memory: optimize qemu_get_ram_ptr and qemu_ram_ptr_length

these two functions consume too much cpu overhead to
find the RAMBlock by ram address.

After this patch, we can pass the RAMBlock pointer
to them so that they don't need to find the RAMBlock
anymore most of the time. We can get better performance
in address translation processing.

Backports commit 3655cb9c7375a595a8051ec677c515b24d5c1fe6 from qemu

qemu/exec.c

@@ -1324,24 +1324,32 @@ void *qemu_get_ram_block_host_ptr(struct uc_struct *uc, ram_addr_t addr)
    It should not be used for general purpose DMA.
    Use cpu_physical_memory_map/cpu_physical_memory_rw instead.
    */
-void *qemu_get_ram_ptr(struct uc_struct *uc, ram_addr_t addr)
+void *qemu_get_ram_ptr(struct uc_struct *uc, RAMBlock *ram_block,
+                       ram_addr_t addr)
 {
-    RAMBlock *block = qemu_get_ram_block(uc, addr);
+    RAMBlock *block = ram_block;
+
+    if (block == NULL) {
+        block = qemu_get_ram_block(uc, addr);
+    }
 
     return ramblock_ptr(block, addr - block->offset);
 }
 
 /* Return a host pointer to guest's ram. Similar to qemu_get_ram_ptr
  * but takes a size argument */
-static void *qemu_ram_ptr_length(struct uc_struct *uc, ram_addr_t addr, hwaddr *size)
+static void *qemu_ram_ptr_length(struct uc_struct *uc, RAMBlock *ram_block,
+                                 ram_addr_t addr, hwaddr *size)
 {
-    RAMBlock *block;
+    RAMBlock *block = ram_block;
     ram_addr_t offset_inside_block;
     if (*size == 0) {
         return NULL;
     }
 
-    block = qemu_get_ram_block(uc, addr);
+    if (block == NULL) {
+        block = qemu_get_ram_block(uc, addr);
+    }
     offset_inside_block = addr - block->offset;
     *size = MIN(*size, block->max_length - offset_inside_block);
 
@@ -1569,13 +1577,13 @@ static void notdirty_mem_write(struct uc_struct* uc, void *opaque, hwaddr ram_ad
     }
     switch (size) {
     case 1:
-        stb_p(qemu_get_ram_ptr(uc, ram_addr), val);
+        stb_p(qemu_get_ram_ptr(uc, NULL, ram_addr), val);
         break;
     case 2:
-        stw_p(qemu_get_ram_ptr(uc, ram_addr), val);
+        stw_p(qemu_get_ram_ptr(uc, NULL, ram_addr), val);
         break;
     case 4:
-        stl_p(qemu_get_ram_ptr(uc, ram_addr), val);
+        stl_p(qemu_get_ram_ptr(uc, NULL, ram_addr), val);
         break;
     default:
         abort();
@@ -1924,7 +1932,7 @@ static MemTxResult address_space_write_continue(AddressSpace *as, hwaddr addr,
         } else {
             addr1 += memory_region_get_ram_addr(mr);
             /* RAM case */
-            ptr = qemu_get_ram_ptr(mr->uc, addr1);
+            ptr = qemu_get_ram_ptr(mr->uc, mr->ram_block, addr1);
             memcpy(ptr, buf, l);
             invalidate_and_set_dirty(mr, addr1, l);
         }
@@ -2021,7 +2029,7 @@ MemTxResult address_space_read_continue(AddressSpace *as, hwaddr addr,
             }
         } else {
             /* RAM case */
-            ptr = qemu_get_ram_ptr(mr->uc, mr->ram_addr + addr1);
+            ptr = qemu_get_ram_ptr(mr->uc, mr->ram_block, mr->ram_addr + addr1);
             memcpy(buf, ptr, l);
         }
 
@@ -2107,7 +2115,7 @@ static inline void cpu_physical_memory_write_rom_internal(AddressSpace *as,
         } else {
             addr1 += memory_region_get_ram_addr(mr);
             /* ROM/RAM case */
-            ptr = qemu_get_ram_ptr(as->uc, addr1);
+            ptr = qemu_get_ram_ptr(mr->uc, mr->ram_block, addr1);
             switch (type) {
                 case WRITE_DATA:
                     memcpy(ptr, buf, l);
@@ -2235,7 +2243,7 @@ void *address_space_map(AddressSpace *as,
 
     memory_region_ref(mr);
     *plen = done;
-    return qemu_ram_ptr_length(as->uc, raddr + base, plen);
+    return qemu_ram_ptr_length(as->uc, mr->ram_block, raddr + base, plen);
 }
 
