Re: [PATCH] new interface: memory_object_get_proxy

[Sergey Bugaev]

On Mon, Nov 1, 2021 at 12:32 PM Joan Lledó <jlledom@mailfence.com> wrote:
> +kern_return_t
> +memory_object_get_proxy (task_t task, const vm_offset_t address,
> +                        vm_prot_t max_protection, vm_offset_t len,
> +                        ipc_port_t *port)
> +{
> +  kern_return_t err;

Super minor nitpick: this variable is typically named 'kr' or 'ret',
it would seem. 'err' is more of a userspace thing with glibc's
error_t.

> +  vm_map_lock_read(task->map);
> +  if (!vm_map_lookup_entry(task->map, address, &tmp_entry)) {
> +    if ((entry = tmp_entry->vme_next) == vm_map_to_entry(task->map)) {
> +      vm_map_unlock_read(task->map);
> +      return(KERN_NO_SPACE);
> +    }
> +  } else {
> +    entry = tmp_entry;
> +  }
> +
> +  /* Limit the allowed protection and range to the entry ones */
> +  if (len > entry->vme_end - entry->vme_start)
> +    return(KERN_INVALID_ARGUMENT);

You also need to unlock the map here.

> +  pager = ipc_port_copy_send(entry->object.vm_object->pager);
> +  offset = entry->offset;
> +  start = 0;
> +
> +  vm_map_unlock_read(task->map);
> +
> +  err = memory_object_create_proxy(task->itk_space, max_protection,
> +                                   &pager, 1,
> +                                   &offset, 1,
> +                                   &start, 1,
> +                                   &len, 1, port);
> +  if (err)
> +    ipc_port_release_send(pager);
> +
> +  return err;

Yes, this looks correct now, at least as far as I understand kernel
internals. Thank you! :)

Sergey