[PATCH v15 00/12] Parallel CPU bringup for x86_64

[PATCH v15 00/12] Parallel CPU bringup for x86_64

[Usama Arif]

This version includes the following changes over v14:
- use vendor instead of sev_es_active function to check if parallel_smp is to
  be done when platform has CC_ATTR_GUEST_STATE_ENCRYPT.
- Call smpboot_restore_warm_reset_vector incase any of the steps in
  native_cpu_up fail.
- Reset stale stack and kasan unpoison in bringup_cpu (Pulled into
  idle_thread_get function)
- Release the trampoline_lock a bit earlier.

Thanks,
Usama

Changes across versions:
v2: Cut it back to just INIT/SIPI/SIPI in parallel for now, nothing more
v3: Clean up x2apic patch, add MTRR optimisation, lock topology update
    in preparation for more parallelisation.
v4: Fixes to the real mode parallelisation patch spotted by SeanC, to
    avoid scribbling on initial_gs in common_cpu_up(), and to allow all
    24 bits of the physical X2APIC ID to be used. That patch still needs
    a Signed-off-by from its original author, who once claimed not to
    remember writing it at all. But now we've fixed it, hopefully he'll
    admit it now :)
v5: rebase to v6.1 and remeasure performance, disable parallel bringup
    for AMD CPUs.
v6: rebase to v6.2-rc6, disabled parallel boot on amd as a cpu bug and
    reused timer calibration for secondary CPUs.
v7: [David Woodhouse] iterate over all possible CPUs to find any existing
    cluster mask in alloc_clustermask. (patch 1/9)
    Keep parallel AMD support enabled in AMD, using APIC ID in CPUID leaf
    0x0B (for x2APIC mode) or CPUID leaf 0x01 where 8 bits are sufficient.
    Included sanity checks for APIC id from 0x0B. (patch 6/9)
    Removed patch for reusing timer calibration for secondary CPUs.
    commit message and code improvements.
v8: Fix CPU0 hotplug by setting up the initial_gs, initial_stack and
    early_gdt_descr.
    Drop trampoline lock and bail if APIC ID not found in find_cpunr.
    Code comments improved and debug prints added.
v9: Drop patch to avoid repeated saves of MTRR at boot time.
    rebased and retested at v6.2-rc8.
    added kernel doc for no_parallel_bringup and made do_parallel_bringup
    __ro_after_init.
v10: Fixed suspend/resume not working with parallel smpboot.
     rebased and retested to 6.2.
     fixed checkpatch errors.
v11: Added patches from Brian Gerst to remove the global variables initial_gs,
     initial_stack, and early_gdt_descr from the 64-bit boot code
     (https://lore.kernel.org/all/20230222221301.245890-1-brgerst@gmail.com/).
v12: Fixed compilation errors, acquire tr_lock for every stack setup in
     trampoline_64.S.
     Rearranged commits for a cleaner git history.
v13: Fix build error with CONFIG_FORCE_NR_CPUS.
     Commit message improved, typos fixed and extra comments added.
v14: Enable parallel bringup for SEV-ES guests.
v15: use vendor parallel_smp when platform has CC_ATTR_GUEST_STATE_ENCRYPT.
     Call smpboot_restore_warm_reset_vector incase any of the steps in
     native_cpu_up fail.
     Reset stale stack and kasan unpoison in bringup_cpu
     Release trampoline_lock a bit earlier.
 
Brian Gerst (3):
  x86/smpboot: Remove initial_stack on 64-bit
  x86/smpboot: Remove early_gdt_descr on 64-bit
  x86/smpboot: Remove initial_gs

David Woodhouse (9):
  x86/apic/x2apic: Allow CPU cluster_mask to be populated in parallel
  cpu/hotplug: Move idle_thread_get() to <linux/smpboot.h>
  cpu/hotplug: Add dynamic parallel bringup states before
    CPUHP_BRINGUP_CPU
  x86/smpboot: Reference count on smpboot_setup_warm_reset_vector()
  x86/smpboot: Split up native_cpu_up into separate phases and document
    them
  x86/smpboot: Support parallel startup of secondary CPUs
  x86/smpboot: Send INIT/SIPI/SIPI to secondary CPUs in parallel
  x86/smpboot: Serialize topology updates for secondary bringup
  x86/smpboot: Allow parallel bringup for SEV-ES

 .../admin-guide/kernel-parameters.txt         |   3 +
 arch/x86/coco/core.c                          |   5 +
 arch/x86/include/asm/coco.h                   |   1 +
 arch/x86/include/asm/cpu.h                    |   1 +
 arch/x86/include/asm/processor.h              |   6 +-
 arch/x86/include/asm/realmode.h               |   4 +-
 arch/x86/include/asm/sev-common.h             |   3 +
 arch/x86/include/asm/smp.h                    |  18 +-
 arch/x86/include/asm/topology.h               |   2 -
 arch/x86/kernel/acpi/sleep.c                  |  30 +-
 arch/x86/kernel/apic/apic.c                   |   2 +-
 arch/x86/kernel/apic/x2apic_cluster.c         | 126 +++---
 arch/x86/kernel/asm-offsets.c                 |   1 +
 arch/x86/kernel/cpu/common.c                  |   6 +-
 arch/x86/kernel/cpu/topology.c                |   3 +-
 arch/x86/kernel/head_64.S                     | 167 ++++++--
 arch/x86/kernel/smpboot.c                     | 372 +++++++++++++-----
 arch/x86/realmode/init.c                      |   3 +
 arch/x86/realmode/rm/trampoline_64.S          |  27 +-
 arch/x86/xen/smp_pv.c                         |   4 +-
 arch/x86/xen/xen-head.S                       |   2 +-
 include/linux/cpuhotplug.h                    |   2 +
 include/linux/smpboot.h                       |  10 +
 kernel/cpu.c                                  |  44 ++-
 kernel/smpboot.c                              |  11 +-
 kernel/smpboot.h                              |   2 -
 26 files changed, 641 insertions(+), 214 deletions(-)

-- 
2.25.1

[PATCH v15 01/12] x86/apic/x2apic: Allow CPU cluster_mask to be populated in parallel

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

Each of the sibling CPUs in a cluster uses the same clustermask. The first
CPU in a cluster will need a new clustermask allocated, while subsequent
siblings will use the same clustermask as the first.

However, the CPU being brought up cannot yet perform memory allocations
at the point that this occurs in init_x2apic_ldr().

So at present, the alloc_clustermask() function allocates a clustermask
just in case it's needed, storing it in the global cluster_hotplug_mask.
A CPU which is the first sibling of a cluster will "take" it from there
and set cluster_hotplug_mask to NULL, in order for alloc_clustermask()
to allocate a new one before bringing up the next CPU.

To facilitate parallel bringup of CPUs in future, switch to a model
where alloc_clustermask() prepopulates the clustermask in the per_cpu
data for each present CPU in the cluster in advance. All that the CPU
needs to do for itself in init_x2apic_ldr() is set its own bit in that
mask.

The 'node' and 'clusterid' members of struct cluster_mask are thus
redundant, and it can become a simple struct cpumask instead.

Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/kernel/apic/x2apic_cluster.c | 126 +++++++++++++++++---------
 1 file changed, 82 insertions(+), 44 deletions(-)

diff --git a/arch/x86/kernel/apic/x2apic_cluster.c b/arch/x86/kernel/apic/x2apic_cluster.c
index e696e22d0531..b2b2b7f3e03f 100644
--- a/arch/x86/kernel/apic/x2apic_cluster.c
+++ b/arch/x86/kernel/apic/x2apic_cluster.c
@@ -9,11 +9,7 @@
 
 #include "local.h"
 
-struct cluster_mask {
-	unsigned int	clusterid;
-	int		node;
-	struct cpumask	mask;
-};
+#define apic_cluster(apicid) ((apicid) >> 4)
 
 /*
  * __x2apic_send_IPI_mask() possibly needs to read
@@ -23,8 +19,7 @@ struct cluster_mask {
 static u32 *x86_cpu_to_logical_apicid __read_mostly;
 
 static DEFINE_PER_CPU(cpumask_var_t, ipi_mask);
-static DEFINE_PER_CPU_READ_MOSTLY(struct cluster_mask *, cluster_masks);
-static struct cluster_mask *cluster_hotplug_mask;
+static DEFINE_PER_CPU_READ_MOSTLY(struct cpumask *, cluster_masks);
 
 static int x2apic_acpi_madt_oem_check(char *oem_id, char *oem_table_id)
 {
@@ -60,10 +55,10 @@ __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 
 	/* Collapse cpus in a cluster so a single IPI per cluster is sent */
 	for_each_cpu(cpu, tmpmsk) {
-		struct cluster_mask *cmsk = per_cpu(cluster_masks, cpu);
+		struct cpumask *cmsk = per_cpu(cluster_masks, cpu);
 
 		dest = 0;
-		for_each_cpu_and(clustercpu, tmpmsk, &cmsk->mask)
+		for_each_cpu_and(clustercpu, tmpmsk, cmsk)
 			dest |= x86_cpu_to_logical_apicid[clustercpu];
 
 		if (!dest)
@@ -71,7 +66,7 @@ __x2apic_send_IPI_mask(const struct cpumask *mask, int vector, int apic_dest)
 
 		__x2apic_send_IPI_dest(dest, vector, APIC_DEST_LOGICAL);
 		/* Remove cluster CPUs from tmpmask */
-		cpumask_andnot(tmpmsk, tmpmsk, &cmsk->mask);
+		cpumask_andnot(tmpmsk, tmpmsk, cmsk);
 	}
 
 	local_irq_restore(flags);
@@ -105,55 +100,98 @@ static u32 x2apic_calc_apicid(unsigned int cpu)
 
 static void init_x2apic_ldr(void)
 {
-	struct cluster_mask *cmsk = this_cpu_read(cluster_masks);
-	u32 cluster, apicid = apic_read(APIC_LDR);
-	unsigned int cpu;
+	struct cpumask *cmsk = this_cpu_read(cluster_masks);
 
-	x86_cpu_to_logical_apicid[smp_processor_id()] = apicid;
+	BUG_ON(!cmsk);
 
-	if (cmsk)
-		goto update;
-
-	cluster = apicid >> 16;
-	for_each_online_cpu(cpu) {
-		cmsk = per_cpu(cluster_masks, cpu);
-		/* Matching cluster found. Link and update it. */
-		if (cmsk && cmsk->clusterid == cluster)
-			goto update;
+	cpumask_set_cpu(smp_processor_id(), cmsk);
+}
+
+/*
+ * As an optimisation during boot, set the cluster_mask for all present
+ * CPUs at once, to prevent each of them having to iterate over the others
+ * to find the existing cluster_mask.
+ */
+static void prefill_clustermask(struct cpumask *cmsk, unsigned int cpu, u32 cluster)
+{
+	int cpu_i;
+
+	for_each_present_cpu(cpu_i) {
+		struct cpumask **cpu_cmsk = &per_cpu(cluster_masks, cpu_i);
+		u32 apicid = apic->cpu_present_to_apicid(cpu_i);
+
+		if (apicid == BAD_APICID || cpu_i == cpu || apic_cluster(apicid) != cluster)
+			continue;
+
+		if (WARN_ON_ONCE(*cpu_cmsk == cmsk))
+			continue;
+
+		BUG_ON(*cpu_cmsk);
+		*cpu_cmsk = cmsk;
 	}
-	cmsk = cluster_hotplug_mask;
-	cmsk->clusterid = cluster;
-	cluster_hotplug_mask = NULL;
-update:
-	this_cpu_write(cluster_masks, cmsk);
-	cpumask_set_cpu(smp_processor_id(), &cmsk->mask);
 }
 
-static int alloc_clustermask(unsigned int cpu, int node)
+static int alloc_clustermask(unsigned int cpu, u32 cluster, int node)
 {
+	struct cpumask *cmsk = NULL;
+	unsigned int cpu_i;
+
+	/*
+	 * At boot time, the CPU present mask is stable. The cluster mask is
+	 * allocated for the first CPU in the cluster and propagated to all
+	 * present siblings in the cluster. If the cluster mask is already set
+	 * on entry to this function for a given CPU, there is nothing to do.
+	 */
 	if (per_cpu(cluster_masks, cpu))
 		return 0;
+
+	if (system_state < SYSTEM_RUNNING)
+		goto alloc;
+
 	/*
-	 * If a hotplug spare mask exists, check whether it's on the right
-	 * node. If not, free it and allocate a new one.
+	 * On post boot hotplug for a CPU which was not present at boot time,
+	 * iterate over all possible CPUs (even those which are not present
+	 * any more) to find any existing cluster mask.
 	 */
-	if (cluster_hotplug_mask) {
-		if (cluster_hotplug_mask->node == node)
-			return 0;
-		kfree(cluster_hotplug_mask);
+	for_each_possible_cpu(cpu_i) {
+		u32 apicid = apic->cpu_present_to_apicid(cpu_i);
+
+		if (apicid != BAD_APICID && apic_cluster(apicid) == cluster) {
+			cmsk = per_cpu(cluster_masks, cpu_i);
+			/*
+			 * If the cluster is already initialized, just store
+			 * the mask and return. There's no need to propagate.
+			 */
+			if (cmsk) {
+				per_cpu(cluster_masks, cpu) = cmsk;
+				return 0;
+			}
+		}
 	}
-
-	cluster_hotplug_mask = kzalloc_node(sizeof(*cluster_hotplug_mask),
-					    GFP_KERNEL, node);
-	if (!cluster_hotplug_mask)
+	/*
+	 * No CPU in the cluster has ever been initialized, so fall through to
+	 * the boot time code which will also populate the cluster mask for any
+	 * other CPU in the cluster which is (now) present.
+	 */
+alloc:
+	cmsk = kzalloc_node(sizeof(*cmsk), GFP_KERNEL, node);
+	if (!cmsk)
 		return -ENOMEM;
-	cluster_hotplug_mask->node = node;
+	per_cpu(cluster_masks, cpu) = cmsk;
+	prefill_clustermask(cmsk, cpu, cluster);
+
 	return 0;
 }
 
 static int x2apic_prepare_cpu(unsigned int cpu)
 {
-	if (alloc_clustermask(cpu, cpu_to_node(cpu)) < 0)
+	u32 phys_apicid = apic->cpu_present_to_apicid(cpu);
+	u32 cluster = apic_cluster(phys_apicid);
+	u32 logical_apicid = (cluster << 16) | (1 << (phys_apicid & 0xf));
+
+	x86_cpu_to_logical_apicid[cpu] = logical_apicid;
+
+	if (alloc_clustermask(cpu, cluster, cpu_to_node(cpu)) < 0)
 		return -ENOMEM;
 	if (!zalloc_cpumask_var(&per_cpu(ipi_mask, cpu), GFP_KERNEL))
 		return -ENOMEM;
@@ -162,10 +200,10 @@ static int x2apic_prepare_cpu(unsigned int cpu)
 
 static int x2apic_dead_cpu(unsigned int dead_cpu)
 {
-	struct cluster_mask *cmsk = per_cpu(cluster_masks, dead_cpu);
+	struct cpumask *cmsk = per_cpu(cluster_masks, dead_cpu);
 
 	if (cmsk)
-		cpumask_clear_cpu(dead_cpu, &cmsk->mask);
+		cpumask_clear_cpu(dead_cpu, cmsk);
 	free_cpumask_var(per_cpu(ipi_mask, dead_cpu));
 	return 0;
 }
-- 
2.25.1

[PATCH v15 02/12] cpu/hotplug: Move idle_thread_get() to <linux/smpboot.h>

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

Instead of relying purely on the special-case wrapper in bringup_cpu()
to pass the idle thread to __cpu_up(), expose idle_thread_get() so that
the architecture code can obtain it directly when necessary.

This will be useful when the existing __cpu_up() is split into multiple
phases, only *one* of which will actually need the idle thread.

If the architecture code is to register its new pre-bringup states with
the cpuhp core, having a special-case wrapper to pass extra arguments is
non-trivial and it's easier just to let the arch register its function
pointer to be invoked with the standard API.

To reduce duplication, move the shadow stack reset and kasan unpoisoning
into idle_thread_get() too.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 include/linux/smpboot.h | 10 ++++++++++
 kernel/cpu.c            | 13 +++----------
 kernel/smpboot.c        | 11 ++++++++++-
 kernel/smpboot.h        |  2 --
 4 files changed, 23 insertions(+), 13 deletions(-)

diff --git a/include/linux/smpboot.h b/include/linux/smpboot.h
index 9d1bc65d226c..df6417703e4c 100644
--- a/include/linux/smpboot.h
+++ b/include/linux/smpboot.h
@@ -5,6 +5,16 @@
 #include <linux/types.h>
 
 struct task_struct;
+
+#ifdef CONFIG_GENERIC_SMP_IDLE_THREAD
+struct task_struct *idle_thread_get(unsigned int cpu, bool unpoison);
+#else
+static inline struct task_struct *idle_thread_get(unsigned int cpu, bool unpoison)
+{
+	return NULL;
+}
+#endif
+
 /* Cookie handed to the thread_fn*/
 struct smpboot_thread_data;
 
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 6c0a92ca6bb5..6b3dccb4a888 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -31,7 +31,6 @@
 #include <linux/smpboot.h>
 #include <linux/relay.h>
 #include <linux/slab.h>
-#include <linux/scs.h>
 #include <linux/percpu-rwsem.h>
 #include <linux/cpuset.h>
 #include <linux/random.h>
@@ -588,15 +587,9 @@ static int bringup_wait_for_ap(unsigned int cpu)
 
 static int bringup_cpu(unsigned int cpu)
 {
-	struct task_struct *idle = idle_thread_get(cpu);
+	struct task_struct *idle = idle_thread_get(cpu, true);
 	int ret;
 
-	/*
-	 * Reset stale stack state from the last time this CPU was online.
-	 */
-	scs_task_reset(idle);
-	kasan_unpoison_task_stack(idle);
-
 	/*
 	 * Some architectures have to walk the irq descriptors to
 	 * setup the vector space for the cpu which comes online.
@@ -614,7 +607,7 @@ static int bringup_cpu(unsigned int cpu)
 
 static int finish_cpu(unsigned int cpu)
 {
-	struct task_struct *idle = idle_thread_get(cpu);
+	struct task_struct *idle = idle_thread_get(cpu, false);
 	struct mm_struct *mm = idle->active_mm;
 
 	/*
@@ -1378,7 +1371,7 @@ static int _cpu_up(unsigned int cpu, int tasks_frozen, enum cpuhp_state target)
 
 	if (st->state == CPUHP_OFFLINE) {
 		/* Let it fail before we try to bring the cpu up */
-		idle = idle_thread_get(cpu);
+		idle = idle_thread_get(cpu, false);
 		if (IS_ERR(idle)) {
 			ret = PTR_ERR(idle);
 			goto out;
diff --git a/kernel/smpboot.c b/kernel/smpboot.c
index 2c7396da470c..24e81c725e7b 100644
--- a/kernel/smpboot.c
+++ b/kernel/smpboot.c
@@ -11,6 +11,7 @@
 #include <linux/slab.h>
 #include <linux/sched.h>
 #include <linux/sched/task.h>
+#include <linux/scs.h>
 #include <linux/export.h>
 #include <linux/percpu.h>
 #include <linux/kthread.h>
@@ -27,12 +28,20 @@
  */
 static DEFINE_PER_CPU(struct task_struct *, idle_threads);
 
-struct task_struct *idle_thread_get(unsigned int cpu)
+struct task_struct *idle_thread_get(unsigned int cpu, bool unpoison)
 {
 	struct task_struct *tsk = per_cpu(idle_threads, cpu);
 
 	if (!tsk)
 		return ERR_PTR(-ENOMEM);
+
+	if (unpoison) {
+		/*
+		 * Reset stale stack state from last time this CPU was online.
+		 */
+		scs_task_reset(tsk);
+		kasan_unpoison_task_stack(tsk);
+	}
 	return tsk;
 }
 
diff --git a/kernel/smpboot.h b/kernel/smpboot.h
index 34dd3d7ba40b..60c609318ad6 100644
--- a/kernel/smpboot.h
+++ b/kernel/smpboot.h
@@ -5,11 +5,9 @@
 struct task_struct;
 
 #ifdef CONFIG_GENERIC_SMP_IDLE_THREAD
-struct task_struct *idle_thread_get(unsigned int cpu);
 void idle_thread_set_boot_cpu(void);
 void idle_threads_init(void);
 #else
-static inline struct task_struct *idle_thread_get(unsigned int cpu) { return NULL; }
 static inline void idle_thread_set_boot_cpu(void) { }
 static inline void idle_threads_init(void) { }
 #endif
-- 
2.25.1

[PATCH v15 03/12] cpu/hotplug: Add dynamic parallel bringup states before CPUHP_BRINGUP_CPU

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

There is often significant latency in the early stages of CPU bringup,
and time is wasted by waking each CPU (e.g. with SIPI/INIT/INIT on x86)
and then waiting for it to respond before moving on to the next.

Allow a platform to register a set of pre-bringup CPUHP states to which
each CPU can be stepped in parallel, thus absorbing some of that latency.

There is a subtlety here: even with an empty CPUHP_BP_PARALLEL_DYN step,
this means that *all* CPUs are brought through the prepare states and to
CPUHP_BP_PREPARE_DYN before any of them are taken to CPUHP_BRINGUP_CPU
and then are allowed to run for themselves to CPUHP_ONLINE.

So any combination of prepare/start calls which depend on A-B ordering
for each CPU in turn, such as the X2APIC code which used to allocate a
cluster mask 'just in case' and store it in a global variable in the
prep stage, then potentially consume that preallocated structure from
the AP and set the global pointer to NULL to be reallocated in
CPUHP_X2APIC_PREPARE for the next CPU... would explode horribly.

Any platform enabling the CPUHP_BP_PARALLEL_DYN steps must be reviewed
and tested to ensure that such issues do not exist, and the existing
behaviour of bringing CPUs to CPUHP_BP_PREPARE_DYN and then immediately
to CPUHP_BRINGUP_CPU and CPUHP_ONLINE only one at a time does not change
unless such a state is registered.

Note that the new parallel stages do *not* yet bring each AP to the
CPUHP_BRINGUP_CPU state at the same time, only to the new states which
exist before it. The final loop in bringup_nonboot_cpus() is untouched,
bringing each AP in turn from the final PARALLEL_DYN state (or all the
way from CPUHP_OFFLINE) to CPUHP_BRINGUP_CPU and then waiting for that
AP to do its own processing and reach CPUHP_ONLINE before releasing the
next.

Parallelising that part by bringing them all to CPUHP_BRINGUP_CPU
and then waiting for them all is an exercise for the future.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 include/linux/cpuhotplug.h |  2 ++
 kernel/cpu.c               | 31 +++++++++++++++++++++++++++++--
 2 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index 6c6859bfc454..e5a73ae6ccc0 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -133,6 +133,8 @@ enum cpuhp_state {
 	CPUHP_MIPS_SOC_PREPARE,
 	CPUHP_BP_PREPARE_DYN,
 	CPUHP_BP_PREPARE_DYN_END		= CPUHP_BP_PREPARE_DYN + 20,
+	CPUHP_BP_PARALLEL_DYN,
+	CPUHP_BP_PARALLEL_DYN_END		= CPUHP_BP_PARALLEL_DYN + 4,
 	CPUHP_BRINGUP_CPU,
 
 	/*
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 6b3dccb4a888..6ccc64defd47 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1497,8 +1497,30 @@ int bringup_hibernate_cpu(unsigned int sleep_cpu)
 
 void bringup_nonboot_cpus(unsigned int setup_max_cpus)
 {
+	unsigned int n = setup_max_cpus - num_online_cpus();
 	unsigned int cpu;
 
+	/*
+	 * An architecture may have registered parallel pre-bringup states to
+	 * which each CPU may be brought in parallel. For each such state,
+	 * bring N CPUs to it in turn before the final round of bringing them
+	 * online.
+	 */
+	if (n > 0) {
+		enum cpuhp_state st = CPUHP_BP_PARALLEL_DYN;
+
+		while (st <= CPUHP_BP_PARALLEL_DYN_END && cpuhp_hp_states[st].name) {
+			int i = n;
+
+			for_each_present_cpu(cpu) {
+				cpu_up(cpu, st);
+				if (!--i)
+					break;
+			}
+			st++;
+		}
+	}
+
 	for_each_present_cpu(cpu) {
 		if (num_online_cpus() >= setup_max_cpus)
 			break;
@@ -1875,6 +1897,10 @@ static int cpuhp_reserve_state(enum cpuhp_state state)
 		step = cpuhp_hp_states + CPUHP_BP_PREPARE_DYN;
 		end = CPUHP_BP_PREPARE_DYN_END;
 		break;
+	case CPUHP_BP_PARALLEL_DYN:
+		step = cpuhp_hp_states + CPUHP_BP_PARALLEL_DYN;
+		end = CPUHP_BP_PARALLEL_DYN_END;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -1899,14 +1925,15 @@ static int cpuhp_store_callbacks(enum cpuhp_state state, const char *name,
 	/*
 	 * If name is NULL, then the state gets removed.
 	 *
-	 * CPUHP_AP_ONLINE_DYN and CPUHP_BP_PREPARE_DYN are handed out on
+	 * CPUHP_AP_ONLINE_DYN and CPUHP_BP_P*_DYN are handed out on
 	 * the first allocation from these dynamic ranges, so the removal
 	 * would trigger a new allocation and clear the wrong (already
 	 * empty) state, leaving the callbacks of the to be cleared state
 	 * dangling, which causes wreckage on the next hotplug operation.
 	 */
 	if (name && (state == CPUHP_AP_ONLINE_DYN ||
-		     state == CPUHP_BP_PREPARE_DYN)) {
+		     state == CPUHP_BP_PREPARE_DYN ||
+		     state == CPUHP_BP_PARALLEL_DYN)) {
 		ret = cpuhp_reserve_state(state);
 		if (ret < 0)
 			return ret;
-- 
2.25.1

[PATCH v15 04/12] x86/smpboot: Reference count on smpboot_setup_warm_reset_vector()

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

When bringing up a secondary CPU from do_boot_cpu(), the warm reset flag
is set in CMOS and the starting IP for the trampoline written inside the
BDA at 0x467. Once the CPU is running, the CMOS flag is unset and the
value in the BDA cleared.

To allow for parallel bringup of CPUs, add a reference count to track the
number of CPUs currently bring brought up, and clear the state only when
the count reaches zero.

Since the RTC spinlock is required to write to the CMOS, it can be used
for mutual exclusion on the refcount too.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/kernel/smpboot.c | 19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 55cad72715d9..3a793772a2aa 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -121,17 +121,20 @@ int arch_update_cpu_topology(void)
 	return retval;
 }
 
+
+static unsigned int smpboot_warm_reset_vector_count;
+
 static inline void smpboot_setup_warm_reset_vector(unsigned long start_eip)
 {
 	unsigned long flags;
 
 	spin_lock_irqsave(&rtc_lock, flags);
-	CMOS_WRITE(0xa, 0xf);
+	if (!smpboot_warm_reset_vector_count++) {
+		CMOS_WRITE(0xa, 0xf);
+		*((volatile unsigned short *)phys_to_virt(TRAMPOLINE_PHYS_HIGH)) = start_eip >> 4;
+		*((volatile unsigned short *)phys_to_virt(TRAMPOLINE_PHYS_LOW)) = start_eip & 0xf;
+	}
 	spin_unlock_irqrestore(&rtc_lock, flags);
-	*((volatile unsigned short *)phys_to_virt(TRAMPOLINE_PHYS_HIGH)) =
-							start_eip >> 4;
-	*((volatile unsigned short *)phys_to_virt(TRAMPOLINE_PHYS_LOW)) =
-							start_eip & 0xf;
 }
 
 static inline void smpboot_restore_warm_reset_vector(void)
@@ -143,10 +146,12 @@ static inline void smpboot_restore_warm_reset_vector(void)
 	 * to default values.
 	 */
 	spin_lock_irqsave(&rtc_lock, flags);
-	CMOS_WRITE(0, 0xf);
+	if (!--smpboot_warm_reset_vector_count) {
+		CMOS_WRITE(0, 0xf);
+		*((volatile u32 *)phys_to_virt(TRAMPOLINE_PHYS_LOW)) = 0;
+	}
 	spin_unlock_irqrestore(&rtc_lock, flags);
 
-	*((volatile u32 *)phys_to_virt(TRAMPOLINE_PHYS_LOW)) = 0;
 }
 
 /*
-- 
2.25.1

[PATCH v15 05/12] x86/smpboot: Split up native_cpu_up into separate phases and document them

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

There are four logical parts to what native_cpu_up() does on the BSP (or
on the controlling CPU for a later hotplug):

 1) Wake the AP by sending the INIT/SIPI/SIPI sequence.

 2) Wait for the AP to make it as far as wait_for_master_cpu() which
    sets that CPU's bit in cpu_initialized_mask, then sets the bit in
    cpu_callout_mask to let the AP proceed through cpu_init().

 3) Wait for the AP to finish cpu_init() and get as far as the
    smp_callin() call, which sets that CPU's bit in cpu_callin_mask.

 4) Perform the TSC synchronization and wait for the AP to actually
    mark itself online in cpu_online_mask.

In preparation to allow these phases to operate in parallel on multiple
APs, split them out into separate functions and document the interactions
a little more clearly in both the BSP and AP code paths.

No functional change intended.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/kernel/smpboot.c | 179 ++++++++++++++++++++++++++------------
 1 file changed, 125 insertions(+), 54 deletions(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 3a793772a2aa..fa1e35b0747c 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -204,6 +204,10 @@ static void smp_callin(void)
 
 	wmb();
 
+	/*
+	 * This runs the AP through all the cpuhp states to its target
+	 * state (CPUHP_ONLINE in the case of serial bringup).
+	 */
 	notify_cpu_starting(cpuid);
 
 	/*
@@ -231,17 +235,32 @@ static void notrace start_secondary(void *unused)
 	load_cr3(swapper_pg_dir);
 	__flush_tlb_all();
 #endif
+	/*
+	 * Sync point with do_wait_cpu_initialized(). Before proceeding through
+	 * cpu_init(), the AP will call wait_for_master_cpu() which sets its
+	 * own bit in cpu_initialized_mask and then waits for the BSP to set
+	 * its bit in cpu_callout_mask to release it.
+	 */
 	cpu_init_secondary();
 	rcu_cpu_starting(raw_smp_processor_id());
 	x86_cpuinit.early_percpu_clock_init();
+
+	/*
+	 * Sync point with do_wait_cpu_callin(). The AP doesn't wait here
+	 * but just sets the bit to let the controlling CPU (BSP) know that
+	 * it's got this far.
+	 */
 	smp_callin();
 
 	enable_start_cpu0 = 0;
 
 	/* otherwise gcc will move up smp_processor_id before the cpu_init */
 	barrier();
+
 	/*
-	 * Check TSC synchronization with the boot CPU:
+	 * Check TSC synchronization with the boot CPU (or whichever CPU
+	 * is controlling the bringup). It will do its part of this from
+	 * do_wait_cpu_online(), making it an implicit sync point.
 	 */
 	check_tsc_sync_target();
 
@@ -254,6 +273,7 @@ static void notrace start_secondary(void *unused)
 	 * half valid vector space.
 	 */
 	lock_vector_lock();
+	/* Sync point with do_wait_cpu_online() */
 	set_cpu_online(smp_processor_id(), true);
 	lapic_online();
 	unlock_vector_lock();
@@ -1083,7 +1103,6 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle,
 	unsigned long start_ip = real_mode_header->trampoline_start;
 
 	unsigned long boot_error = 0;
-	unsigned long timeout;
 
 #ifdef CONFIG_X86_64
 	/* If 64-bit wakeup method exists, use the 64-bit mode trampoline IP */
@@ -1144,55 +1163,94 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle,
 		boot_error = wakeup_cpu_via_init_nmi(cpu, start_ip, apicid,
 						     cpu0_nmi_registered);
 
-	if (!boot_error) {
-		/*
-		 * Wait 10s total for first sign of life from AP
-		 */
-		boot_error = -1;
-		timeout = jiffies + 10*HZ;
-		while (time_before(jiffies, timeout)) {
-			if (cpumask_test_cpu(cpu, cpu_initialized_mask)) {
-				/*
-				 * Tell AP to proceed with initialization
-				 */
-				cpumask_set_cpu(cpu, cpu_callout_mask);
-				boot_error = 0;
-				break;
-			}
-			schedule();
-		}
-	}
+	return boot_error;
+}
 
-	if (!boot_error) {
-		/*
-		 * Wait till AP completes initial initialization
-		 */
-		while (!cpumask_test_cpu(cpu, cpu_callin_mask)) {
-			/*
-			 * Allow other tasks to run while we wait for the
-			 * AP to come online. This also gives a chance
-			 * for the MTRR work(triggered by the AP coming online)
-			 * to be completed in the stop machine context.
-			 */
-			schedule();
-		}
+static int do_wait_cpu_cpumask(unsigned int cpu, const struct cpumask *mask)
+{
+	unsigned long timeout;
+
+	/*
+	 * Wait up to 10s for the CPU to report in.
+	 */
+	timeout = jiffies + 10*HZ;
+	while (time_before(jiffies, timeout)) {
+		if (cpumask_test_cpu(cpu, mask))
+			return 0;
+
+		schedule();
 	}
+	return -1;
+}
 
-	if (x86_platform.legacy.warm_reset) {
-		/*
-		 * Cleanup possible dangling ends...
-		 */
-		smpboot_restore_warm_reset_vector();
+/*
+ * Bringup step two: Wait for the target AP to reach cpu_init_secondary()
+ * and thus wait_for_master_cpu(), then set cpu_callout_mask to allow it
+ * to proceed.  The AP will then proceed past setting its 'callin' bit
+ * and end up waiting in check_tsc_sync_target() until we reach
+ * do_wait_cpu_online() to tend to it.
+ */
+static int do_wait_cpu_initialized(unsigned int cpu)
+{
+	/*
+	 * Wait for first sign of life from AP.
+	 */
+	if (do_wait_cpu_cpumask(cpu, cpu_initialized_mask))
+		return -1;
+
+	cpumask_set_cpu(cpu, cpu_callout_mask);
+	return 0;
+}
+
+/*
+ * Bringup step three: Wait for the target AP to reach smp_callin().
+ * The AP is not waiting for us here so we don't need to parallelise
+ * this step. Not entirely clear why we care about this, since we just
+ * proceed directly to TSC synchronization which is the next sync
+ * point with the AP anyway.
+ */
+static int do_wait_cpu_callin(unsigned int cpu)
+{
+	/*
+	 * Wait till AP completes initial initialization.
+	 */
+	return do_wait_cpu_cpumask(cpu, cpu_callin_mask);
+}
+
+/*
+ * Bringup step four: Synchronize the TSC and wait for the target AP
+ * to reach set_cpu_online() in start_secondary().
+ */
+static int do_wait_cpu_online(unsigned int cpu)
+{
+	unsigned long flags;
+
+	/*
+	 * Check TSC synchronization with the AP (keep irqs disabled
+	 * while doing so):
+	 */
+	local_irq_save(flags);
+	check_tsc_sync_source(cpu);
+	local_irq_restore(flags);
+
+	/*
+	 * Wait for the AP to mark itself online. Not entirely
+	 * clear why we care, since the generic cpuhp code will
+	 * wait for it to each CPUHP_AP_ONLINE_IDLE before going
+	 * ahead with the rest of the bringup anyway.
+	 */
+	while (!cpu_online(cpu)) {
+		cpu_relax();
+		touch_nmi_watchdog();
 	}
 
-	return boot_error;
+	return 0;
 }
 
-int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
+static int do_cpu_up(unsigned int cpu, struct task_struct *tidle)
 {
 	int apicid = apic->cpu_present_to_apicid(cpu);
 	int cpu0_nmi_registered = 0;
-	unsigned long flags;
 	int err, ret = 0;
 
 	lockdep_assert_irqs_enabled();
@@ -1239,19 +1297,6 @@ int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
 		goto unreg_nmi;
 	}
 
-	/*
-	 * Check TSC synchronization with the AP (keep irqs disabled
-	 * while doing so):
-	 */
-	local_irq_save(flags);
-	check_tsc_sync_source(cpu);
-	local_irq_restore(flags);
-
-	while (!cpu_online(cpu)) {
-		cpu_relax();
-		touch_nmi_watchdog();
-	}
-
 unreg_nmi:
 	/*
 	 * Clean up the nmi handler. Do this after the callin and callout sync
@@ -1263,6 +1308,32 @@ int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
 	return ret;
 }
 
+int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
+{
+	int ret;
+
+	ret = do_cpu_up(cpu, tidle);
+	if (ret)
+		goto out;
+
+	ret = do_wait_cpu_initialized(cpu);
+	if (ret)
+		goto out;
+
+	ret = do_wait_cpu_callin(cpu);
+	if (ret)
+		goto out;
+
+	ret = do_wait_cpu_online(cpu);
+
+ out:
+	/* Cleanup possible dangling ends... */
+	if (x86_platform.legacy.warm_reset)
+		smpboot_restore_warm_reset_vector();
+
+	return ret;
+}
+
 /**
  * arch_disable_smp_support() - disables SMP support for x86 at runtime
  */
-- 
2.25.1

[PATCH v15 06/12] x86/smpboot: Remove initial_stack on 64-bit

[Usama Arif]

From: Brian Gerst <brgerst@gmail.com>

In order to facilitate parallel startup, start to eliminate some of the
global variables passing information to CPUs in the startup path.

However, start by introducing one more: smpboot_control. For now this
merely holds the CPU# of the CPU which is coming up. Each CPU can then
find its own per-cpu data, and everything else it needs can be found
from there, allowing the other global variables to be removed.

First to be removed is initial_stack. Each CPU can load %rsp from its
current_task->thread.sp instead. That is already set up with the correct
idle thread for APs. Set up the .sp field in INIT_THREAD on x86 so that
the BSP also finds a suitable stack pointer in the static per-cpu data
when coming up on first boot.

On resume from S3, the CPU needs a temporary stack because its idle task
is already active. Instead of setting initial_stack, the sleep code can
simply set its own current->thread.sp to point to the temporary stack.
Nobody else cares about ->thread.sp for a thread which is currently on
a CPU, because the true value is actually in the %rsp register. Which
is restored with the rest of the CPU context in do_suspend_lowlevel().

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Usama Arif <usama.arif@bytedance.com>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/include/asm/processor.h |  6 ++++-
 arch/x86/include/asm/smp.h       |  5 +++-
 arch/x86/kernel/acpi/sleep.c     | 20 +++++++++++++--
 arch/x86/kernel/asm-offsets.c    |  1 +
 arch/x86/kernel/head_64.S        | 43 +++++++++++++++++++++-----------
 arch/x86/kernel/smpboot.c        |  7 +++++-
 arch/x86/xen/xen-head.S          |  2 +-
 7 files changed, 63 insertions(+), 21 deletions(-)

diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 4e35c66edeb7..bdde7316e75b 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -648,7 +648,11 @@ static inline void spin_lock_prefetch(const void *x)
 #define KSTK_ESP(task)		(task_pt_regs(task)->sp)
 
 #else
-#define INIT_THREAD { }
+extern unsigned long __end_init_task[];
+
+#define INIT_THREAD {							    \
+	.sp	= (unsigned long)&__end_init_task - sizeof(struct pt_regs), \
+}
 
 extern unsigned long KSTK_ESP(struct task_struct *task);
 
diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index b4dbb20dab1a..bf2c51df9e0b 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -199,5 +199,8 @@ extern void nmi_selftest(void);
 #define nmi_selftest() do { } while (0)
 #endif
 
-#endif /* __ASSEMBLY__ */
+extern unsigned int smpboot_control;
+
+#endif /* !__ASSEMBLY__ */
+
 #endif /* _ASM_X86_SMP_H */
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 3b7f4cdbf2e0..1b4c43d0819a 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -111,13 +111,29 @@ int x86_acpi_suspend_lowlevel(void)
 	saved_magic = 0x12345678;
 #else /* CONFIG_64BIT */
 #ifdef CONFIG_SMP
-	initial_stack = (unsigned long)temp_stack + sizeof(temp_stack);
+	/*
+	 * As each CPU starts up, it will find its own stack pointer
+	 * from its current_task->thread.sp. Typically that will be
+	 * the idle thread for a newly-started AP, or even the boot
+	 * CPU which will find it set to &init_task in the static
+	 * per-cpu data.
+	 *
+	 * Make the resuming CPU use the temporary stack at startup
+	 * by setting current->thread.sp to point to that. The true
+	 * %rsp will be restored with the rest of the CPU context,
+	 * by do_suspend_lowlevel(). And unwinders don't care about
+	 * the abuse of ->thread.sp because it's a dead variable
+	 * while the thread is running on the CPU anyway; the true
+	 * value is in the actual %rsp register.
+	 */
+	current->thread.sp = (unsigned long)temp_stack + sizeof(temp_stack);
 	early_gdt_descr.address =
 			(unsigned long)get_cpu_gdt_rw(smp_processor_id());
 	initial_gs = per_cpu_offset(smp_processor_id());
+	smpboot_control = smp_processor_id();
 #endif
 	initial_code = (unsigned long)wakeup_long64;
-       saved_magic = 0x123456789abcdef0L;
+	saved_magic = 0x123456789abcdef0L;
 #endif /* CONFIG_64BIT */
 
 	/*
diff --git a/arch/x86/kernel/asm-offsets.c b/arch/x86/kernel/asm-offsets.c
index 82c783da16a8..797ae1a15c91 100644
--- a/arch/x86/kernel/asm-offsets.c
+++ b/arch/x86/kernel/asm-offsets.c
@@ -108,6 +108,7 @@ static void __used common(void)
 	OFFSET(TSS_sp1, tss_struct, x86_tss.sp1);
 	OFFSET(TSS_sp2, tss_struct, x86_tss.sp2);
 	OFFSET(X86_top_of_stack, pcpu_hot, top_of_stack);
+	OFFSET(X86_current_task, pcpu_hot, current_task);
 #ifdef CONFIG_CALL_DEPTH_TRACKING
 	OFFSET(X86_call_depth, pcpu_hot, call_depth);
 #endif
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 222efd4a09bc..cc1b145055ac 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -61,8 +61,8 @@ SYM_CODE_START_NOALIGN(startup_64)
 	 * tables and then reload them.
 	 */
 
-	/* Set up the stack for verify_cpu(), similar to initial_stack below */
-	leaq	(__end_init_task - FRAME_SIZE)(%rip), %rsp
+	/* Set up the stack for verify_cpu() */
+	leaq	(__end_init_task - PTREGS_SIZE)(%rip), %rsp
 
 	leaq	_text(%rip), %rdi
 
@@ -241,6 +241,24 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	UNWIND_HINT_EMPTY
 	ANNOTATE_NOENDBR // above
 
+#ifdef CONFIG_SMP
+	movl	smpboot_control(%rip), %ecx
+
+	/* Get the per cpu offset for the given CPU# which is in ECX */
+	movq	__per_cpu_offset(,%rcx,8), %rdx
+#else
+	xorl	%edx, %edx /* zero-extended to clear all of RDX */
+#endif /* CONFIG_SMP */
+
+	/*
+	 * Setup a boot time stack - Any secondary CPU will have lost its stack
+	 * by now because the cr3-switch above unmaps the real-mode stack.
+	 *
+	 * RDX contains the per-cpu offset
+	 */
+	movq	pcpu_hot + X86_current_task(%rdx), %rax
+	movq	TASK_threadsp(%rax), %rsp
+
 	/*
 	 * We must switch to a new descriptor in kernel space for the GDT
 	 * because soon the kernel won't have access anymore to the userspace
@@ -275,12 +293,6 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	movl	initial_gs+4(%rip),%edx
 	wrmsr
 
-	/*
-	 * Setup a boot time stack - Any secondary CPU will have lost its stack
-	 * by now because the cr3-switch above unmaps the real-mode stack
-	 */
-	movq initial_stack(%rip), %rsp
-
 	/* Setup and Load IDT */
 	pushq	%rsi
 	call	early_setup_idt
@@ -372,7 +384,11 @@ SYM_CODE_END(secondary_startup_64)
 SYM_CODE_START(start_cpu0)
 	ANNOTATE_NOENDBR
 	UNWIND_HINT_EMPTY
-	movq	initial_stack(%rip), %rsp
+
+	/* Find the idle task stack */
+	movq	PER_CPU_VAR(pcpu_hot) + X86_current_task, %rcx
+	movq	TASK_threadsp(%rcx), %rsp
+
 	jmp	.Ljump_to_C_code
 SYM_CODE_END(start_cpu0)
 #endif
@@ -420,12 +436,6 @@ SYM_DATA(initial_gs,	.quad INIT_PER_CPU_VAR(fixed_percpu_data))
 #ifdef CONFIG_AMD_MEM_ENCRYPT
 SYM_DATA(initial_vc_handler,	.quad handle_vc_boot_ghcb)
 #endif
-
-/*
- * The FRAME_SIZE gap is a convention which helps the in-kernel unwinder
- * reliably detect the end of the stack.
- */
-SYM_DATA(initial_stack, .quad init_thread_union + THREAD_SIZE - FRAME_SIZE)
 	__FINITDATA
 
 	__INIT
@@ -660,6 +670,9 @@ SYM_DATA_END(level1_fixmap_pgt)
 SYM_DATA(early_gdt_descr,		.word GDT_ENTRIES*8-1)
 SYM_DATA_LOCAL(early_gdt_descr_base,	.quad INIT_PER_CPU_VAR(gdt_page))
 
+	.align 16
+SYM_DATA(smpboot_control,		.long 0)
+
 	.align 16
 /* This must match the first entry in level2_kernel_pgt */
 SYM_DATA(phys_base, .quad 0x0)
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index fa1e35b0747c..ab8decac6294 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -1112,7 +1112,12 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle,
 	idle->thread.sp = (unsigned long)task_pt_regs(idle);
 	early_gdt_descr.address = (unsigned long)get_cpu_gdt_rw(cpu);
 	initial_code = (unsigned long)start_secondary;
-	initial_stack  = idle->thread.sp;
+
+	if (IS_ENABLED(CONFIG_X86_32)) {
+		initial_stack  = idle->thread.sp;
+	} else {
+		smpboot_control = cpu;
+	}
 
 	/* Enable the espfix hack for this CPU */
 	init_espfix_ap(cpu);
diff --git a/arch/x86/xen/xen-head.S b/arch/x86/xen/xen-head.S
index ffaa62167f6e..6bd391476656 100644
--- a/arch/x86/xen/xen-head.S
+++ b/arch/x86/xen/xen-head.S
@@ -49,7 +49,7 @@ SYM_CODE_START(startup_xen)
 	ANNOTATE_NOENDBR
 	cld
 
-	mov initial_stack(%rip), %rsp
+	leaq	(__end_init_task - PTREGS_SIZE)(%rip), %rsp
 
 	/* Set up %gs.
 	 *
-- 
2.25.1

[PATCH v15 07/12] x86/smpboot: Remove early_gdt_descr on 64-bit

[Usama Arif]

From: Brian Gerst <brgerst@gmail.com>

Build the GDT descriptor on the stack instead.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Usama Arif <usama.arif@bytedance.com>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/kernel/acpi/sleep.c |  2 --
 arch/x86/kernel/head_64.S    | 11 ++++++-----
 arch/x86/kernel/smpboot.c    |  2 +-
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 1b4c43d0819a..de89bb4719d0 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -127,8 +127,6 @@ int x86_acpi_suspend_lowlevel(void)
 	 * value is in the actual %rsp register.
 	 */
 	current->thread.sp = (unsigned long)temp_stack + sizeof(temp_stack);
-	early_gdt_descr.address =
-			(unsigned long)get_cpu_gdt_rw(smp_processor_id());
 	initial_gs = per_cpu_offset(smp_processor_id());
 	smpboot_control = smp_processor_id();
 #endif
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index cc1b145055ac..a5b46c2fba05 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -265,7 +265,12 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	 * addresses where we're currently running on. We have to do that here
 	 * because in 32bit we couldn't load a 64bit linear address.
 	 */
-	lgdt	early_gdt_descr(%rip)
+	subq	$16, %rsp
+	movw	$(GDT_SIZE-1), (%rsp)
+	leaq	gdt_page(%rdx), %rax
+	movq	%rax, 2(%rsp)
+	lgdt	(%rsp)
+	addq	$16, %rsp
 
 	/* set up data segments */
 	xorl %eax,%eax
@@ -667,10 +672,6 @@ SYM_DATA_END(level1_fixmap_pgt)
 	.data
 	.align 16
 
-SYM_DATA(early_gdt_descr,		.word GDT_ENTRIES*8-1)
-SYM_DATA_LOCAL(early_gdt_descr_base,	.quad INIT_PER_CPU_VAR(gdt_page))
-
-	.align 16
 SYM_DATA(smpboot_control,		.long 0)
 
 	.align 16
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index ab8decac6294..67224e61310c 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -1110,10 +1110,10 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle,
 		start_ip = real_mode_header->trampoline_start64;
 #endif
 	idle->thread.sp = (unsigned long)task_pt_regs(idle);
-	early_gdt_descr.address = (unsigned long)get_cpu_gdt_rw(cpu);
 	initial_code = (unsigned long)start_secondary;
 
 	if (IS_ENABLED(CONFIG_X86_32)) {
+		early_gdt_descr.address = (unsigned long)get_cpu_gdt_rw(cpu);
 		initial_stack  = idle->thread.sp;
 	} else {
 		smpboot_control = cpu;
-- 
2.25.1

[PATCH v15 08/12] x86/smpboot: Remove initial_gs

[Usama Arif]

