[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] util/mmap-alloc: add comments, assertions
From: |
Paolo Bonzini |
Subject: |
Re: [Qemu-devel] [PATCH] util/mmap-alloc: add comments, assertions |
Date: |
Tue, 29 Sep 2015 12:53:17 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.2.0 |
On 29/09/2015 12:51, Michael S. Tsirkin wrote:
> Document RAM guard page logic within mmap-alloc.
>
> Signed-off-by: Michael S. Tsirkin <address@hidden>
> ---
>
> Paolo, can you pls confirm this is what you had in mind?
>
> util/mmap-alloc.c | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/util/mmap-alloc.c b/util/mmap-alloc.c
> index 05c8b4b..d978399 100644
> --- a/util/mmap-alloc.c
> +++ b/util/mmap-alloc.c
> @@ -12,9 +12,14 @@
> #include <qemu/mmap-alloc.h>
> #include <sys/types.h>
> #include <sys/mman.h>
> +#include <assert.h>
>
> void *qemu_ram_mmap(int fd, size_t size, size_t align)
> {
> + /*
> + * Note: this always allocates at least one extra page of virtual address
> + * space, even if size is already aligned.
> + */
> size_t total = size + align;
> void *ptr = mmap(0, total, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE, -1,
> 0);
> size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
> @@ -24,6 +29,11 @@ void *qemu_ram_mmap(int fd, size_t size, size_t align)
> return NULL;
> }
>
> + /* Make sure align is a power of 2 */
> + assert(!(align & (align - 1)));
> + /* Always align to host page size */
> + assert(align >= getpagesize());
> +
> ptr1 = mmap(ptr + offset, size, PROT_READ | PROT_WRITE,
> MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE, fd, 0);
> if (ptr1 == MAP_FAILED) {
> @@ -37,6 +47,11 @@ void *qemu_ram_mmap(int fd, size_t size, size_t align)
> if (offset > 0) {
> munmap(ptr - offset, offset);
> }
> +
> + /*
> + * Leave a single PROT_NONE page allocated after the RAM block, to serve
> as
> + * a guard page guarding against potential buffer overflows.
> + */
> if (total > size + getpagesize()) {
> munmap(ptr + size + getpagesize(), total - size - getpagesize());
> }
> @@ -47,6 +62,7 @@ void *qemu_ram_mmap(int fd, size_t size, size_t align)
> void qemu_ram_munmap(void *ptr, size_t size)
> {
> if (ptr) {
> + /* Unmap both the RAM block and the guard page */
> munmap(ptr, size + getpagesize());
> }
> }
>
Yes, thanks!
Paolo