| Age | Commit message (Collapse) | Author |
|---|
|
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: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
The pci channel state is currently uninitialized, thus there are two ways
of indicating that "everything's OK": 0 and 1. This is a bit of a burden.
If a devce driver wants to check if the pci channel is in a working or a
disconnected state, the driver writer must perform checks similar to
if((pdev->error_state != 0) &&
(pdev->error_state != pci_channel_io_normal)) {
whatever();
}
which is rather akward. The first check is needed because stuct pci_dev is
inited to all-zeros. The scond is needed because the error recovery will
set the state to pci_channel_io_normal (which is not zero).
This patch fixes this awkwardness.
Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
pci_walk_bus has a race with pci_destroy_dev. When cb is called
in pci_walk_bus, pci_destroy_dev might unlink the dev pointed by next.
Later on in the next loop, pointer next becomes NULL and cause
kernel panic.
Below patch against 2.6.17-rc4 fixes it by changing pci_bus_lock (spin_lock)
to pci_bus_sem (rw_semaphore).
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
When we detect a 64-bit pre-set address in a BAR on a 32-bit platform,
we disable it and treat it as if it had been unset, thus allowing the
general address assignment code to assign a new address to it when the
device is enabled. This can happen either if the firmware assigns
64-bit addresses; additionally, some cards have been found "in the
wild" which do not come out of reset with all the BAR registers set to
zero.
Unfortunately, the patch that implemented this tested the low part of
the address instead of the high part of the address. This patch fixes
that.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
[pci] Ignore pre-set 64-bit BARs on 32-bit platforms
Currently, Linux always rejects a device which has a pre-set 64-bit
address on a 32-bit platform. On systems which do not do PCI
initialization in firmware, this causes some devices which don't
correctly power up with all BARs zero to fail.
This patch makes the kernel automatically zero out such an address
(thus treating it as if it had not been set at all, meaning it will
assign an address if necessary).
I have done this only for devices, not bridges. It seems potentially
hazardous to do for bridges.
Signed-off-by: H. Peter Anvin <hpa@c2micro.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
this patch converts drivers/pci to kzalloc usage.
Compile tested with allyes config.
Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
"In some cases, especially on modern laptops with a lot of PCI and cardbus
bridges, we're unable to assign correct secondary/subordinate bus numbers
to all cardbus bridges due to BIOS limitations unless we are using
"pci=assign-busses" boot option." -- Ivan Kokshaysky (from a patch comment)
Without it, Cardbus cards inserted are never seen by PCI because the parent
PCI-PCI Bridge of the Cardbus bridge will not pass and translate Type 1 PCI
configuration cycles correctly and the system will fail to find and
initialise the PCI devices in the system.
Reference: PCI-PCI Bridges: PCI Configuration Cycles and PCI Bus Numbering:
http://www.science.unitn.it/~fiorella/guidelinux/tlk/node72.html
The reason for this is that:
``All PCI busses located behind a PCI-PCI bridge must reside between the
secondary bus number and the subordinate bus number (inclusive).''
"pci=assign-busses" makes pcibios_assign_all_busses return 1 and this
turns on PCI renumbering during PCI probing.
Alan suggested to use DMI automatically set assign-busses on problem systems.
The only question for me was where to put it. I put it directly before
scanning PCI bus into pcibios_scan_root() because it's called from legacy,
acpi and numa and so it can be one place for all systems and configurations
which may need it.
AMD64 Laptops are also affected and fixed by assign-busses, and the code is
also incuded from arch/x86_64/pci/ that place will also work for x86_64
kernels, I only ifdef'-ed the x86-only Laptop in this example.
Affected and known or assumed to be fixed with it are (found by googling):
* ASUS Z71V and L3s
* Samsung X20
* Compaq R3140us and all Compaq R3000 series laptops with TI1620 Controller,
also Compaq R4000 series (from a kernel.org bugreport)
* HP zv5000z (AMD64 3700+, known that fixup_parent_subordinate_busnr fixes it)
* HP zv5200z
* IBM ThinkPad 240
* An IBM ThinkPad (1.8 GHz Pentium M) debugged by Pavel Machek
gives the correspondig message which detects the possible problem.
* MSI S260 / Medion SIM 2100 MD 95600
The patch also expands the "try pci=assign-busses" warning so testers will
help us to update the DMI table.
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
It turns out AMD 8131 quirk only affects MSI for devices behind the 8131 bridge.
Handle this by adding a flags field in pci_bus, inherited from parent to child.
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
from pci_scan_bridge.
