linux-pci.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Bjorn Helgaas <helgaas@kernel.org>
To: "Koehrer Mathias (ETAS/ESW5)" <mathias.koehrer@etas.com>
Cc: "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
	"linux-pci@vger.kernel.org" <linux-pci@vger.kernel.org>,
	"bhelgaas@google.com" <bhelgaas@google.com>,
	"hjk@hansjkoch.de" <hjk@hansjkoch.de>
Subject: Re: [PATCH] Extending kernel option pci=resource_alignment to be able to specify PCI device/vendor IDs
Date: Tue, 21 Jun 2016 17:00:03 -0500	[thread overview]
Message-ID: <20160621220003.GC7603@localhost> (raw)
In-Reply-To: <22400b8828ad44ddbccb876cc5ca3b11@FE-MBX1012.de.bosch.com>

On Tue, Jun 07, 2016 at 02:24:17PM +0000, Koehrer Mathias (ETAS/ESW5) wrote:
> Some uio based PCI drivers (e.g. uio_cif) do not work if the assigned 
> PCI memory resources are not page aligned.

Right.  PCI BARs must be aligned on their size.  BARs that are smaller
than a page are not required by PCI to be page-aligned.  Therefore, a
single page may contain several small BARs, or it may contain a small
BAR and some unallocated space.

Both are potential issues because mmap works on a page granularity.
If a user can mmap a page that contains several small BARs, he may be
able to access to more devices than he should.  If a user can mmap a
page that contains unallocated space, she may be able to cause PCI
errors.

You have a device with a 128-byte BAR at 0xeae4f400.  That doesn't
work with uio_cif because b65502879556 ("uio: we cannot mmap unaligned
page contents"), which appeared in v3.13, makes the mmap fail if the
starting address is not page-aligned.

Your patch works around that by moving BARs so they are page-aligned.
I think this is OK, but of course, there's nothing you can do about
the fact that the BAR is smaller than a page, and there might be other
things after the BAR in the same page.  I think that's a problem, and
I wouldn't be surprised if we eventually disallow mmap of any BAR
smaller than a page.

I don't know the history of UIO mmap, and I do see the comment in
vm_iomap_memory() about how we've historically allowed non
page-aligned mmap because I/O memory often has smaller alignment, but
it seems like a safety issue to me, so I'm kind of surprised that we
allow it.

In any case, I think I'll apply this patch for v4.8 because it seems
like a reasonable extension of the existing resource_alignment
support.

