Use _CRS for PCI resource allocation

This patch resolves an issue where incorrect PCI memory and i/o ranges
are being assigned to hotplugged PCI devices on some IBM systems.  The
resource mis-allocation not only makes the PCI device unuseable but
often makes the entire system unuseable due to resulting machine checks.

The hotplug capable PCI slots on the affected systems are not located
under a standard P2P bridge but are instead located under PCI root
bridges or subtractive decode P2P bridges.  For example, the IBM x3850
contains 2 hotplug capable PCI-X slots and 4 hotplug capable PCIe slots
with the PCI-X slots each located under a PCI root bridge and the PCIe
slots each located under a subtractive decode P2P bridge.

The current i386/x86_64 PCI resource allocation code does not use _CRS
returned resource information.  No other resource information source is
available for slots that are not below a standard P2P bridge so
incorrect ranges are being allocated from e820 hole causing the bad
result.

This patch causes the kernel to use _CRS returned resource info.  It is
roughly based on a change provided by Matthew Wilcox for the ia64 kernel
in 2005.  Due to possible buggy BIOS factor and possible yet to be
discovered kernel issues the function is disabled by default and can be
enabled with pci=use_crs.
Signed-off-by: NGary Hade <gary.hade@us.ibm.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Skip ISA ioresource alignment on some systems

To conserve limited PCI i/o resource on some IBM multi-node systems, the
BIOS allocates (via _CRS) and expects the kernel to use addresses in
ranges currently excluded by pcibios_align_resource() [i386/pci/i386.c].
This change allows the kernel to use the currently excluded address
ranges on the IBM x3800, x3850, and x3950.
Signed-off-by: NGary Hade <gary.hade@us.ibm.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Some broken devices have been discovered to require %al/%ax/%eax registers
for MMIO config space accesses.  Modify mmconfig.c to use these registers
explicitly (rather than modify the global readb/writeb/etc inlines).

AK: also changed i386 to always use eax
AK: moved change to extended space probing to different patch
AK: reworked with inlines according to Linus' requirements.
AK: improve comments.
Signed-off-by: Ndean gaudet <dean@arctic.org>
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Currently, unreachable_devices() compares value of mmconfig and value
of conf1. But it doesn't check the device is reachable or not.
Signed-off-by: NOGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: NAndi Kleen <ak@suse.de>

This just cleans up.
Signed-off-by: NOGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: NAndi Kleen <ak@suse.de>

i386 and x86-64 pci mmconfig code have a lot in common.  So share what's
shareable between the two.
Signed-off-by: NOlivier Galibert <galibert@pobox.com>
Signed-off-by: NAndi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>

This reverts commit 53e4d30d.

It was found that it caused unneeded problems (see
http://bugzilla.kernel.org/show_bug.cgi?id=7082 for details of one such
issue.
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Problem:
New Dell PowerEdge servers have 2 embedded ethernet ports, which are
labeled NIC1 and NIC2 on the chassis, in the BIOS setup screens, and
in the printed documentation.  Assuming no other add-in ethernet ports
in the system, Linux 2.4 kernels name these eth0 and eth1
respectively.  Many people have come to expect this naming.  Linux 2.6
kernels name these eth1 and eth0 respectively (backwards from
expectations).  I also have reports that various Sun and HP servers
have similar behavior.


Root cause:
Linux 2.4 kernels walk the pci_devices list, which happens to be
sorted in breadth-first order (or pcbios_find_device order on i386,
which most often is breadth-first also).  2.6 kernels have both the
pci_devices list and the pci_bus_type.klist_devices list, the latter
is what is walked at driver load time to match the pci_id tables; this
klist happens to be in depth-first order.

On systems where, for physical routing reasons, NIC1 appears on a
lower bus number than NIC2, but NIC2's bridge is discovered first in
the depth-first ordering, NIC2 will be discovered before NIC1.  If the
list were sorted breadth-first, NIC1 would be discovered before NIC2.

A PowerEdge 1955 system has the following topology which easily
exhibits the difference between depth-first and breadth-first device
lists.

-[0000:00]-+-00.0  Intel Corporation 5000P Chipset Memory Controller Hub
           +-02.0-[0000:03-08]--+-00.0-[0000:04-07]--+-00.0-[0000:05-06]----00.0-[0000:06]----00.0  Broadcom Corporation NetXtreme II BCM5708S Gigabit Ethernet (labeled NIC2, 2.4 kernel name eth1, 2.6 kernel name eth0)
           +-1c.0-[0000:01-02]----00.0-[0000:02]----00.0  Broadcom Corporation NetXtreme II BCM5708S Gigabit Ethernet (labeled NIC1, 2.4 kernel name eth0, 2.6 kernel name eth1)


Other factors, such as device driver load order and the presence of
PCI slots at various points in the bus hierarchy further complicate
this problem; I'm not trying to solve those here, just restore the
device order, and thus basic behavior, that 2.4 kernels had.


Solution:

The solution can come in multiple steps.

Suggested fix #1: kernel
Patch below optionally sorts the two device lists into breadth-first
ordering to maintain compatibility with 2.4 kernels.  It adds two new
command line options:
  pci=bfsort
  pci=nobfsort
to force the sort order, or not, as you wish.  It also adds DMI checks
for the specific Dell systems which exhibit "backwards" ordering, to
make them "right".


