> -----Original Message----- > From: Andrew Cooper <andrew.coop...@citrix.com> > Sent: 25 November 2020 11:31 > To: Paul Durrant <p...@xen.org>; xen-devel@lists.xenproject.org > Cc: Durrant, Paul <pdurr...@amazon.co.uk>; Elnikety, Eslam > <elnik...@amazon.com>; Christian Lindig > <christian.lin...@citrix.com>; David Scott <d...@recoil.org>; Ian Jackson > <i...@xenproject.org>; Wei > Liu <w...@xen.org>; George Dunlap <george.dun...@citrix.com>; Jan Beulich > <jbeul...@suse.com>; Julien > Grall <jul...@xen.org>; Stefano Stabellini <sstabell...@kernel.org> > Subject: RE: [EXTERNAL] [PATCH v4 1/3] domctl: introduce a new domain create > flag, > XEN_DOMCTL_CDF_disable_fifo, ... > > CAUTION: This email originated from outside of the organization. Do not click > links or open > attachments unless you can confirm the sender and know the content is safe. > > > > On 24/11/2020 19:17, Paul Durrant wrote: > > diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h > > index 666aeb71bf1b..70701c59d053 100644 > > --- a/xen/include/public/domctl.h > > +++ b/xen/include/public/domctl.h > > @@ -70,9 +70,11 @@ struct xen_domctl_createdomain { > > #define XEN_DOMCTL_CDF_iommu (1U<<_XEN_DOMCTL_CDF_iommu) > > #define _XEN_DOMCTL_CDF_nested_virt 6 > > #define XEN_DOMCTL_CDF_nested_virt (1U << _XEN_DOMCTL_CDF_nested_virt) > > +#define _XEN_DOMCTL_CDF_disable_fifo 7 > > +#define XEN_DOMCTL_CDF_disable_fifo (1U << _XEN_DOMCTL_CDF_disable_fifo) > > The sense is backwards. It should be a "permit the use of FIFO" > control. If the code had been written this way to begin with, the bug > you found wouldn't have existed. > > Given that there is not currently a way to disable FIFO, you can > probably do without an enumeration of whether the hypervisor supports it > or not. >
Ok, I can reverse the sense. I found another one that we ought to control in a similar way... the per-cpu evtchn upcalls. AFAIK only the Windows PV drivers make use of it (and I can arrange to squash that with a registry flag) but it really falls into the same category as FIFO... so maybe we need a separate bit-field for these sorts of thing? Paul > ~Andrew