Merge tag 'x86-urgent-2024-08-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:

 - Fix 32-bit PTI for real.

   pti_clone_entry_text() is called twice, once before initcalls so that
   initcalls can use the user-mode helper and then again after text is
   set read only. Setting read only on 32-bit might break up the PMD
   mapping, which makes the second invocation of pti_clone_entry_text()
   find the mappings out of sync and failing.

   Allow the second call to split the existing PMDs in the user mapping
   and synchronize with the kernel mapping.

 - Don't make acpi_mp_wake_mailbox read-only after init as the mail box
   must be writable in the case that CPU hotplug operations happen after
   boot. Otherwise the attempt to start a CPU crashes with a write to
   read only memory.

 - Add a missing sanity check in mtrr_save_state() to ensure that the
   fixed MTRR MSRs are supported.

   Otherwise mtrr_save_state() ends up in a #GP, which is fixed up, but
   the WARN_ON() can bring systems down when panic on warn is set.

* tag 'x86-urgent-2024-08-11' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mtrr: Check if fixed MTRRs exist before saving them
  x86/paravirt: Fix incorrect virt spinlock setting on bare metal
  x86/acpi: Remove __ro_after_init from acpi_mp_wake_mailbox
  x86/mm: Fix PTI for i386 some more

5 files changed, 41 insertions, 27 deletions

diff --git a/arch/x86/include/asm/qspinlock.h b/arch/x86/include/asm/qspinlock.h
index a053c1293975..68da67df304d 100644
--- a/arch/x86/include/asm/qspinlock.h
+++ b/arch/x86/include/asm/qspinlock.h
@@ -66,13 +66,15 @@ static inline bool vcpu_is_preempted(long cpu)
 
 #ifdef CONFIG_PARAVIRT
 /*
- * virt_spin_lock_key - enables (by default) the virt_spin_lock() hijack.
+ * virt_spin_lock_key - disables by default the virt_spin_lock() hijack.
  *
- * Native (and PV wanting native due to vCPU pinning) should disable this key.
- * It is done in this backwards fashion to only have a single direction change,
- * which removes ordering between native_pv_spin_init() and HV setup.
+ * Native (and PV wanting native due to vCPU pinning) should keep this key
+ * disabled. Native does not touch the key.
+ *
+ * When in a guest then native_pv_lock_init() enables the key first and
+ * KVM/XEN might conditionally disable it later in the boot process again.
  */
-DECLARE_STATIC_KEY_TRUE(virt_spin_lock_key);
+DECLARE_STATIC_KEY_FALSE(virt_spin_lock_key);
 
 /*
  * Shortcut for the queued_spin_lock_slowpath() function that allows
diff --git a/arch/x86/kernel/acpi/madt_wakeup.c b/arch/x86/kernel/acpi/madt_wakeup.c
index 6cfe762be28b..d5ef6215583b 100644
--- a/arch/x86/kernel/acpi/madt_wakeup.c
+++ b/arch/x86/kernel/acpi/madt_wakeup.c
@@ -19,7 +19,7 @@
 static u64 acpi_mp_wake_mailbox_paddr __ro_after_init;
 
 /* Virtual address of the Multiprocessor Wakeup Structure mailbox */
-static struct acpi_madt_multiproc_wakeup_mailbox *acpi_mp_wake_mailbox __ro_after_init;
+static struct acpi_madt_multiproc_wakeup_mailbox *acpi_mp_wake_mailbox;
 
 static u64 acpi_mp_pgd __ro_after_init;
 static u64 acpi_mp_reset_vector_paddr __ro_after_init;
diff --git a/arch/x86/kernel/cpu/mtrr/mtrr.c b/arch/x86/kernel/cpu/mtrr/mtrr.c
index 767bf1c71aad..2a2fc14955cd 100644
--- a/arch/x86/kernel/cpu/mtrr/mtrr.c
+++ b/arch/x86/kernel/cpu/mtrr/mtrr.c
@@ -609,7 +609,7 @@ void mtrr_save_state(void)
 {
 	int first_cpu;
 
-	if (!mtrr_enabled())
+	if (!mtrr_enabled() || !mtrr_state.have_fixed)
 		return;
 
 	first_cpu = cpumask_first(cpu_online_mask);
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 5358d43886ad..fec381533555 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -51,13 +51,12 @@ DEFINE_ASM_FUNC(pv_native_irq_enable, "sti", .noinstr.text);
 DEFINE_ASM_FUNC(pv_native_read_cr2, "mov %cr2, %rax", .noinstr.text);
 #endif
 
-DEFINE_STATIC_KEY_TRUE(virt_spin_lock_key);
+DEFINE_STATIC_KEY_FALSE(virt_spin_lock_key);
 
 void __init native_pv_lock_init(void)
 {
-	if (IS_ENABLED(CONFIG_PARAVIRT_SPINLOCKS) &&
-	    !boot_cpu_has(X86_FEATURE_HYPERVISOR))
-		static_branch_disable(&virt_spin_lock_key);
+	if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
+		static_branch_enable(&virt_spin_lock_key);
 }
 
 static void native_tlb_remove_table(struct mmu_gather *tlb, void *table)
diff --git a/arch/x86/mm/pti.c b/arch/x86/mm/pti.c
index bfdf5f45b137..851ec8f1363a 100644
--- a/arch/x86/mm/pti.c
+++ b/arch/x86/mm/pti.c
@@ -241,7 +241,7 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address)
  *
  * Returns a pointer to a PTE on success, or NULL on failure.
  */
-static pte_t *pti_user_pagetable_walk_pte(unsigned long address)
+static pte_t *pti_user_pagetable_walk_pte(unsigned long address, bool late_text)
 {
 	gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
 	pmd_t *pmd;
@@ -251,10 +251,15 @@ static pte_t *pti_user_pagetable_walk_pte(unsigned long address)
 	if (!pmd)
 		return NULL;
 
