From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Return-Path: Subject: Re: [PATCH v1 1/3] powerpc/kernel: Split up pci_bus_add_device To: Bjorn Helgaas Cc: bhelgaas@google.com, benh@kernel.crashing.org, paulus@samba.org, mpe@ellerman.id.au, linux-pci@vger.kernel.org, linuxppc-dev@lists.ozlabs.org, "Juan J . Alvarez" References: <20170918192651.78404-1-bryantly@linux.vnet.ibm.com> <20170918192651.78404-2-bryantly@linux.vnet.ibm.com> <20170921204326.GJ22312@bhelgaas-glaptop.roam.corp.google.com> From: "Bryant G. Ly" Date: Fri, 22 Sep 2017 08:46:26 -0500 MIME-Version: 1.0 In-Reply-To: <20170921204326.GJ22312@bhelgaas-glaptop.roam.corp.google.com> Content-Type: multipart/alternative; boundary="------------0B9EC22EE5D642BE41404790" Message-Id: <1a13b02a-4e32-adff-1083-8bf538cbdf9b@linux.vnet.ibm.com> List-ID: This is a multi-part message in MIME format. --------------0B9EC22EE5D642BE41404790 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit On 9/21/17 3:43 PM, Bjorn Helgaas wrote: > On Mon, Sep 18, 2017 at 02:26:49PM -0500, Bryant G. Ly wrote: >> When enabling SR-IOV one might want to have their >> own version of starting device drivers for the VFs. >> This patch allows for SR-IOV callers to use >> pci_bus_add_virtfn_device instead of generic >> pci_bus_add_device. >> >> When enabling SR-IOV in PSeries architecture the >> dynamic VFs created within the sriov_configure sysfs call >> will not load the device driver as firmware will load >> the device node when the VF device is assigned to the >> logical partition. So we needed a way to overwrite the >> way device driver matching is done for virtual functions >> on powervm. >> >> Signed-off-by: Bryant G. Ly >> Signed-off-by: Juan J. Alvarez >> --- >> drivers/pci/bus.c | 51 ++++++++++++++++++++++++++++++++++++++++++--------- >> drivers/pci/iov.c | 2 +- >> include/linux/pci.h | 3 +++ >> 3 files changed, 46 insertions(+), 10 deletions(-) >> >> diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c >> index bc56cf19afd3..86daf62c4048 100644 >> --- a/drivers/pci/bus.c >> +++ b/drivers/pci/bus.c >> @@ -302,16 +302,9 @@ void __weak pcibios_resource_survey_bus(struct pci_bus *bus) { } >> >> void __weak pcibios_bus_add_device(struct pci_dev *pdev) { } >> >> -/** >> - * pci_bus_add_device - start driver for a single device >> - * @dev: device to add >> - * >> - * This adds add sysfs entries and start device drivers >> - */ >> -void pci_bus_add_device(struct pci_dev *dev) >> -{ >> - int retval; >> >> +void pci_bus_add_sysfs_entries(struct pci_dev *dev) >> +{ >> /* >> * Can not put in pci_device_add yet because resources >> * are not assigned yet for some devices. >> @@ -321,6 +314,11 @@ void pci_bus_add_device(struct pci_dev *dev) >> pci_create_sysfs_dev_files(dev); >> pci_proc_attach_device(dev); >> pci_bridge_d3_update(dev); >> +} >> + >> +void pci_bus_match_device_driver(struct pci_dev *dev) >> +{ >> + int retval; >> >> dev->match_driver = true; >> retval = device_attach(&dev->dev); >> @@ -333,6 +331,41 @@ void pci_bus_add_device(struct pci_dev *dev) >> >> dev->is_added = 1; >> } >> + >> +#ifdef CONFIG_PCI_IOV >> +void __weak pci_bus_match_virtfn_driver(struct pci_dev *dev) >> +{ >> + pci_bus_match_device_driver(dev); >> +} >> + >> +/** >> + * pci_bus_add_virtfn_device - start driver for a virtual function device >> + * @dev: device to add >> + * >> + * This adds add sysfs entries and start device drivers for >> + * virtual function devices >> + * >> + */ >> +void pci_bus_add_virtfn_device(struct pci_dev *pdev) >> +{ >> + pci_bus_add_sysfs_entries(pdev); >> + pci_bus_match_virtfn_driver(pdev); >> +} >> +EXPORT_SYMBOL_GPL(pci_bus_add_virtfn_device); > Is there any way we can avoid adding this new interface? I don't > really want a new global exported symbol just to support this highly > platform-specific functionality. > > If we could figure out a way to set dev->match_driver to false instead > of true in pci_bus_add_device(), it *looks* like that would be enough > to accomplish what you need. Maybe that could be done via a pcibios > hook or a new function pointer in struct pci_host_bridge, similar to > the swizzle_irq pointer? Yes! I will remove the first two patches in the series and only utilize the third patch but addingpdev->match_driver = -1; under pseries_pcibios_bus_add_device, which is a new function I added for machine dependent bus add of a device. The patch will then require https://patchwork.kernel.org/patch/9882915/ which basically allows for match_driver state to not get changed later under the pci_bus_match due to allowing for the match_driver to be a int vs a bool. I will push up another patch removing the first two patches and adding that extra line. -Bryant >> +#endif >> + >> +/** >> + * pci_bus_add_device - start driver for a single device >> + * @dev: device to add >> + * >> + * This adds add sysfs entries and start device drivers >> + */ >> +void pci_bus_add_device(struct pci_dev *dev) >> +{ >> + pci_bus_add_sysfs_entries(dev); >> + pci_bus_match_device_driver(dev); >> +} >> + >> EXPORT_SYMBOL_GPL(pci_bus_add_device); >> >> /** >> diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c >> index ac41c8be9200..16cc72545847 100644 >> --- a/drivers/pci/iov.c >> +++ b/drivers/pci/iov.c >> @@ -162,7 +162,7 @@ int pci_iov_add_virtfn(struct pci_dev *dev, int id, int reset) >> >> pci_device_add(virtfn, virtfn->bus); >> >> - pci_bus_add_device(virtfn); >> + pci_bus_add_virtfn_device(virtfn); >> sprintf(buf, "virtfn%u", id); >> rc = sysfs_create_link(&dev->dev.kobj, &virtfn->dev.kobj, buf); >> if (rc) >> diff --git a/include/linux/pci.h b/include/linux/pci.h >> index f68c58a93dd0..39f5c0b4bf23 100644 >> --- a/include/linux/pci.h >> +++ b/include/linux/pci.h >> @@ -911,6 +911,9 @@ struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn); >> void pci_device_add(struct pci_dev *dev, struct pci_bus *bus); >> unsigned int pci_scan_child_bus(struct pci_bus *bus); >> void pci_bus_add_device(struct pci_dev *dev); >> +#ifdef CONFIG_PCI_IOV >> +void pci_bus_add_virtfn_device(struct pci_dev *dev); >> +#endif >> void pci_read_bridge_bases(struct pci_bus *child); >> struct resource *pci_find_parent_resource(const struct pci_dev *dev, >> struct resource *res); >> -- >> 2.11.0 (Apple Git-81) >> --------------0B9EC22EE5D642BE41404790 Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: 7bit



