| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20231002115903.660850472@linutronix.de>
Date: Mon, 2 Oct 2023 14:00:08 +0200 (CEST)
From: Thomas Gleixner <tglx@...utronix.de>
To: LKML <linux-kernel@...r.kernel.org>
Cc: x86@...nel.org, Borislav Petkov <bp@...en8.de>,
"Chang S. Bae" <chang.seok.bae@...el.com>,
Arjan van de Ven <arjan@...ux.intel.com>,
Nikolay Borisov <nik.borisov@...e.com>
Subject: [patch V4 28/30] x86/microcode: Handle "offline" CPUs correctly
From: Thomas Gleixner <tglx@...utronix.de>
Offline CPUs need to be parked in a safe loop when microcode update is in
progress on the primary CPU. Currently offline CPUs are parked in
'mwait_play_dead()', and for Intel CPUs, its not a safe instruction, because
'mwait' instruction can be patched in the new microcode update that can
cause instability.
- Adds a new microcode state 'UCODE_OFFLINE' to report status on per-cpu
basis.
- Force NMI on the offline CPUs.
Wakeup offline CPUs while the update is in progress and then return them
back to 'mwait_play_dead()' after microcode update is complete.
Signed-off-by: Thomas Gleixner <tglx@...utronix.de>
---
arch/x86/include/asm/microcode.h | 1
arch/x86/kernel/cpu/microcode/core.c | 112 +++++++++++++++++++++++++++++--
arch/x86/kernel/cpu/microcode/internal.h | 1
arch/x86/kernel/nmi.c | 5 +
4 files changed, 113 insertions(+), 6 deletions(-)
---
--- a/arch/x86/include/asm/microcode.h
+++ b/arch/x86/include/asm/microcode.h
@@ -71,6 +71,7 @@ static inline u32 intel_get_microcode_re
#endif /* !CONFIG_CPU_SUP_INTEL */
bool microcode_nmi_handler(void);
+void microcode_offline_nmi_handler(void);
#ifdef CONFIG_MICROCODE_LATE_LOADING
DECLARE_STATIC_KEY_FALSE(microcode_nmi_handler_enable);
--- a/arch/x86/kernel/cpu/microcode/core.c
+++ b/arch/x86/kernel/cpu/microcode/core.c
@@ -301,8 +301,9 @@ struct microcode_ctrl {
DEFINE_STATIC_KEY_FALSE(microcode_nmi_handler_enable);
static DEFINE_PER_CPU(struct microcode_ctrl, ucode_ctrl);
+static atomic_t late_cpus_in, offline_in_nmi;
static unsigned int loops_per_usec;
-static atomic_t late_cpus_in;
+static cpumask_t cpu_offline_mask;
static noinstr bool wait_for_cpus(atomic_t *cnt)
{
@@ -410,7 +411,7 @@ static noinstr void load_secondary(unsig
instrumentation_end();
}
-static void load_primary(unsigned int cpu)
+static void __load_primary(unsigned int cpu)
{
struct cpumask *secondaries = topology_sibling_cpumask(cpu);
enum sibling_ctrl ctrl;
@@ -445,6 +446,67 @@ static void load_primary(unsigned int cp
}
}
+static bool kick_offline_cpus(unsigned int nr_offl)
+{
+ unsigned int cpu, timeout;
+
+ for_each_cpu(cpu, &cpu_offline_mask) {
+ /* Enable the rendevouz handler and send NMI */
+ per_cpu(ucode_ctrl.nmi_enabled, cpu) = true;
+ apic_send_nmi_to_offline_cpu(cpu);
+ }
+
+ /* Wait for them to arrive */
+ for (timeout = 0; timeout < (USEC_PER_SEC / 2); timeout++) {
+ if (atomic_read(&offline_in_nmi) == nr_offl)
+ return true;
+ udelay(1);
+ }
+ /* Let the others time out */
+ return false;
+}
+
+static void release_offline_cpus(void)
+{
+ unsigned int cpu;
+
+ for_each_cpu(cpu, &cpu_offline_mask)
+ per_cpu(ucode_ctrl.ctrl, cpu) = SCTRL_DONE;
+}
+
+static void load_primary(unsigned int cpu)
+{
+ unsigned int nr_offl = cpumask_weight(&cpu_offline_mask);
+ bool proceed = true;
+
+ /* Kick soft-offlined SMT siblings if required */
+ if (!cpu && nr_offl)
+ proceed = kick_offline_cpus(nr_offl);
+
+ /* If the soft-offlined CPUs did not respond, abort */
+ if (proceed)
+ __load_primary(cpu);
+
+ /* Unconditionally release soft-offlined SMT siblings if required */
+ if (!cpu && nr_offl)
+ release_offline_cpus();
+}
+
+/*
+ * Minimal stub rendevouz handler for soft-offlined CPUs which participate
+ * in the NMI rendevouz to protect against a concurrent NMI on affected
+ * CPUs.
