kvm: Avoid COW if KVM MMU is asynchronous

Avi Kivity wrote:
> Suggest wrapping in a function and hiding it deep inside kvm-all.c.
>

Done in v2:

---------->

If the KVM MMU is asynchronous (kernel does not support MMU_NOTIFIER),
we have to avoid COW for the guest memory. Otherwise we risk serious
breakage when guest pages change there physical locations due to COW
after fork. Seen when forking smbd during runtime via -smb.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

diff --git a/exec.c b/exec.c
index 2effd04f62cc20cf39ae222e3bbf7952c54c632b..c64938124b81db6530048ca0088db5745e32a39f 100644 (file)
--- a/exec.c
+++ b/exec.c
@@ -2484,6 +2484,9 @@ ram_addr_t qemu_ram_alloc(ram_addr_t size)
 
     last_ram_offset += size;
 
+    if (kvm_enabled())
+        kvm_setup_guest_memory(new_block->host, size);
+
     return new_block->offset;
 }
 

diff --git a/kvm-all.c b/kvm-all.c
index 0da5c7aa8de60e75722504b2e5a064dd8b92e8a7..36659a95e8d5d718f97a9bb1bae20c6d0e5977c1 100644 (file)
--- a/kvm-all.c
+++ b/kvm-all.c
@@ -773,6 +773,24 @@ int kvm_has_sync_mmu(void)
     return 0;
 }
 
+void kvm_setup_guest_memory(void *start, size_t size)
+{
+    if (!kvm_has_sync_mmu()) {
+#ifdef MADV_DONTFORK
+        int ret = madvise(start, size, MADV_DONTFORK);
+
+        if (ret) {
+            perror("madvice");
+            exit(1);
+        }
+#else
+        fprintf(stderr,
+                "Need MADV_DONTFORK in absence of synchronous KVM MMU\n");
+        exit(1);
+#endif
+    }
+}
+
 #ifdef KVM_CAP_SET_GUEST_DEBUG
 struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *env,
                                                  target_ulong pc)

diff --git a/kvm.h b/kvm.h
index 803a874113426f0a5fec9a3b18c6d2bc0d2dcbc7..0ea24266be2ac65e0371d8115442055b4e11c76d 100644 (file)
--- a/kvm.h
+++ b/kvm.h
@@ -48,6 +48,8 @@ int kvm_log_stop(target_phys_addr_t phys_addr, ram_addr_t size);
 
 int kvm_has_sync_mmu(void);
 
+void kvm_setup_guest_memory(void *start, size_t size);
+
 int kvm_coalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);
 int kvm_uncoalesce_mmio_region(target_phys_addr_t start, ram_addr_t size);