x86/mm/pti: Add functions to clone kernel PMDs
authorAndy Lutomirski <luto@kernel.org>
Mon, 4 Dec 2017 14:07:42 +0000 (15:07 +0100)
committerIngo Molnar <mingo@kernel.org>
Sat, 23 Dec 2017 20:13:00 +0000 (21:13 +0100)
Provide infrastructure to:

 - find a kernel PMD for a mapping which must be visible to user space for
   the entry/exit code to work.

 - walk an address range and share the kernel PMD with it.

This reuses a small part of the original KAISER patches to populate the
user space page table.

[ tglx: Made it universally usable so it can be used for any kind of shared
mapping. Add a mechanism to clear specific bits in the user space
visible PMD entry. Folded Andys simplifactions ]

Originally-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bpetkov@suse.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Laight <David.Laight@aculab.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: Eduardo Valentin <eduval@amazon.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: aliguori@amazon.com
Cc: daniel.gruss@iaik.tugraz.at
Cc: hughd@google.com
Cc: keescook@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/mm/pti.c

index 69a983365392961358ddbb719a9ab7fe5f4e2965..d58bcee470fcb6d527217d0717bfeb8618698a54 100644 (file)
 #undef pr_fmt
 #define pr_fmt(fmt)     "Kernel/User page tables isolation: " fmt
 
+/* Backporting helper */
+#ifndef __GFP_NOTRACK
+#define __GFP_NOTRACK  0
+#endif
+
 static void __init pti_print_if_insecure(const char *reason)
 {
        if (boot_cpu_has_bug(X86_BUG_CPU_INSECURE))
@@ -137,6 +142,128 @@ pgd_t __pti_set_user_pgd(pgd_t *pgdp, pgd_t pgd)
        return pgd;
 }
 
+/*
+ * Walk the user copy of the page tables (optionally) trying to allocate
+ * page table pages on the way down.
+ *
+ * Returns a pointer to a P4D on success, or NULL on failure.
+ */
+static p4d_t *pti_user_pagetable_walk_p4d(unsigned long address)
+{
+       pgd_t *pgd = kernel_to_user_pgdp(pgd_offset_k(address));
+       gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
+
+       if (address < PAGE_OFFSET) {
+               WARN_ONCE(1, "attempt to walk user address\n");
+               return NULL;
+       }
+
+       if (pgd_none(*pgd)) {
+               unsigned long new_p4d_page = __get_free_page(gfp);
+               if (!new_p4d_page)
+                       return NULL;
+
+               if (pgd_none(*pgd)) {
+                       set_pgd(pgd, __pgd(_KERNPG_TABLE | __pa(new_p4d_page)));
+                       new_p4d_page = 0;
+               }
+               if (new_p4d_page)
+                       free_page(new_p4d_page);
+       }
+       BUILD_BUG_ON(pgd_large(*pgd) != 0);
+
+       return p4d_offset(pgd, address);
+}
+
+/*
+ * Walk the user copy of the page tables (optionally) trying to allocate
+ * page table pages on the way down.
+ *
+ * Returns a pointer to a PMD on success, or NULL on failure.
+ */
+static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address)
+{
+       gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
+       p4d_t *p4d = pti_user_pagetable_walk_p4d(address);
+       pud_t *pud;
+
+       BUILD_BUG_ON(p4d_large(*p4d) != 0);
+       if (p4d_none(*p4d)) {
+               unsigned long new_pud_page = __get_free_page(gfp);
+               if (!new_pud_page)
+                       return NULL;
+
+               if (p4d_none(*p4d)) {
+                       set_p4d(p4d, __p4d(_KERNPG_TABLE | __pa(new_pud_page)));
+                       new_pud_page = 0;
+               }
+               if (new_pud_page)
+                       free_page(new_pud_page);
+       }
+
+       pud = pud_offset(p4d, address);
+       /* The user page tables do not use large mappings: */
+       if (pud_large(*pud)) {
+               WARN_ON(1);
+               return NULL;
+       }
+       if (pud_none(*pud)) {
+               unsigned long new_pmd_page = __get_free_page(gfp);
+               if (!new_pmd_page)
+                       return NULL;
+
+               if (pud_none(*pud)) {
+                       set_pud(pud, __pud(_KERNPG_TABLE | __pa(new_pmd_page)));
+                       new_pmd_page = 0;
+               }
+               if (new_pmd_page)
+                       free_page(new_pmd_page);
+       }
+
+       return pmd_offset(pud, address);
+}
+
+static void __init
+pti_clone_pmds(unsigned long start, unsigned long end, pmdval_t clear)
+{
+       unsigned long addr;
+
+       /*
+        * Clone the populated PMDs which cover start to end. These PMD areas
+        * can have holes.
+        */
+       for (addr = start; addr < end; addr += PMD_SIZE) {
+               pmd_t *pmd, *target_pmd;
+               pgd_t *pgd;
+               p4d_t *p4d;
+               pud_t *pud;
+
+               pgd = pgd_offset_k(addr);
+               if (WARN_ON(pgd_none(*pgd)))
+                       return;
+               p4d = p4d_offset(pgd, addr);
+               if (WARN_ON(p4d_none(*p4d)))
+                       return;
+               pud = pud_offset(p4d, addr);
+               if (pud_none(*pud))
+                       continue;
+               pmd = pmd_offset(pud, addr);
+               if (pmd_none(*pmd))
+                       continue;
+
+               target_pmd = pti_user_pagetable_walk_pmd(addr);
+               if (WARN_ON(!target_pmd))
+                       return;
+
+               /*
+                * Copy the PMD.  That is, the kernelmode and usermode
+                * tables will share the last-level page tables of this
+                * address range
+                */
+               *target_pmd = pmd_clear_flags(*pmd, clear);
+       }
+}
+
 /*
  * Initialize kernel page table isolation
  */