From: Bjorn Helgaas <helgaas@kernel.org>
To: Hannes Reinecke <hare@suse.de>
Cc: Bjorn Helgaas <bhelgaas@google.com>,
linux-pci@vger.kernel.org,
Jordan Hargrave <Jordan_Hargrave@dell.com>,
Babu Moger <babu.moger@oracle.com>,
Alexander Duyck <alexander.duyck@gmail.com>
Subject: Re: [PATCH v4 00/11] PCI VPD access fixes
Date: Tue, 23 Feb 2016 18:45:42 -0600 [thread overview]
Message-ID: <20160224004542.GA13026@localhost> (raw)
In-Reply-To: <56CCF9AE.9080601@suse.de>
On Wed, Feb 24, 2016 at 08:30:38AM +0800, Hannes Reinecke wrote:
> On 02/24/2016 06:36 AM, Bjorn Helgaas wrote:
> > On Wed, Feb 24, 2016 at 05:48:33AM +0800, Hannes Reinecke wrote:
> >> On 02/23/2016 10:07 PM, Bjorn Helgaas wrote:
> >>> Hi Hannes,
> >>>
> >>> Thanks for taking a look at the rest of these.
> >>>
> >>> On Mon, Feb 22, 2016 at 06:46:23PM -0600, Bjorn Helgaas wrote:
> >>>> Hi Hannes,
> >>>>
> >>>> This is a revision of your v3 series:
> >>>> http://lkml.kernel.org/r/1455525722-122040-1-git-send-email-hare@suse.de
> >>>>
> >>>> Here's the description from your v3 posting:
> >>>>
> >>>> the current PCI VPD page access assumes that the entire possible VPD
> >>>> data is readable. However, the spec only guarantees a VPD data up to
> >>>> the 'end' marker, with everything beyond that being undefined.
> >>>> This causes a system lockup on certain devices.
> >>>>
> >>>> With this patch we always set the VPD sysfs attribute size to '0', and
> >>>> calculate the available VPD size on the first access.
> >>>> If no valid data can be read an I/O error is returned.
> >>>
> >>> Just to see if I have this right: the VPD file size in sysfs will
> >>> always appear as zero, regardless of whether it has been read or
> >>> written, right? I don't think the user-visible size should change.
> >>>
> >> That is correct.
> >> As the actual size is evaluated on the first access, we don't have it
> >> available when creating the sysfs attribute itself.
> >> And when using the nominal size of 32k some bright program might try to
> >> jump to somewhere in the middle of the data, which will make calculating
> >> the validity of this horribly complex.
> >> Setting it to '0' is an easy way of avoiding this kinda games.
> >>
> >> So yes, there will be a user-visible change, but it shouldn't affect the
> >> programs accessing this attribute.
> >> lspci works happily with these changes
> >
> > What is the user-visible change? Here's what I'm thinking. If we do
> > this:
> >
> > ls -l /sys/.../vpd
> > dd if=/sys/.../vpd bs=1 count=1
> > ls -l /sys/.../vpd
> >
> > Do we see different sizes from the two "ls" invocations? My thought
> > is that we should see '0' both times, because I don't really think
> > that output should change depending on previous actions of this user
> > or other users.
> >
> Originally we have:
>
> # ls -l 0000:07:00.0/vpd
> -rw------- 1 root root 32768 Feb 24 01:29 0000:07:00.0/vpd
>
> and with this patchset we have:
>
> # ls -l 0000:07:00.0/vpd
> -rw------- 1 root root 0 Feb 24 01:29 0000:07:00.0/vpd
>
> So only programs doing a 'stat' on the device node will see a difference.
Oh, I think I see: you mean there's a user-visible difference between
the current tree and the tree with your patches applied.
I was hoping that on a single kernel, the "vpd" attribute size was
always the same, regardless of whether anybody had read or written it.
If we always report zero size for all "vpd" files, I think that's OK.
Bjorn
next prev parent reply other threads:[~2016-02-24 0:45 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-02-23 0:46 [PATCH v4 00/11] PCI VPD access fixes Bjorn Helgaas
2016-02-23 0:46 ` [PATCH v4 01/11] PCI: Update VPD definitions Bjorn Helgaas
2016-02-23 0:46 ` [PATCH v4 02/11] PCI: Allow access to VPD attributes with size 0 Bjorn Helgaas
2016-02-23 0:46 ` [PATCH v4 03/11] PCI: Use bitfield instead of bool for struct pci_vpd_pci22.busy Bjorn Helgaas
2016-02-23 12:19 ` Hannes Reinecke
2016-02-23 0:46 ` [PATCH v4 04/11] PCI: Determine actual VPD size on first access Bjorn Helgaas
2016-02-23 0:47 ` [PATCH v4 05/11] FIXME need bugzilla link Bjorn Helgaas
2016-02-23 0:47 ` [PATCH v4 06/11] PCI: Move pci_read_vpd() and pci_write_vpd() close to other VPD code Bjorn Helgaas
2016-02-23 12:20 ` Hannes Reinecke
2016-02-23 0:47 ` [PATCH v4 07/11] PCI: Move pci_vpd_release() from header file to pci/access.c Bjorn Helgaas
2016-02-23 12:21 ` Hannes Reinecke
2016-02-23 0:47 ` [PATCH v4 08/11] PCI: Remove struct pci_vpd_ops.release function pointer Bjorn Helgaas
2016-02-23 12:22 ` Hannes Reinecke
2016-02-23 0:47 ` [PATCH v4 09/11] PCI: Rename VPD symbols to remove unnecessary "pci22" Bjorn Helgaas
2016-02-23 12:23 ` Hannes Reinecke
2016-02-23 0:47 ` [PATCH v4 10/11] PCI: Fold struct pci_vpd_pci22 into struct pci_vpd Bjorn Helgaas
2016-02-23 12:24 ` Hannes Reinecke
2016-02-23 0:47 ` [PATCH v4 11/11] PCI: Sleep rather than busy-wait for VPD access completion Bjorn Helgaas
2016-02-23 12:25 ` Hannes Reinecke
2016-02-23 14:07 ` [PATCH v4 00/11] PCI VPD access fixes Bjorn Helgaas
2016-02-23 21:48 ` Hannes Reinecke
2016-02-23 22:36 ` Bjorn Helgaas
2016-02-24 0:30 ` Hannes Reinecke
2016-02-24 0:45 ` Bjorn Helgaas [this message]
2016-02-23 17:11 ` Bjorn Helgaas
2016-02-24 4:52 ` Seymour, Shane M
2016-02-29 22:36 ` Babu Moger
2016-02-29 17:27 ` Babu Moger
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20160224004542.GA13026@localhost \
--to=helgaas@kernel.org \
--cc=Jordan_Hargrave@dell.com \
--cc=alexander.duyck@gmail.com \
--cc=babu.moger@oracle.com \
--cc=bhelgaas@google.com \
--cc=hare@suse.de \
--cc=linux-pci@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).