linux-kernel - Re: [PATCH] PCI: Move ATS declarations to linux/pci.h

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: <20190902211100.GH7013@google.com>
Date:   Mon, 2 Sep 2019 16:11:00 -0500
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     Christoph Hellwig <hch@...radead.org>
Cc:     Krzysztof Wilczynski <kw@...ux.com>, Will Deacon <will@...nel.org>,
        Joerg Roedel <joro@...tes.org>, linux-kernel@...r.kernel.org,
        iommu@...ts.linux-foundation.org, linux-pci@...r.kernel.org,
        Robin Murphy <robin.murphy@....com>,
        David Woodhouse <dwmw2@...radead.org>,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH] PCI: Move ATS declarations to linux/pci.h

On Fri, Aug 30, 2019 at 09:18:40AM -0700, Christoph Hellwig wrote:
> On Fri, Aug 30, 2019 at 05:07:56PM +0200, Krzysztof Wilczynski wrote:
> > Move ATS function prototypes from include/linux/pci-ats.h to
> > include/linux/pci.h so users only need to include <linux/pci.h>:
> 
> Why is that so important?  Very few PCI(e) device drivers use ATS,
> so keeping it out of everyones include hell doesn't seem all bad.

This was my idea, and it wasn't a good one, sorry.

The ATS, PRI, and PASID interfaces are all sort of related and are
used only by the IOMMU drivers, so it probably makes sense to put them
all together.  Right now the ATS stuff is in linux/pci.h and PRI/PASID
stuff is in linux/pci-ats.h.  Maybe the right thing would be to move
the ATS stuff to pci-ats.h.

I previously moved it from pci-ats.h to pci.h with ff9bee895c4d ("PCI:
Move ATS declarations to linux/pci.h so they're all together") with
the excuse of putting the external ATS interfaces next to
pci_ats_init().  But that really looks like it was a mistake because
pci_ats_init() is a PCI-internal thing and its declaration should
probably be in drivers/pci/pci.h instead.

There's also a useless "struct pci_ats" forward declaration in
linux/pci.h that I should have removed with d544d75ac96a ("PCI: Embed
ATS info directly into struct pci_dev").

Bjorn