From: Brian Gerst <brgerst@gmail.com>

Given its CPU#, each CPU can find its own per-cpu offset, and directly set
GSBASE accordingly. The global variable can be eliminated.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Usama Arif <usama.arif@bytedance.com>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/include/asm/realmode.h |  1 -
 arch/x86/kernel/acpi/sleep.c    |  1 -
 arch/x86/kernel/head_64.S       | 22 ++++++++--------------
 arch/x86/kernel/smpboot.c       |  2 --
 4 files changed, 8 insertions(+), 18 deletions(-)

diff --git a/arch/x86/include/asm/realmode.h b/arch/x86/include/asm/realmode.h
index a336feef0af1..f6a1737c77be 100644
--- a/arch/x86/include/asm/realmode.h
+++ b/arch/x86/include/asm/realmode.h
@@ -59,7 +59,6 @@ extern struct real_mode_header *real_mode_header;
 extern unsigned char real_mode_blob_end[];
 
 extern unsigned long initial_code;
-extern unsigned long initial_gs;
 extern unsigned long initial_stack;
 #ifdef CONFIG_AMD_MEM_ENCRYPT
 extern unsigned long initial_vc_handler;
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index de89bb4719d0..1328c221af30 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -127,7 +127,6 @@ int x86_acpi_suspend_lowlevel(void)
 	 * value is in the actual %rsp register.
 	 */
 	current->thread.sp = (unsigned long)temp_stack + sizeof(temp_stack);
-	initial_gs = per_cpu_offset(smp_processor_id());
 	smpboot_control = smp_processor_id();
 #endif
 	initial_code = (unsigned long)wakeup_long64;
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index a5b46c2fba05..6a8238702eab 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -66,18 +66,10 @@ SYM_CODE_START_NOALIGN(startup_64)
 
 	leaq	_text(%rip), %rdi
 
-	/*
-	 * initial_gs points to initial fixed_percpu_data struct with storage for
-	 * the stack protector canary. Global pointer fixups are needed at this
-	 * stage, so apply them as is done in fixup_pointer(), and initialize %gs
-	 * such that the canary can be accessed at %gs:40 for subsequent C calls.
-	 */
+	/* Setup GSBASE to allow stack canary access for C code */
 	movl	$MSR_GS_BASE, %ecx
-	movq	initial_gs(%rip), %rax
-	movq	$_text, %rdx
-	subq	%rdx, %rax
-	addq	%rdi, %rax
-	movq	%rax, %rdx
+	leaq	INIT_PER_CPU_VAR(fixed_percpu_data)(%rip), %rdx
+	movl	%edx, %eax
 	shrq	$32,  %rdx
 	wrmsr
 
@@ -294,8 +286,11 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	 * the per cpu areas are set up.
 	 */
 	movl	$MSR_GS_BASE,%ecx
-	movl	initial_gs(%rip),%eax
-	movl	initial_gs+4(%rip),%edx
+#ifndef CONFIG_SMP
+	leaq	INIT_PER_CPU_VAR(fixed_percpu_data)(%rip), %rdx
+#endif
+	movl	%edx, %eax
+	shrq	$32, %rdx
 	wrmsr
 
 	/* Setup and Load IDT */
@@ -437,7 +432,6 @@ SYM_CODE_END(vc_boot_ghcb)
 	__REFDATA
 	.balign	8
 SYM_DATA(initial_code,	.quad x86_64_start_kernel)
-SYM_DATA(initial_gs,	.quad INIT_PER_CPU_VAR(fixed_percpu_data))
 #ifdef CONFIG_AMD_MEM_ENCRYPT
 SYM_DATA(initial_vc_handler,	.quad handle_vc_boot_ghcb)
 #endif
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 67224e61310c..28d1643eee99 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -1084,8 +1084,6 @@ int common_cpu_up(unsigned int cpu, struct task_struct *idle)
 #ifdef CONFIG_X86_32
 	/* Stack for startup_32 can be just as for start_secondary onwards */
 	per_cpu(pcpu_hot.top_of_stack, cpu) = task_top_of_stack(idle);
-#else
-	initial_gs = per_cpu_offset(cpu);
 #endif
 	return 0;
 }
-- 
2.25.1

[PATCH v15 09/12] x86/smpboot: Support parallel startup of secondary CPUs

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

Rework the real-mode startup code to allow for APs to be brought up in
parallel. This is in two parts:

1. Introduce a bit-spinlock to prevent them from all using the real
   mode stack at the same time.

2. Avoid needing to use the global smpboot_control variable to pass
   each AP its CPU#.

To achieve the latter, export the cpuid_to_apicid[] array so that each
AP can find its own CPU# by searching therein based on its APIC ID.

Introduce flags in the top bits of smpboot_control which indicate methods
by which an AP should find its CPU#. For a serialized bringup, the CPU#
is explicitly passed in the low bits of smpboot_control as before. For
parallel mode there are flags directing the AP to find its APIC ID in
CPUID leaf 0x0b (for X2APIC mode) or CPUID leaf 0x01 where 8 bits are
sufficient, then perform the cpuid_to_apicid[] lookup with that.

Parallel startup may be disabled by a command line option, and also if:
 • AMD SEV-ES is in use, since the AP may not use CPUID that early.
 • X2APIC is enabled, but CPUID leaf 0xb is not present and correct.
 • X2APIC is not enabled but not even CPUID leaf 0x01 exists.

Aside from the fact that APs will now look up their CPU# via the
newly-exported cpuid_to_apicid[] table, there is no behavioural change
intended yet, since new parallel CPUHP states have not — yet — been
added.

[ tglx: Initial proof of concept patch with bitlock and APIC ID lookup ]
[ dwmw2: Rework and testing, commit message, CPUID 0x1 and CPU0 support ]
[ seanc: Fix stray override of initial_gs in common_cpu_up() ]
[ Oleksandr Natalenko: reported suspend/resume issue fixed in
  x86_acpi_suspend_lowlevel ]
Co-developed-by: Thomas Gleixner <tglx@linutronix.de>
Co-developed-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 .../admin-guide/kernel-parameters.txt         |  3 +
 arch/x86/include/asm/cpu.h                    |  1 +
 arch/x86/include/asm/realmode.h               |  3 +
 arch/x86/include/asm/smp.h                    |  6 ++
 arch/x86/kernel/acpi/sleep.c                  |  9 ++-
 arch/x86/kernel/apic/apic.c                   |  2 +-
 arch/x86/kernel/cpu/topology.c                |  3 +-
 arch/x86/kernel/head_64.S                     | 65 +++++++++++++++++++
 arch/x86/kernel/smpboot.c                     | 50 +++++++++++++-
 arch/x86/realmode/init.c                      |  3 +
 arch/x86/realmode/rm/trampoline_64.S          | 27 ++++++--
 11 files changed, 163 insertions(+), 9 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 6cfa6e3996cf..7bb7020f97e2 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3819,6 +3819,9 @@
 
 	nomodule	Disable module load
 
+	no_parallel_bringup
+			[X86,SMP] Disable parallel bring-up of secondary cores.
+
 	nopat		[X86] Disable PAT (page attribute table extension of
 			pagetables) support.
 
diff --git a/arch/x86/include/asm/cpu.h b/arch/x86/include/asm/cpu.h
index 78796b98a544..ef8ba318dca1 100644
--- a/arch/x86/include/asm/cpu.h
+++ b/arch/x86/include/asm/cpu.h
@@ -97,5 +97,6 @@ static inline bool intel_cpu_signatures_match(unsigned int s1, unsigned int p1,
 extern u64 x86_read_arch_cap_msr(void);
 int intel_find_matching_signature(void *mc, unsigned int csig, int cpf);
 int intel_microcode_sanity_check(void *mc, bool print_err, int hdr_type);
+int check_extended_topology_leaf(int leaf);
 
 #endif /* _ASM_X86_CPU_H */
diff --git a/arch/x86/include/asm/realmode.h b/arch/x86/include/asm/realmode.h
index f6a1737c77be..87e5482acd0d 100644
--- a/arch/x86/include/asm/realmode.h
+++ b/arch/x86/include/asm/realmode.h
@@ -52,6 +52,7 @@ struct trampoline_header {
 	u64 efer;
 	u32 cr4;
 	u32 flags;
+	u32 lock;
 #endif
 };
 
@@ -64,6 +65,8 @@ extern unsigned long initial_stack;
 extern unsigned long initial_vc_handler;
 #endif
 
+extern u32 *trampoline_lock;
+
 extern unsigned char real_mode_blob[];
 extern unsigned char real_mode_relocs[];
 
diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index bf2c51df9e0b..1cf4f1e57570 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -203,4 +203,10 @@ extern unsigned int smpboot_control;
 
 #endif /* !__ASSEMBLY__ */
 
+/* Control bits for startup_64 */
+#define STARTUP_APICID_CPUID_0B	0x80000000
+#define STARTUP_APICID_CPUID_01	0x40000000
+
+#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | STARTUP_APICID_CPUID_0B)
+
 #endif /* _ASM_X86_SMP_H */
diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
index 1328c221af30..6dfecb27b846 100644
--- a/arch/x86/kernel/acpi/sleep.c
+++ b/arch/x86/kernel/acpi/sleep.c
@@ -16,6 +16,7 @@
 #include <asm/cacheflush.h>
 #include <asm/realmode.h>
 #include <asm/hypervisor.h>
+#include <asm/smp.h>
 
 #include <linux/ftrace.h>
 #include "../../realmode/rm/wakeup.h"
@@ -127,7 +128,13 @@ int x86_acpi_suspend_lowlevel(void)
 	 * value is in the actual %rsp register.
 	 */
 	current->thread.sp = (unsigned long)temp_stack + sizeof(temp_stack);
-	smpboot_control = smp_processor_id();
+	/*
+	 * Ensure the CPU knows which one it is when it comes back, if
+	 * it isn't in parallel mode and expected to work that out for
+	 * itself.
+	 */
+	if (!(smpboot_control & STARTUP_PARALLEL_MASK))
+		smpboot_control = smp_processor_id();
 #endif
 	initial_code = (unsigned long)wakeup_long64;
 	saved_magic = 0x123456789abcdef0L;
diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
index 20d9a604da7c..ac1d7e5da1f2 100644
--- a/arch/x86/kernel/apic/apic.c
+++ b/arch/x86/kernel/apic/apic.c
@@ -2377,7 +2377,7 @@ static int nr_logical_cpuids = 1;
 /*
  * Used to store mapping between logical CPU IDs and APIC IDs.
  */
-static int cpuid_to_apicid[] = {
+int cpuid_to_apicid[] = {
 	[0 ... NR_CPUS - 1] = -1,
 };
 
diff --git a/arch/x86/kernel/cpu/topology.c b/arch/x86/kernel/cpu/topology.c
index 5e868b62a7c4..4373442e500a 100644
--- a/arch/x86/kernel/cpu/topology.c
+++ b/arch/x86/kernel/cpu/topology.c
@@ -9,6 +9,7 @@
 #include <asm/apic.h>
 #include <asm/memtype.h>
 #include <asm/processor.h>
+#include <asm/cpu.h>
 
 #include "cpu.h"
 
@@ -32,7 +33,7 @@ EXPORT_SYMBOL(__max_die_per_package);
 /*
  * Check if given CPUID extended topology "leaf" is implemented
  */
-static int check_extended_topology_leaf(int leaf)
+int check_extended_topology_leaf(int leaf)
 {
 	unsigned int eax, ebx, ecx, edx;
 
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 6a8238702eab..65bca47d84a1 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -25,6 +25,7 @@
 #include <asm/export.h>
 #include <asm/nospec-branch.h>
 #include <asm/fixmap.h>
+#include <asm/smp.h>
 
 /*
  * We are not able to switch in one step to the final KERNEL ADDRESS SPACE
@@ -234,8 +235,61 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	ANNOTATE_NOENDBR // above
 
 #ifdef CONFIG_SMP
+	/*
+	 * For parallel boot, the APIC ID is retrieved from CPUID, and then
+	 * used to look up the CPU number.  For booting a single CPU, the
+	 * CPU number is encoded in smpboot_control.
+	 *
+	 * Bit 31	STARTUP_APICID_CPUID_0B flag (use CPUID 0x0b)
+	 * Bit 30	STARTUP_APICID_CPUID_01 flag (use CPUID 0x01)
+	 * Bit 0-24	CPU# if STARTUP_APICID_CPUID_xx flags are not set
+	 */
 	movl	smpboot_control(%rip), %ecx
+	testl	$STARTUP_APICID_CPUID_0B, %ecx
+	jnz	.Luse_cpuid_0b
+	testl	$STARTUP_APICID_CPUID_01, %ecx
+	jnz	.Luse_cpuid_01
+	andl	$0x0FFFFFFF, %ecx
+	jmp	.Lsetup_cpu
+
+.Luse_cpuid_01:
+	mov	$0x01, %eax
+	cpuid
+	mov	%ebx, %edx
+	shr	$24, %edx
+	jmp	.Lsetup_AP
 
+.Luse_cpuid_0b:
+	mov	$0x0B, %eax
+	xorl	%ecx, %ecx
+	cpuid
+
+.Lsetup_AP:
+	/* EDX contains the APIC ID of the current CPU */
+	xorq	%rcx, %rcx
+	leaq	cpuid_to_apicid(%rip), %rbx
+
+.Lfind_cpunr:
+	cmpl	(%rbx,%rcx,4), %edx
+	jz	.Lsetup_cpu
+	inc	%ecx
+#ifdef CONFIG_FORCE_NR_CPUS
+	cmpl	$NR_CPUS, %ecx
+#else
+	cmpl	nr_cpu_ids(%rip), %ecx
+#endif
+	jb	.Lfind_cpunr
+
+	/*  APIC ID not found in the table. Drop the trampoline lock and bail. */
+	movq	trampoline_lock(%rip), %rax
+	lock
+	btrl	$0, (%rax)
+
+1:	cli
+	hlt
+	jmp	1b
+
+.Lsetup_cpu:
 	/* Get the per cpu offset for the given CPU# which is in ECX */
 	movq	__per_cpu_offset(,%rcx,8), %rdx
 #else
@@ -264,6 +318,14 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	lgdt	(%rsp)
 	addq	$16, %rsp
 
+	/* Drop the realmode protection. For the boot CPU the pointer is NULL! */
+	movq	trampoline_lock(%rip), %rax
+	testq	%rax, %rax
+	jz	.Lsetup_data_segments
+	lock
+	btrl	$0, (%rax)
+
+.Lsetup_data_segments:
 	/* set up data segments */
 	xorl %eax,%eax
 	movl %eax,%ds
@@ -293,6 +355,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	shrq	$32, %rdx
 	wrmsr
 
+.Lsetup_idt:
 	/* Setup and Load IDT */
 	pushq	%rsi
 	call	early_setup_idt
@@ -435,6 +498,8 @@ SYM_DATA(initial_code,	.quad x86_64_start_kernel)
 #ifdef CONFIG_AMD_MEM_ENCRYPT
 SYM_DATA(initial_vc_handler,	.quad handle_vc_boot_ghcb)
 #endif
+
+SYM_DATA(trampoline_lock, .quad 0);
 	__FINITDATA
 
 	__INIT
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 28d1643eee99..a9e48946fc89 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -797,6 +797,16 @@ static int __init cpu_init_udelay(char *str)
 }
 early_param("cpu_init_udelay", cpu_init_udelay);
 
+static bool do_parallel_bringup __ro_after_init = true;
+
+static int __init no_parallel_bringup(char *str)
+{
+	do_parallel_bringup = false;
+
+	return 0;
+}
+early_param("no_parallel_bringup", no_parallel_bringup);
+
 static void __init smp_quirk_init_udelay(void)
 {
 	/* if cmdline changed it from default, leave it alone */
@@ -1113,7 +1123,7 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle,
 	if (IS_ENABLED(CONFIG_X86_32)) {
 		early_gdt_descr.address = (unsigned long)get_cpu_gdt_rw(cpu);
 		initial_stack  = idle->thread.sp;
-	} else {
+	} else if (!do_parallel_bringup) {
 		smpboot_control = cpu;
 	}
 
@@ -1473,6 +1483,41 @@ void __init smp_prepare_cpus_common(void)
 	set_cpu_sibling_map(0);
 }
 
