[Qemu-devel] [PATCH] memory: size can be different from ptr_size

Marc-André Lureau posted 1 patch 7 years, 2 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20180905082919.6986-1-marcandre.lureau@redhat.com
Test docker-clang@ubuntu failed
Test checkpatch passed
include/exec/memory.h  |  8 ++++++--
exec.c                 |  3 +++
hw/display/g364fb.c    |  2 +-
hw/vfio/common.c       |  3 ++-
hw/virtio/vhost-user.c |  2 +-
memory.c               | 10 ++++++----
6 files changed, 19 insertions(+), 9 deletions(-)
[Qemu-devel] [PATCH] memory: size can be different from ptr_size
Posted by Marc-André Lureau 7 years, 2 months ago
memory_region_init_ram*_ptr() take only the size of the MemoryRegion,
and allocate a RAMBlock with the same size. However, it may be
convenient to expose a smaller MemoryRegion (for ex: a few bytes) than
the RAMBlock (which must be a multiple of TARGET_PAGE_SIZE).

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 include/exec/memory.h  |  8 ++++++--
 exec.c                 |  3 +++
 hw/display/g364fb.c    |  2 +-
 hw/vfio/common.c       |  3 ++-
 hw/virtio/vhost-user.c |  2 +-
 memory.c               | 10 ++++++----
 6 files changed, 19 insertions(+), 9 deletions(-)

diff --git a/include/exec/memory.h b/include/exec/memory.h
index eb4f2fb249..03f257829b 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -700,6 +700,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
  *        must be unique within any device
  * @size: size of the region.
  * @ptr: memory to be mapped; must contain at least @size bytes.
+ * @ptr_size: size of @ptr buffer
  *
  * Note that this function does not do anything to cause the data in the
  * RAM memory region to be migrated; that is the responsibility of the caller.
@@ -708,7 +709,8 @@ void memory_region_init_ram_ptr(MemoryRegion *mr,
                                 struct Object *owner,
                                 const char *name,
                                 uint64_t size,
-                                void *ptr);
+                                void *ptr,
+                                uint64_t ptr_size);
 
 /**
  * memory_region_init_ram_device_ptr:  Initialize RAM device memory region from
@@ -727,6 +729,7 @@ void memory_region_init_ram_ptr(MemoryRegion *mr,
  * @name: the name of the region.
  * @size: size of the region.
  * @ptr: memory to be mapped; must contain at least @size bytes.
+ * @ptr_size: size of @ptr buffer
  *
  * Note that this function does not do anything to cause the data in the
  * RAM memory region to be migrated; that is the responsibility of the caller.
@@ -736,7 +739,8 @@ void memory_region_init_ram_device_ptr(MemoryRegion *mr,
                                        struct Object *owner,
                                        const char *name,
                                        uint64_t size,
-                                       void *ptr);
+                                       void *ptr,
+                                       uint64_t ptr_size);
 
 /**
  * memory_region_init_alias: Initialize a memory region that aliases all or a
diff --git a/exec.c b/exec.c
index 6826c8337d..fcea614e79 100644
--- a/exec.c
+++ b/exec.c
@@ -2361,6 +2361,9 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
 RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
                                    MemoryRegion *mr, Error **errp)
 {
+    assert(size >= TARGET_PAGE_SIZE);
+    assert(size % TARGET_PAGE_SIZE == 0);
+
     return qemu_ram_alloc_internal(size, size, NULL, host, false,
                                    false, mr, errp);
 }
diff --git a/hw/display/g364fb.c b/hw/display/g364fb.c
index fbc2b2422d..f4f5643761 100644
--- a/hw/display/g364fb.c
+++ b/hw/display/g364fb.c
@@ -475,7 +475,7 @@ static void g364fb_init(DeviceState *dev, G364State *s)
 
