Re: [PATCH v7 4/5] iommu: Use EINVAL for incompatible device/domain in ->attach_dev

From: Jason Gunthorpe <jgg@nvidia.com>
To: Will Deacon <will@kernel.org>
Cc: Nicolin Chen <nicolinc@nvidia.com>,
	joro@8bytes.org, robin.murphy@arm.com, kevin.tian@intel.com,
	quic_jjohnson@quicinc.com, suravee.suthikulpanit@amd.com,
	robdclark@gmail.com, dwmw2@infradead.org,
	baolu.lu@linux.intel.com, yong.wu@mediatek.com,
	matthias.bgg@gmail.com, orsonzhai@gmail.com,
	baolin.wang@linux.alibaba.com, zhang.lyra@gmail.com,
	thierry.reding@gmail.com, vdumpa@nvidia.com,
	jonathanh@nvidia.com, jean-philippe@linaro.org,
	tglx@linutronix.de, shameerali.kolothum.thodi@huawei.com,
	christophe.jaillet@wanadoo.fr, yangyicong@hisilicon.com,
	yangyingliang@huawei.com, quic_saipraka@quicinc.com,
	jon@solid-run.com, iommu@lists.linux.dev,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-arm-msm@vger.kernel.org,
	linux-mediatek@lists.infradead.org, linux-tegra@vger.kernel.org,
	virtualization@lists.linux-foundation.org
Subject: Re: [PATCH v7 4/5] iommu: Use EINVAL for incompatible device/domain in ->attach_dev
Date: Tue, 8 Nov 2022 09:23:10 -0400	[thread overview]
Message-ID: <Y2pYPnjMqrwDDwB/@nvidia.com> (raw)
In-Reply-To: <20221108132041.GB22816@willie-the-truck>

On Tue, Nov 08, 2022 at 01:20:42PM +0000, Will Deacon wrote:
> On Mon, Nov 07, 2022 at 04:14:32PM -0800, Nicolin Chen wrote:
> > On Mon, Nov 07, 2022 at 03:26:45PM +0000, Will Deacon wrote:
> > 
> > > > diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> > > > index ba47c73f5b8c..01fd7df16cb9 100644
> > > > --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> > > > +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
> > > > @@ -2430,23 +2430,14 @@ static int arm_smmu_attach_dev(struct iommu_domain *domain, struct device *dev)
> > > >                       goto out_unlock;
> > > >               }
> > > >       } else if (smmu_domain->smmu != smmu) {
> > > > -             dev_err(dev,
> > > > -                     "cannot attach to SMMU %s (upstream of %s)\n",
> > > > -                     dev_name(smmu_domain->smmu->dev),
> > > > -                     dev_name(smmu->dev));
> > > > -             ret = -ENXIO;
> > > > +             ret = -EINVAL;
> > > >               goto out_unlock;
> > > >       } else if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1 &&
> > > >                  master->ssid_bits != smmu_domain->s1_cfg.s1cdmax) {
> > > > -             dev_err(dev,
> > > > -                     "cannot attach to incompatible domain (%u SSID bits != %u)\n",
> > > > -                     smmu_domain->s1_cfg.s1cdmax, master->ssid_bits);
> > > >               ret = -EINVAL;
> > > >               goto out_unlock;
> > > >       } else if (smmu_domain->stage == ARM_SMMU_DOMAIN_S1 &&
> > > >                  smmu_domain->stall_enabled != master->stall_enabled) {
> > > > -             dev_err(dev, "cannot attach to stall-%s domain\n",
> > > > -                     smmu_domain->stall_enabled ? "enabled" : "disabled");
> > > >               ret = -EINVAL;
> > > >               goto out_unlock;
> > > >       }
> > 
> > > I think it would be helpful to preserve these messages using
> > > dev_err_ratelimited() so that attach failure can be diagnosed without
> > > having to hack the messages back into the driver.
> > 
> > Thank you for the review.
> > 
> > The change is already picked up last week. Yet, I can add prints
> > back with a followup patch, if no one has a problem with that.
> 
> Sorry, I fell behind with upstream so I got to this late. A patch on top
> would be fantastic!
> 
> > Also, I am not quite sure what the use case would be to have an
> > error print. Perhaps dev_dbg() would be more fitting if it is
> > just for diagnosis?
> 
> Sure, that works for me. I think the messages are useful for folks
> triggering this path e.g. via sysfs but if they're limited to debug I think
> that's better than removing them altogether.

I suspsect it has to be dbg - vfio/iommufd will probably trigger these
messages as it probes for domains that are compatible - eg certainly
the first one. Even if it is a "once" it would still emit a confusing
message for a normal occurance.

This is why they were removed in the first place..

Jason