linux-kernel - Re: [PATCH V7 3/4] PCI: vmd: Add vmd_device

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] [day] [month] [year] [list]

Message-ID: <ddd1c452d8843fa137bc294b04ee5195967e4b67.camel@linux.intel.com>
Date:   Mon, 31 Oct 2022 08:40:49 -0700
From:   "David E. Box" <david.e.box@...ux.intel.com>
To:     Christoph Hellwig <hch@...radead.org>,
        Bjorn Helgaas <helgaas@...nel.org>
Cc:     nirmal.patel@...ux.intel.com, jonathan.derrick@...ux.dev,
        lorenzo.pieralisi@....com, kw@...ux.com, robh@...nel.org,
        bhelgaas@...gle.com, michael.a.bottini@...el.com,
        rafael@...nel.org, me@...ityamohan.in, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH V7 3/4] PCI: vmd: Add vmd_device_data

On Mon, 2022-10-31 at 00:04 -0700, Christoph Hellwig wrote:
> On Fri, Oct 28, 2022 at 02:13:08PM -0500, Bjorn Helgaas wrote:
> > It looks like these devices come in families where several device IDs
> > share the same features.  I think this would be more readable if you
> > defined each family outside this table and simply referenced the
> > family here.  E.g., you could do something like:
> > 
> >   static struct vmd_device_data vmd_v1 = {
> >     .features = VMD_FEAT_HAS_MEMBAR_SHADOW_VSCAP |
> > 		VMD_FEAT_HAS_BUS_RESTRICTIONS |
> > 		VMD_FEAT_OFFSET_FIRST_VECTOR,
> >   };
> > 
> >   {PCI_VDEVICE(INTEL, PCI_DEVICE_ID_INTEL_VMD_9A0B),
> >     .driver_data = (kernel_ulong_t) &vmd_v1,
> > 
> > Then you can add VMD_FEAT_BIOS_PM_QUIRK and the .ltr value in one place
> > instead of repeating it a half dozen times.
> 
> I wonder why we need the ltr field at all.  For those that set it
> is always the same value, so it could just be a quirk flag to set it.

Yeah, this makes sense particularly since this isn't intended as a permanent
fix. I'll get rid of it.

> 
> Tat being said I think thegrouping makes a lot of sense, but I'd just
> do it with a #define for the set of common quirk flags.

Works for me. I'll create a VMD_FEATS_CLIENT group but I'll keep the ltr quirk
separate since future client systems won't be using it.

David