+ */
+void noinstr microcode_offline_nmi_handler(void)
+{
+ if (!raw_cpu_read(ucode_ctrl.nmi_enabled))
+ return;
+ raw_cpu_write(ucode_ctrl.nmi_enabled, false);
+ raw_cpu_write(ucode_ctrl.result, UCODE_OFFLINE);
+ raw_atomic_inc(&offline_in_nmi);
+ wait_for_ctrl();
+}
+
static noinstr bool microcode_update_handler(void)
{
unsigned int cpu = raw_smp_processor_id();
@@ -500,6 +562,7 @@ static int load_cpus_stopped(void *unuse
static int load_late_stop_cpus(void)
{
unsigned int cpu, updated = 0, failed = 0, timedout = 0, siblings = 0;
+ unsigned int nr_offl, offline = 0;
int old_rev = boot_cpu_data.microcode;
struct cpuinfo_x86 prev_info;
@@ -507,6 +570,7 @@ static int load_late_stop_cpus(void)
pr_err("You should switch to early loading, if possible.\n");
atomic_set(&late_cpus_in, num_online_cpus());
+ atomic_set(&offline_in_nmi, 0);
loops_per_usec = loops_per_jiffy / (TICK_NSEC / 1000);
/*
@@ -529,6 +593,7 @@ static int load_late_stop_cpus(void)
case UCODE_UPDATED: updated++; break;
case UCODE_TIMEOUT: timedout++; break;
case UCODE_OK: siblings++; break;
+ case UCODE_OFFLINE: offline++; break;
default: failed++; break;
}
}
@@ -540,6 +605,13 @@ static int load_late_stop_cpus(void)
/* Nothing changed. */
if (!failed && !timedout)
return 0;
+
+ nr_offl = cpumask_weight(&cpu_offline_mask);
+ if (offline < nr_offl) {
+ pr_warn("%u offline siblings did not respond.\n",
+ nr_offl - atomic_read(&offline_in_nmi));
+ return -EIO;
+ }
pr_err("update failed: %u CPUs failed %u CPUs timed out\n",
failed, timedout);
return -EIO;
@@ -573,19 +645,49 @@ static int load_late_stop_cpus(void)
* modern CPUs uses MWAIT, which is also not guaranteed to be safe
* against a microcode update which affects MWAIT.
*
- * 2) Initialize the per CPU control structure
+ * As soft-offlined CPUs still react on NMIs, the SMT sibling
+ * restriction can be lifted when the vendor driver signals to use NMI
+ * for rendevouz and the APIC provides a mechanism to send an NMI to a
+ * soft-offlined CPU. The soft-offlined CPUs are then able to
+ * participate in the rendezvouz in a trivial stub handler.
+ *
+ * 2) Initialize the per CPU control structure and create a cpumask
+ * which contains "offline"; secondary threads, so they can be handled
+ * correctly by a control CPU.
*/
static bool setup_cpus(void)
{
struct microcode_ctrl ctrl = { .ctrl = SCTRL_WAIT, .result = -1, };
+ bool allow_smt_offline;
unsigned int cpu;
+ allow_smt_offline = microcode_ops->nmi_safe ||
+ (microcode_ops->use_nmi && apic->nmi_to_offline_cpu);
+
+ cpumask_clear(&cpu_offline_mask);
+
for_each_cpu_and(cpu, cpu_present_mask, &cpus_booted_once_mask) {
+ /*
+ * Offline CPUs sit in one of the play_dead() functions
+ * with interrupts disabled, but they still react on NMIs
+ * and execute arbitrary code. Also MWAIT being updated
+ * while the offline CPU sits there is not necessarily safe
+ * on all CPU variants.
+ *
+ * Mark them in the offline_cpus mask which will be handled
+ * by CPU0 later in the update process.
+ *
+ * Ensure that the primary thread is online so that it is
+ * guaranteed that all cores are updated.
+ */
if (!cpu_online(cpu)) {
- if (topology_is_primary_thread(cpu) || !microcode_ops->nmi_safe) {
- pr_err("CPU %u not online\n", cpu);
+ if (topology_is_primary_thread(cpu) || !allow_smt_offline) {
+ pr_err("CPU %u not online, loading aborted\n", cpu);
return false;
}
+ cpumask_set_cpu(cpu, &cpu_offline_mask);
+ per_cpu(ucode_ctrl, cpu) = ctrl;
+ continue;
}
/*
--- a/arch/x86/kernel/cpu/microcode/internal.h
+++ b/arch/x86/kernel/cpu/microcode/internal.h
@@ -17,6 +17,7 @@ enum ucode_state {
UCODE_NFOUND,
UCODE_ERROR,
UCODE_TIMEOUT,
+ UCODE_OFFLINE,
};
struct microcode_ops {
--- a/arch/x86/kernel/nmi.c
+++ b/arch/x86/kernel/nmi.c
@@ -502,8 +502,11 @@ DEFINE_IDTENTRY_RAW(exc_nmi)
if (IS_ENABLED(CONFIG_NMI_CHECK_CPU))
raw_atomic_long_inc(&nsp->idt_calls);
- if (IS_ENABLED(CONFIG_SMP) && arch_cpu_is_offline(smp_processor_id()))
+ if (IS_ENABLED(CONFIG_SMP) && arch_cpu_is_offline(smp_processor_id())) {
+ if (microcode_nmi_handler_enabled())
+ microcode_offline_nmi_handler();
return;
+ }
if (this_cpu_read(nmi_state) != NMI_NOT_RUNNING) {
this_cpu_write(nmi_state, NMI_LATCHED);
Powered by blists - more mailing lists