On 01/26/2015 06:59 PM, Bjorn Helgaas wrote:
On Mon, Jan 26, 2015 at 5:25 PM, Murali Karicheri<m-karicheri2@xxxxxx> wrote:
On 01/23/2015 06:41 PM, Bjorn Helgaas wrote:
On Fri, Jan 23, 2015 at 05:32:37PM -0500, Murali Karicheri wrote:
Add of_pci_dma_configure() to allow updating the dma configuration
of the pci device using the configuration from DT of the parent of
the root bridge device.
-- Cut ---
Signed-off-by: Murali Karicheri<m-karicheri2@xxxxxx>
---
drivers/of/of_pci.c | 39 +++++++++++++++++++++++++++++++++++++++
include/linux/of_pci.h | 12 ++++++++++++
2 files changed, 51 insertions(+)
diff --git a/drivers/of/of_pci.c b/drivers/of/of_pci.c
index 88471d3..34878c9 100644
--- a/drivers/of/of_pci.c
+++ b/drivers/of/of_pci.c
@@ -2,6 +2,7 @@
#include<linux/export.h>
#include<linux/of.h>
#include<linux/of_address.h>
+#include<linux/of_device.h>
#include<linux/of_pci.h>
#include<linux/slab.h>
@@ -229,6 +230,44 @@ parse_failed:
return err;
}
EXPORT_SYMBOL_GPL(of_pci_get_host_bridge_resources);
+
+/**
+ * of_get_pci_root_bridge_parent - get the OF node of the root bridge's
parent
+ * @dev: ptr to pci_dev struct of the pci device
+ *
+ * This function will traverse the bus up to the root bus starting with
+ * the child and return the OF node ptr to root bridge device's parent
device.
+ */
+struct device_node *of_get_pci_root_bridge_parent(struct pci_dev *dev)
I'm not an OF person, but this interface seems like it might be too
special-purpose. Maybe it would be enough to add
"of_get_pci_root_bridge()", and the caller could do this:
struct device *bridge = of_get_pci_root_bridge(dev);
struct device_node *parent_np = bridge->parent->of_node;
Also, the name "of_get_..." suggests that it increments a refcount, as
of_get_parent() does. But you aren't doing anything with the refcount.
But I guess an "of_get_pci_root_bridge()" isn't doing anything OF-related,
so maybe we should just add a "pci_get_host_bridge(struct pci_dev *)"
to PCI instead.
Bjorn,
Thanks for the comment.
I think adding pci_get_host_bridge() is a good idea. There is already
similar function in host-bridge.c. I have added this function re-using
existing function find_pci_root_bus(). See the incremental diff below after
this change. Does this look good?
I like the implementation, but I think either we need to take a
reference on the host bridge, or change the name to something like
"pci_find_host_bridge()", because using "_get_" is conventional for
functions that acquire a reference.
Since host bridges are hot-pluggable, at least in theory, I vote for
taking a reference. Then of course, you'd have to add code to drop
the reference when you're finished with it.
Bjorn,
Thanks. I agree with your suggestion even though the convention is not
followed fully :) of_pci_get_devfn(), of_get_pci_domain_nr(),
of_pci_get_host_bridge_resources() are some of those functions not
following the convention. I plan to change the function as below. Also
want to name functions as pci_get/put_host_bridge_device() as existing
function find_pci_host_bridge() is actually returning ptr to struct
pci_host_bridge vs the new function returning ptr to device. Here are
the new functions and how they will be used. Please review and respond
so that I can avoid a re-spin.
in linux/include/pci.h add the prototypes of
pci_get/put_host_bridge_device().
in drivers/pci/host-bridge.c add two new functions.
struct device *pci_get_host_bridge_device(struct pci_dev *dev)
{
struct pci_bus *root_bus = find_pci_root_bus(dev->bus);
struct device *bridge = root_bus->bridge;
kobject_get(&bridge->kobj);
return bridge;
}
void pci_put_host_bridge_device(struct pci_dev *dev)
{
struct pci_bus *root_bus = find_pci_root_bus(dev->bus);
struct device *bridge = root_bus->bridge;
kobject_put(&bridge->kobj);
}
drivers/of/of_pci.c
void of_pci_dma_configure(struct pci_dev *pci_dev)
{
struct device *dev = &pci_dev->dev;
struct device *bridge = pci_get_host_bridge_device(pci_dev);
of_dma_configure(dev, bridge->parent->of_node);
pci_put_host_bridge_device(pci_dev);
}
Murali
Bjorn
--
Murali Karicheri
Linux Kernel, Texas Instruments
--
Murali Karicheri
Linux Kernel, Texas Instruments
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