+/*
+ * We can do 64-bit AP bringup in parallel if the CPU reports its APIC
+ * ID in CPUID (either leaf 0x0B if we need the full APIC ID in X2APIC
+ * mode, or leaf 0x01 if 8 bits are sufficient). Otherwise it's too
+ * hard. And not for SEV-ES guests because they can't use CPUID that
+ * early.
+ */
+static bool prepare_parallel_bringup(void)
+{
+	if (IS_ENABLED(CONFIG_X86_32) || cc_platform_has(CC_ATTR_GUEST_STATE_ENCRYPT))
+		return false;
+
+	if (x2apic_mode) {
+		if (boot_cpu_data.cpuid_level < 0x0b)
+			return false;
+
+		if (check_extended_topology_leaf(0x0b) != 0) {
+			pr_info("Disabling parallel bringup because CPUID 0xb looks untrustworthy\n");
+			return false;
+		}
+
+		pr_debug("Using CPUID 0xb for parallel CPU startup\n");
+		smpboot_control = STARTUP_APICID_CPUID_0B;
+	} else {
+		/* Without X2APIC, what's in CPUID 0x01 should suffice. */
+		if (boot_cpu_data.cpuid_level < 0x01)
+			return false;
+
+		pr_debug("Using CPUID 0x1 for parallel CPU startup\n");
+		smpboot_control = STARTUP_APICID_CPUID_01;
+	}
+
+	return true;
+}
+
 /*
  * Prepare for SMP bootup.
  * @max_cpus: configured maximum number of CPUs, It is a legacy parameter
@@ -1513,6 +1558,9 @@ void __init native_smp_prepare_cpus(unsigned int max_cpus)
 
 	speculative_store_bypass_ht_init();
 
+	if (do_parallel_bringup)
+		do_parallel_bringup = prepare_parallel_bringup();
+
 	snp_set_wakeup_secondary_cpu();
 }
 
diff --git a/arch/x86/realmode/init.c b/arch/x86/realmode/init.c
index af565816d2ba..788e5559549f 100644
--- a/arch/x86/realmode/init.c
+++ b/arch/x86/realmode/init.c
@@ -154,6 +154,9 @@ static void __init setup_real_mode(void)
 
 	trampoline_header->flags = 0;
 
+	trampoline_lock = &trampoline_header->lock;
+	*trampoline_lock = 0;
+
 	trampoline_pgd = (u64 *) __va(real_mode_header->trampoline_pgd);
 
 	/* Map the real mode stub as virtual == physical */
diff --git a/arch/x86/realmode/rm/trampoline_64.S b/arch/x86/realmode/rm/trampoline_64.S
index e38d61d6562e..2dfb1c400167 100644
--- a/arch/x86/realmode/rm/trampoline_64.S
+++ b/arch/x86/realmode/rm/trampoline_64.S
@@ -37,6 +37,24 @@
 	.text
 	.code16
 
+.macro LOAD_REALMODE_ESP
+	/*
+	 * Make sure only one CPU fiddles with the realmode stack
+	 */
+.Llock_rm\@:
+	btl	$0, tr_lock
+	jnc	2f
+	pause
+	jmp	.Llock_rm\@
+2:
+	lock
+	btsl	$0, tr_lock
+	jc	.Llock_rm\@
+
+	# Setup stack
+	movl	$rm_stack_end, %esp
+.endm
+
 	.balign	PAGE_SIZE
 SYM_CODE_START(trampoline_start)
 	cli			# We should be safe anyway
@@ -49,8 +67,7 @@ SYM_CODE_START(trampoline_start)
 	mov	%ax, %es
 	mov	%ax, %ss
 
-	# Setup stack
-	movl	$rm_stack_end, %esp
+	LOAD_REALMODE_ESP
 
 	call	verify_cpu		# Verify the cpu supports long mode
 	testl   %eax, %eax		# Check for return code
@@ -93,8 +110,7 @@ SYM_CODE_START(sev_es_trampoline_start)
 	mov	%ax, %es
 	mov	%ax, %ss
 
-	# Setup stack
-	movl	$rm_stack_end, %esp
+	LOAD_REALMODE_ESP
 
 	jmp	.Lswitch_to_protected
 SYM_CODE_END(sev_es_trampoline_start)
@@ -177,7 +193,7 @@ SYM_CODE_START(pa_trampoline_compat)
 	 * In compatibility mode.  Prep ESP and DX for startup_32, then disable
 	 * paging and complete the switch to legacy 32-bit mode.
 	 */
-	movl	$rm_stack_end, %esp
+	LOAD_REALMODE_ESP
 	movw	$__KERNEL_DS, %dx
 
 	movl	$(CR0_STATE & ~X86_CR0_PG), %eax
@@ -241,6 +257,7 @@ SYM_DATA_START(trampoline_header)
 	SYM_DATA(tr_efer,		.space 8)
 	SYM_DATA(tr_cr4,		.space 4)
 	SYM_DATA(tr_flags,		.space 4)
+	SYM_DATA(tr_lock,		.space 4)
 SYM_DATA_END(trampoline_header)
 
 #include "trampoline_common.S"
-- 
2.25.1

[PATCH v15 10/12] x86/smpboot: Send INIT/SIPI/SIPI to secondary CPUs in parallel

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

When the APs can find their own APIC ID without assistance, perform the
AP bringup in parallel.

Register a CPUHP_BP_PARALLEL_DYN stage "x86/cpu:kick" which just calls
do_boot_cpu() to deliver INIT/SIPI/SIPI to each AP in turn before the
normal native_cpu_up() does the rest of the hand-holding.

The APs will then take turns through the real mode code (which has its
own bitlock for exclusion) until they make it to their own stack, then
proceed through the first few lines of start_secondary() and execute
these parts in parallel:

 start_secondary()
    -> cr4_init()
    -> (some 32-bit only stuff so not in the parallel cases)
    -> cpu_init_secondary()
       -> cpu_init_exception_handling()
       -> cpu_init()
          -> wait_for_master_cpu()

At this point they wait for the BSP to set their bit in cpu_callout_mask
(from do_wait_cpu_initialized()), and release them to continue through
the rest of cpu_init() and beyond.

This reduces the time taken for bringup on my 28-thread Haswell system
from about 120ms to 80ms. On a socket 96-thread Skylake it takes the
bringup time from 500ms to 100ms.

There is more speedup to be had by doing the remaining parts in parallel
too — especially notify_cpu_starting() in which the AP takes itself
through all the stages from CPUHP_BRINGUP_CPU to CPUHP_ONLINE. But those
require careful auditing to ensure they are reentrant, before we can go
that far.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/kernel/smpboot.c | 21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index a9e48946fc89..69b56c597949 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -57,6 +57,7 @@
 #include <linux/pgtable.h>
 #include <linux/overflow.h>
 #include <linux/stackprotector.h>
+#include <linux/smpboot.h>
 
 #include <asm/acpi.h>
 #include <asm/cacheinfo.h>
@@ -992,7 +993,8 @@ static void announce_cpu(int cpu, int apicid)
 		node_width = num_digits(num_possible_nodes()) + 1; /* + '#' */
 
 	if (cpu == 1)
-		printk(KERN_INFO "x86: Booting SMP configuration:\n");
+		printk(KERN_INFO "x86: Booting SMP configuration in %s:\n",
+		       do_parallel_bringup ? "parallel" : "series");
 
 	if (system_state < SYSTEM_RUNNING) {
 		if (node != current_node) {
@@ -1325,9 +1327,12 @@ int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
 {
 	int ret;
 
-	ret = do_cpu_up(cpu, tidle);
-	if (ret)
-		goto out;
+	/* If parallel AP bringup isn't enabled, perform the first steps now. */
+	if (!do_parallel_bringup) {
+		ret = do_cpu_up(cpu, tidle);
+		if (ret)
+			goto out;
+	}
 
 	ret = do_wait_cpu_initialized(cpu);
 	if (ret)
@@ -1347,6 +1352,12 @@ int native_cpu_up(unsigned int cpu, struct task_struct *tidle)
 	return ret;
 }
 
+/* Bringup step one: Send INIT/SIPI to the target AP */
+static int native_cpu_kick(unsigned int cpu)
+{
+	return do_cpu_up(cpu, idle_thread_get(cpu, true));
+}
+
 /**
  * arch_disable_smp_support() - disables SMP support for x86 at runtime
  */
@@ -1515,6 +1526,8 @@ static bool prepare_parallel_bringup(void)
 		smpboot_control = STARTUP_APICID_CPUID_01;
 	}
 
+	cpuhp_setup_state_nocalls(CPUHP_BP_PARALLEL_DYN, "x86/cpu:kick",
+				  native_cpu_kick, NULL);
 	return true;
 }
 
-- 
2.25.1

[PATCH v15 11/12] x86/smpboot: Serialize topology updates for secondary bringup

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

The toplogy update is performed by the AP via smp_callin() after the BSP
has called do_wait_cpu_initialized(), setting the AP's bit in
cpu_callout_mask to allow it to proceed.

In preparation to enable further parallelism of AP bringup, add locking to
serialize the update even if multiple APs are (in future) permitted to
proceed through the next stages of bringup in parallel.

Without such ordering (and with that future extra parallelism), confusion
ensues:

[    1.360149] x86: Booting SMP configuration:
[    1.360221] .... node  #0, CPUs:        #1  #2  #3  #4  #5  #6  #7  #8  #9 #10 #11 #12 #13 #14 #15 #16 #17 #18 #19 #20 #21 #22 #23
[    1.366225] .... node  #1, CPUs:   #24 #25 #26 #27 #28 #29 #30 #31 #32 #33 #34 #35 #36 #37 #38 #39 #40 #41 #42 #43 #44 #45 #46 #47
[    1.370219] .... node  #0, CPUs:   #48 #49 #50 #51 #52 #53 #54 #55 #56 #57 #58 #59 #60 #61 #62 #63 #64 #65 #66 #67 #68 #69 #70 #71
[    1.378226] .... node  #1, CPUs:   #72 #73 #74 #75 #76 #77 #78 #79 #80 #81 #82 #83 #84 #85 #86 #87 #88 #89 #90 #91 #92 #93 #94 #95
[    1.382037] Brought 96 CPUs to x86/cpu:kick in 72232606 cycles
[    0.104104] smpboot: CPU 26 Converting physical 0 to logical die 1
[    0.104104] smpboot: CPU 27 Converting physical 1 to logical package 2
[    0.104104] smpboot: CPU 24 Converting physical 1 to logical package 3
[    0.104104] smpboot: CPU 27 Converting physical 0 to logical die 2
[    0.104104] smpboot: CPU 25 Converting physical 1 to logical package 4
[    1.385609] Brought 96 CPUs to x86/cpu:wait-init in 9269218 cycles
[    1.395285] Brought CPUs online in 28930764 cycles
[    1.395469] smp: Brought up 2 nodes, 96 CPUs
[    1.395689] smpboot: Max logical packages: 2
[    1.396222] smpboot: Total of 96 processors activated (576000.00 BogoMIPS)

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
Tested-by: Paul E. McKenney <paulmck@kernel.org>
Tested-by: Kim Phillips <kim.phillips@amd.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
---
 arch/x86/include/asm/smp.h      |  4 +-
 arch/x86/include/asm/topology.h |  2 -
 arch/x86/kernel/cpu/common.c    |  6 +--
 arch/x86/kernel/smpboot.c       | 73 ++++++++++++++++++++-------------
 arch/x86/xen/smp_pv.c           |  4 +-
 5 files changed, 48 insertions(+), 41 deletions(-)

diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index 1cf4f1e57570..defe76ee9e64 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -48,8 +48,6 @@ struct smp_ops {
 };
 
 /* Globals due to paravirt */
-extern void set_cpu_sibling_map(int cpu);
-
 #ifdef CONFIG_SMP
 extern struct smp_ops smp_ops;
 
@@ -137,7 +135,7 @@ void native_send_call_func_single_ipi(int cpu);
 void x86_idle_thread_init(unsigned int cpu, struct task_struct *idle);
 
 void smp_store_boot_cpu_info(void);
-void smp_store_cpu_info(int id);
+void smp_store_cpu_info(int id, bool force_single_core);
 
 asmlinkage __visible void smp_reboot_interrupt(void);
 __visible void smp_reschedule_interrupt(struct pt_regs *regs);
diff --git a/arch/x86/include/asm/topology.h b/arch/x86/include/asm/topology.h
index 458c891a8273..4bccbd949a99 100644
--- a/arch/x86/include/asm/topology.h
+++ b/arch/x86/include/asm/topology.h
@@ -136,8 +136,6 @@ static inline int topology_max_smt_threads(void)
 	return __max_smt_threads;
 }
 
-int topology_update_package_map(unsigned int apicid, unsigned int cpu);
-int topology_update_die_map(unsigned int dieid, unsigned int cpu);
 int topology_phys_to_logical_pkg(unsigned int pkg);
 int topology_phys_to_logical_die(unsigned int die, unsigned int cpu);
 bool topology_is_primary_thread(unsigned int cpu);
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index f3cc7699e1e1..06d7f9e55d45 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -1771,7 +1771,7 @@ static void generic_identify(struct cpuinfo_x86 *c)
  * Validate that ACPI/mptables have the same information about the
  * effective APIC id and update the package map.
  */
-static void validate_apic_and_package_id(struct cpuinfo_x86 *c)
+static void validate_apic_id(struct cpuinfo_x86 *c)
 {
 #ifdef CONFIG_SMP
 	unsigned int apicid, cpu = smp_processor_id();
@@ -1782,8 +1782,6 @@ static void validate_apic_and_package_id(struct cpuinfo_x86 *c)
 		pr_err(FW_BUG "CPU%u: APIC id mismatch. Firmware: %x APIC: %x\n",
 		       cpu, apicid, c->initial_apicid);
 	}
-	BUG_ON(topology_update_package_map(c->phys_proc_id, cpu));
-	BUG_ON(topology_update_die_map(c->cpu_die_id, cpu));
 #else
 	c->logical_proc_id = 0;
 #endif
@@ -1974,7 +1972,7 @@ void identify_secondary_cpu(struct cpuinfo_x86 *c)
 #ifdef CONFIG_X86_32
 	enable_sep_cpu();
 #endif
-	validate_apic_and_package_id(c);
+	validate_apic_id(c);
 	x86_spec_ctrl_setup_ap();
 	update_srbds_msr();
 
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 69b56c597949..0960fdec43c2 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -180,16 +180,12 @@ static void smp_callin(void)
 	apic_ap_setup();
 
 	/*
-	 * Save our processor parameters. Note: this information
-	 * is needed for clock calibration.
-	 */
-	smp_store_cpu_info(cpuid);
-
-	/*
+	 * Save our processor parameters and update topology.
+	 * Note: this information is needed for clock calibration.
 	 * The topology information must be up to date before
 	 * calibrate_delay() and notify_cpu_starting().
 	 */
-	set_cpu_sibling_map(raw_smp_processor_id());
+	smp_store_cpu_info(cpuid, false);
 
 	ap_init_aperfmperf();
 
@@ -243,6 +239,12 @@ static void notrace start_secondary(void *unused)
 	 * its bit in cpu_callout_mask to release it.
 	 */
 	cpu_init_secondary();
+
+	/*
+	 * Even though notify_cpu_starting() will do this, it does so too late
+	 * as the AP may already have triggered lockdep splats by then. See
+	 * commit 29368e093 ("x86/smpboot:  Move rcu_cpu_starting() earlier").
+	 */
 	rcu_cpu_starting(raw_smp_processor_id());
 	x86_cpuinit.early_percpu_clock_init();
 
@@ -351,7 +353,7 @@ EXPORT_SYMBOL(topology_phys_to_logical_die);
  * @pkg:	The physical package id as retrieved via CPUID
  * @cpu:	The cpu for which this is updated
  */
-int topology_update_package_map(unsigned int pkg, unsigned int cpu)
+static int topology_update_package_map(unsigned int pkg, unsigned int cpu)
 {
 	int new;
 
@@ -374,7 +376,7 @@ int topology_update_package_map(unsigned int pkg, unsigned int cpu)
  * @die:	The die id as retrieved via CPUID
  * @cpu:	The cpu for which this is updated
  */
-int topology_update_die_map(unsigned int die, unsigned int cpu)
+static int topology_update_die_map(unsigned int die, unsigned int cpu)
 {
 	int new;
 
@@ -405,25 +407,7 @@ void __init smp_store_boot_cpu_info(void)
 	c->initialized = true;
 }
 
-/*
- * The bootstrap kernel entry code has set these up. Save them for
- * a given CPU
- */
-void smp_store_cpu_info(int id)
-{
-	struct cpuinfo_x86 *c = &cpu_data(id);
-
-	/* Copy boot_cpu_data only on the first bringup */
-	if (!c->initialized)
-		*c = boot_cpu_data;
-	c->cpu_index = id;
-	/*
-	 * During boot time, CPU0 has this setup already. Save the info when
-	 * bringing up AP or offlined CPU0.
-	 */
-	identify_secondary_cpu(c);
-	c->initialized = true;
-}
+static arch_spinlock_t topology_lock = __ARCH_SPIN_LOCK_UNLOCKED;
 
 static bool
 topology_same_node(struct cpuinfo_x86 *c, struct cpuinfo_x86 *o)
@@ -629,7 +613,7 @@ static struct sched_domain_topology_level x86_topology[] = {
  */
 static bool x86_has_numa_in_package;
 
-void set_cpu_sibling_map(int cpu)
+static void set_cpu_sibling_map(int cpu)
 {
 	bool has_smt = smp_num_siblings > 1;
 	bool has_mp = has_smt || boot_cpu_data.x86_max_cores > 1;
@@ -708,6 +692,37 @@ void set_cpu_sibling_map(int cpu)
 	}
 }
 
+/*
+ * The bootstrap kernel entry code has set these up. Save them for
+ * a given CPU
+ */
+void smp_store_cpu_info(int id, bool force_single_core)
+{
+	struct cpuinfo_x86 *c = &cpu_data(id);
+
+	/* Copy boot_cpu_data only on the first bringup */
+	if (!c->initialized)
+		*c = boot_cpu_data;
+	c->cpu_index = id;
+	/*
+	 * During boot time, CPU0 has this setup already. Save the info when
+	 * bringing up AP or offlined CPU0.
+	 */
+	identify_secondary_cpu(c);
+
+	arch_spin_lock(&topology_lock);
+	BUG_ON(topology_update_package_map(c->phys_proc_id, id));
+	BUG_ON(topology_update_die_map(c->cpu_die_id, id));
+	c->initialized = true;
+
+	/* For Xen PV */
+	if (force_single_core)
+		c->x86_max_cores = 1;
+
+	set_cpu_sibling_map(id);
+	arch_spin_unlock(&topology_lock);
+}
+
 /* maps the cpu to the sched domain representing multi-core */
 const struct cpumask *cpu_coregroup_mask(int cpu)
 {
diff --git a/arch/x86/xen/smp_pv.c b/arch/x86/xen/smp_pv.c
index 6175f2c5c822..09f94f940689 100644
--- a/arch/x86/xen/smp_pv.c
+++ b/arch/x86/xen/smp_pv.c
@@ -71,9 +71,7 @@ static void cpu_bringup(void)
 		xen_enable_syscall();
 	}
 	cpu = smp_processor_id();
-	smp_store_cpu_info(cpu);
-	cpu_data(cpu).x86_max_cores = 1;
-	set_cpu_sibling_map(cpu);
+	smp_store_cpu_info(cpu, true);
 
 	speculative_store_bypass_ht_init();
 
-- 
2.25.1

[PATCH v15 12/12] x86/smpboot: Allow parallel bringup for SEV-ES

[Usama Arif]

From: David Woodhouse <dwmw@amazon.co.uk>

Enable parallel bringup for SEV-ES guests. The APs can't actually
execute the CPUID instruction directly during early startup, but they
can make the GHCB call directly instead, just as the VC trap handler
would do.