> On Mon, Feb 13, 2006 at 05:13:21PM -0800, David S. Miller wrote:
> >
> > In drivers/pci/probe.c:pci_scan_bridge(), if this is not the first
> > pass (pass != 0) we don't restore the PCI_BRIDGE_CONTROL_REGISTER and
> > thus leave PCI_BRIDGE_CTL_MASTER_ABORT off:
> >
> > int __devinit pci_scan_bridge(struct pci_bus *bus, struct pci_dev * dev, int max, int pass)
> > {
> > ...
> > /* Disable MasterAbortMode during probing to avoid reporting
> > of bus errors (in some architectures) */
> > pci_read_config_word(dev, PCI_BRIDGE_CONTROL, &bctl);
> > pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
> > bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
> > ...
> > if ((buses & 0xffff00) && !pcibios_assign_all_busses() && !is_cardbus) {
> > unsigned int cmax, busnr;
> > /*
> > * Bus already configured by firmware, process it in the first
> > * pass and just note the configuration.
> > */
> > if (pass)
> > return max;
> > ...
> > }
> >
> > pci_write_config_word(dev, PCI_BRIDGE_CONTROL, bctl);
> > ...
> >
> > This doesn't seem intentional.
Agreed, looks like an accident. The patch [1] originally came from Kip
Walker (Broadcom back then) between 2.6.0-test3 and 2.6.0-test4. As I
recall it was supposed to fix an issue with with PCI aborts being
signalled by the PCI bridge of the Broadcom BCM1250 family of SOCs when
probing behind pci_scan_bridge. It is undeseriable to disable
PCI_BRIDGE_CTL_MASTER_ABORT in pci_{read,write)_config_* and the
behaviour wasn't considered a bug in need of a workaround, so this was
put in probe.c.
I don't have an affected system at hand, so can't really test but I
propose something like the below patch.
[1] http://www.linux-mips.org/git?p=linux.git;a=commit;h=599457e0cb702a31a3247ea6a5d9c6c99c4cf195
[PCI] Avoid leaving MASTER_ABORT disabled permanently when returning from pci_scan_bridge.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
After you find the maximum value of the subordinate buses below the child
bus, you must fix the parent's subordinate bus number again, otherwise
it may be too small.
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
The powerpc PCI code sets up the PCI tree without doing config space
accesses in most cases, from the firmware tree. However, it still wants
to call pci_cfg_space_size() under some conditions, thus it needs to
be made non-static (though I don't see a point to export it to modules).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
Add a warning if a child bus may be inaccessible because the
parent bridge has wrong secondary or subordinate bus numbers.
Note that this may or may not happen on "transparent" bridges,
as can be seen in bug #5557.
Also, if we do not fix up the assignment of bus numbers, try to
make use of the bus number space available.
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
I've implemented a quirk to take advantage of the 1KB I/O space
granularity option on the Intel P64H2 PCI Bridge. I had to change
probe.c because it sets the resource start and end to be aligned on 4k
boundaries (after the quirk sets them to 1k boundaries). I've tested
this patch on a Unisys ES7000-600 both with and without the 1KB option
enabled. I also tested this on a 2 processor Dell box that doesn't have
a P64H2 to make sure there were no negative affects there.
Signed-off-by: Dan Yeisley <dan.yeisley@unisys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
Call pci_read_irq() for bridges too, so that the pin value
is stored for bridges that require interrupts.
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
Store the value of the INTERRUPT_PIN in the pci_dev structure
so that it can be retrieved later.
Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
PCI: add descriptions for missing function parameters.
Eliminate all kernel-doc warnings here.
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
I believe the change that broke things is introduction of
pci_fixup_parent_subordinate_busnr().
The patch here does two things:
- hunk #1 should fix the problems you've seen when you boot without
additional "pci" kernel options;
- hunk #2 supposedly fixes boot with "pci=assign-busses" option which
otherwise hangs Acer TM81xx machines as reported.
Please try this with and without "pci=assign-busses". If it boots,
I'd like to see 'lspci -vvx' for both cases.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This function expects an unsigned 32-bit type as its third argument:
static u32 pci_size(u32 base, u32 maxbase, u32 mask)
However, given these definitions:
#define PCI_BASE_ADDRESS_MEM_MASK (~0x0fUL)
#define PCI_ROM_ADDRESS_MASK (~0x7ffUL)
these two calls in drivers/pci/probe.c are problematic for architectures
for which a UL is not equivalent to a u32:
sz = pci_size(l, sz, PCI_BASE_ADDRESS_MEM_MASK);
sz = pci_size(l, sz, PCI_ROM_ADDRESS_MASK);
Hence the below compile warning when building for ARCH=ppc64:
drivers/pci/probe.c: In function `pci_read_bases':
/.../probe.c:168: warning: large integer implicitly truncated to unsigned type
/.../probe.c:218: warning: large integer implicitly truncated to unsigned type
Here is a simple fix.
