linux-kernel - RE: [PATCH v1 3/3] iommu/vt-d: Fix ATS support check for integrated endpoints

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <DS0PR11MB6373F96B334179FC4C406D04DC8AA@DS0PR11MB6373.namprd11.prod.outlook.com>
Date: Fri, 9 May 2025 11:12:44 +0000
From: "Wang, Wei W" <wei.w.wang@...el.com>
To: "Liu, Yi L" <yi.l.liu@...el.com>, "dwmw2@...radead.org"
	<dwmw2@...radead.org>, "baolu.lu@...ux.intel.com" <baolu.lu@...ux.intel.com>,
	"Tian, Kevin" <kevin.tian@...el.com>, "jroedel@...e.de" <jroedel@...e.de>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"iommu@...ts.linux.dev" <iommu@...ts.linux.dev>
Subject: RE: [PATCH v1 3/3] iommu/vt-d: Fix ATS support check for integrated
 endpoints

On Friday, May 9, 2025 5:32 PM, Liu, Yi L wrote:
2025/5/9 22:00, Wei Wang wrote:
> > The VT-d spec states that "SATC reporting structure identifies devices
> > that have address translation cache and that is validated per
> > requirements described in the 'Device TLB in System-on-Chip (SoC)
> Integrated Devices'
> > section. It is recommended that system software enable ATC for this
> > device". It is possible for an integrated device to have PCI ATC
> > capability implemented but not validated per the requirements, and
> > thus not appear in the SATC structure as recommended for ATS enablement.
> >
> > The current implementation checks ATS support for integrated endpoints
> > in two places. First, it verifies if the integrated endpoint device is
> > listed in SATC. If not, it proceeds to the second check that always
> > returns true for integrated devices. This could result in endpoint
> > devices not recommended in SATC presenting "supported = true" to the
> caller.
> >
> > Add integrated_device_ats_supported() for the integrated device ATS
> > check in a single location, which improves readability. The above
> > issue is also fixed in the function via returning false in that case.
> 
> if it is a fix. A Fixes tag is needed.

Yeah, will add it.

> 
> > Signed-off-by: Wei Wang <wei.w.wang@...el.com>
> > ---
> >   drivers/iommu/intel/iommu.c | 42 +++++++++++++++++++++++++------------
> >   1 file changed, 29 insertions(+), 13 deletions(-)
> >
> > diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c
> > index 2778bfe14f36..39abcf4e0f8f 100644
> > --- a/drivers/iommu/intel/iommu.c
> > +++ b/drivers/iommu/intel/iommu.c
> > @@ -2760,6 +2760,34 @@ static struct dmar_satc_unit
> *dmar_find_matched_satc_unit(struct pci_dev *dev)
> >   	return satcu;
> >   }
> >
> > +static bool integrated_device_ats_supported(struct pci_dev *dev,
> > +struct intel_iommu *iommu) {
> > +	struct dmar_satc_unit *satcu = dmar_find_matched_satc_unit(dev);
> > +
> > +	/*
> > +	 * This device supports ATS as it is in SATC table. When IOMMU is in
> > +	 * legacy mode, enabling ATS is done automatically by HW for the
> device
> > +	 * that requires ATS, hence OS should not enable this device ATS to
> > +	 * avoid duplicated TLB invalidation.
> > +	 */
> > +	if (satcu)
> > +		return !(satcu->atc_required && !sm_supported(iommu));
> > +
> > +	/*
> > +	 * The integrated device isn't enumerated in the SATC structure. For
> > +	 * example, it has ATS PCI capability implemented but not validated
> per
> > +	 * the requirements described in the VT-d specification, specifically
> > +	 * in the "Device TLB in System-on-Chip (SoC) Integrated Devices"
> > +	 * section. Therefore, it does not appear in the SATC structure. Return
> > +	 * false in this case.
> > +	 *
> > +	 * On older machines that do not support SATC (i.e., no SATC structure
> > +	 * present), ATS is considered to be "always" supported for integrated
> > +	 * endpoints.
> > +	 */
> > +	return !list_empty(&dmar_satc_units);
> 
> shouldn't it be "return list_empty(&dmar_satc_units);"?

Right, thanks for the catch up.