[Qemu-devel] [PATCH v2] exec.c: Ensure right alignment also for file backed ram

[Qemu-devel] [PATCH v2] exec.c: Ensure right alignment also for file backed ram

[Dominik Dingel]

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>

---

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               |  8 +++++---
 include/qemu/osdep.h | 12 ++++++++++++
 util/oslib-posix.c   | 12 ------------
 3 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/exec.c b/exec.c
index c4f9036..1ae98e4 100644
--- a/exec.c
+++ b/exec.c
@@ -1241,6 +1241,7 @@ static void *file_ram_alloc(RAMBlock *block,
     void *area;
     int fd = -1;
     int64_t page_size;
+    int64_t alignment;
 
     if (kvm_enabled() && !kvm_has_sync_mmu()) {
         error_setg(errp,
@@ -1296,7 +1297,8 @@ static void *file_ram_alloc(RAMBlock *block,
     }
 
     page_size = qemu_fd_getpagesize(fd);
-    block->mr->align = page_size;
+    alignment = MAX(page_size, QEMU_VMALLOC_ALIGN);
+    block->mr->align = alignment;
 
     if (memory < page_size) {
         error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
@@ -1305,7 +1307,7 @@ static void *file_ram_alloc(RAMBlock *block,
         goto error;
     }
 
-    memory = ROUND_UP(memory, page_size);
+    memory = ROUND_UP(memory, alignment);
 
     /*
      * ftruncate is not supported by hugetlbfs in older
@@ -1317,7 +1319,7 @@ 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, alignment, 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..a472372 100644
--- a/include/qemu/osdep.h
+++ b/include/qemu/osdep.h
@@ -247,6 +247,18 @@ void qemu_anon_ram_free(void *ptr, size_t size);
 
 #endif
 
+#if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
+   /* 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 20ca141..4adde93 100644
--- a/util/oslib-posix.c
+++ b/util/oslib-posix.c
@@ -26,18 +26,6 @@
  * THE SOFTWARE.
  */
 
-#if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
-   /* 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>
-- 
2.6.6

Re: [Qemu-devel] [PATCH v2] exec.c: Ensure right alignment also for file backed ram

[Paolo Bonzini]

On 07/04/2016 23:31, Dominik Dingel wrote:
> diff --git a/exec.c b/exec.c
> index c4f9036..1ae98e4 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -1241,6 +1241,7 @@ static void *file_ram_alloc(RAMBlock *block,
>      void *area;
>      int fd = -1;
>      int64_t page_size;
> +    int64_t alignment;
>  
>      if (kvm_enabled() && !kvm_has_sync_mmu()) {
>          error_setg(errp,
> @@ -1296,7 +1297,8 @@ static void *file_ram_alloc(RAMBlock *block,
>      }
>  
>      page_size = qemu_fd_getpagesize(fd);
> -    block->mr->align = page_size;
> +    alignment = MAX(page_size, QEMU_VMALLOC_ALIGN);
> +    block->mr->align = alignment;
>  
>      if (memory < page_size) {
>          error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
> @@ -1305,7 +1307,7 @@ static void *file_ram_alloc(RAMBlock *block,
>          goto error;
>      }
>  
> -    memory = ROUND_UP(memory, page_size);
> +    memory = ROUND_UP(memory, alignment);

I think this change is not necessary either; it is enough to change the
qemu_ram_mmap below.

Paolo

>      /*
>       * ftruncate is not supported by hugetlbfs in older
> @@ -1317,7 +1319,7 @@ 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, alignment, 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..a472372 100644
> --- a/include/qemu/osdep.h
> +++ b/include/qemu/osdep.h
> @@ -247,6 +247,18 @@ void qemu_anon_ram_free(void *ptr, size_t size);
>  
>  #endif
>  
> +#if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
> +   /* 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 20ca141..4adde93 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -26,18 +26,6 @@
>   * THE SOFTWARE.
>   */
>  
> -#if defined(__linux__) && (defined(__x86_64__) || defined(__arm__))
> -   /* 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>
> 

Re: [Qemu-devel] [PATCH v2] exec.c: Ensure right alignment also for file backed ram

[Dominik Dingel]

On Tue, 12 Apr 2016 23:45:18 +0200
Paolo Bonzini <pbonzini@redhat.com> wrote:

> 
> 
> On 07/04/2016 23:31, Dominik Dingel wrote:
> > diff --git a/exec.c b/exec.c
> > index c4f9036..1ae98e4 100644
> > --- a/exec.c
> > +++ b/exec.c
> > @@ -1241,6 +1241,7 @@ static void *file_ram_alloc(RAMBlock *block,
> >      void *area;
> >      int fd = -1;
> >      int64_t page_size;
> > +    int64_t alignment;
> >  
> >      if (kvm_enabled() && !kvm_has_sync_mmu()) {
> >          error_setg(errp,
> > @@ -1296,7 +1297,8 @@ static void *file_ram_alloc(RAMBlock *block,
> >      }
> >  
> >      page_size = qemu_fd_getpagesize(fd);
> > -    block->mr->align = page_size;
> > +    alignment = MAX(page_size, QEMU_VMALLOC_ALIGN);
> > +    block->mr->align = alignment;
> >  
> >      if (memory < page_size) {
> >          error_setg(errp, "memory size 0x" RAM_ADDR_FMT " must be equal to "
> > @@ -1305,7 +1307,7 @@ static void *file_ram_alloc(RAMBlock *block,
> >          goto error;
> >      }
> >  
> > -    memory = ROUND_UP(memory, page_size);
> > +    memory = ROUND_UP(memory, alignment);
> 
> I think this change is not necessary either; it is enough to change the
> qemu_ram_mmap below.

You are right "memory" will never be back propagated to anything related to memory slots.
On which we have such a size constraint.

Thanks,
	Dominik