[PATCH PTI v3 07/10] x86/pti: Map the vsyscall page if needed

[Andy Lutomirski <luto@kernel.org>]

Make VSYSCALLs work fully in PTI mode.

Signed-off-by: Andy Lutomirski <luto@kernel.org>
---
 arch/x86/entry/vsyscall/vsyscall_64.c |  6 ++--
 arch/x86/include/asm/pgtable.h        |  6 +++-
 arch/x86/include/asm/pgtable_64.h     |  9 +++--
 arch/x86/include/asm/vsyscall.h       |  1 +
 arch/x86/mm/pti.c                     | 63 +++++++++++++++++++++++++++++++++++
 5 files changed, 78 insertions(+), 7 deletions(-)

diff --git a/arch/x86/entry/vsyscall/vsyscall_64.c b/arch/x86/entry/vsyscall/vsyscall_64.c
index a06f2ae09ad6..e4a6fe8354f0 100644
--- a/arch/x86/entry/vsyscall/vsyscall_64.c
+++ b/arch/x86/entry/vsyscall/vsyscall_64.c
@@ -343,14 +343,14 @@ int in_gate_area_no_mm(unsigned long addr)
  * vsyscalls but leave the page not present.  If so, we skip calling
  * this.
  */
-static void __init set_vsyscall_pgtable_user_bits(void)
+void __init set_vsyscall_pgtable_user_bits(pgd_t *root)
 {
 	pgd_t *pgd;
 	p4d_t *p4d;
 	pud_t *pud;
 	pmd_t *pmd;
 
-	pgd = pgd_offset_k(VSYSCALL_ADDR);
+	pgd = pgd_offset_pgd(root, VSYSCALL_ADDR);
 	pgd->pgd |= _PAGE_USER;
 	p4d = p4d_offset(pgd, VSYSCALL_ADDR);
 #if CONFIG_PGTABLE_LEVELS >= 5
@@ -372,7 +372,7 @@ void __init map_vsyscall(void)
 			     vsyscall_mode == NATIVE
 			     ? PAGE_KERNEL_VSYSCALL
 			     : PAGE_KERNEL_VVAR);
-		set_vsyscall_pgtable_user_bits();
+		set_vsyscall_pgtable_user_bits(swapper_pg_dir);
 	}
 
 	BUILD_BUG_ON((unsigned long)__fix_to_virt(VSYSCALL_PAGE) !=
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 83c0c77e7365..a8a8fc15ca16 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -920,7 +920,11 @@ static inline int pgd_none(pgd_t pgd)
  * pgd_offset() returns a (pgd_t *)
  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
  */
-#define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
+#define pgd_offset_pgd(pgd, address) (pgd + pgd_index((address)))
+/*
+ * a shortcut to get a pgd_t in a given mm
+ */
+#define pgd_offset(mm, address) pgd_offset_pgd((mm)->pgd, (address))
 /*
  * a shortcut which implies the use of the kernel's pgd, instead
  * of a process's
diff --git a/arch/x86/include/asm/pgtable_64.h b/arch/x86/include/asm/pgtable_64.h
index be8d086de927..a2fb3f8bc985 100644
--- a/arch/x86/include/asm/pgtable_64.h
+++ b/arch/x86/include/asm/pgtable_64.h
@@ -220,11 +220,14 @@ static inline pgd_t pti_set_user_pgd(pgd_t *pgdp, pgd_t pgd)
 		 * the wrong CR3.
 		 *
 		 * As exceptions, we don't set NX if:
-		 *  - this is EFI or similar, the kernel may execute from it
+		 *  - _PAGE_USER is not set.  This could be an executable
+		 *     EFI runtime mapping or something similar, and the kernel
+		 *     may execute from it
 		 *  - we don't have NX support
-		 *  - we're clearing the PGD (i.e. pgd.pgd == 0).
+		 *  - we're clearing the PGD (i.e. the new pgd is not present).
 		 */
-		if ((pgd.pgd & _PAGE_USER) && (__supported_pte_mask & _PAGE_NX))
+		if ((pgd.pgd & (_PAGE_USER|_PAGE_PRESENT)) == (_PAGE_USER|_PAGE_PRESENT) &&
+		    (__supported_pte_mask & _PAGE_NX))
 			pgd.pgd |= _PAGE_NX;
 	} else {
 		/*
diff --git a/arch/x86/include/asm/vsyscall.h b/arch/x86/include/asm/vsyscall.h
index d9a7c659009c..b986b2ca688a 100644
--- a/arch/x86/include/asm/vsyscall.h
+++ b/arch/x86/include/asm/vsyscall.h
@@ -7,6 +7,7 @@
 
