1 files changed, 40 insertions, 22 deletions

diff --git a/arch/x86/mm/pti.c b/arch/x86/mm/pti.c
index 2e69abf4f852..c0c40b67524e 100644
--- a/arch/x86/mm/pti.c
+++ b/arch/x86/mm/pti.c
@@ -98,6 +98,11 @@ void __init pti_check_boottime_disable(void)
 		return;
 
 	setup_force_cpu_cap(X86_FEATURE_PTI);
+
+	if (cpu_feature_enabled(X86_FEATURE_INVLPGB)) {
+		pr_debug("PTI enabled, disabling INVLPGB\n");
+		setup_clear_cpu_cap(X86_FEATURE_INVLPGB);
+	}
 }
 
 static int __init pti_parse_cmdline(char *arg)
@@ -132,7 +137,7 @@ pgd_t __pti_set_user_pgtbl(pgd_t *pgdp, pgd_t pgd)
 	 * Top-level entries added to init_mm's usermode pgd after boot
 	 * will not be automatically propagated to other mms.
 	 */
-	if (!pgdp_maps_userspace(pgdp))
+	if (!pgdp_maps_userspace(pgdp) || (pgd.pgd & _PAGE_NOPTISHADOW))
 		return pgd;
 
 	/*
@@ -185,7 +190,7 @@ static p4d_t *pti_user_pagetable_walk_p4d(unsigned long address)
 
 		set_pgd(pgd, __pgd(_KERNPG_TABLE | __pa(new_p4d_page)));
 	}
-	BUILD_BUG_ON(pgd_leaf(*pgd) != 0);
+	BUILD_BUG_ON(pgd_leaf(*pgd));
 
 	return p4d_offset(pgd, address);
 }
@@ -206,7 +211,7 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address)
 	if (!p4d)
 		return NULL;
 
-	BUILD_BUG_ON(p4d_leaf(*p4d) != 0);
+	BUILD_BUG_ON(p4d_leaf(*p4d));
 	if (p4d_none(*p4d)) {
 		unsigned long new_pud_page = __get_free_page(gfp);
 		if (WARN_ON_ONCE(!new_pud_page))
@@ -241,7 +246,7 @@ 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 +256,15 @@ 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 +293,7 @@ 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 +311,7 @@ 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;
 
@@ -374,14 +384,14 @@ pti_clone_pgtable(unsigned long start, unsigned long end,
 			 */
 			*target_pmd = *pmd;
 
-			addr += PMD_SIZE;
+			addr = round_up(addr + 1, PMD_SIZE);
 
 		} else if (level == PTI_CLONE_PTE) {
 
 			/* Walk the page-table down to the pte level */
 			pte = pte_offset_kernel(pmd, addr);
 			if (pte_none(*pte)) {
-				addr += PAGE_SIZE;
+				addr = round_up(addr + 1, PAGE_SIZE);
 				continue;
 			}
 
@@ -390,7 +400,7 @@ 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;
 
@@ -401,7 +411,7 @@ pti_clone_pgtable(unsigned long start, unsigned long end,
 			/* Clone the PTE */
 			*target_pte = *pte;
 
-			addr += PAGE_SIZE;
+			addr = round_up(addr + 1, PAGE_SIZE);
 
 		} else {
 			BUG();
@@ -452,7 +462,7 @@ 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 +485,7 @@ 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 +502,11 @@ 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_CLONE_PMD);
+			  PTI_LEVEL_KERNEL_IMAGE, late);
 }
 
 /*
@@ -571,7 +581,7 @@ 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 +648,15 @@ 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 +673,11 @@ 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();