linux-kernel - Re: [RFC PATCH 2/3] PCI: Pass available resources into pci_create

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]

Date:	Wed, 24 Aug 2011 08:28:18 -0600
From:	Bjorn Helgaas <bhelgaas@...gle.com>
To:	Deng-Cheng Zhu <dengcheng.zhu@...il.com>
Cc:	jbarnes@...tuousgeek.org, ralf@...ux-mips.org,
	linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org,
	linux-mips@...ux-mips.org, eyal@...s.com, zenon@...s.com
Subject: Re: [RFC PATCH 2/3] PCI: Pass available resources into pci_create_bus()

On Wed, Aug 24, 2011 at 12:24 AM, Deng-Cheng Zhu
<dengcheng.zhu@...il.com> wrote:
> Currently, after pci_create_bus(), resources available on the bus could be
> handled by pci_scan_child_bus(). The problem is that, in
> pci_scan_child_bus(), before calling arch-dependent pcibios_fixup_bus(),
> PCI quirks will probably conflict (while doing pci_claim_resource()) with
> resources like system controller's I/O resource that have not yet been
> added to the bus. So, add those resources right before returning the newly
> created bus in pci_create_bus().

I like this approach a lot.  Thanks for working it up.  It's a nice
small change with very little impact to other architectures, and you
have a nice clear changelog.  You might mention something about the
fact that by default, the bus starts out with all of ioport_resource
and iomem_resource -- that will mean something to people who know how
host bridges work.

> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
> index 8473727..7735fe7 100644
> --- a/drivers/pci/probe.c
> +++ b/drivers/pci/probe.c
> @@ -1516,7 +1516,8 @@ unsigned int __devinit pci_scan_child_bus(struct pci_bus *bus)
>  }
>
>  struct pci_bus * pci_create_bus(struct device *parent,
> -               int bus, struct pci_ops *ops, void *sysdata)
> +               int bus, struct pci_ops *ops, void *sysdata,
> +               struct pci_bus_resource *bus_res)
>  {
>        int error;
>        struct pci_bus *b, *b2;
> @@ -1570,8 +1571,14 @@ struct pci_bus * pci_create_bus(struct device *parent,
>        pci_create_legacy_files(b);
>
>        b->number = b->secondary = bus;
> -       b->resource[0] = &ioport_resource;
> -       b->resource[1] = &iomem_resource;
> +
> +       /* Add initial resources to the bus */
> +       if (bus_res != NULL) {
> +               list_add_tail(&b->resources, &bus_res->list);
> +       } else {
> +               pci_bus_add_resource(b, &ioport_resource, 0);
> +               pci_bus_add_resource(b, &iomem_resource, 0);
> +       }

Using pci_bus_add_resource() here *seems* like it should be the right
thing, but I don't think it will work correctly.

The problem is that struct pci_bus has both a table of resources
(bus->resource[]) *and* a list (bus->resources).
pci_bus_add_resource() always puts the new resource on the list, but
various arch code still references the table directly, e.g., sparc has
"pbus->resource[0] = &pbm->io_space" in pcibios_fixup_bus().

As written, I think this patch will break sparc because the host
bridge will end up with both pbm->io_space (in the table) and
ioport_resource (in the list).

I think something like this would work, though:

    if (bus_res)
        list_add_tail(&b->resources, &bus_res->list);
    else {
        b->resource[0] = &ioport_resource;
        b->resource[1] = &iomem_resource;
    }

Bjorn
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/