Re: [Qemu-devel] [PATCH 3/6] kvm: workaround a possible KVM bug when using KVM_MEM_READONLY

[Jan Kiszka]

On 2013-04-28 10:32, Jordan Justen wrote:
> On a Linux 3.8.0 based kernel, I occasionally saw a situation
> where the memory region would continue to trap on memory
> read even though KVM_MEM_READONLY was set.

Only 3.8.0? Did you bisect the issue down to the causing commit? Is it
fixed in later versions?

Jan

> 
> I found that if I set the slot to a size of 0, and before
> setting the slot, it would then behave as expected.
> 
> Signed-off-by: Jordan Justen <address@hidden>
> Cc: Xiao Guangrong <address@hidden>
> ---
>  kvm-all.c |    7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/kvm-all.c b/kvm-all.c
> index 95e6bf2..e2ddbcb 100644
> --- a/kvm-all.c
> +++ b/kvm-all.c
> @@ -205,6 +205,13 @@ static int kvm_set_user_memory_region(KVMState *s, 
> KVMSlot *slot)
>      if (s->migration_log) {
>          mem.flags |= KVM_MEM_LOG_DIRTY_PAGES;
>      }
> +    if (mem.flags & KVM_MEM_READONLY && mem.memory_size != 0) {
> +        /* Workaround an issue with setting a READONLY slot. Set the
> +         * slot size to 0 before setting the slot to the desired value. */
> +        mem.memory_size = 0;
> +        kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
> +        mem.memory_size = slot->memory_size;
> +    }
>      return kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
>  }
>  
> 
-- 
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux