Re: [Qemu-devel] [PATCH] migration: cache memory region ram ptr

[Peter Lieven <pl@kamp.de>]

or even this:

diff --git a/include/exec/memory.h b/include/exec/memory.h
index 1d55ad9..3003875 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -161,6 +161,7 @@ struct MemoryRegion {
     unsigned ioeventfd_nb;
     MemoryRegionIoeventfd *ioeventfds;
     NotifierList iommu_notify;
+    void *ram_ptr;
 };
 
 /**
diff --git a/memory.c b/memory.c
index 3f1df23..6c1e413 100644
--- a/memory.c
+++ b/memory.c
@@ -862,6 +862,7 @@ void memory_region_init(MemoryRegion *mr,
     mr->ioeventfd_nb = 0;
     mr->ioeventfds = NULL;
     mr->flush_coalesced_mmio = false;
+    mr->ram_ptr = NULL;
 }
 
 static uint64_t unassigned_mem_read(void *opaque, hwaddr addr,
@@ -1241,17 +1242,23 @@ void memory_region_reset_dirty(MemoryRegion *mr, hwaddr addr,
     cpu_physical_memory_reset_dirty(mr->ram_addr + addr, size, client);
 }
 
-void *memory_region_get_ram_ptr(MemoryRegion *mr)
+static void *memory_region_get_ram_ptr2(MemoryRegion *mr)
 {
     if (mr->alias) {
         return memory_region_get_ram_ptr(mr->alias) + mr->alias_offset;
     }
-
     assert(mr->terminates);
-
     return qemu_get_ram_ptr(mr->ram_addr & TARGET_PAGE_MASK);
 }
 
+inline void *memory_region_get_ram_ptr(MemoryRegion *mr)
+{
+    if (!mr->ram_ptr) {
+        mr->ram_ptr = memory_region_get_ram_ptr2(mr);
+    }
+    return mr->ram_ptr;
+}
+
 static void memory_region_update_coalesced_range_as(MemoryRegion *mr, AddressSpace *as)
 {
     FlatView *view;


Am 10.05.2014 18:32, schrieb Peter Lieven:
> Am 10.05.2014 17:33, schrieb Paolo Bonzini:
>> Il 10/05/2014 12:51, Peter Lieven ha scritto:
>>> we currently look up the ram ptr for each single page. Cache
>>> the pointer while we operate on the same block.
>> Why don't you instead cache the result in the MemoryRegion, so that memory_region_get_ram_ptr becomes a simple, inline field access?
> This seems to work. Wondering if it has other side implications. Basic tests like booting vServers and migration work. What about XEN?
>
> diff --git a/include/exec/memory.h b/include/exec/memory.h
> index 1d55ad9..3003875 100644
> --- a/include/exec/memory.h
> +++ b/include/exec/memory.h
> @@ -161,6 +161,7 @@ struct MemoryRegion {
>      unsigned ioeventfd_nb;
>      MemoryRegionIoeventfd *ioeventfds;
>      NotifierList iommu_notify;
> +    void *ram_ptr;
>  };
>  
>  /**
> diff --git a/memory.c b/memory.c
> index 3f1df23..78d4032 100644
> --- a/memory.c
> +++ b/memory.c
> @@ -862,6 +862,7 @@ void memory_region_init(MemoryRegion *mr,
>      mr->ioeventfd_nb = 0;
>      mr->ioeventfds = NULL;
>      mr->flush_coalesced_mmio = false;
> +    mr->ram_ptr = NULL;
>  }
>  
>  static uint64_t unassigned_mem_read(void *opaque, hwaddr addr,
> @@ -1249,7 +1250,12 @@ void *memory_region_get_ram_ptr(MemoryRegion *mr)
>  
>      assert(mr->terminates);
>  
> -    return qemu_get_ram_ptr(mr->ram_addr & TARGET_PAGE_MASK);
> +    if (mr->ram_ptr) {
> +        return mr->ram_ptr;
> +    }
> +
> +    mr->ram_ptr = qemu_get_ram_ptr(mr->ram_addr & TARGET_PAGE_MASK);
> +    return mr->ram_ptr;
>  }
>  
>  static void memory_region_update_coalesced_range_as(MemoryRegion *mr, AddressSpace *as)
>
>
> Peter
>
>> Paolo
>>
>>> Signed-off-by: Peter Lieven <pl@kamp.de>
>>> ---
>>>  arch_init.c |   23 ++++++++++++++++-------
>>>  1 file changed, 16 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/arch_init.c b/arch_init.c
>>> index 582b716..ce338aa 100644
>>> --- a/arch_init.c
>>> +++ b/arch_init.c
>>> @@ -594,13 +594,19 @@ static int ram_save_block(QEMUFile *f, bool last_stage)
>>>                  ram_bulk_stage = false;
>>>              }
>>>          } else {
>>> +            static uint8_t *ram_ptr;
>>>              int ret;
>>>              uint8_t *p;
>>>              bool send_async = true;
>>> -            int cont = (block == last_sent_block) ?
>>> -                RAM_SAVE_FLAG_CONTINUE : 0;
>>> +            int cont = 0;
>>>
>>> -            p = memory_region_get_ram_ptr(mr) + offset;
>>> +            if (block != last_sent_block) {
>>> +                ram_ptr = memory_region_get_ram_ptr(mr);
>>> +            } else {
>>> +                cont = RAM_SAVE_FLAG_CONTINUE;
>>> +            }
>>> +
>>> +            p = ram_ptr + offset;
>>>
>>>              /* In doubt sent page as normal */
>>>              bytes_sent = -1;
>>> @@ -990,16 +996,17 @@ static inline void *host_from_stream_offset(QEMUFile *f,
>>>                                              int flags)
>>>  {
>>>      static RAMBlock *block = NULL;
>>> +    static uint8_t *ram_ptr;
>>>      char id[256];
>>>      uint8_t len;
>>>
>>>      if (flags & RAM_SAVE_FLAG_CONTINUE) {
>>> -        if (!block) {
>>> +        if (!block || !ram_ptr) {
>>>              fprintf(stderr, "Ack, bad migration stream!\n");
>>>              return NULL;
>>>          }
>>>
>>> -        return memory_region_get_ram_ptr(block->mr) + offset;
>>> +        return ram_ptr + offset;
>>>      }
>>>
>>>      len = qemu_get_byte(f);
>>> @@ -1007,8 +1014,10 @@ static inline void *host_from_stream_offset(QEMUFile *f,
>>>      id[len] = 0;
>>>
>>>      QTAILQ_FOREACH(block, &ram_list.blocks, next) {
>>> -        if (!strncmp(id, block->idstr, sizeof(id)))
>>> -            return memory_region_get_ram_ptr(block->mr) + offset;
>>> +        if (!strncmp(id, block->idstr, sizeof(id))) {
>>> +            ram_ptr = memory_region_get_ram_ptr(block->mr);
>>> +            return ram_ptr + offset;
>>> +        }
>>>      }
>>>
>>>      fprintf(stderr, "Can't find block %s!\n", id);
>>>