| Age | Commit message (Collapse) | Author |
|---|
|
The IOQ code uses structures which are designed to be shared between
disparate systems, such as 32-bit and 64-bit, as well as Linux and Windows.
Since IOQ is primarily intended to be used by qemu/kvm, which support
virtual guests with a different CPU architecture than the host, clearly
define the endianness for the shared structures.
The endianness is defined to be little-endian, to avoid byte swapping in
the most common case: x86.
Note that the cookie member was not changed to have a fixed endianness.
This is because it is only intended for use by one side of the IOQ.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
|
|
The sparse utility gave tons of warnings about signed bitfields. A simple
inspection shows that they are all used as booleans, so convert them to the
correct type.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
|
|
The sparse tool caught many missing static declarations. Add them.
There are still many address space errors, but I'm unsure about the best
way to fix these.
Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
|
|
We should technically enable the busmaster bit, even though its not
actually used on the host side
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
This enables the guest to automatically load the appropriate driver
when vbus devices are detected.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
We cannot rely on create_irq() being available in all circumstances,
therefore lets remove it from the code for now until we can figure
out a way to support it with config options.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
DEVADD events currently perform a GFP_KERNEL allocation for the device
object in interrupt context. This is technically illegal, although we
have gotten away with it to date by sheer luck that the allocation
never tried to swap or otherwise sleep.
Lets fix this properly by making sure that we only allocated the space
for the device object using GFP_KERNEL from process-context. We achieve
this by generating a temporary GFP_ATOMIC relay for the event and
deferring the actual device allocation/registration to process context.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
This will allow us to view the vector statistics in a uniform way, as
well as pave the way for supporting irqbalance in the future.
The VBUS PCI-BRIDGE will present itself as an irqchip device to the
kernel proper. Each shm-signal that is created by a specific driver
will show up as a "VBUS-edge" IRQ in /proc/interrupts. For instane,
here is example output from running a venet device as "eth1" (see
vectors 28-31):
vbus-guest:/home/ghaskins # cat /proc/interrupts
CPU0
0: 89 IO-APIC-edge timer
1: 6 IO-APIC-edge i8042
4: 1620 IO-APIC-edge serial
6: 2 IO-APIC-edge floppy
7: 0 IO-APIC-edge parport0
8: 0 IO-APIC-edge rtc0
9: 0 IO-APIC-fasteoi acpi
10: 0 IO-APIC-fasteoi virtio1
12: 90 IO-APIC-edge i8042
14: 3474 IO-APIC-edge ata_piix
15: 7778 IO-APIC-edge ata_piix
24: 185849 PCI-MSI-edge vbus
25: 0 PCI-MSI-edge virtio0-config
26: 210 PCI-MSI-edge virtio0-input
27: 28 PCI-MSI-edge virtio0-output
28: 167062 VBUS-edge eth1-evq
29: 0 VBUS-edge eth1-pageq
30: 161593 VBUS-edge eth1-rx
31: 0 VBUS-edge eth1-tx
NMI: 0 Non-maskable interrupts
LOC: 72010 Local timer interrupts
SPU: 0 Spurious interrupts
CNT: 0 Performance counter interrupts
PND: 0 Performance pending work
RES: 0 Rescheduling interrupts
CAL: 0 Function call interrupts
TLB: 0 TLB shootdowns
TRM: 0 Thermal event interrupts
THR: 0 Threshold APIC interrupts
MCE: 0 Machine check exceptions
MCP: 20 Machine check polls
ERR: 0
MIS: 0
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
This will allow the signals to be displayed to the end-user in some
meaningful way later in the series, such as for statistics, etc.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
The functions vbus_pci_open, _ioq_init, eventq_init,
vbus_pci_eventq_register are
annoted with __init and referenced from a function __devinit
vbus_pci_probe, which
causes a section mismatch.So annote those functions with __devinit.
We were warned by the following warnings:
LD drivers/vbus/vbus-pcibridge.o
WARNING: drivers/vbus/vbus-pcibridge.o(.devinit.text+0x19e): Section
mismatch in reference from the function vbus_pci_probe() to the
function .init.text:vbus_pci_open()
The function __devinit vbus_pci_probe() references
a function __init vbus_pci_open().
If vbus_pci_open is only used by vbus_pci_probe then
annotate vbus_pci_open with a matching annotation.
WARNING: drivers/vbus/vbus-pcibridge.o(.devinit.text+0x1f4): Section
mismatch in reference from the function vbus_pci_probe() to the
function .init.text:_ioq_init()
The function __devinit vbus_pci_probe() references
a function __init _ioq_init().
If _ioq_init is only used by vbus_pci_probe then
annotate _ioq_init with a matching annotation.
WARNING: drivers/vbus/vbus-pcibridge.o(.devinit.text+0x240): Section
mismatch in reference from the function vbus_pci_probe() to the
function .init.text:eventq_init()
The function __devinit vbus_pci_probe() references
a function __init eventq_init().
If eventq_init is only used by vbus_pci_probe then
annotate eventq_init with a matching annotation.
WARNING: drivers/vbus/vbus-pcibridge.o(.devinit.text+0x45a): Section
mismatch in reference from the function vbus_pci_probe() to the
function .init.text:vbus_pci_eventq_register()
The function __devinit vbus_pci_probe() references
a function __init vbus_pci_eventq_register().
If vbus_pci_eventq_register is only used by vbus_pci_probe then
annotate vbus_pci_eventq_register with a matching annotation.
Signed-off-by: Rakib Mullick <rakib.mullick@gmail.com>
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
When CONFIG_VBUS_PROXY=[ym] and VBUS_PCIBRIDGE=n:
ERROR: "ioq_init" [drivers/vbus/vbus-proxy.ko] undefined!
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
It will be a common pattern to map an IOQ over the VBUS shared-memory
interfaces. Therefore, we provide a helper function to generalize
the allocation and registration of an IOQ to make this use case
simple and easy.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
This patch adds a pci-based bridge driver to interface between the a
host VBUS and the guest's vbus-proxy bus model. It completes the guest
side notion of a "vbus-connector", and requires a cooresponding
host-side connector (in this case, the pci-bridge model) to comlete
the connection.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
|
This will generally be used for hypervisors to publish any host-side
virtual devices up to a guest. The guest will have the opportunity
to consume any devices present on the vbus-proxy as if they were
platform devices, similar to existing buses like PCI.
Signed-off-by: Gregory Haskins <ghaskins@novell.com>
|
