qemu: address todo comment in exec.c

exec.c has a comment 'XXX: optimize' for lduw_phys/stw_phys,
so let's do it, along the lines of stl_phys.

The reason to address 16 bit accesses specifically is that virtio relies
on these accesses to be done atomically, using memset as we do now
breaks this assumption, which is reported to cause qemu with kvm
to read wrong index values under stress.

https://bugzilla.redhat.com/show_bug.cgi?id=525323

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

1 files changed, 60 insertions, 7 deletions

diff --git a/exec.c b/exec.c
index cddaf366b..e980788c4 100644
--- a/exec.c
+++ b/exec.c
@@ -3715,12 +3715,36 @@ uint32_t ldub_phys(target_phys_addr_t addr)
     return val;
 }
 
-/* XXX: optimize */
+/* warning: addr must be aligned */
 uint32_t lduw_phys(target_phys_addr_t addr)
 {
-    uint16_t val;
-    cpu_physical_memory_read(addr, (uint8_t *)&val, 2);
-    return tswap16(val);
+    int io_index;
+    uint8_t *ptr;
+    uint64_t val;
+    unsigned long pd;
+    PhysPageDesc *p;
+
+    p = phys_page_find(addr >> TARGET_PAGE_BITS);
+    if (!p) {
+        pd = IO_MEM_UNASSIGNED;
+    } else {
+        pd = p->phys_offset;
+    }
+
+    if ((pd & ~TARGET_PAGE_MASK) > IO_MEM_ROM &&
+        !(pd & IO_MEM_ROMD)) {
+        /* I/O case */
+        io_index = (pd >> IO_MEM_SHIFT) & (IO_MEM_NB_ENTRIES - 1);
+        if (p)
+            addr = (addr & ~TARGET_PAGE_MASK) + p->region_offset;
+        val = io_mem_read[io_index][1](io_mem_opaque[io_index], addr);
+    } else {
+        /* RAM case */
+        ptr = qemu_get_ram_ptr(pd & TARGET_PAGE_MASK) +
+            (addr & ~TARGET_PAGE_MASK);
+        val = lduw_p(ptr);
+    }
+    return val;
 }
 
 /* warning: addr must be aligned. The ram page is not masked as dirty
@@ -3837,11 +3861,40 @@ void stb_phys(target_phys_addr_t addr, uint32_t val)
     cpu_physical_memory_write(addr, &v, 1);
 }
 
-/* XXX: optimize */
+/* warning: addr must be aligned */
 void stw_phys(target_phys_addr_t addr, uint32_t val)
 {
-    uint16_t v = tswap16(val);
-    cpu_physical_memory_write(addr, (const uint8_t *)&v, 2);
+    int io_index;
+    uint8_t *ptr;
+    unsigned long pd;
+    PhysPageDesc *p;
+
+    p = phys_page_find(addr >> TARGET_PAGE_BITS);
+    if (!p) {
+        pd = IO_MEM_UNASSIGNED;
+    } else {
+        pd = p->phys_offset;
+    }
+
+    if ((pd & ~TARGET_PAGE_MASK) != IO_MEM_RAM) {
+        io_index = (pd >> IO_MEM_SHIFT) & (IO_MEM_NB_ENTRIES - 1);
+        if (p)
+            addr = (addr & ~TARGET_PAGE_MASK) + p->region_offset;
+        io_mem_write[io_index][1](io_mem_opaque[io_index], addr, val);
+    } else {
+        unsigned long addr1;
+        addr1 = (pd & TARGET_PAGE_MASK) + (addr & ~TARGET_PAGE_MASK);
+        /* RAM case */
+        ptr = qemu_get_ram_ptr(addr1);
+        stw_p(ptr, val);
+        if (!cpu_physical_memory_is_dirty(addr1)) {
+            /* invalidate code */
+            tb_invalidate_phys_page_range(addr1, addr1 + 2, 0);
+            /* set dirty bit */
+            cpu_physical_memory_set_dirty_flags(addr1,
+                (0xff & ~CODE_DIRTY_FLAG));
+        }
+    }
 }
 
 /* XXX: optimize */