lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:	Sun, 16 Nov 2014 19:30:56 -0800
From:	Yinghai Lu <yinghai@...nel.org>
To:	Thomas Gleixner <tglx@...utronix.de>
Cc:	Kees Cook <keescook@...omium.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Ingo Molnar <mingo@...hat.com>,
	"H. Peter Anvin" <hpa@...or.com>,
	"the arch/x86 maintainers" <x86@...nel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Andy Lutomirski <luto@...capital.net>,
	Yasuaki Ishimatsu <isimatu.yasuaki@...fujitsu.com>,
	Wang Nan <wangnan0@...wei.com>,
	David Vrabel <david.vrabel@...rix.com>
Subject: Re: [PATCH v2] x86, mm: set NX across entire PMD at boot

On Sun, Nov 16, 2014 at 1:26 PM, Thomas Gleixner <tglx@...utronix.de> wrote:
> On Sat, 15 Nov 2014, Yinghai Lu wrote:
>> +static pmd_t *last_pmd;
>>  /*
>>   * The head.S code sets up the kernel high mapping:
>>   *
>> @@ -408,9 +409,26 @@ void __init cleanup_highmap(void)
>>              continue;
>>          if (vaddr < (unsigned long) _text || vaddr > end)
>>              set_pmd(pmd, __pmd(0));
>> +        else
>> +            last_pmd = pmd;
>
> Why do you need to store this? You can compute this.

I'm not quite sure about the xen path.

>
>> +static void __init cleanup_highmap_tail(unsigned long addr)
>
> Brilliant stuff. mark_rodata_ro() is called AFTER free_initmem() which
> will free exactly that code.

I missed that.

Please check this one that address three problems that you pointed out.

Subject: [PATCH v2] x86, 64bit: cleanup highmap tail near partial 2M range

1. should use _brk_end instead of &_end in mark_rodata_ro().
   _brk_end can move up to &_end, i.e. to __brk_limit.  It's safe to
   use _brk_end when mark_rodata_ro() is called because extend_brk()
   is gone already at that point.
2. [_brk_end, pm_end) page range is already converted mem. and
   is not wasted.
3. add cleanup_highmap_tail for [_brk_end, pm_end).

Kernel Layout:
[    0.000000]    .brk: [0x0437c000-0x043a1fff]

Actually used brk:
[    0.272959] memblock_reserve: [0x0000000437c000-0x00000004382fff]
flags 0x0 BRK

Before patch:
---[ High Kernel Mapping ]---
...
0xffffffff83400000-0xffffffff84200000          14M     RW         PSE GLB NX pmd
0xffffffff84200000-0xffffffff843a2000        1672K     RW             GLB NX pte
0xffffffff843a2000-0xffffffff84400000         376K     RW             GLB x  pte
0xffffffff84400000-0xffffffffa0000000         444M                           pmd
After patch:
---[ High Kernel Mapping ]---
...
0xffffffff83400000-0xffffffff84200000          14M     RW         PSE GLB NX pmd
0xffffffff84200000-0xffffffff84383000        1548K     RW             GLB NX pte
0xffffffff84383000-0xffffffff84400000         500K                           pte
0xffffffff84400000-0xffffffffa0000000         444M                           pmd

-v2: according to tglx
 caculate the pmd postion instead of passing last_pmd.
 cleanup_highmap_tail could not have __init, as it is called in mark_rodata_ro
 and mark_rodata_ro is called after free_initmem.
 highmap_end_pfn should keep PMD_SIZE alignment on !CONFIG_DEBUG_RODATA

Signed-off-by: Yinghai Lu <yinghai@...nel.org>

---
 arch/x86/mm/init_64.c  |   22 +++++++++++++++++++++-
 arch/x86/mm/pageattr.c |    4 ++++
 2 files changed, 25 insertions(+), 1 deletion(-)

Index: linux-2.6/arch/x86/mm/init_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init_64.c
+++ linux-2.6/arch/x86/mm/init_64.c
@@ -411,6 +411,23 @@ void __init cleanup_highmap(void)
     }
 }

+static void cleanup_highmap_tail(unsigned long addr)
+{
+    int i;
+    pgd_t *pgd;
+    pud_t *pud;
+    pmd_t *pmd;
+    pte_t *pte;
+
+    pgd = pgd_offset_k(addr);
+    pud = (pud_t *)pgd_page_vaddr(*pgd) + pud_index(addr);
+    pmd = (pmd_t *)pud_page_vaddr(*pud) + pmd_index(addr);
+    pte = (pte_t *)pmd_page_vaddr(*pmd) + pte_index(addr);
+
+    for (i = pte_index(addr); i < PTRS_PER_PTE; i++, pte++)
+        set_pte(pte, __pte(0));
+}
+
 static unsigned long __meminit
 phys_pte_init(pte_t *pte_page, unsigned long addr, unsigned long end,
           pgprot_t prot)
@@ -1124,7 +1141,8 @@ void mark_rodata_ro(void)
     unsigned long end = (unsigned long) &__end_rodata_hpage_align;
     unsigned long text_end = PFN_ALIGN(&__stop___ex_table);
     unsigned long rodata_end = PFN_ALIGN(&__end_rodata);
-    unsigned long all_end = PFN_ALIGN(&_end);
+    unsigned long all_end = PFN_ALIGN(_brk_end);
+    unsigned long pmd_end = roundup(all_end, PMD_SIZE);

     printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
            (end - start) >> 10);
@@ -1137,6 +1155,8 @@ void mark_rodata_ro(void)
      * should also be not-executable.
      */
     set_memory_nx(rodata_start, (all_end - rodata_start) >> PAGE_SHIFT);
+    if (all_end < pmd_end)
+        cleanup_highmap_tail(all_end);

     rodata_test();

Index: linux-2.6/arch/x86/mm/pageattr.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/pageattr.c
+++ linux-2.6/arch/x86/mm/pageattr.c
@@ -100,7 +100,11 @@ static inline unsigned long highmap_star

 static inline unsigned long highmap_end_pfn(void)
 {
+#ifdef CONFIG_DEBUG_RODATA
+    return __pa_symbol(PFN_ALIGN(_brk_end)) >> PAGE_SHIFT;
+#else
     return __pa_symbol(roundup(_brk_end, PMD_SIZE)) >> PAGE_SHIFT;
+#endif
 }

 #endif

View attachment "nx_after_end_v2.patch" of type "text/x-patch" (3817 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