 #ifdef CONFIG_X86_VSYSCALL_EMULATION
 extern void map_vsyscall(void);
+extern void set_vsyscall_pgtable_user_bits(pgd_t *root);
 
 /*
  * Called on instruction fetch fault in vsyscall page.
diff --git a/arch/x86/mm/pti.c b/arch/x86/mm/pti.c
index e01c4aa3ec73..b984e2311969 100644
--- a/arch/x86/mm/pti.c
+++ b/arch/x86/mm/pti.c
@@ -38,6 +38,7 @@
 
 #include <asm/cpufeature.h>
 #include <asm/hypervisor.h>
+#include <asm/vsyscall.h>
 #include <asm/cmdline.h>
 #include <asm/pti.h>
 #include <asm/pgtable.h>
@@ -145,6 +146,48 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address)
 	return pmd_offset(pud, address);
 }
 
+/*
+ * Walk the shadow copy of the page tables (optionally) trying to allocate
+ * page table pages on the way down.  Does not support large pages.
+ *
+ * Note: this is only used when mapping *new* kernel data into the
+ * user/shadow page tables.  It is never used for userspace data.
+ *
+ * Returns a pointer to a PTE on success, or NULL on failure.
+ */
+static pte_t *pti_user_pagetable_walk_pte(unsigned long address)
+{
+	gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
+	pmd_t *pmd = pti_user_pagetable_walk_pmd(address);
+	pte_t *pte;
+
+	/* We can't do anything sensible if we hit a large mapping. */
+	if (pmd_large(*pmd)) {
+		WARN_ON(1);
+		return NULL;
+	}
+
+	if (pmd_none(*pmd)) {
+		unsigned long new_pte_page = __get_free_page(gfp);
+		if (!new_pte_page)
+			return NULL;
+
+		if (pmd_none(*pmd)) {
+			set_pmd(pmd, __pmd(_KERNPG_TABLE | __pa(new_pte_page)));
+			new_pte_page = 0;
+		}
+		if (new_pte_page)
+			free_page(new_pte_page);
+	}
+
+	pte = pte_offset_kernel(pmd, address);
+	if (pte_flags(*pte) & _PAGE_USER) {
+		WARN_ONCE(1, "attempt to walk to user pte\n");
+		return NULL;
+	}
+	return pte;
+}
+
 static void __init
 pti_clone_pmds(unsigned long start, unsigned long end, pmdval_t clear)
 {
@@ -205,6 +248,25 @@ static void __init pti_setup_espfix64(void)
 #endif
 }
 
+static void __init pti_setup_vsyscall(void)
+{
+#ifdef CONFIG_X86_VSYSCALL_EMULATION
+	pte_t *pte, *target_pte;
+	unsigned int level;
+
+	pte = lookup_address(VSYSCALL_ADDR, &level);
+	if (!pte || WARN_ON(level != PG_LEVEL_4K))
+		return;
+
+	target_pte = pti_user_pagetable_walk_pte(VSYSCALL_ADDR);
+	if (WARN_ON(!target_pte))
+		return;
+
+	*target_pte = *pte;
+	set_vsyscall_pgtable_user_bits(kernel_to_user_pgdp(swapper_pg_dir));
+#endif
+}
+
 /*
  * Clone the populated PMDs of the user shared fixmaps into the user space
  * visible page table.
@@ -244,4 +306,5 @@ void __init pti_init(void)
 	pti_clone_user_shared();
 	pti_clone_entry_text();
 	pti_setup_espfix64();
+	pti_setup_vsyscall();
 }
-- 
2.13.6