     memory_region_init_io(&s->mem_ctrl, NULL, &g364fb_ctrl_ops, s, "ctrl", 0x180000);
     memory_region_init_ram_ptr(&s->mem_vram, NULL, "vram",
-                               s->vram_size, s->vram);
+                               s->vram_size, s->vram, s->vram_size);
     vmstate_register_ram(&s->mem_vram, dev);
     memory_region_set_log(&s->mem_vram, true, DIRTY_MEMORY_VGA);
 }
diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index 7c185e5a2e..e5dfbbafe3 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -854,7 +854,8 @@ int vfio_region_mmap(VFIORegion *region)
         memory_region_init_ram_device_ptr(&region->mmaps[i].mem,
                                           memory_region_owner(region->mem),
                                           name, region->mmaps[i].size,
-                                          region->mmaps[i].mmap);
+                                          region->mmaps[i].mmap,
+                                          region->mmaps[i].size);
         g_free(name);
         memory_region_add_subregion(region->mem, region->mmaps[i].offset,
                                     &region->mmaps[i].mem);
diff --git a/hw/virtio/vhost-user.c b/hw/virtio/vhost-user.c
index b041343632..3abe960f29 100644
--- a/hw/virtio/vhost-user.c
+++ b/hw/virtio/vhost-user.c
@@ -935,7 +935,7 @@ static int vhost_user_slave_handle_vring_host_notifier(struct vhost_dev *dev,
     name = g_strdup_printf("vhost-user/host-notifier@%p mmaps[%d]",
                            user, queue_idx);
     memory_region_init_ram_device_ptr(&n->mr, OBJECT(vdev), name,
-                                      page_size, addr);
+                                      page_size, addr, page_size);
     g_free(name);
 
     if (virtio_queue_set_host_notifier_mr(vdev, queue_idx, &n->mr, true)) {
diff --git a/memory.c b/memory.c
index 9b73892768..07f8458d38 100644
--- a/memory.c
+++ b/memory.c
@@ -1587,7 +1587,8 @@ void memory_region_init_ram_ptr(MemoryRegion *mr,
                                 Object *owner,
                                 const char *name,
                                 uint64_t size,
-                                void *ptr)
+                                void *ptr,
+                                uint64_t ptr_size)
 {
     memory_region_init(mr, owner, name, size);
     mr->ram = true;
@@ -1597,16 +1598,17 @@ void memory_region_init_ram_ptr(MemoryRegion *mr,
 
     /* qemu_ram_alloc_from_ptr cannot fail with ptr != NULL.  */
     assert(ptr != NULL);
-    mr->ram_block = qemu_ram_alloc_from_ptr(size, ptr, mr, &error_fatal);
+    mr->ram_block = qemu_ram_alloc_from_ptr(ptr_size, ptr, mr, &error_fatal);
 }
 
 void memory_region_init_ram_device_ptr(MemoryRegion *mr,
                                        Object *owner,
                                        const char *name,
                                        uint64_t size,
-                                       void *ptr)
+                                       void *ptr,
+                                       uint64_t ptr_size)
 {
-    memory_region_init_ram_ptr(mr, owner, name, size, ptr);
+    memory_region_init_ram_ptr(mr, owner, name, size, ptr, ptr_size);
     mr->ram_device = true;
     mr->ops = &ram_device_mem_ops;
     mr->opaque = mr;
-- 
2.19.0.rc1


Re: [Qemu-devel] [PATCH] memory: size can be different from ptr_size
Posted by Juan Quintela 7 years, 2 months ago
Marc-André Lureau <marcandre.lureau@redhat.com> wrote:
> memory_region_init_ram*_ptr() take only the size of the MemoryRegion,
> and allocate a RAMBlock with the same size. However, it may be
> convenient to expose a smaller MemoryRegion (for ex: a few bytes) than
> the RAMBlock (which must be a multiple of TARGET_PAGE_SIZE).
>
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>  include/exec/memory.h  |  8 ++++++--
>  exec.c                 |  3 +++
>  hw/display/g364fb.c    |  2 +-
>  hw/vfio/common.c       |  3 ++-
>  hw/virtio/vhost-user.c |  2 +-
>  memory.c               | 10 ++++++----
>  6 files changed, 19 insertions(+), 9 deletions(-)
>
> diff --git a/include/exec/memory.h b/include/exec/memory.h
> index eb4f2fb249..03f257829b 100644
> --- a/include/exec/memory.h
> +++ b/include/exec/memory.h
> @@ -700,6 +700,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
>   *        must be unique within any device
>   * @size: size of the region.
>   * @ptr: memory to be mapped; must contain at least @size bytes.
                                                       ^^^^^

