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@xxxxxxxx>
>
> ---
> 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@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