 /* Unmaps a memory region previously mapped by address_space_map().
@@ -2308,9 +2316,10 @@ static inline uint32_t address_space_ldl_internal(AddressSpace *as, hwaddr addr,
 #endif
     } else {
         /* RAM case */
-        ptr = qemu_get_ram_ptr(as->uc, (memory_region_get_ram_addr(mr)
-                    & TARGET_PAGE_MASK)
-                + addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block,
+                               (memory_region_get_ram_addr(mr)
+                               & TARGET_PAGE_MASK)
+                               + addr1);
         switch (endian) {
         case DEVICE_LITTLE_ENDIAN:
             val = ldl_le_p(ptr);
@@ -2395,9 +2404,10 @@ static inline uint64_t address_space_ldq_internal(AddressSpace *as, hwaddr addr,
 #endif
     } else {
         /* RAM case */
-        ptr = qemu_get_ram_ptr(as->uc, (memory_region_get_ram_addr(mr)
-                    & TARGET_PAGE_MASK)
-                + addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block,
+                               (memory_region_get_ram_addr(mr)
+                               & TARGET_PAGE_MASK)
+                               + addr1);
         switch (endian) {
         case DEVICE_LITTLE_ENDIAN:
             val = ldq_le_p(ptr);
@@ -2502,9 +2512,10 @@ static inline uint32_t address_space_lduw_internal(AddressSpace *as,
 #endif
     } else {
         /* RAM case */
-        ptr = qemu_get_ram_ptr(as->uc, (memory_region_get_ram_addr(mr)
-                    & TARGET_PAGE_MASK)
-                + addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block,
+                               (memory_region_get_ram_addr(mr)
+                               & TARGET_PAGE_MASK)
+                               + addr1);
         switch (endian) {
         case DEVICE_LITTLE_ENDIAN:
             val = lduw_le_p(ptr);
@@ -2578,7 +2589,7 @@ void address_space_stl_notdirty(AddressSpace *as, hwaddr addr, uint32_t val,
         r = memory_region_dispatch_write(mr, addr1, val, 4, attrs);
     } else {
         addr1 += memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK;
-        ptr = qemu_get_ram_ptr(as->uc, addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block, addr1);
         stl_p(ptr, val);
         r = MEMTX_OK;
     }
@@ -2621,7 +2632,7 @@ static inline void address_space_stl_internal(AddressSpace *as,
     } else {
         /* RAM case */
         addr1 += memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK;
-        ptr = qemu_get_ram_ptr(as->uc, addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block, addr1);
         switch (endian) {
         case DEVICE_LITTLE_ENDIAN:
             stl_le_p(ptr, val);
@@ -2723,7 +2734,7 @@ static inline void address_space_stw_internal(AddressSpace *as,
     } else {
         /* RAM case */
         addr1 += memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK;
-        ptr = qemu_get_ram_ptr(as->uc, addr1);
+        ptr = qemu_get_ram_ptr(as->uc, mr->ram_block, addr1);
         switch (endian) {
         case DEVICE_LITTLE_ENDIAN:
             stw_le_p(ptr, val);

qemu/include/exec/memory.h

@@ -1108,7 +1108,8 @@ MemTxResult address_space_read_continue(AddressSpace *as, hwaddr addr,
 
 MemTxResult address_space_read_full(AddressSpace *as, hwaddr addr,
                                     MemTxAttrs attrs, uint8_t *buf, int len);
-void *qemu_get_ram_ptr(struct uc_struct *uc, ram_addr_t addr);
+void *qemu_get_ram_ptr(struct uc_struct *uc, RAMBlock *ram_block,
+                       ram_addr_t addr);
 
 static inline bool memory_access_is_direct(MemoryRegion *mr, bool is_write)
 {
@@ -1151,7 +1152,7 @@ MemTxResult address_space_read(AddressSpace *as, hwaddr addr, MemTxAttrs attrs,
             mr = address_space_translate(as, addr, &addr1, &l, false);
             if (len == l && memory_access_is_direct(mr, false)) {
                 addr1 += memory_region_get_ram_addr(mr);
-                ptr = qemu_get_ram_ptr(addr1);
+                ptr = qemu_get_ram_ptr(mr->uc, mr->ram_block, addr1);
                 memcpy(buf, ptr, len);
             } else {
                 result = address_space_read_continue(as, addr, attrs, buf, len,

qemu/memory.c

@@ -1373,7 +1373,7 @@ void *memory_region_get_ram_ptr(MemoryRegion *mr)
 
     assert(mr->terminates);
 
-    return qemu_get_ram_ptr(mr->uc, mr->ram_addr & TARGET_PAGE_MASK);
+    return qemu_get_ram_ptr(mr->uc, mr->ram_block, mr->ram_addr & TARGET_PAGE_MASK);
 }
 
 bool memory_region_test_and_clear_dirty(MemoryRegion *mr, hwaddr addr,