Re: [Qemu-devel] [PATCH v12 11/21] migration: Create multifd packet

[Dr. David Alan Gilbert]

* Juan Quintela (address@hidden) wrote:
> "Dr. David Alan Gilbert" <address@hidden> wrote:
> > * Juan Quintela (address@hidden) wrote:
> >> "Dr. David Alan Gilbert" <address@hidden> wrote:
> >> > * Juan Quintela (address@hidden) wrote:
> >> > I think that needs validating to ensure that the source didn't
> >> > send us junk and cause us to overwrite after the end of block->host
> >> 
> >>         if (offset > block->used_length) {
> >>             error_setg(errp, "multifd: offest too long %" PRId64
> >>                        " (max %" PRId64 ")",
> >>                        offset, block->max_length);
> >>             return -1;
> >>         }
> >> ??
> >
> > It's probably  (offset + TARGET_PAGE_SIZE) that needs checking
> > but it needs doing in a wrap-safe way.
> >
> 
>         if ((offset + TARGET_PAGE_SIZE) < offset) {
>             error_setg(errp, "multifd: offset %" PRId64 " wraps around"
>                        " with offset: %" PRId64, offset, block->max_length);
>             return -1;
>         }
>         if ((offset + TARGET_PAGE_SIZE) > block->used_length) {
>             error_setg(errp, "multifd: offset too long %" PRId64
>                        " (max %" PRId64 ")",
>                        offset, block->max_length);
>             return -1;
>         }

How about:
   if (offset > (block->used_length - TARGET_PAGE_SIZE)) {
    ....
   }
  (*assuming that block->used_length is always at least a
TARGET_PAGE_SIZE ?)

Dave

> Sometimes I wonder how is that we don't have
> 
> ramblock_contains_range(ramblock, start, size);
> 
> But well, c'est la vie.
> 
> Later, Juan.
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK