[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <BN9PR11MB527640B28F8350DDEC2D26F48C0F9@BN9PR11MB5276.namprd11.prod.outlook.com>
Date: Thu, 24 Nov 2022 01:08:33 +0000
From: "Tian, Kevin" <kevin.tian@...el.com>
To: Thomas Gleixner <tglx@...utronix.de>,
LKML <linux-kernel@...r.kernel.org>
CC: "x86@...nel.org" <x86@...nel.org>, Joerg Roedel <joro@...tes.org>,
"Will Deacon" <will@...nel.org>,
"linux-pci@...r.kernel.org" <linux-pci@...r.kernel.org>,
Bjorn Helgaas <bhelgaas@...gle.com>,
"Lorenzo Pieralisi" <lorenzo.pieralisi@....com>,
Marc Zyngier <maz@...nel.org>,
"Greg Kroah-Hartman" <gregkh@...uxfoundation.org>,
Jason Gunthorpe <jgg@...lanox.com>,
"Jiang, Dave" <dave.jiang@...el.com>,
Alex Williamson <alex.williamson@...hat.com>,
"Williams, Dan J" <dan.j.williams@...el.com>,
Logan Gunthorpe <logang@...tatee.com>,
"Raj, Ashok" <ashok.raj@...el.com>, Jon Mason <jdmason@...zu.us>,
Allen Hubbe <allenbh@...il.com>,
"Ahmed S. Darwish" <darwi@...utronix.de>
Subject: RE: [patch V2 12/33] PCI/MSI: Add support for per device MSI[X]
domains
> From: Thomas Gleixner <tglx@...utronix.de>
> Sent: Thursday, November 24, 2022 5:50 AM
>
> >> This is general PCI MSI logic. So an open related to my rely to patch02,
> >> is it correct for PCI core to assume that the real parent imposes all the
> >> restrictions and there is no need to further go down the hierarchy?
> >
> > That was my working assumption and it turned out to be correct with both
> > x86 and ARM.
>
> As a follow up, I went through some of the other architectures,
> especially the places which have extra limitations and it turns out that
> the restriction comes always from the direct parent.
>
> If that ever changes then we need a callback which lets us evaluate the
> resulting capabilities through the hierarchy. That's nothing which can
> be evaluated directly.
>
> Just look at the x86 hierarchy with IR. IR allows multi PCI-MSI, but the
> vector domain does not. Who is right? That's a decision which is made in
> the particular hierarchy.
make sense.
>
> For now it's valid that the direct MSI parent has the proper set
> available.
>
probably above can be marked out in the comment.
Powered by blists - more mailing lists