyin_update(addr);
- if (error)
+ if (error) {
return error;
+ }
/* Farm out the updates */
xcpu_update();