Signed-off-by: Amos Waterland <apw@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
pcibus_to_cpumask expands into more than just an initialiser so gcc
moans about code before variable declarations.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
This patch makes some small rearrangements of the PCI probing code in
order to make it possible for arch code to set up the PCI tree
without needing to duplicate code from the PCI layer unnecessarily.
PPC64 will use this to set up the PCI tree from the Open Firmware
device tree, which we need to do on logically-partitioned pSeries
systems.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
- support PCI PM CAP version 3 (as defined in PCI PM Interface Spec v1.2)
- pci/probe.c sets the PM state initially to 4 which is D3cold. add a
PCI_UNKNOWN
- minor cleanups
Signed-off-by: Daniel Ritz <daniel.ritz@gmx.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
This patch removes CONFIG_PCI_NAMES.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
The setup-bus code doesn't work correctly for configurations
with more than one display adapter in the same PCI domain.
This stuff actually is a leftover of an early 2.4 PCI setup code
and apparently it stopped working after some "bridge_ctl" changes.
So the best thing we can do is just to remove it and rely on the fact
that any firmware *has* to configure VGA port forwarding for the boot
display device properly.
But then we need to ensure that the bus->bridge_ctl will always
contain valid information collected at the probe time, therefore
the following change in pci_scan_bridge() is needed.
Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
With the number of PCI bus resources increased to 8, we can
handle the subtractive decode PCI-PCI bridge like a normal
bridge, taking into account standard PCI-PCI bridge windows
(resources 0-2). This helps to avoid problems with peer-to-peer DMA
behind such bridges, poor performance for MMIO ranges outside bridge
windows and prefetchable vs. non-prefetchable memory issues.
To reflect the fact that such bridges do forward all addresses to
the secondary bus (transparency), remaining bus resources 3-7 are
linked to resources 0-4 of the primary bus. These resources will be
used as fallback by resource management code if allocation from
standard bridge windows fails for some reason.
Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Acked-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
When the cardbus bridge is behind another bridge change the routing
in the parent bridge for new cards. This fixes Cardbus on various AMD64
laptops.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
on creation
When a pci child bus is created, add it to the parent's children list
immediately rather than waiting till pci_bus_add_devices(). For hot-plug
bridges/devices, pci_bus_add_devices() may be called much later, after they
have been properly configured. In the meantime, this allows us to use the
normal pci bus search functions for the hot-plug bridges/buses.
Signed-off-by: Rajesh Shah <rajesh.shah@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
device lists
With root bridge and pci bridge hot-plug, new buses and devices can be added
or removed at run time. Protect the pci bus and device lists with the pci
lock when doing so.
Signed-off-by: Rajesh Shah <rajesh.shah@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
bridge
When hot-plugging a root bridge, as we try to assign bus numbers we may find
that the hotplugged hieratchy has more PCI to PCI bridges (i.e. bus
requirements) than available. Make sure we don't step over an existing bus
when that happens.
Signed-off-by: Rajesh Shah <rajesh.shah@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
When you hot-plug a (root) bridge hierarchy, it may have p2p bridges and
devices attached to it that have not been configured by firmware. In this
case, we need to configure the devices before starting them. This patch
separates device start from device scan so that we can introduce the
configuration step in the middle.
I kept the existing semantics for pci_scan_bus() since there are a huge number
of callers to that function.
Also, I have no way of testing the changes I made to the parisc files, so this
needs review by those folks. Sorry for the massive cross-post, this touches
files in many different places.
Signed-off-by: Rajesh Shah <rajesh.shah@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
|
|
On 64-bit machines, PCI_BASE_ADDRESS_MEM_MASK and other mask constants
passed to pci_size() are 64-bit (for example ~0x0fUL). However, pci_size
does comparisons between the u32 arguments and the mask, which will fail
even though any result from pci_size is still just 32-bit.
Changing the mask argument to u32 seems the obvious thing to do, since all
arithmetic in the function is 32-bit and having a larger mask makes no
sense.
This triggered on a PPC64 system here where an adapter (VGA, as it
happened) had a memory region base of 0xfe000000 and a sz of the same,
matching the if (max == maxbase ...) test at the bottom of pci_size but
failing the mask comparison. Quite a corner case which I guess explains
why we haven't seen it until now.
Signed-off-by: Olof Johansson <olof@lixom.net>
Acked-by: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
|
|
Signed-off-by: Greg 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!
|
