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 <address@hidden>
---
exec.c | 14 ++++++++++++++
1 files changed, 14 insertions(+), 0 deletions(-)
diff --git a/exec.c b/exec.c
index 23013fb..ce0603e 100644
--- a/exec.c
+++ b/exec.c
@@ -2477,6 +2477,20 @@ ram_addr_t qemu_ram_alloc(ram_addr_t size)
new_block->next = ram_blocks;
ram_blocks = new_block;
+ if (kvm_enabled() && !kvm_has_sync_mmu()) {
+#ifdef MADV_DONTFORK
+ int ret = madvise(new_block->host, 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
+ }
+