x86: Fix improper large page preservation

This patch fixes a bug in try_preserve_large_page() which may
result in improper large page preservation and improper
application of page attributes to the memory area outside of the
original change request.

More specifically, the problem manifests itself when set_memory_*()
is called for several pages at the beginning of the large page and
try_preserve_large_page() erroneously concludes that the change can
be applied to whole large page.

The fix consists of 3 parts:

  1. Addition of "required" protection attributes in
     static_protections(), so .data and .bss can be guaranteed to
     stay "RW"

  2. static_protections() is now called for every small
     page within large page to determine compatibility of new
     protection attributes (instead of just small pages within the
     requested range).

  3. Large page can be preserved only if attribute change is
     large-page-aligned and covers whole large page.

 -v1: Try_preserve_large_page() patch for Linux 2.6.34-rc2
 -v2: Replaced pfn check with address check for kernel rw-data

Signed-off-by: Siarhei Liakh <sliakh.lkml@gmail.com>
Signed-off-by: Xuxian Jiang <jiang@cs.ncsu.edu>
Reviewed-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: James Morris <jmorris@namei.org>
Cc: Andi Kleen <ak@muc.de>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Dave Jones <davej@redhat.com>
Cc: Kees Cook <kees.cook@canonical.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <4CE2F7F3.8030809@free.fr>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index 532e7933d606fdbdde77aacdb2de746f249d2bf2..6f2a6b6deb6b1342faa0b02942eb40d93df13f60 100644 (file)
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -255,6 +255,7 @@ static inline pgprot_t static_protections(pgprot_t prot, unsigned long address,
                                   unsigned long pfn)
 {
        pgprot_t forbidden = __pgprot(0);
+       pgprot_t required = __pgprot(0);
 
        /*
         * The BIOS area between 640k and 1Mb needs to be executable for
@@ -278,6 +279,12 @@ static inline pgprot_t static_protections(pgprot_t prot, unsigned long address,
        if (within(pfn, __pa((unsigned long)__start_rodata) >> PAGE_SHIFT,
                   __pa((unsigned long)__end_rodata) >> PAGE_SHIFT))
                pgprot_val(forbidden) |= _PAGE_RW;
+       /*
+        * .data and .bss should always be writable.
+        */
+       if (within(address, (unsigned long)_sdata, (unsigned long)_edata) ||
+           within(address, (unsigned long)__bss_start, (unsigned long)__bss_stop))
+               pgprot_val(required) |= _PAGE_RW;
 
 #if defined(CONFIG_X86_64) && defined(CONFIG_DEBUG_RODATA)
        /*
@@ -317,6 +324,7 @@ static inline pgprot_t static_protections(pgprot_t prot, unsigned long address,
 #endif
 
        prot = __pgprot(pgprot_val(prot) & ~pgprot_val(forbidden));
+       prot = __pgprot(pgprot_val(prot) | pgprot_val(required));
 
        return prot;
 }
@@ -393,7 +401,7 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
 {
        unsigned long nextpage_addr, numpages, pmask, psize, flags, addr, pfn;
        pte_t new_pte, old_pte, *tmp;
-       pgprot_t old_prot, new_prot;
+       pgprot_t old_prot, new_prot, req_prot;
        int i, do_split = 1;
        unsigned int level;
 
@@ -438,10 +446,10 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
         * We are safe now. Check whether the new pgprot is the same:
         */
        old_pte = *kpte;
-       old_prot = new_prot = pte_pgprot(old_pte);
+       old_prot = new_prot = req_prot = pte_pgprot(old_pte);
 
-       pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
-       pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
+       pgprot_val(req_prot) &= ~pgprot_val(cpa->mask_clr);
+       pgprot_val(req_prot) |= pgprot_val(cpa->mask_set);
 
        /*
         * old_pte points to the large page base address. So we need
@@ -450,17 +458,17 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
        pfn = pte_pfn(old_pte) + ((address & (psize - 1)) >> PAGE_SHIFT);
        cpa->pfn = pfn;
 
-       new_prot = static_protections(new_prot, address, pfn);
+       new_prot = static_protections(req_prot, address, pfn);
 
        /*
         * We need to check the full range, whether
         * static_protection() requires a different pgprot for one of
         * the pages in the range we try to preserve:
         */
-       addr = address + PAGE_SIZE;
-       pfn++;
-       for (i = 1; i < cpa->numpages; i++, addr += PAGE_SIZE, pfn++) {
-               pgprot_t chk_prot = static_protections(new_prot, addr, pfn);
+       addr = address & pmask;
+       pfn = pte_pfn(old_pte);
+       for (i = 0; i < (psize >> PAGE_SHIFT); i++, addr += PAGE_SIZE, pfn++) {
+               pgprot_t chk_prot = static_protections(req_prot, addr, pfn);
 
                if (pgprot_val(chk_prot) != pgprot_val(new_prot))
                        goto out_unlock;
@@ -483,7 +491,7 @@ try_preserve_large_page(pte_t *kpte, unsigned long address,
         * that we limited the number of possible pages already to
         * the number of pages in the large page.
         */
-       if (address == (nextpage_addr - psize) && cpa->numpages == numpages) {
+       if (address == (address & pmask) && cpa->numpages == (psize >> PAGE_SHIFT)) {
                /*
                 * The address is aligned and the number of pages
                 * covers the full page.