Suggested fix #2: udev rules from userland
Many people also have the expectation that embedded NICs are always
discovered before add-in NICs (which this patch does not try to do).
Using the PCI IRQ Routing Table provided by system BIOS, it's easy to
determine which PCI devices are embedded, or if add-in, which PCI slot
they're in.  I'm working on a tool that would allow udev to name
ethernet devices in ascending embedded, slot 1 .. slot N order,
subsort by PCI bus/dev/fn breadth-first.  It'll be possible to use it
independent of udev as well for those distributions that don't use
udev in their installers.

Suggested fix #3: system board routing rules
One can constrain the system board layout to put NIC1 ahead of NIC2
regardless of breadth-first or depth-first discovery order.  This adds
a significant level of complexity to board routing, and may not be
possible in all instances (witness the above systems from several
major manufacturers).  I don't want to encourage this particular train
of thought too far, at the expense of not doing #1 or #2 above.


Feedback appreciated.  Patch tested on a Dell PowerEdge 1955 blade
with 2.6.18.

You'll also note I took some liberty and temporarily break the klist
abstraction to simplify and speed up the sort algorithm.  I think
that's both safe and appropriate in this instance.
Signed-off-by: NMatt Domsch <Matt_Domsch@dell.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Some buggy systems can machine check when config space accesses
happen for some non existent devices.  i386/x86-64 do some early
device scans that might trigger this. Allow pci=noearly to disable
this. Also when type 1 is disabling also don't do any early
accesses which are always type1.

This moves the pci= configuration parsing to be a early parameter.
I don't think this can break anything because it only changes
a single global that is only used by PCI.

Cc: gregkh@suse.de
Cc: Trammell Hudson <hudson@osresearch.net>
Signed-off-by: NAndi Kleen <ak@suse.de>

First probe if type1/2 accesses work, but then only initialize them at the end.

This is useful for a later patch that needs this information inbetween.
Signed-off-by: NAndi Kleen <ak@suse.de>

This reverts commits 11012d41 and
40dd2d20, which allowed us to use the
MMIO accesses for PCI config cycles even without the area being marked
reserved in the e820 memory tables.

Those changes were needed for EFI-environment Intel macs, but broke some
newer Intel 965 boards, so for now it's better to revert to our old
2.6.17 behaviour and at least avoid introducing any new breakage.

Andi Kleen has a set of patches that work with both EFI and the broken
Intel 965 boards, which will be applied once they get wider testing.

Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Edgar Hucek <hostmaster@ed-soft.at>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

As a replacement for the earlier removal of the e820 MCFG check
we blacklist the Intel SDV with the original BIOS bug that
motivated that check. On those machines don't use MMCONFIG.

This also adds a new pci=mmconf parameter to override the blacklist.

Cc: Greg KH <gregkh@suse.de>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This patch adds a proper prototype for pcibios_sort() in
arch/i386/pci/pci.h.
Signed-off-by: NAdrian Bunk <bunk@stusta.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

When a PCI device is disabled via pci_disable_device(), it's still
left decoding its BAR resource ranges even though its driver
will have likely released those regions (and may even have
unloaded). pci_enable_device() already explicitly enables
BAR resource decode for the device being enabled. This patch
disables resource decode for the PCI device being disabled,
making it symmetric with the enable call.

I saw this while doing something else, not because of a
problem report. Still, seems to be the correct thing to do.
Signed-off-by: NRajesh Shah <rajesh.shah@intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

I moved it to a separate function which is safer.

This avoids problems with the linker reordering them and the
less useful PCI config space access methods taking priority
over the better ones.

Fixes some problems with broken MMCONFIG

Cc: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

When there is no entry for a bus in MCFG fall back to type1.  This is
especially important on K8 systems where always some devices can't be accessed
using mmconfig (in particular the builtin northbridge doesn't support it for
its own devices)

Cc: <gregkh@suse.de>
Cc: <jgarzik@pobox.com>
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Add reference count and disable ACPI PCI Interrupt Link
when no device still uses it.

Warn when drivers have not released Link at suspend time.

http://bugzilla.kernel.org/show_bug.cgi?id=3469Signed-off-by: NDavid Shaohua Li <shaohua.li@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

I updated this to remove unnecessary variable initialization, make
check_routing be inline only and not __init, switch to strtoul, and
formatting fixes as per Randy Dunlap's recommendations.

I updated this to change pirq_table_addr to a long, and to add a warning
msg if the PIRQ table wasn't found at the specified address, as per thread
with Matthew Wilcox.

In our hardware situation, the BIOS is unable to store or generate it's PIRQ
table in the F0000h-100000h standard range. This patch adds a pci kernel
parameter, pirqaddr to allow the bootloader (or BIOS based loader) to inform
the kernel where the PIRQ table got stored. A beneficial side-effect is that,
if one's BIOS uses a static address each time for it's PIRQ table, then
pirqaddr can be used to avoid the $pirq search through that address block each
time at boot for normal PIRQ BIOSes.
Signed-off-by: NJaya Kumar <jayalk@intworks.biz>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!