-	/* We can't do anything sensible if we hit a large mapping. */
+	/* Large PMD mapping found */
 	if (pmd_leaf(*pmd)) {
-		WARN_ON(1);
-		return NULL;
+		/* Clear the PMD if we hit a large mapping from the first round */
+		if (late_text) {
+			set_pmd(pmd, __pmd(0));
+		} else {
+			WARN_ON_ONCE(1);
+			return NULL;
+		}
 	}
 
 	if (pmd_none(*pmd)) {
@@ -283,7 +288,7 @@ static void __init pti_setup_vsyscall(void)
 	if (!pte || WARN_ON(level != PG_LEVEL_4K) || pte_none(*pte))
 		return;
 
-	target_pte = pti_user_pagetable_walk_pte(VSYSCALL_ADDR);
+	target_pte = pti_user_pagetable_walk_pte(VSYSCALL_ADDR, false);
 	if (WARN_ON(!target_pte))
 		return;
 
@@ -301,7 +306,7 @@ enum pti_clone_level {
 
 static void
 pti_clone_pgtable(unsigned long start, unsigned long end,
-		  enum pti_clone_level level)
+		  enum pti_clone_level level, bool late_text)
 {
 	unsigned long addr;
 
@@ -390,7 +395,7 @@ pti_clone_pgtable(unsigned long start, unsigned long end,
 				return;
 
 			/* Allocate PTE in the user page-table */
-			target_pte = pti_user_pagetable_walk_pte(addr);
+			target_pte = pti_user_pagetable_walk_pte(addr, late_text);
 			if (WARN_ON(!target_pte))
 				return;
 
@@ -452,7 +457,7 @@ static void __init pti_clone_user_shared(void)
 		phys_addr_t pa = per_cpu_ptr_to_phys((void *)va);
 		pte_t *target_pte;
 
-		target_pte = pti_user_pagetable_walk_pte(va);
+		target_pte = pti_user_pagetable_walk_pte(va, false);
 		if (WARN_ON(!target_pte))
 			return;
 
@@ -475,7 +480,7 @@ static void __init pti_clone_user_shared(void)
 	start = CPU_ENTRY_AREA_BASE;
 	end   = start + (PAGE_SIZE * CPU_ENTRY_AREA_PAGES);
 
-	pti_clone_pgtable(start, end, PTI_CLONE_PMD);
+	pti_clone_pgtable(start, end, PTI_CLONE_PMD, false);
 }
 #endif /* CONFIG_X86_64 */
 
@@ -492,11 +497,11 @@ static void __init pti_setup_espfix64(void)
 /*
  * Clone the populated PMDs of the entry text and force it RO.
  */
-static void pti_clone_entry_text(void)
+static void pti_clone_entry_text(bool late)
 {
 	pti_clone_pgtable((unsigned long) __entry_text_start,
 			  (unsigned long) __entry_text_end,
-			  PTI_LEVEL_KERNEL_IMAGE);
+			  PTI_LEVEL_KERNEL_IMAGE, late);
 }
 
 /*
@@ -571,7 +576,7 @@ static void pti_clone_kernel_text(void)
 	 * pti_set_kernel_image_nonglobal() did to clear the
 	 * global bit.
 	 */
-	pti_clone_pgtable(start, end_clone, PTI_LEVEL_KERNEL_IMAGE);
+	pti_clone_pgtable(start, end_clone, PTI_LEVEL_KERNEL_IMAGE, false);
 
 	/*
 	 * pti_clone_pgtable() will set the global bit in any PMDs
@@ -638,8 +643,15 @@ void __init pti_init(void)
 
 	/* Undo all global bits from the init pagetables in head_64.S: */
 	pti_set_kernel_image_nonglobal();
+
 	/* Replace some of the global bits just for shared entry text: */
-	pti_clone_entry_text();
+	/*
+	 * This is very early in boot. Device and Late initcalls can do
+	 * modprobe before free_initmem() and mark_readonly(). This
+	 * pti_clone_entry_text() allows those user-mode-helpers to function,
+	 * but notably the text is still RW.
+	 */
+	pti_clone_entry_text(false);
 	pti_setup_espfix64();
 	pti_setup_vsyscall();
 }
@@ -656,10 +668,11 @@ void pti_finalize(void)
 	if (!boot_cpu_has(X86_FEATURE_PTI))
 		return;
 	/*
-	 * We need to clone everything (again) that maps parts of the
-	 * kernel image.
+	 * This is after free_initmem() (all initcalls are done) and we've done
+	 * mark_readonly(). Text is now NX which might've split some PMDs
+	 * relative to the early clone.
 	 */
-	pti_clone_entry_text();
+	pti_clone_entry_text(true);
 	pti_clone_kernel_text();
 
 	debug_checkwx_user();