Re: [PATCH for-4.22] EFI: Fix boot from a device without a file system

[Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>]

On 5/19/26 3:45 PM, Andrew Cooper wrote:
> On 19/05/2026 2:41 pm, Marek Marczykowski-Górecki wrote:
> > On Tue, May 19, 2026 at 03:06:57PM +0200, Szymon Acedański wrote:
> > > When netbooting a unified Xen kernel image (via GRUB chainloader),
> > > the resulting loaded_image->DeviceHandle does not support
> > > SIMPLE_FILE_SYSTEM_PROTOCOL.
> > >
> > > Instead of crashing via noreturn PrintErrMesg(), print a message
> > > via PrintStr() and return NULL from get_parent_handle().
> > It's worth noting this isn't the first instance of returning NULL from
> > get_parent_handle(). The return value is used only as an argument
> > to read_file() (sometimes indirectly), and if it gets to be called with
> > NULL, read_file() will terminate execution via PrintErrMesg(). But with
> > unified Xen image, the intention is to not call read_file() at all, only
> > read_section(), so tolerating get_parent_handle() failure in this case
> > is desired. Keeping the message in place will ease debugging if
> > read_file() will actually be called later.
> >
> > Acked-by: Marek Marczykowski-Górecki <marmarek@xxxxxxxxxxxxxxxxxxxxxx>
> >
> > As a side note, a slightly better approach would be to call
> > get_parent_handle() lazily (on the first call to read_file()?). But it's
> > a bigger change that I feel may be too late for in Xen 4.22.
> >
> > Also, adding cc: Oleksii for release ack.
> Bugfixes are still fine to go in.
>
> Personally, I think moving the call to get_parent_handle() is a better
> fix, and fine for 4.22 even at this juncture.  ARM already does
> something along these lines in allocate_module_file().
I agree that we could consider suggested better approach and I am okay 
with having this patch (or its new version) as part of 4.22.
Thanks.

~ Oleksii