> By using the kernel option "pci=resource_alignment" it is possible to force
> single PCI boards to use page alignment for their memory resources.
> However, this is fairly cumbersome if multiple of these boards are in use as 
> the specification of the cards has to be done via PCI bus/slot/function number
> which might change e.g. by adding another board.
> This patch extends the kernel option "pci=resource_alignment" to allow to
> specify the relevant boards via PCI device/vendor (and subdevice/subvendor) ids.
> The specification of the devices via device/vendor is indicated by a leading
> string "pci:" as argument to "pci=resource_alignment".
> The format of the specification is
>   pci:<vendor>:<device>[:<subvendor>:<subdevice>]
> 
> Signed-off-by: Mathias Koehrer <mathias.koehrer@etas.com>
> 
> ---
>  Documentation/kernel-parameters.txt |    2 +
>  drivers/pci/pci.c                   |   66 +++++++++++++++++++++++++-----------
>  2 files changed, 49 insertions(+), 19 deletions(-)
> 
> Index: linux-4.7-rc1/Documentation/kernel-parameters.txt
> ===================================================================
> --- linux-4.7-rc1.orig/Documentation/kernel-parameters.txt
> +++ linux-4.7-rc1/Documentation/kernel-parameters.txt
> @@ -2998,6 +2998,8 @@ bytes respectively. Such letter suffixes
>  		resource_alignment=
>  				Format:
>  				[<order of align>@][<domain>:]<bus>:<slot>.<func>[; ...]
> +				[<order of align>@]pci:<vendor>:<device>\
> +						[:<subvendor>:<subdevice>][; ...] 
>  				Specifies alignment and device to reassign
>  				aligned memory resources.
>  				If <order of align> is not specified,
> Index: linux-4.7-rc1/drivers/pci/pci.c
> ===================================================================
> --- linux-4.7-rc1.orig/drivers/pci/pci.c
> +++ linux-4.7-rc1/drivers/pci/pci.c
> @@ -4755,6 +4755,7 @@ static DEFINE_SPINLOCK(resource_alignmen
>  static resource_size_t pci_specified_resource_alignment(struct pci_dev *dev)
>  {
>  	int seg, bus, slot, func, align_order, count;
> +	unsigned short vendor, device, subsystem_vendor, subsystem_device;
>  	resource_size_t align = 0;
>  	char *p;
>  
> @@ -4768,28 +4769,55 @@ static resource_size_t pci_specified_res
>  		} else {
>  			align_order = -1;
>  		}
> -		if (sscanf(p, "%x:%x:%x.%x%n",
> -			&seg, &bus, &slot, &func, &count) != 4) {
> -			seg = 0;
> -			if (sscanf(p, "%x:%x.%x%n",
> -					&bus, &slot, &func, &count) != 3) {
> -				/* Invalid format */
> -				printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: %s\n",
> -					p);
> +		if (strncmp(p, "pci:", 4) == 0) {
> +			/* PCI vendor/device (subvendor/subdevice) ids are specified */
> +			p += 4;
> +			if (sscanf(p, "%hx:%hx:%hx:%hx%n",
> +				&vendor, &device, &subsystem_vendor, &subsystem_device, &count) != 4) {
> +				if (sscanf(p, "%hx:%hx%n", &vendor, &device, &count) != 2) {
> +					printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: pci:%s\n",
> +						p);
> +					break;
> +				}
> +				subsystem_vendor = subsystem_device = 0;
> +			}
> +			p += count;
> +			if ((!vendor || (vendor == dev->vendor)) &&
> +				(!device || (device == dev->device)) &&
> +				(!subsystem_vendor || (subsystem_vendor == dev->subsystem_vendor)) &&
> +				(!subsystem_device || (subsystem_device == dev->subsystem_device))) {
> +				if (align_order == -1)
> +					align = PAGE_SIZE;
> +				else
> +					align = 1 << align_order;
> +				/* Found */
>  				break;
>  			}
>  		}
> -		p += count;
> -		if (seg == pci_domain_nr(dev->bus) &&
> -			bus == dev->bus->number &&
> -			slot == PCI_SLOT(dev->devfn) &&
> -			func == PCI_FUNC(dev->devfn)) {
> -			if (align_order == -1)
> -				align = PAGE_SIZE;
> -			else
> -				align = 1 << align_order;
> -			/* Found */
> -			break;
> +		else {
> +			if (sscanf(p, "%x:%x:%x.%x%n",
> +				&seg, &bus, &slot, &func, &count) != 4) {
> +				seg = 0;
> +				if (sscanf(p, "%x:%x.%x%n",
> +						&bus, &slot, &func, &count) != 3) {
> +					/* Invalid format */
> +					printk(KERN_ERR "PCI: Can't parse resource_alignment parameter: %s\n",
> +						p);
> +					break;
> +				}
> +			}
> +			p += count;
> +			if (seg == pci_domain_nr(dev->bus) &&
> +				bus == dev->bus->number &&
> +				slot == PCI_SLOT(dev->devfn) &&
> +				func == PCI_FUNC(dev->devfn)) {
> +				if (align_order == -1)
> +					align = PAGE_SIZE;
> +				else
> +					align = 1 << align_order;
> +				/* Found */
> +				break;
> +			}
>  		}
>  		if (*p != ';' && *p != ',') {
>  			/* End of param or invalid format */
> --
> To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

  parent reply	other threads:[~2016-06-21 22:00 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-07 14:24 [PATCH] Extending kernel option pci=resource_alignment to be able to specify PCI device/vendor IDs Koehrer Mathias (ETAS/ESW5)
2016-06-07 14:32 ` gregkh
2016-06-21 22:00 ` Bjorn Helgaas [this message]
2016-06-21 22:04 ` Bjorn Helgaas
  -- strict thread matches above, loose matches on Subject: below --
2016-06-08  6:14 Koehrer Mathias (ETAS/ESW5)
2016-06-09 12:40 Koehrer Mathias (ETAS/ESW5)
2016-06-09 16:12 ` gregkh
2016-08-08  7:39 Koehrer Mathias (ETAS/ESW5)
2016-08-08 14:01 ` Bjorn Helgaas

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=20160621220003.GC7603@localhost \
    --to=helgaas@kernel.org \
    --cc=bhelgaas@google.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=hjk@hansjkoch.de \
    --cc=linux-pci@vger.kernel.org \
    --cc=mathias.koehrer@etas.com \
    /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).