On 29/01/2021 14:59, Roger Pau Monné wrote: > On Fri, Jan 29, 2021 at 11:38:43AM +0100, Manuel Bouyer wrote: >> On Thu, Jan 28, 2021 at 12:01:52PM +0000, Andrew Cooper wrote: >>> With the Xen side of this interface (soon to be) fixed to return real sizes, >>> userspace needs to be able to make the query. >>> >>> Introduce xenforeignmemory_resource_size() for the purpose, bumping the >>> library minor version. >>> >>> Update both all osdep_xenforeignmemory_map_resource() implementations to >>> understand size requests, skip the mmap() operation, and copy back the >>> nr_frames field. >>> >>> For NetBSD, also fix up the ioctl() error path to issue an unmap(), which >>> was >>> overlooked by c/s 4a64e2bb39 "libs/foreignmemory: Implement on NetBSD". >>> [....] >>> diff --git a/tools/libs/foreignmemory/netbsd.c >>> b/tools/libs/foreignmemory/netbsd.c >>> index d26566f601..4ae60aafdd 100644 >>> --- a/tools/libs/foreignmemory/netbsd.c >>> +++ b/tools/libs/foreignmemory/netbsd.c >>> @@ -132,6 +132,10 @@ int osdep_xenforeignmemory_map_resource( >>> }; >>> int rc; >>> >>> + if ( !fres->addr && !fres->nr_frames ) >>> + /* Request for resource size. Skip mmap(). */ >>> + goto skip_mmap; >>> + >>> fres->addr = mmap(fres->addr, fres->nr_frames << PAGE_SHIFT, >>> fres->prot, fres->flags | MAP_ANON | MAP_SHARED, -1, >>> 0); >> What happens if fres->addr is not NULL and nr_frames is 0 ? > mmap would return MAP_FAILED and errno == EINVAL in that case AFAICT > on Linux and FreeBSD. NetBSD mmap man page doesn't seem to mention > what happens in that case, so the comments below apply to Linux and > FreeBSD. Maybe we need to handle this differently for NetBSD? > >> Is it supposed to happen ? > I think that's fine. Calling osdep_xenforeignmemory_map_resource with > nr_frames == 0 is pointing to a bug in the caller, so returning error > should be fine. > >> Should we assert that fres->addr is NULL when >> nr_frames is 0 ? Or force fres->addr to NULL when nr_frames is 0 ? > Doesn't really matter, mmap will return EINVAL if nr_frames == 0 > regardless of the value of addr.
mmap() of 0 is an unconditional failure. So sayeth POSIX. For the size request, we don't mmap(), and a pointer of 0 is the signal to Xen. For the regular mapping, we support both NULL (let the kernel choose), and non-NULL (I want my mapping here please), just like regular mmap(). ~Andrew
