Re: [Qemu-devel] [PATCH] exec.c: Ensure right alignment also for file backed ram

[Paolo Bonzini <pbonzini@redhat.com>]

On 25/04/2016 13:55, Dominik Dingel wrote:
> While in the anonymous ram case we already take care of the right alignment
> such an alignment gurantee does not exist for file backed ram allocation.
> 
> Instead, pagesize is used for alignment. On s390 this is not enough for gmap,
> as we need to satisfy an alignment up to segments.
> 
> Reported-by: Halil Pasic <pasic@linux.vnet.ibm.com>
> Signed-off-by: Dominik Dingel <dingel@linux.vnet.ibm.com>

Looks good now, thanks.

Cc: qemu-stable <qemu-stable@nongnu.org>

Paolo

> ---
> 
> v2 -> v3:
> Skipping additional variable and just use alignment of memory region.
> As memory will not be backpropagated it is enough to round up to page_sizes.
> 
> v1 -> v2:
> While enforcing alignments we allow memory sizes on page_size.
> On mmap the memory size will be round up to alignments.
> 
> I thought about moving this alignment into qemu_ram_mmap but the result
> was a lot of code churn, the other possibility was to create an additional
> define ending up with two defines with the same semantics.
> ---
>  exec.c               |  5 +++--
>  include/qemu/osdep.h | 13 +++++++++++++
>  util/oslib-posix.c   | 13 -------------
>  3 files changed, 16 insertions(+), 15 deletions(-)
> 
> diff --git a/exec.c b/exec.c
> index c4f9036..fc75266 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1296,7 +1296,7 @@ static void *file_ram_alloc(RAMBlock *block,
>      }
>  
>      page_size = qemu_fd_getpagesize(fd);
> -    block->mr->align = page_size;
> +    block->mr->align = MAX(page_size, QEMU_VMALLOC_ALIGN);
>  
>      if (memory < page_size) {
>          error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
> @@ -1317,7 +1317,8 @@ static void *file_ram_alloc(RAMBlock *block,
>          perror("ftruncate");
>      }
>  
> -    area = qemu_ram_mmap(fd, memory, page_size, block->flags & RAM_SHARED);
> +    area = qemu_ram_mmap(fd, memory, block->mr->align,
> +                         block->flags & RAM_SHARED);
>      if (area == MAP_FAILED) {
>          error_setg_errno(errp, errno,
>                           "unable to map backing store for guest RAM");
> diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h
> index 408783f..783270f 100644
> --- a/include/qemu/osdep.h
> +++ b/include/qemu/osdep.h
> @@ -247,6 +247,19 @@ void qemu_anon_ram_free(void *ptr, size_t size);
>  
>  #endif
>  
> +#if defined(__linux__) && \
> +    (defined(__x86_64__) || defined(__arm__) || defined(__aarch64__))
> +   /* Use 2 MiB alignment so transparent hugepages can be used by KVM.
> +      Valgrind does not support alignments larger than 1 MiB,
> +      therefore we need special code which handles running on Valgrind. */
> +#  define QEMU_VMALLOC_ALIGN (512 * 4096)
> +#elif defined(__linux__) && defined(__s390x__)
> +   /* Use 1 MiB (segment size) alignment so gmap can be used by KVM. */
> +#  define QEMU_VMALLOC_ALIGN (256 * 4096)
> +#else
> +#  define QEMU_VMALLOC_ALIGN getpagesize()
> +#endif
> +
>  int qemu_madvise(void *addr, size_t len, int advice);
>  
>  int qemu_open(const char *name, int flags, ...);
> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
> index 6cc4b8f..4adde93 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -26,19 +26,6 @@
>   * THE SOFTWARE.
>   */
>  
> -#if defined(__linux__) && \
> -    (defined(__x86_64__) || defined(__arm__) || defined(__aarch64__))
> -   /* Use 2 MiB alignment so transparent hugepages can be used by KVM.
> -      Valgrind does not support alignments larger than 1 MiB,
> -      therefore we need special code which handles running on Valgrind. */
> -#  define QEMU_VMALLOC_ALIGN (512 * 4096)
> -#elif defined(__linux__) && defined(__s390x__)
> -   /* Use 1 MiB (segment size) alignment so gmap can be used by KVM. */
> -#  define QEMU_VMALLOC_ALIGN (256 * 4096)
> -#else
> -#  define QEMU_VMALLOC_ALIGN getpagesize()
> -#endif
> -
>  #include "qemu/osdep.h"
>  #include <termios.h>
>  #include <termios.h>
>