Thanks to Sabin for talking me through the way this works.

Suggested-by: Sabin Rapan <sabrapan@amazon.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Usama Arif <usama.arif@bytedance.com>
---
 arch/x86/coco/core.c              |  5 ++++
 arch/x86/include/asm/coco.h       |  1 +
 arch/x86/include/asm/sev-common.h |  3 +++
 arch/x86/include/asm/smp.h        |  5 +++-
 arch/x86/kernel/head_64.S         | 30 ++++++++++++++++++++++++
 arch/x86/kernel/smpboot.c         | 39 ++++++++++++++++++++++++++-----
 6 files changed, 76 insertions(+), 7 deletions(-)

diff --git a/arch/x86/coco/core.c b/arch/x86/coco/core.c
index 49b44f881484..0bab38efb15a 100644
--- a/arch/x86/coco/core.c
+++ b/arch/x86/coco/core.c
@@ -129,6 +129,11 @@ u64 cc_mkdec(u64 val)
 }
 EXPORT_SYMBOL_GPL(cc_mkdec);
 
+enum cc_vendor cc_get_vendor(void)
+{
+	return vendor;
+}
+
 __init void cc_set_vendor(enum cc_vendor v)
 {
 	vendor = v;
diff --git a/arch/x86/include/asm/coco.h b/arch/x86/include/asm/coco.h
index 3d98c3a60d34..0428d9712c96 100644
--- a/arch/x86/include/asm/coco.h
+++ b/arch/x86/include/asm/coco.h
@@ -12,6 +12,7 @@ enum cc_vendor {
 };
 
 void cc_set_vendor(enum cc_vendor v);
+enum cc_vendor cc_get_vendor(void);
 void cc_set_mask(u64 mask);
 
 #ifdef CONFIG_ARCH_HAS_CC_PLATFORM
diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h
index b8357d6ecd47..f25df4bd318e 100644
--- a/arch/x86/include/asm/sev-common.h
+++ b/arch/x86/include/asm/sev-common.h
@@ -70,6 +70,7 @@
 	/* GHCBData[63:12] */				\
 	(((u64)(v) & GENMASK_ULL(63, 12)) >> 12)
 
+#ifndef __ASSEMBLY__
 /*
  * SNP Page State Change Operation
  *
@@ -160,6 +161,8 @@ struct snp_psc_desc {
 
 #define GHCB_RESP_CODE(v)		((v) & GHCB_MSR_INFO_MASK)
 
+#endif /* __ASSEMBLY__ */
+
 /*
  * Error codes related to GHCB input that can be communicated back to the guest
  * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2.
diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
index defe76ee9e64..1584f04a7007 100644
--- a/arch/x86/include/asm/smp.h
+++ b/arch/x86/include/asm/smp.h
@@ -204,7 +204,10 @@ extern unsigned int smpboot_control;
 /* Control bits for startup_64 */
 #define STARTUP_APICID_CPUID_0B	0x80000000
 #define STARTUP_APICID_CPUID_01	0x40000000
+#define STARTUP_APICID_SEV_ES	0x20000000
 
-#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | STARTUP_APICID_CPUID_0B)
+#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | \
+			       STARTUP_APICID_CPUID_0B | \
+			       STARTUP_APICID_SEV_ES)
 
 #endif /* _ASM_X86_SMP_H */
diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
index 65bca47d84a1..79a17e705f4e 100644
--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -26,6 +26,7 @@
 #include <asm/nospec-branch.h>
 #include <asm/fixmap.h>
 #include <asm/smp.h>
+#include <asm/sev-common.h>
 
 /*
  * We are not able to switch in one step to the final KERNEL ADDRESS SPACE
@@ -242,6 +243,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	 *
 	 * Bit 31	STARTUP_APICID_CPUID_0B flag (use CPUID 0x0b)
 	 * Bit 30	STARTUP_APICID_CPUID_01 flag (use CPUID 0x01)
+	 * Bit 29	STARTUP_APICID_SEV_ES flag (CPUID 0x0b via GHCB MSR)
 	 * Bit 0-24	CPU# if STARTUP_APICID_CPUID_xx flags are not set
 	 */
 	movl	smpboot_control(%rip), %ecx
@@ -249,6 +251,10 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	jnz	.Luse_cpuid_0b
 	testl	$STARTUP_APICID_CPUID_01, %ecx
 	jnz	.Luse_cpuid_01
+#ifdef CONFIG_AMD_MEM_ENCRYPT
+	testl	$STARTUP_APICID_SEV_ES, %ecx
+	jnz	.Luse_sev_cpuid_0b
+#endif
 	andl	$0x0FFFFFFF, %ecx
 	jmp	.Lsetup_cpu
 
@@ -259,6 +265,30 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	shr	$24, %edx
 	jmp	.Lsetup_AP
 
+#ifdef CONFIG_AMD_MEM_ENCRYPT
+.Luse_sev_cpuid_0b:
+	/* Set the GHCB MSR to request CPUID 0xB_EDX */
+	movl	$MSR_AMD64_SEV_ES_GHCB, %ecx
+	movl	$(GHCB_CPUID_REQ_EDX << 30) | GHCB_MSR_CPUID_REQ, %eax
+	movl	$0x0B, %edx
+	wrmsr
+
+	/* Perform GHCB MSR protocol */
+	rep; vmmcall		/* vmgexit */
+
+	/*
+	 * Get the result. After the RDMSR:
+	 *   EAX should be 0xc0000005
+	 *   EDX should have the CPUID register value and since EDX
+	 *   is the target register, no need to move the result.
+	 */
+	rdmsr
+	andl	$GHCB_MSR_INFO_MASK, %eax
+	cmpl	$GHCB_MSR_CPUID_RESP, %eax
+	jne	1f
+	jmp	.Lsetup_AP
+#endif
+
 .Luse_cpuid_0b:
 	mov	$0x0B, %eax
 	xorl	%ecx, %ecx
diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 0960fdec43c2..895395787afc 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -85,6 +85,7 @@
 #include <asm/hw_irq.h>
 #include <asm/stackprotector.h>
 #include <asm/sev.h>
+#include <asm/coco.h>
 
 /* representing HT siblings of each logical CPU */
 DEFINE_PER_CPU_READ_MOSTLY(cpumask_var_t, cpu_sibling_map);
@@ -1513,15 +1514,36 @@ void __init smp_prepare_cpus_common(void)
  * We can do 64-bit AP bringup in parallel if the CPU reports its APIC
  * ID in CPUID (either leaf 0x0B if we need the full APIC ID in X2APIC
  * mode, or leaf 0x01 if 8 bits are sufficient). Otherwise it's too
- * hard. And not for SEV-ES guests because they can't use CPUID that
- * early.
+ * hard.
  */
 static bool prepare_parallel_bringup(void)
 {
-	if (IS_ENABLED(CONFIG_X86_32) || cc_platform_has(CC_ATTR_GUEST_STATE_ENCRYPT))
+	bool has_sev_es = false;
+
+	if (IS_ENABLED(CONFIG_X86_32))
 		return false;
 
-	if (x2apic_mode) {
+	/*
+	 * Encrypted guests other than SEV-ES (in the future) will need to
+	 * implement an early way of finding the APIC ID, since they will
+	 * presumably block direct CPUID too. Be kind to our future selves
+	 * by warning here instead of just letting them break. Parallel
+	 * startup doesn't have to be in the first round of enabling patches
+	 * for any such technology.
+	 */
+	if (cc_platform_has(CC_ATTR_GUEST_STATE_ENCRYPT)) {
+		switch (cc_get_vendor()) {
+		case CC_VENDOR_AMD:
+			has_sev_es = true;
+			break;
+
+		default:
+			pr_info("Disabling parallel bringup due to guest state encryption\n");
+			return false;
+		}
+	}
+
+	if (x2apic_mode || has_sev_es) {
 		if (boot_cpu_data.cpuid_level < 0x0b)
 			return false;
 
@@ -1530,8 +1552,13 @@ static bool prepare_parallel_bringup(void)
 			return false;
 		}
 
-		pr_debug("Using CPUID 0xb for parallel CPU startup\n");
-		smpboot_control = STARTUP_APICID_CPUID_0B;
+		if (has_sev_es) {
+			pr_debug("Using SEV-ES CPUID 0xb for parallel CPU startup\n");
+			smpboot_control = STARTUP_APICID_SEV_ES;
+		} else {
+			pr_debug("Using CPUID 0xb for parallel CPU startup\n");
+			smpboot_control = STARTUP_APICID_CPUID_0B;
+		}
 	} else {
 		/* Without X2APIC, what's in CPUID 0x01 should suffice. */
 		if (boot_cpu_data.cpuid_level < 0x01)
-- 
2.25.1

Re: [PATCH v15 12/12] x86/smpboot: Allow parallel bringup for SEV-ES

[Tom Lendacky]

On 3/16/23 17:21, Usama Arif wrote:
> From: David Woodhouse <dwmw@amazon.co.uk>
> 
> Enable parallel bringup for SEV-ES guests. The APs can't actually
> execute the CPUID instruction directly during early startup, but they
> can make the GHCB call directly instead, just as the VC trap handler
> would do.
> 
> Thanks to Sabin for talking me through the way this works.
> 
> Suggested-by: Sabin Rapan <sabrapan@amazon.com>
> Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
> Signed-off-by: Usama Arif <usama.arif@bytedance.com>

Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>

> ---
>   arch/x86/coco/core.c              |  5 ++++
>   arch/x86/include/asm/coco.h       |  1 +
>   arch/x86/include/asm/sev-common.h |  3 +++
>   arch/x86/include/asm/smp.h        |  5 +++-
>   arch/x86/kernel/head_64.S         | 30 ++++++++++++++++++++++++
>   arch/x86/kernel/smpboot.c         | 39 ++++++++++++++++++++++++++-----
>   6 files changed, 76 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/coco/core.c b/arch/x86/coco/core.c
> index 49b44f881484..0bab38efb15a 100644
> --- a/arch/x86/coco/core.c
> +++ b/arch/x86/coco/core.c
> @@ -129,6 +129,11 @@ u64 cc_mkdec(u64 val)
>   }
>   EXPORT_SYMBOL_GPL(cc_mkdec);
>   
> +enum cc_vendor cc_get_vendor(void)
> +{
> +	return vendor;
> +}
> +
>   __init void cc_set_vendor(enum cc_vendor v)
>   {
>   	vendor = v;
> diff --git a/arch/x86/include/asm/coco.h b/arch/x86/include/asm/coco.h
> index 3d98c3a60d34..0428d9712c96 100644
> --- a/arch/x86/include/asm/coco.h
> +++ b/arch/x86/include/asm/coco.h
> @@ -12,6 +12,7 @@ enum cc_vendor {
>   };
>   
>   void cc_set_vendor(enum cc_vendor v);
> +enum cc_vendor cc_get_vendor(void);
>   void cc_set_mask(u64 mask);
>   
>   #ifdef CONFIG_ARCH_HAS_CC_PLATFORM
> diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h
> index b8357d6ecd47..f25df4bd318e 100644
> --- a/arch/x86/include/asm/sev-common.h
> +++ b/arch/x86/include/asm/sev-common.h
> @@ -70,6 +70,7 @@
>   	/* GHCBData[63:12] */				\
>   	(((u64)(v) & GENMASK_ULL(63, 12)) >> 12)
>   
> +#ifndef __ASSEMBLY__
>   /*
>    * SNP Page State Change Operation
>    *
> @@ -160,6 +161,8 @@ struct snp_psc_desc {
>   
>   #define GHCB_RESP_CODE(v)		((v) & GHCB_MSR_INFO_MASK)
>   
> +#endif /* __ASSEMBLY__ */
> +
>   /*
>    * Error codes related to GHCB input that can be communicated back to the guest
>    * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2.
> diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
> index defe76ee9e64..1584f04a7007 100644
> --- a/arch/x86/include/asm/smp.h
> +++ b/arch/x86/include/asm/smp.h
> @@ -204,7 +204,10 @@ extern unsigned int smpboot_control;
>   /* Control bits for startup_64 */
>   #define STARTUP_APICID_CPUID_0B	0x80000000
>   #define STARTUP_APICID_CPUID_01	0x40000000
> +#define STARTUP_APICID_SEV_ES	0x20000000
>   
> -#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | STARTUP_APICID_CPUID_0B)
> +#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | \
> +			       STARTUP_APICID_CPUID_0B | \
> +			       STARTUP_APICID_SEV_ES)
>   
>   #endif /* _ASM_X86_SMP_H */
> diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
> index 65bca47d84a1..79a17e705f4e 100644
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -26,6 +26,7 @@
>   #include <asm/nospec-branch.h>
>   #include <asm/fixmap.h>
>   #include <asm/smp.h>
> +#include <asm/sev-common.h>
>   
>   /*
>    * We are not able to switch in one step to the final KERNEL ADDRESS SPACE
> @@ -242,6 +243,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>   	 *
>   	 * Bit 31	STARTUP_APICID_CPUID_0B flag (use CPUID 0x0b)
>   	 * Bit 30	STARTUP_APICID_CPUID_01 flag (use CPUID 0x01)
> +	 * Bit 29	STARTUP_APICID_SEV_ES flag (CPUID 0x0b via GHCB MSR)
>   	 * Bit 0-24	CPU# if STARTUP_APICID_CPUID_xx flags are not set
>   	 */
>   	movl	smpboot_control(%rip), %ecx
> @@ -249,6 +251,10 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>   	jnz	.Luse_cpuid_0b
>   	testl	$STARTUP_APICID_CPUID_01, %ecx
>   	jnz	.Luse_cpuid_01
> +#ifdef CONFIG_AMD_MEM_ENCRYPT
> +	testl	$STARTUP_APICID_SEV_ES, %ecx
> +	jnz	.Luse_sev_cpuid_0b
> +#endif
>   	andl	$0x0FFFFFFF, %ecx
>   	jmp	.Lsetup_cpu
>   
> @@ -259,6 +265,30 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>   	shr	$24, %edx
>   	jmp	.Lsetup_AP
>   
> +#ifdef CONFIG_AMD_MEM_ENCRYPT
> +.Luse_sev_cpuid_0b:
> +	/* Set the GHCB MSR to request CPUID 0xB_EDX */
> +	movl	$MSR_AMD64_SEV_ES_GHCB, %ecx
> +	movl	$(GHCB_CPUID_REQ_EDX << 30) | GHCB_MSR_CPUID_REQ, %eax
> +	movl	$0x0B, %edx
> +	wrmsr
> +
> +	/* Perform GHCB MSR protocol */
> +	rep; vmmcall		/* vmgexit */
> +
> +	/*
> +	 * Get the result. After the RDMSR:
> +	 *   EAX should be 0xc0000005
> +	 *   EDX should have the CPUID register value and since EDX
> +	 *   is the target register, no need to move the result.
> +	 */
> +	rdmsr
> +	andl	$GHCB_MSR_INFO_MASK, %eax
> +	cmpl	$GHCB_MSR_CPUID_RESP, %eax
> +	jne	1f
> +	jmp	.Lsetup_AP
> +#endif
> +
>   .Luse_cpuid_0b:
>   	mov	$0x0B, %eax
>   	xorl	%ecx, %ecx
> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
> index 0960fdec43c2..895395787afc 100644
> --- a/arch/x86/kernel/smpboot.c
> +++ b/arch/x86/kernel/smpboot.c
> @@ -85,6 +85,7 @@
>   #include <asm/hw_irq.h>
>   #include <asm/stackprotector.h>
>   #include <asm/sev.h>
> +#include <asm/coco.h>
>   
>   /* representing HT siblings of each logical CPU */
>   DEFINE_PER_CPU_READ_MOSTLY(cpumask_var_t, cpu_sibling_map);
> @@ -1513,15 +1514,36 @@ void __init smp_prepare_cpus_common(void)
>    * We can do 64-bit AP bringup in parallel if the CPU reports its APIC
>    * ID in CPUID (either leaf 0x0B if we need the full APIC ID in X2APIC
>    * mode, or leaf 0x01 if 8 bits are sufficient). Otherwise it's too
> - * hard. And not for SEV-ES guests because they can't use CPUID that
> - * early.
> + * hard.
>    */
>   static bool prepare_parallel_bringup(void)
>   {
> -	if (IS_ENABLED(CONFIG_X86_32) || cc_platform_has(CC_ATTR_GUEST_STATE_ENCRYPT))
> +	bool has_sev_es = false;
> +
> +	if (IS_ENABLED(CONFIG_X86_32))
>   		return false;
>   
> -	if (x2apic_mode) {
> +	/*
> +	 * Encrypted guests other than SEV-ES (in the future) will need to
> +	 * implement an early way of finding the APIC ID, since they will
> +	 * presumably block direct CPUID too. Be kind to our future selves
> +	 * by warning here instead of just letting them break. Parallel
> +	 * startup doesn't have to be in the first round of enabling patches
> +	 * for any such technology.
> +	 */
> +	if (cc_platform_has(CC_ATTR_GUEST_STATE_ENCRYPT)) {
> +		switch (cc_get_vendor()) {
> +		case CC_VENDOR_AMD:
> +			has_sev_es = true;
> +			break;
> +
> +		default:
> +			pr_info("Disabling parallel bringup due to guest state encryption\n");
> +			return false;
> +		}
> +	}
> +
> +	if (x2apic_mode || has_sev_es) {
>   		if (boot_cpu_data.cpuid_level < 0x0b)
>   			return false;
>   
> @@ -1530,8 +1552,13 @@ static bool prepare_parallel_bringup(void)
>   			return false;
>   		}
>   
> -		pr_debug("Using CPUID 0xb for parallel CPU startup\n");
> -		smpboot_control = STARTUP_APICID_CPUID_0B;
> +		if (has_sev_es) {
> +			pr_debug("Using SEV-ES CPUID 0xb for parallel CPU startup\n");
> +			smpboot_control = STARTUP_APICID_SEV_ES;
> +		} else {
> +			pr_debug("Using CPUID 0xb for parallel CPU startup\n");
> +			smpboot_control = STARTUP_APICID_CPUID_0B;
> +		}
>   	} else {
>   		/* Without X2APIC, what's in CPUID 0x01 should suffice. */
>   		if (boot_cpu_data.cpuid_level < 0x01)

