Re: GNTTABOP_unmap_grant_ref doc

[David Vrabel <dvrabel@cantab.net>]

On 24/07/13 13:49, David Scott wrote:
> Hi,
> 
> I've been working on the Mirage (mini-OS + OCaml runtime) blkback
> implementation. It seems to be working much more happily now but I was a
> bit surprised by the GNTTABOP_unmap_grant_ref doc -- maybe I misread it.
> 
> In ./xen/include/public/grant_table.h it says
> 
> /*
>  * GNTTABOP_unmap_grant_ref: Destroy one or more grant-reference mappings
>  * tracked by <handle>. If <host_addr> or <dev_bus_addr> is zero, that
>  * field is ignored. If non-zero, they must refer to a device/host mapping
>  * that is tracked by <handle>
>  * NOTES:
>  *  1. The call may fail in an undefined manner if either mapping is not
>  *     tracked by <handle>.
>  *  3. After executing a batch of unmaps, it is guaranteed that no stale
>  *     mappings will remain in the device or host TLBs.
>  */
> 
> When I read
>   "If <host_addr> or <dev_bus_addr> is zero, that field is ignored"
>   "The call may fail in an undefined manner if either mapping is not
> tracked by <handle>"
> 
> I thought, "great, I'll not bother tracking anything except the handle
> and set both fields to zero". However if I set host_addr to zero then my
> frontend complains bitterly
> 
> gnttab_stubs.c: WARNING: g.e. 53 still in use! (19)
> 
> If I store the host_addr and fill it in during the unmap, everything
> seems to work.
> 
> So did I misread the doc? Or did it mean exclusive-or i.e. you must fill
> in either one of these two fields? Or have I done something stupid
> elsewhere (always possible)?

If host_addr is non-zero, a host (cpu) mapping is cleared.  If
dev_bus_addr is non-zero a device (iommu) mapping is cleared.  If both
are zero, GNTTAB_unmap_grant_ref is a no-op.

David