Re: [PATCH V4] migration: ram block cpr blockers

[Steven Sistare <steven.sistare@oracle.com>]

On 2/25/2025 4:10 PM, David Hildenbrand wrote:
>> +    /* Ram device is remapped in new QEMU */
>> +    if (memory_region_is_ram_device(mr)) {
>> +        return true;
>> +    }
>> +
>> +    /* Named files are remapped in new QEMU, same contents if shared (no COW) */
>> +    if (qemu_ram_is_shared(rb) && qemu_ram_is_named_file(rb)) {
>> +        return true;
>> +    }
>> +
>> +    /* A file descriptor is remapped in new QEMU */
>> +    if (rb->fd >= 0 && qemu_ram_is_shared(rb)) {
>> +        return true;
>> +    }
> 
> 
> Sorry, I was not fast enough to reply to your v3 reply.
> 
> This is now essentially:
> 
> if (qemu_ram_is_shared(rb) &&
>      (qemu_ram_is_named_file(rb) || rb->fd >= 0)) {
>      return true;
> }
> 
> But what is the purpose of the "name file" check then, if rb->fd essentially allows for any files?
> 
> So either the "fd >= 0" check is insufficient or the qemu_ram_is_named_file() check us superfluous.

Yup, qemu_ram_is_named_file always has an fd at this point, so that check is gratuitous.
The preservation mechanism is different, which is why I thought of it as a different
case, but that is just a commenting issue.

V5 will be:

static bool ram_is_cpr_compatible(RAMBlock *rb)
{
     MemoryRegion *mr = rb->mr;

     if (!mr || !memory_region_is_ram(mr)) {
         return true;
     }

     /* Ram device is remapped in new QEMU */
     if (memory_region_is_ram_device(mr)) {
         return true;
     }

     /*
      * A file descriptor is passed to new QEMU and remapped, or its backing
      * file is reopened and mapped.  It must be shared to avoid COW.
      */
     if (rb->fd >= 0 && qemu_ram_is_shared(rb)) {
         return true;
     }

     return false;
}

- Steve