[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <BN9PR11MB5276A8B4E2466BE080CA9E9B8CB39@BN9PR11MB5276.namprd11.prod.outlook.com>
Date: Tue, 21 Jun 2022 05:48:18 +0000
From: "Tian, Kevin" <kevin.tian@...el.com>
To: Baolu Lu <baolu.lu@...ux.intel.com>,
Joerg Roedel <joro@...tes.org>,
"Raj, Ashok" <ashok.raj@...el.com>
CC: "Qiang, Chenyi" <chenyi.qiang@...el.com>,
"Liu, Yi L" <yi.l.liu@...el.com>,
"Pan, Jacob jun" <jacob.jun.pan@...el.com>,
"iommu@...ts.linux-foundation.org" <iommu@...ts.linux-foundation.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"stable@...r.kernel.org" <stable@...r.kernel.org>
Subject: RE: [PATCH 1/1] iommu/vt-d: Fix RID2PASID setup failure
> From: Baolu Lu <baolu.lu@...ux.intel.com>
> Sent: Tuesday, June 21, 2022 12:28 PM
>
> On 2022/6/21 11:46, Tian, Kevin wrote:
> >> From: Baolu Lu <baolu.lu@...ux.intel.com>
> >> Sent: Tuesday, June 21, 2022 11:39 AM
> >>
> >> On 2022/6/21 10:54, Tian, Kevin wrote:
> >>>> From: Lu Baolu <baolu.lu@...ux.intel.com>
> >>>> Sent: Monday, June 20, 2022 4:17 PM
> >>>> @@ -2564,7 +2564,7 @@ static int domain_add_dev_info(struct
> >>>> dmar_domain *domain, struct device *dev)
> >>>> ret = intel_pasid_setup_second_level(iommu,
> >>>> domain,
> >>>> dev, PASID_RID2PASID);
> >>>> spin_unlock_irqrestore(&iommu->lock, flags);
> >>>> - if (ret) {
> >>>> + if (ret && ret != -EBUSY) {
> >>>> dev_err(dev, "Setup RID2PASID failed\n");
> >>>> dmar_remove_one_dev_info(dev);
> >>>> return ret;
> >>>> --
> >>>> 2.25.1
> >>>
> >>> It's cleaner to avoid this error at the first place, i.e. only do the
> >>> setup when the first device is attached to the pasid table.
> >>
> >> The logic that identifies the first device might introduce additional
> >> unnecessary complexity. Devices that share a pasid table are rare. I
> >> even prefer to give up sharing tables so that the code can be
> >> simpler.:-)
> >>
> >
> > It's not that complex if you simply move device_attach_pasid_table()
> > out of intel_pasid_alloc_table(). Then do the setup if
> > list_empty(&pasid_table->dev) and then attach device to the
> > pasid table in domain_add_dev_info().
>
> The pasid table is part of the device, hence a better place to
> allocate/free the pasid table is in the device probe/release paths.
> Things will become more complicated if we change relationship between
> device and it's pasid table when attaching/detaching a domain. That's
> the reason why I thought it was additional complexity.
>
If you do want to follow current route it’s still cleaner to check
whether the pasid entry has pointed to the domain in the individual
setup function instead of blindly returning -EBUSY and then ignoring
it even if a real busy condition occurs. The setup functions can
just return zero for this benign alias case.
Powered by blists - more mailing lists