this comment gets wrong with your patches

> + * @ptr_size: size of @ptr buffer

> diff --git a/exec.c b/exec.c
> index 6826c8337d..fcea614e79 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -2361,6 +2361,9 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
>  RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
>                                     MemoryRegion *mr, Error **errp)
>  {
> +    assert(size >= TARGET_PAGE_SIZE);
> +    assert(size % TARGET_PAGE_SIZE == 0);
> +
>      return qemu_ram_alloc_internal(size, size, NULL, host, false,
>                                     false, mr, errp);
>  }

ok with this bit.

But how about to change instead to:

void memory_region_init_ram_ptr(MemoryRegion *mr,
                                Object *owner,
                                const char *name,
                                uint64_t size,
                                void *ptr)
{
    uint64_t real_size = ROUND_UP_TARGET_PAGE_SIZE(size);
    memory_region_init(mr, owner, name, real_size);
    mr->ram = true;
    mr->terminates = true;
    mr->destructor = memory_region_destructor_ram;
    mr->dirty_log_mask = tcg_enabled() ? (1 << DIRTY_MEMORY_CODE) : 0;

    /* qemu_ram_alloc_from_ptr cannot fail with ptr != NULL.  */
    assert(ptr != NULL);
    mr->ram_block = qemu_ram_alloc_from_ptr(size, ptr, mr, &error_fatal);
}

For a suitable ROUND_UP_TARGET_PAGE_SIZE() macro.  You get the idea.

And memory_region_device_ram_ptr() don't even need a change.
We need to adjust the comments, but it looks like an easier patch to me, no?

> diff --git a/hw/display/g364fb.c b/hw/display/g364fb.c
> index fbc2b2422d..f4f5643761 100644
> --- a/hw/display/g364fb.c
> +++ b/hw/display/g364fb.c
> @@ -475,7 +475,7 @@ static void g364fb_init(DeviceState *dev, G364State *s)
>  
>      memory_region_init_io(&s->mem_ctrl, NULL, &g364fb_ctrl_ops, s, "ctrl", 0x180000);
>      memory_region_init_ram_ptr(&s->mem_vram, NULL, "vram",
> -                               s->vram_size, s->vram);
> +                               s->vram_size, s->vram, s->vram_size);

Having to change all the devices that use the function with exactly the
same parameter looks weird to me.

What do you think?

Later, Juan.

Re: [Qemu-devel] [PATCH] memory: size can be different from ptr_size
Posted by Marc-André Lureau 7 years, 2 months ago
Hi

On Wed, Sep 5, 2018 at 1:37 PM, Juan Quintela <quintela@redhat.com> wrote:
> Marc-André Lureau <marcandre.lureau@redhat.com> wrote:
>> memory_region_init_ram*_ptr() take only the size of the MemoryRegion,
>> and allocate a RAMBlock with the same size. However, it may be
>> convenient to expose a smaller MemoryRegion (for ex: a few bytes) than
>> the RAMBlock (which must be a multiple of TARGET_PAGE_SIZE).
>>
>> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
>> ---
>>  include/exec/memory.h  |  8 ++++++--
>>  exec.c                 |  3 +++
>>  hw/display/g364fb.c    |  2 +-
>>  hw/vfio/common.c       |  3 ++-
>>  hw/virtio/vhost-user.c |  2 +-
>>  memory.c               | 10 ++++++----
>>  6 files changed, 19 insertions(+), 9 deletions(-)
>>
>> diff --git a/include/exec/memory.h b/include/exec/memory.h
>> index eb4f2fb249..03f257829b 100644
>> --- a/include/exec/memory.h
>> +++ b/include/exec/memory.h
>> @@ -700,6 +700,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
>>   *        must be unique within any device
>>   * @size: size of the region.
>>   * @ptr: memory to be mapped; must contain at least @size bytes.
>                                                        ^^^^^
>
> this comment gets wrong with your patches

why? it must contain at least @size (And preferrably exactly @ptr_size)


>
>> + * @ptr_size: size of @ptr buffer
>
>> diff --git a/exec.c b/exec.c
>> index 6826c8337d..fcea614e79 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -2361,6 +2361,9 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
>>  RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
>>                                     MemoryRegion *mr, Error **errp)
>>  {
>> +    assert(size >= TARGET_PAGE_SIZE);
>> +    assert(size % TARGET_PAGE_SIZE == 0);
>> +
>>      return qemu_ram_alloc_internal(size, size, NULL, host, false,
>>                                     false, mr, errp);
>>  }
>
> ok with this bit.
>
> But how about to change instead to:
>
> void memory_region_init_ram_ptr(MemoryRegion *mr,
>                                 Object *owner,
>                                 const char *name,
>                                 uint64_t size,
>                                 void *ptr)
> {
>     uint64_t real_size = ROUND_UP_TARGET_PAGE_SIZE(size);
>     memory_region_init(mr, owner, name, real_size);
>     mr->ram = true;
>     mr->terminates = true;
>     mr->destructor = memory_region_destructor_ram;
>     mr->dirty_log_mask = tcg_enabled() ? (1 << DIRTY_MEMORY_CODE) : 0;
>
>     /* qemu_ram_alloc_from_ptr cannot fail with ptr != NULL.  */
>     assert(ptr != NULL);
>     mr->ram_block = qemu_ram_alloc_from_ptr(size, ptr, mr, &error_fatal);
> }
>
> For a suitable ROUND_UP_TARGET_PAGE_SIZE() macro.  You get the idea.
>
> And memory_region_device_ram_ptr() don't even need a change.
> We need to adjust the comments, but it looks like an easier patch to me, no?

That's a good suggestion, it puts a bit more responsability into
caller side, put that's fair.

>
>> diff --git a/hw/display/g364fb.c b/hw/display/g364fb.c
>> index fbc2b2422d..f4f5643761 100644
>> --- a/hw/display/g364fb.c
>> +++ b/hw/display/g364fb.c
>> @@ -475,7 +475,7 @@ static void g364fb_init(DeviceState *dev, G364State *s)
>>
>>      memory_region_init_io(&s->mem_ctrl, NULL, &g364fb_ctrl_ops, s, "ctrl", 0x180000);
>>      memory_region_init_ram_ptr(&s->mem_vram, NULL, "vram",
>> -                               s->vram_size, s->vram);
>> +                               s->vram_size, s->vram, s->vram_size);
>
> Having to change all the devices that use the function with exactly the
> same parameter looks weird to me.
>
> What do you think?

I'll update the patch, thanks

Re: [Qemu-devel] [PATCH] memory: size can be different from ptr_size
Posted by Marc-André Lureau 7 years, 2 months ago
Hi

