linux-kernel - Re: [PATCH 3/6] pci: add PCI quirk cmdmem fixup for Intel DSA 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]

Date:   Thu, 2 Apr 2020 00:39:40 -0700
From:   Christoph Hellwig <hch@...radead.org>
To:     Dan Williams <dan.j.williams@...el.com>
Cc:     Christoph Hellwig <hch@...radead.org>,
        Dave Jiang <dave.jiang@...el.com>,
        Vinod Koul <vkoul@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        "H. Peter Anvin" <hpa@...or.com>,
        Bjorn Helgaas <bhelgaas@...gle.com>,
        Greg KH <gregkh@...uxfoundation.org>,
        Arnd Bergmann <arnd@...db.de>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        X86 ML <x86@...nel.org>, dmaengine@...r.kernel.org,
        "Raj, Ashok" <ashok.raj@...el.com>,
        Fenghua Yu <fenghua.yu@...el.com>, linux-pci@...r.kernel.org,
        "Luck, Tony" <tony.luck@...el.com>, Jing Lin <jing.lin@...el.com>,
        Sanjay K Kumar <sanjay.k.kumar@...el.com>
Subject: Re: [PATCH 3/6] pci: add PCI quirk cmdmem fixup for Intel DSA device

On Wed, Apr 01, 2020 at 07:20:59PM -0700, Dan Williams wrote:
> On Wed, Apr 1, 2020 at 12:19 AM Christoph Hellwig <hch@...radead.org> wrote:
> >
> > On Mon, Mar 30, 2020 at 02:27:06PM -0700, Dave Jiang wrote:
> > > Since there is no standard way that defines a PCI device that receives
> > > descriptors or commands with synchronous write operations, add quirk to set
> > > cmdmem for the Intel accelerator device that supports it.
> >
> > Why do we need a quirk for this?  Even assuming a flag is needed in
> > struct pci_dev (and I don't really understand why that is needed to
> > start with), it could be set in ->probe.
> 
> The consideration in my mind is whether this new memory type and
> instruction combination warrants a new __attribute__((noderef,
> address_space(X))) separate from __iomem. If it stays a single device
> concept layered on __iomem then yes, I agree it can all live locally
> in the driver. However, when / if this facility becomes wider spread,
> as the PCI ECR in question is trending, it might warrant general
> annotation.
> 
> The enqcmds instruction does not operate on typical x86 mmio
> addresses, only these special device portals offer the ability to have
> non-posted writes with immediate results in the cpu condition code
> flags.

But that is not what this series does at all.  And I think it makes
sense to wait until it gains adoption to think about a different address
space.  In this series we could just trivially kill patches 2-4 and make
it much easier to understand.