Re: [PATCH v15 02/12] cpu/hotplug: Move idle_thread_get() to <linux/smpboot.h>

[Borislav Petkov]

On Thu, Mar 16, 2023 at 10:20:59PM +0000, Usama Arif wrote:
> From: David Woodhouse <dwmw@amazon.co.uk>
> 
> Instead of relying purely on the special-case wrapper in bringup_cpu()
> to pass the idle thread to __cpu_up(), expose idle_thread_get() so that
> the architecture code can obtain it directly when necessary.
> 
> This will be useful when the existing __cpu_up() is split into multiple
> phases, only *one* of which will actually need the idle thread.
> 
> If the architecture code is to register its new pre-bringup states with
> the cpuhp core, having a special-case wrapper to pass extra arguments is
> non-trivial and it's easier just to let the arch register its function
> pointer to be invoked with the standard API.
> 
> To reduce duplication, move the shadow stack reset and kasan unpoisoning

I was wondering what "shadow stack" as that set is not upstream yet. You mean
"shadow call stack" which is apparently something else,
compiler-generated, purely software thing.

> into idle_thread_get() too.

Frankly, I don't think resetting shadow call stack and kasan state
belongs in a function which returns the idle thread. Even more so if you
have to add an @unpoison param which is false sometimes and sometimes
true, depending on where you call the function.

I think you should have a helper

	tsk_reset_stacks(struct task_struct *tsk);

or so which is called where @unpoison == true instead of having a getter
function do something unrelated too.

Thx.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH v15 02/12] cpu/hotplug: Move idle_thread_get() to <linux/smpboot.h>

[David Woodhouse]

[-- Attachment #1: Type: text/plain, Size: 2576 bytes --]

On Sun, 2023-03-19 at 17:34 +0100, Borislav Petkov wrote:
> Frankly, I don't think resetting shadow call stack and kasan state
> belongs in a function which returns the idle thread. Even more so if you
> have to add an @unpoison param which is false sometimes and sometimes
> true, depending on where you call the function.
> 
> I think you should have a helper
> 
>         tsk_reset_stacks(struct task_struct *tsk);
> 
> or so which is called where @unpoison == true instead of having a getter
> function do something unrelated too.

Yeah, I see that but my primary concern was that I didn't want callers
to be able to *forget* it, which is what happened the first time.

I don't feel so bad about the getter function actually making the
object *usable* as well as getting it. We don't have to remember to
make a separate function call to unpoison a pointer newly returned from
kmalloc either.

I *think* the 'unpoison' arg is purely an optimisation anyway. Because
those operations are, or at least *could* be, idempotent. It's just
that it's a bit pointless doing them from the finish_cpu() call.

But actually, I think we can have it both ways. There's an early call
to idle_thread_get(cpu) in _cpu_up(), with the comment /* Let it fail
before we try to bring the cpu up */

With an adjustment to the comment, we could do the unpoison and reset
the shadow call stack there — in common code where no architecture can
forget to do it — and still leave idle_thread_get() doing precisely
just the 'get'.

In _cpu_up() the call to idle_thread_get() only happens if the CPU in
question is starting from CPUHP_OFFLINE but I think that's sufficient?

diff --git a/kernel/cpu.c b/kernel/cpu.c
index 6c0a92ca6bb5..43e0a77f21e8 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -591,12 +591,6 @@ static int bringup_cpu(unsigned int cpu)
 	struct task_struct *idle = idle_thread_get(cpu);
 	int ret;
 
-	/*
-	 * Reset stale stack state from the last time this CPU was online.
-	 */
-	scs_task_reset(idle);
-	kasan_unpoison_task_stack(idle);
-
 	/*
 	 * Some architectures have to walk the irq descriptors to
 	 * setup the vector space for the cpu which comes online.
@@ -1383,6 +1377,12 @@ static int _cpu_up(unsigned int cpu, int tasks_frozen, enum cpuhp_state target)
 			ret = PTR_ERR(idle);
 			goto out;
 		}
+
+		/*
+		 * Reset stale stack state from the last time this CPU was online.
+		 */
+		scs_task_reset(idle);
+		kasan_unpoison_task_stack(idle);
 	}
 
 	cpuhp_tasks_frozen = tasks_frozen;


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5965 bytes --]

Re: [PATCH v15 03/12] cpu/hotplug: Add dynamic parallel bringup states before CPUHP_BRINGUP_CPU

[Thomas Gleixner]

On Thu, Mar 16 2023 at 22:21, Usama Arif wrote:
> diff --git a/kernel/cpu.c b/kernel/cpu.c
> index 6b3dccb4a888..6ccc64defd47 100644
> --- a/kernel/cpu.c
> +++ b/kernel/cpu.c
> @@ -1497,8 +1497,30 @@ int bringup_hibernate_cpu(unsigned int sleep_cpu)
>  
>  void bringup_nonboot_cpus(unsigned int setup_max_cpus)
>  {
> +	unsigned int n = setup_max_cpus - num_online_cpus();
>  	unsigned int cpu;
>  
> +	/*
> +	 * An architecture may have registered parallel pre-bringup states to
> +	 * which each CPU may be brought in parallel. For each such state,
> +	 * bring N CPUs to it in turn before the final round of bringing them
> +	 * online.
> +	 */
> +	if (n > 0) {
> +		enum cpuhp_state st = CPUHP_BP_PARALLEL_DYN;
> +
> +		while (st <= CPUHP_BP_PARALLEL_DYN_END && cpuhp_hp_states[st].name) {
> +			int i = n;
> +
> +			for_each_present_cpu(cpu) {
> +				cpu_up(cpu, st);
> +				if (!--i)
> +					break;
> +			}
> +			st++;
> +		}
> +	}
> +
>  	for_each_present_cpu(cpu) {
>  		if (num_online_cpus() >= setup_max_cpus)
>  			break;

This causes a subtle issue. The bringup loop above moves all CPUs to
cpuhp_state == CPUHP_BP_PARALLEL_DYN_END. So the serial bootup will
start from there and bring them fully up.

Now if a bringup fails, then the rollback will only go back down to
CPUHP_BP_PARALLEL_DYN_END, which means that the control CPU won't do any
cleanups below CPUHP_BP_PARALLEL_DYN_END.

That 'fail' is a common case for SMT soft disable via the 'nosmt'
command line parameter. Due to the marvelous MCE broadcast 'feature' we
need to bringup the SMT siblings at least to the CPUHP_AP_ONLINE_IDLE
state once and then roll them back.

While this is not necessarily a fatal problem, it's changing behaviour
and with quite some of the details hidden in the (then not issued)
teardown callbacks might cause some hard to decode subtle surprises.

So that second for_each_present_cpu() loop needs to check the return
value of cpu_up() and issue a full rollback to CPUHP_OFFLINE in case of
fail.

Thanks,

        tglx

Re: [PATCH v15 04/12] x86/smpboot: Reference count on smpboot_setup_warm_reset_vector()

[Borislav Petkov]

On Thu, Mar 16, 2023 at 10:21:01PM +0000, Usama Arif wrote:
> From: David Woodhouse <dwmw@amazon.co.uk>
> 
> When bringing up a secondary CPU from do_boot_cpu(), the warm reset flag
> is set in CMOS and the starting IP for the trampoline written inside the
> BDA at 0x467. Once the CPU is running, the CMOS flag is unset and the
> value in the BDA cleared.
> 
> To allow for parallel bringup of CPUs, add a reference count to track the
> number of CPUs currently bring brought up, and clear the state only when

s/bring //

> the count reaches zero.

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette

Re: [PATCH v15 04/12] x86/smpboot: Reference count on smpboot_setup_warm_reset_vector()

[David Woodhouse]

On 21 March 2023 11:41:15 GMT, Borislav Petkov <bp@alien8.de> wrote:
>On Thu, Mar 16, 2023 at 10:21:01PM +0000, Usama Arif wrote:
>> From: David Woodhouse <dwmw@amazon.co.uk>
>> 
>> When bringing up a secondary CPU from do_boot_cpu(), the warm reset flag
>> is set in CMOS and the starting IP for the trampoline written inside the
>> BDA at 0x467. Once the CPU is running, the CMOS flag is unset and the
>> value in the BDA cleared.
>> 
>> To allow for parallel bringup of CPUs, add a reference count to track the
>> number of CPUs currently bring brought up, and clear the state only when
>
>s/bring //

"being", I think.

Re: [PATCH v15 09/12] x86/smpboot: Support parallel startup of secondary CPUs

[Brian Gerst]