On Wed, Sep 5, 2018 at 7:51 PM, Marc-André Lureau
<marcandre.lureau@redhat.com> wrote:
> Hi
>
> On Wed, Sep 5, 2018 at 1:37 PM, Juan Quintela <quintela@redhat.com> wrote:
>> Marc-André Lureau <marcandre.lureau@redhat.com> wrote:
>>> memory_region_init_ram*_ptr() take only the size of the MemoryRegion,
>>> and allocate a RAMBlock with the same size. However, it may be
>>> convenient to expose a smaller MemoryRegion (for ex: a few bytes) than
>>> the RAMBlock (which must be a multiple of TARGET_PAGE_SIZE).
>>>
>>> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
>>> ---
>>>  include/exec/memory.h  |  8 ++++++--
>>>  exec.c                 |  3 +++
>>>  hw/display/g364fb.c    |  2 +-
>>>  hw/vfio/common.c       |  3 ++-
>>>  hw/virtio/vhost-user.c |  2 +-
>>>  memory.c               | 10 ++++++----
>>>  6 files changed, 19 insertions(+), 9 deletions(-)
>>>
>>> diff --git a/include/exec/memory.h b/include/exec/memory.h
>>> index eb4f2fb249..03f257829b 100644
>>> --- a/include/exec/memory.h
>>> +++ b/include/exec/memory.h
>>> @@ -700,6 +700,7 @@ void memory_region_init_ram_from_fd(MemoryRegion *mr,
>>>   *        must be unique within any device
>>>   * @size: size of the region.
>>>   * @ptr: memory to be mapped; must contain at least @size bytes.
>>                                                        ^^^^^
>>
>> this comment gets wrong with your patches
>
> why? it must contain at least @size (And preferrably exactly @ptr_size)
>
>
>>
>>> + * @ptr_size: size of @ptr buffer
>>
>>> diff --git a/exec.c b/exec.c
>>> index 6826c8337d..fcea614e79 100644
>>> --- a/exec.c
>>> +++ b/exec.c
>>> @@ -2361,6 +2361,9 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
>>>  RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
>>>                                     MemoryRegion *mr, Error **errp)
>>>  {
>>> +    assert(size >= TARGET_PAGE_SIZE);
>>> +    assert(size % TARGET_PAGE_SIZE == 0);
>>> +
>>>      return qemu_ram_alloc_internal(size, size, NULL, host, false,
>>>                                     false, mr, errp);
>>>  }
>>
>> ok with this bit.
>>
>> But how about to change instead to:
>>
>> void memory_region_init_ram_ptr(MemoryRegion *mr,
>>                                 Object *owner,
>>                                 const char *name,
>>                                 uint64_t size,
>>                                 void *ptr)
>> {
>>     uint64_t real_size = ROUND_UP_TARGET_PAGE_SIZE(size);
>>     memory_region_init(mr, owner, name, real_size);
>>     mr->ram = true;
>>     mr->terminates = true;
>>     mr->destructor = memory_region_destructor_ram;
>>     mr->dirty_log_mask = tcg_enabled() ? (1 << DIRTY_MEMORY_CODE) : 0;
>>
>>     /* qemu_ram_alloc_from_ptr cannot fail with ptr != NULL.  */
>>     assert(ptr != NULL);
>>     mr->ram_block = qemu_ram_alloc_from_ptr(size, ptr, mr, &error_fatal);
>> }
>>
>> For a suitable ROUND_UP_TARGET_PAGE_SIZE() macro.  You get the idea.
>>
>> And memory_region_device_ram_ptr() don't even need a change.
>> We need to adjust the comments, but it looks like an easier patch to me, no?
>
> That's a good suggestion, it puts a bit more responsability into
> caller side, put that's fair.
>
>>
>>> diff --git a/hw/display/g364fb.c b/hw/display/g364fb.c
>>> index fbc2b2422d..f4f5643761 100644
>>> --- a/hw/display/g364fb.c
>>> +++ b/hw/display/g364fb.c
>>> @@ -475,7 +475,7 @@ static void g364fb_init(DeviceState *dev, G364State *s)
>>>
>>>      memory_region_init_io(&s->mem_ctrl, NULL, &g364fb_ctrl_ops, s, "ctrl", 0x180000);
>>>      memory_region_init_ram_ptr(&s->mem_vram, NULL, "vram",
>>> -                               s->vram_size, s->vram);
>>> +                               s->vram_size, s->vram, s->vram_size);
>>
>> Having to change all the devices that use the function with exactly the
>> same parameter looks weird to me.
>>
>> What do you think?
>
> I'll update the patch, thanks

Actually the patch is a bit pointless, since qemu_ram_alloc_internal()
already HOST_PAGE_ALIGN(size). And host page is supposed to be larger
than target page. So, I can simply over-allocate the ram ptr/device
buffer in the TPM PPI patch.

thanks