Re: [PATCH v2] powerpc: Do not make the entire heap executable

From: Denys Vlasenko
Date: Tue Aug 09 2016 - 15:07:28 EST




On 08/08/2016 09:07 PM, Kees Cook wrote:
On Mon, Aug 8, 2016 at 7:55 AM, Denys Vlasenko <dvlasenk@xxxxxxxxxx> wrote:
On 32-bit powerps the ELF PLT sections of binaries (built with --bss-plt,
or with a toolchain which defaults to it) look like this:

[17] .sbss NOBITS 0002aff8 01aff8 000014 00 WA 0 0 4
[18] .plt NOBITS 0002b00c 01aff8 000084 00 WAX 0 0 4
[19] .bss NOBITS 0002b090 01aff8 0000a4 00 WA 0 0 4

Which results in an ELF load header:

Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
LOAD 0x019c70 0x00029c70 0x00029c70 0x01388 0x014c4 RWE 0x10000

This is all correct, the load region containing the PLT is marked as
executable. Note that the PLT starts at 0002b00c but the file mapping ends at
0002aff8, so the PLT falls in the 0 fill section described by the load header,
and after a page boundary.

Unfortunately the generic ELF loader ignores the X bit in the load headers
when it creates the 0 filled non-file backed mappings. It assumes all of these
mappings are RW BSS sections, which is not the case for PPC.

gcc/ld has an option (--secure-plt) to not do this, this is said to incur
a small performance penalty.

Currently, to support 32-bit binaries with PLT in BSS kernel maps *entire
brk area* with executable rights for all binaries, even --secure-plt ones.

Stop doing that.

Teach the ELF loader to check the X bit in the relevant load header
and create 0 filled anonymous mappings that are executable
if the load header requests that.

The patch was originally posted in 2012 by Jason Gunthorpe
and apparently ignored:

https://lkml.org/lkml/2012/9/30/138

Lightly run-tested.

Signed-off-by: Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
CC: Benjamin Herrenschmidt <benh@xxxxxxxxxxxxxxxxxxx>
CC: Paul Mackerras <paulus@xxxxxxxxx>
CC: Kees Cook <keescook@xxxxxxxxxxxx>
CC: Oleg Nesterov <oleg@xxxxxxxxxx>,
CC: Michael Ellerman <mpe@xxxxxxxxxxxxxx>
CC: Florian Weimer <fweimer@xxxxxxxxxx>
CC: linux-mm@xxxxxxxxx,
CC: linuxppc-dev@xxxxxxxxxxxxxxxx
CC: linux-kernel@xxxxxxxxxxxxxxx
---
Changes since v1:
* wrapped lines to not exceed 79 chars
* improved comment
* expanded CC list

arch/powerpc/include/asm/page.h | 10 +------
arch/powerpc/include/asm/page_32.h | 2 --
arch/powerpc/include/asm/page_64.h | 4 ---
fs/binfmt_elf.c | 56 ++++++++++++++++++++++++++++++--------
4 files changed, 45 insertions(+), 27 deletions(-)

diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h
index 56398e7..42d7ea1 100644
--- a/arch/powerpc/include/asm/page.h
+++ b/arch/powerpc/include/asm/page.h
@@ -225,15 +225,7 @@ extern long long virt_phys_offset;
#endif
#endif

-/*
- * Unfortunately the PLT is in the BSS in the PPC32 ELF ABI,
- * and needs to be executable. This means the whole heap ends
- * up being executable.
- */
-#define VM_DATA_DEFAULT_FLAGS32 (VM_READ | VM_WRITE | VM_EXEC | \
- VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
-
-#define VM_DATA_DEFAULT_FLAGS64 (VM_READ | VM_WRITE | \
+#define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | \
VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)

#ifdef __powerpc64__
diff --git a/arch/powerpc/include/asm/page_32.h b/arch/powerpc/include/asm/page_32.h
index 6a8e179..6113fa8 100644
--- a/arch/powerpc/include/asm/page_32.h
+++ b/arch/powerpc/include/asm/page_32.h
@@ -9,8 +9,6 @@
#endif
#endif

-#define VM_DATA_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS32
-
#ifdef CONFIG_NOT_COHERENT_CACHE
#define ARCH_DMA_MINALIGN L1_CACHE_BYTES
#endif
diff --git a/arch/powerpc/include/asm/page_64.h b/arch/powerpc/include/asm/page_64.h
index dd5f071..52d8e9c 100644
--- a/arch/powerpc/include/asm/page_64.h
+++ b/arch/powerpc/include/asm/page_64.h
@@ -159,10 +159,6 @@ do { \

#endif /* !CONFIG_HUGETLB_PAGE */

-#define VM_DATA_DEFAULT_FLAGS \
- (is_32bit_task() ? \
- VM_DATA_DEFAULT_FLAGS32 : VM_DATA_DEFAULT_FLAGS64)
-
/*
* This is the default if a program doesn't have a PT_GNU_STACK
* program header entry. The PPC64 ELF ABI has a non executable stack
diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c
index a7a28110..50006d0 100644
--- a/fs/binfmt_elf.c
+++ b/fs/binfmt_elf.c
@@ -91,14 +91,25 @@ static struct linux_binfmt elf_format = {

#define BAD_ADDR(x) ((unsigned long)(x) >= TASK_SIZE)

-static int set_brk(unsigned long start, unsigned long end)
+static int set_brk(unsigned long start, unsigned long end, int prot)
{
start = ELF_PAGEALIGN(start);
end = ELF_PAGEALIGN(end);
if (end > start) {
- int error = vm_brk(start, end - start);
- if (error)
- return error;
+ /* Map the non-file portion of the last load header. If the
+ header is requesting these pages to be executeable then
+ we have to honour that, otherwise assume they are bss. */
+ if (prot & PROT_EXEC) {
+ unsigned long addr;
+ addr = vm_mmap(0, start, end - start, prot,
+ MAP_PRIVATE | MAP_FIXED, 0);
+ if (BAD_ADDR(addr))
+ return addr;
+ } else {
+ int error = vm_brk(start, end - start);
+ if (error)
+ return error;
+ }

Rather than repeating this logic twice, I think this should be a
helper that both sections can call.

I propose to teach vm_brk() to be able to optionally set some VM_foo bits,
such as VM_EXEC. Then there will be no need for conditional vm_mmap/vm_brk here.

Patch v3 is on the way.