On Thu, Mar 16, 2023 at 6:21 PM Usama Arif <usama.arif@bytedance.com> wrote:
>
> From: David Woodhouse <dwmw@amazon.co.uk>
>
> Rework the real-mode startup code to allow for APs to be brought up in
> parallel. This is in two parts:
>
> 1. Introduce a bit-spinlock to prevent them from all using the real
>    mode stack at the same time.
>
> 2. Avoid needing to use the global smpboot_control variable to pass
>    each AP its CPU#.
>
> To achieve the latter, export the cpuid_to_apicid[] array so that each
> AP can find its own CPU# by searching therein based on its APIC ID.
>
> Introduce flags in the top bits of smpboot_control which indicate methods
> by which an AP should find its CPU#. For a serialized bringup, the CPU#
> is explicitly passed in the low bits of smpboot_control as before. For
> parallel mode there are flags directing the AP to find its APIC ID in
> CPUID leaf 0x0b (for X2APIC mode) or CPUID leaf 0x01 where 8 bits are
> sufficient, then perform the cpuid_to_apicid[] lookup with that.
>
> Parallel startup may be disabled by a command line option, and also if:
>  • AMD SEV-ES is in use, since the AP may not use CPUID that early.
>  • X2APIC is enabled, but CPUID leaf 0xb is not present and correct.
>  • X2APIC is not enabled but not even CPUID leaf 0x01 exists.
>
> Aside from the fact that APs will now look up their CPU# via the
> newly-exported cpuid_to_apicid[] table, there is no behavioural change
> intended yet, since new parallel CPUHP states have not — yet — been
> added.
>
> [ tglx: Initial proof of concept patch with bitlock and APIC ID lookup ]
> [ dwmw2: Rework and testing, commit message, CPUID 0x1 and CPU0 support ]
> [ seanc: Fix stray override of initial_gs in common_cpu_up() ]
> [ Oleksandr Natalenko: reported suspend/resume issue fixed in
>   x86_acpi_suspend_lowlevel ]
> Co-developed-by: Thomas Gleixner <tglx@linutronix.de>
> Co-developed-by: Brian Gerst <brgerst@gmail.com>
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> Signed-off-by: Brian Gerst <brgerst@gmail.com>
> Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
> Signed-off-by: Usama Arif <usama.arif@bytedance.com>
> Tested-by: Paul E. McKenney <paulmck@kernel.org>
> Tested-by: Kim Phillips <kim.phillips@amd.com>
> Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
> Tested-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
> ---
>  .../admin-guide/kernel-parameters.txt         |  3 +
>  arch/x86/include/asm/cpu.h                    |  1 +
>  arch/x86/include/asm/realmode.h               |  3 +
>  arch/x86/include/asm/smp.h                    |  6 ++
>  arch/x86/kernel/acpi/sleep.c                  |  9 ++-
>  arch/x86/kernel/apic/apic.c                   |  2 +-
>  arch/x86/kernel/cpu/topology.c                |  3 +-
>  arch/x86/kernel/head_64.S                     | 65 +++++++++++++++++++
>  arch/x86/kernel/smpboot.c                     | 50 +++++++++++++-
>  arch/x86/realmode/init.c                      |  3 +
>  arch/x86/realmode/rm/trampoline_64.S          | 27 ++++++--
>  11 files changed, 163 insertions(+), 9 deletions(-)
>
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index 6cfa6e3996cf..7bb7020f97e2 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -3819,6 +3819,9 @@
>
>         nomodule        Disable module load
>
> +       no_parallel_bringup
> +                       [X86,SMP] Disable parallel bring-up of secondary cores.
> +
>         nopat           [X86] Disable PAT (page attribute table extension of
>                         pagetables) support.
>
> diff --git a/arch/x86/include/asm/cpu.h b/arch/x86/include/asm/cpu.h
> index 78796b98a544..ef8ba318dca1 100644
> --- a/arch/x86/include/asm/cpu.h
> +++ b/arch/x86/include/asm/cpu.h
> @@ -97,5 +97,6 @@ static inline bool intel_cpu_signatures_match(unsigned int s1, unsigned int p1,
>  extern u64 x86_read_arch_cap_msr(void);
>  int intel_find_matching_signature(void *mc, unsigned int csig, int cpf);
>  int intel_microcode_sanity_check(void *mc, bool print_err, int hdr_type);
> +int check_extended_topology_leaf(int leaf);
>
>  #endif /* _ASM_X86_CPU_H */
> diff --git a/arch/x86/include/asm/realmode.h b/arch/x86/include/asm/realmode.h
> index f6a1737c77be..87e5482acd0d 100644
> --- a/arch/x86/include/asm/realmode.h
> +++ b/arch/x86/include/asm/realmode.h
> @@ -52,6 +52,7 @@ struct trampoline_header {
>         u64 efer;
>         u32 cr4;
>         u32 flags;
> +       u32 lock;
>  #endif
>  };
>
> @@ -64,6 +65,8 @@ extern unsigned long initial_stack;
>  extern unsigned long initial_vc_handler;
>  #endif
>
> +extern u32 *trampoline_lock;
> +
>  extern unsigned char real_mode_blob[];
>  extern unsigned char real_mode_relocs[];
>
> diff --git a/arch/x86/include/asm/smp.h b/arch/x86/include/asm/smp.h
> index bf2c51df9e0b..1cf4f1e57570 100644
> --- a/arch/x86/include/asm/smp.h
> +++ b/arch/x86/include/asm/smp.h
> @@ -203,4 +203,10 @@ extern unsigned int smpboot_control;
>
>  #endif /* !__ASSEMBLY__ */
>
> +/* Control bits for startup_64 */
> +#define STARTUP_APICID_CPUID_0B        0x80000000
> +#define STARTUP_APICID_CPUID_01        0x40000000
> +
> +#define STARTUP_PARALLEL_MASK (STARTUP_APICID_CPUID_01 | STARTUP_APICID_CPUID_0B)
> +
>  #endif /* _ASM_X86_SMP_H */
> diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
> index 1328c221af30..6dfecb27b846 100644
> --- a/arch/x86/kernel/acpi/sleep.c
> +++ b/arch/x86/kernel/acpi/sleep.c
> @@ -16,6 +16,7 @@
>  #include <asm/cacheflush.h>
>  #include <asm/realmode.h>
>  #include <asm/hypervisor.h>
> +#include <asm/smp.h>
>
>  #include <linux/ftrace.h>
>  #include "../../realmode/rm/wakeup.h"
> @@ -127,7 +128,13 @@ int x86_acpi_suspend_lowlevel(void)
>          * value is in the actual %rsp register.
>          */
>         current->thread.sp = (unsigned long)temp_stack + sizeof(temp_stack);
> -       smpboot_control = smp_processor_id();
> +       /*
> +        * Ensure the CPU knows which one it is when it comes back, if
> +        * it isn't in parallel mode and expected to work that out for
> +        * itself.
> +        */
> +       if (!(smpboot_control & STARTUP_PARALLEL_MASK))
> +               smpboot_control = smp_processor_id();
>  #endif
>         initial_code = (unsigned long)wakeup_long64;
>         saved_magic = 0x123456789abcdef0L;
> diff --git a/arch/x86/kernel/apic/apic.c b/arch/x86/kernel/apic/apic.c
> index 20d9a604da7c..ac1d7e5da1f2 100644
> --- a/arch/x86/kernel/apic/apic.c
> +++ b/arch/x86/kernel/apic/apic.c
> @@ -2377,7 +2377,7 @@ static int nr_logical_cpuids = 1;
>  /*
>   * Used to store mapping between logical CPU IDs and APIC IDs.
>   */
> -static int cpuid_to_apicid[] = {
> +int cpuid_to_apicid[] = {
>         [0 ... NR_CPUS - 1] = -1,
>  };
>
> diff --git a/arch/x86/kernel/cpu/topology.c b/arch/x86/kernel/cpu/topology.c
> index 5e868b62a7c4..4373442e500a 100644
> --- a/arch/x86/kernel/cpu/topology.c
> +++ b/arch/x86/kernel/cpu/topology.c
> @@ -9,6 +9,7 @@
>  #include <asm/apic.h>
>  #include <asm/memtype.h>
>  #include <asm/processor.h>
> +#include <asm/cpu.h>
>
>  #include "cpu.h"
>
> @@ -32,7 +33,7 @@ EXPORT_SYMBOL(__max_die_per_package);
>  /*
>   * Check if given CPUID extended topology "leaf" is implemented
>   */
> -static int check_extended_topology_leaf(int leaf)
> +int check_extended_topology_leaf(int leaf)
>  {
>         unsigned int eax, ebx, ecx, edx;
>
> diff --git a/arch/x86/kernel/head_64.S b/arch/x86/kernel/head_64.S
> index 6a8238702eab..65bca47d84a1 100644
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -25,6 +25,7 @@
>  #include <asm/export.h>
>  #include <asm/nospec-branch.h>
>  #include <asm/fixmap.h>
> +#include <asm/smp.h>
>
>  /*
>   * We are not able to switch in one step to the final KERNEL ADDRESS SPACE
> @@ -234,8 +235,61 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>         ANNOTATE_NOENDBR // above
>
>  #ifdef CONFIG_SMP
> +       /*
> +        * For parallel boot, the APIC ID is retrieved from CPUID, and then
> +        * used to look up the CPU number.  For booting a single CPU, the
> +        * CPU number is encoded in smpboot_control.
> +        *
> +        * Bit 31       STARTUP_APICID_CPUID_0B flag (use CPUID 0x0b)
> +        * Bit 30       STARTUP_APICID_CPUID_01 flag (use CPUID 0x01)
> +        * Bit 0-24     CPU# if STARTUP_APICID_CPUID_xx flags are not set
> +        */
>         movl    smpboot_control(%rip), %ecx
> +       testl   $STARTUP_APICID_CPUID_0B, %ecx
> +       jnz     .Luse_cpuid_0b
> +       testl   $STARTUP_APICID_CPUID_01, %ecx
> +       jnz     .Luse_cpuid_01
> +       andl    $0x0FFFFFFF, %ecx
> +       jmp     .Lsetup_cpu
> +
> +.Luse_cpuid_01:
> +       mov     $0x01, %eax
> +       cpuid
> +       mov     %ebx, %edx
> +       shr     $24, %edx
> +       jmp     .Lsetup_AP
>
> +.Luse_cpuid_0b:
> +       mov     $0x0B, %eax
> +       xorl    %ecx, %ecx
> +       cpuid
> +
> +.Lsetup_AP:
> +       /* EDX contains the APIC ID of the current CPU */
> +       xorq    %rcx, %rcx
> +       leaq    cpuid_to_apicid(%rip), %rbx
> +
> +.Lfind_cpunr:
> +       cmpl    (%rbx,%rcx,4), %edx
> +       jz      .Lsetup_cpu
> +       inc     %ecx
> +#ifdef CONFIG_FORCE_NR_CPUS
> +       cmpl    $NR_CPUS, %ecx
> +#else
> +       cmpl    nr_cpu_ids(%rip), %ecx
> +#endif
> +       jb      .Lfind_cpunr
> +
> +       /*  APIC ID not found in the table. Drop the trampoline lock and bail. */
> +       movq    trampoline_lock(%rip), %rax
> +       lock
> +       btrl    $0, (%rax)
> +
> +1:     cli
> +       hlt
> +       jmp     1b
> +
> +.Lsetup_cpu:
>         /* Get the per cpu offset for the given CPU# which is in ECX */
>         movq    __per_cpu_offset(,%rcx,8), %rdx
>  #else
> @@ -264,6 +318,14 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>         lgdt    (%rsp)
>         addq    $16, %rsp
>
> +       /* Drop the realmode protection. For the boot CPU the pointer is NULL! */
> +       movq    trampoline_lock(%rip), %rax
> +       testq   %rax, %rax
> +       jz      .Lsetup_data_segments
> +       lock
> +       btrl    $0, (%rax)
> +
> +.Lsetup_data_segments:
>         /* set up data segments */
>         xorl %eax,%eax
>         movl %eax,%ds

This can still go earlier, right after "movq TASK_threadsp(%rax),
%rsp".  The GDT descriptor is placed on the idle thread stack, so it's
safe to drop the lock before it.

--
Brian Gerst

Re: [PATCH v15 09/12] x86/smpboot: Support parallel startup of secondary CPUs

[David Woodhouse]

[-- Attachment #1: Type: text/plain, Size: 2102 bytes --]

On Tue, 2023-03-21 at 14:28 -0400, Brian Gerst wrote:
> 
> > @@ -264,6 +318,14 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
> >          lgdt    (%rsp)
> >          addq    $16, %rsp
> > 
> > +       /* Drop the realmode protection. For the boot CPU the pointer is NULL! */
> > +       movq    trampoline_lock(%rip), %rax
> > +       testq   %rax, %rax
> > +       jz      .Lsetup_data_segments
> > +       lock
> > +       btrl    $0, (%rax)
> > +
> > +.Lsetup_data_segments:
> >          /* set up data segments */
> >          xorl %eax,%eax
> >          movl %eax,%ds
> 
> This can still go earlier, right after "movq TASK_threadsp(%rax),
> %rsp".  The GDT descriptor is placed on the idle thread stack, so it's
> safe to drop the lock before it.


--- a/arch/x86/kernel/head_64.S
+++ b/arch/x86/kernel/head_64.S
@@ -335,6 +335,17 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
        movq    pcpu_hot + X86_current_task(%rdx), %rax
        movq    TASK_threadsp(%rax), %rsp
 
+       /*
+        * Now that this CPU is running on its own stack, drop the realmode
+        * protection. For the boot CPU the pointer is NULL!
+        */
+       movq    trampoline_lock(%rip), %rax
+       testq   %rax, %rax
+       jz      .Lsetup_gdt
+       lock
+       btrl    $0, (%rax)
+
+.Lsetup_gdt:
        /*
         * We must switch to a new descriptor in kernel space for the GDT
         * because soon the kernel won't have access anymore to the userspace
@@ -377,14 +388,6 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
        shrq    $32, %rdx
        wrmsr
 
-       /* Drop the realmode protection. For the boot CPU the pointer is NULL! */
-       movq    trampoline_lock(%rip), %rax
-       testq   %rax, %rax
-       jz      .Lsetup_idt
-       lock
-       btrl    $0, (%rax)
-
-.Lsetup_idt:
        /* Setup and Load IDT */
        pushq   %rsi
        call    early_setup_idt


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5965 bytes --]

Re: [PATCH v15 03/12] cpu/hotplug: Add dynamic parallel bringup states before CPUHP_BRINGUP_CPU

[David Woodhouse]

[-- Attachment #1: Type: text/plain, Size: 2329 bytes --]

On Mon, 2023-03-20 at 15:30 +0100, Thomas Gleixner wrote:
> 
> This causes a subtle issue. The bringup loop above moves all CPUs to
> cpuhp_state == CPUHP_BP_PARALLEL_DYN_END. So the serial bootup will
> start from there and bring them fully up.
> 
> Now if a bringup fails, then the rollback will only go back down to
> CPUHP_BP_PARALLEL_DYN_END, which means that the control CPU won't do any
> cleanups below CPUHP_BP_PARALLEL_DYN_END.
> 
> That 'fail' is a common case for SMT soft disable via the 'nosmt'
> command line parameter. Due to the marvelous MCE broadcast 'feature' we
> need to bringup the SMT siblings at least to the CPUHP_AP_ONLINE_IDLE
> state once and then roll them back.
> 
> While this is not necessarily a fatal problem, it's changing behaviour
> and with quite some of the details hidden in the (then not issued)
> teardown callbacks might cause some hard to decode subtle surprises.
> 
> So that second for_each_present_cpu() loop needs to check the return
> value of cpu_up() and issue a full rollback to CPUHP_OFFLINE in case of
> fail.

@@ -1524,8 +1531,22 @@ void bringup_nonboot_cpus(unsigned int setup_max_cpus)
        for_each_present_cpu(cpu) {
                if (num_online_cpus() >= setup_max_cpus)
                        break;
-               if (!cpu_online(cpu))
-                       cpu_up(cpu, CPUHP_ONLINE);
+               if (!cpu_online(cpu)) {
+                       int ret = cpu_up(cpu, CPUHP_ONLINE);
+
+                       /*
+                        * For the parallel bringup case, roll all the way back
+                        * to CPUHP_OFFLINE on failure; don't leave them in the
+                        * parallel stages. This happens in the nosmt case for
+                        * non-primary threads.
+                        */
+                       if (ret && cpuhp_hp_states[CPUHP_BP_PARALLEL_DYN].name) {
+                               struct cpuhp_cpu_state *st = per_cpu_ptr(&cpuhp_state, cpu);
+                               if (can_rollback_cpu(st))
+                                       WARN_ON(cpuhp_invoke_callback_range(false, cpu, st,
+                                                                           CPUHP_OFFLINE));
+                       }
+               }
        }
 }
 


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5965 bytes --]

Re: [PATCH v15 02/12] cpu/hotplug: Move idle_thread_get() to <linux/smpboot.h>

[David Woodhouse]

[-- Attachment #1: Type: text/plain, Size: 1447 bytes --]

On Mon, 2023-03-20 at 08:17 +0000, David Woodhouse wrote:
> 
> In _cpu_up() the call to idle_thread_get() only happens if the CPU in
> question is starting from CPUHP_OFFLINE but I think that's sufficient?
> 
> diff --git a/kernel/cpu.c b/kernel/cpu.c
> index 6c0a92ca6bb5..43e0a77f21e8 100644
> --- a/kernel/cpu.c
> +++ b/kernel/cpu.c
> @@ -591,12 +591,6 @@ static int bringup_cpu(unsigned int cpu)
>  	struct task_struct *idle = idle_thread_get(cpu);
>  	int ret;
>  
> -	/*
> -	 * Reset stale stack state from the last time this CPU was online.
> -	 */
> -	scs_task_reset(idle);
> -	kasan_unpoison_task_stack(idle);
> -
>  	/*
>  	 * Some architectures have to walk the irq descriptors to
>  	 * setup the vector space for the cpu which comes online.
> @@ -1383,6 +1377,12 @@ static int _cpu_up(unsigned int cpu, int tasks_frozen, enum cpuhp_state target)
>  			ret = PTR_ERR(idle);
>  			goto out;
>  		}
> +
> +		/*
> +		 * Reset stale stack state from the last time this CPU was online.
> +		 */
> +		scs_task_reset(idle);
> +		kasan_unpoison_task_stack(idle);
>  	}
>  
>  	cpuhp_tasks_frozen = tasks_frozen;

I put that into a separate commit, rebased on tip/x86/apic and pushed
it out with the rest of the changes to
https://git.infradead.org/users/dwmw2/linux.git/shortlog/refs/heads/parallel-6.2-rc3-v16

I'll let Usama do another round of testing and repost it (please).

Thanks!

[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5965 bytes --]

Re: [PATCH v15 09/12] x86/smpboot: Support parallel startup of secondary CPUs

[Brian Gerst]

On Tue, Mar 21, 2023 at 3:12 PM David Woodhouse <dwmw2@infradead.org> wrote:
>
> On Tue, 2023-03-21 at 14:28 -0400, Brian Gerst wrote:
> >
> > > @@ -264,6 +318,14 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
> > >          lgdt    (%rsp)
> > >          addq    $16, %rsp
> > >
> > > +       /* Drop the realmode protection. For the boot CPU the pointer is NULL! */
> > > +       movq    trampoline_lock(%rip), %rax
> > > +       testq   %rax, %rax
> > > +       jz      .Lsetup_data_segments
> > > +       lock
> > > +       btrl    $0, (%rax)
> > > +
> > > +.Lsetup_data_segments:
> > >          /* set up data segments */
> > >          xorl %eax,%eax
> > >          movl %eax,%ds
> >
> > This can still go earlier, right after "movq TASK_threadsp(%rax),
> > %rsp".  The GDT descriptor is placed on the idle thread stack, so it's
> > safe to drop the lock before it.
>
>
> --- a/arch/x86/kernel/head_64.S
> +++ b/arch/x86/kernel/head_64.S
> @@ -335,6 +335,17 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>         movq    pcpu_hot + X86_current_task(%rdx), %rax
>         movq    TASK_threadsp(%rax), %rsp
>
> +       /*
> +        * Now that this CPU is running on its own stack, drop the realmode
> +        * protection. For the boot CPU the pointer is NULL!
> +        */
> +       movq    trampoline_lock(%rip), %rax
> +       testq   %rax, %rax
> +       jz      .Lsetup_gdt
> +       lock
> +       btrl    $0, (%rax)
> +
> +.Lsetup_gdt:
>         /*
>          * We must switch to a new descriptor in kernel space for the GDT
>          * because soon the kernel won't have access anymore to the userspace
> @@ -377,14 +388,6 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
>         shrq    $32, %rdx
>         wrmsr
>
> -       /* Drop the realmode protection. For the boot CPU the pointer is NULL! */
> -       movq    trampoline_lock(%rip), %rax
> -       testq   %rax, %rax
> -       jz      .Lsetup_idt
> -       lock
> -       btrl    $0, (%rax)
> -
> -.Lsetup_idt:
>         /* Setup and Load IDT */
>         pushq   %rsi
>         call    early_setup_idt
>

Looks good, thanks.