On 9/21/17 3:43 PM, Bjorn Helgaas wrote:
On Mon, Sep 18, 2017 at 02:26:49PM -0500, Bryant G. Ly wrote:

When enabling SR-IOV one might want to have their
own version of starting device drivers for the VFs.
This patch allows for SR-IOV callers to use
pci_bus_add_virtfn_device instead of generic
pci_bus_add_device.

When enabling SR-IOV in PSeries architecture the
dynamic VFs created within the sriov_configure sysfs call
will not load the device driver as firmware will load
the device node when the VF device is assigned to the
logical partition. So we needed a way to overwrite the
way device driver matching is done for virtual functions
on powervm.

Signed-off-by: Bryant G. Ly <bryantly@linux.vnet.ibm.com>
Signed-off-by: Juan J. Alvarez <jjalvare@us.ibm.com>
---
 drivers/pci/bus.c   | 51 ++++++++++++++++++++++++++++++++++++++++++---------
 drivers/pci/iov.c   |  2 +-
 include/linux/pci.h |  3 +++
 3 files changed, 46 insertions(+), 10 deletions(-)

diff --git a/drivers/pci/bus.c b/drivers/pci/bus.c
index bc56cf19afd3..86daf62c4048 100644
--- a/drivers/pci/bus.c
+++ b/drivers/pci/bus.c
@@ -302,16 +302,9 @@ void __weak pcibios_resource_survey_bus(struct pci_bus *bus) { }
 
 void __weak pcibios_bus_add_device(struct pci_dev *pdev) { }
 
-/**
- * pci_bus_add_device - start driver for a single device
- * @dev: device to add
- *
- * This adds add sysfs entries and start device drivers
- */
-void pci_bus_add_device(struct pci_dev *dev)
-{
-	int retval;
 
+void pci_bus_add_sysfs_entries(struct pci_dev *dev)
+{
 	/*
 	 * Can not put in pci_device_add yet because resources
 	 * are not assigned yet for some devices.
@@ -321,6 +314,11 @@ void pci_bus_add_device(struct pci_dev *dev)
 	pci_create_sysfs_dev_files(dev);
 	pci_proc_attach_device(dev);
 	pci_bridge_d3_update(dev);
+}
+
+void pci_bus_match_device_driver(struct pci_dev *dev)
+{
+	int retval;
 
 	dev->match_driver = true;
 	retval = device_attach(&dev->dev);
@@ -333,6 +331,41 @@ void pci_bus_add_device(struct pci_dev *dev)
 
 	dev->is_added = 1;
 }
+
+#ifdef CONFIG_PCI_IOV
+void __weak pci_bus_match_virtfn_driver(struct pci_dev *dev)
+{
+	pci_bus_match_device_driver(dev);
+}
+
+/**
+ * pci_bus_add_virtfn_device - start driver for a virtual function device
+ * @dev: device to add
+ *
+ * This adds add sysfs entries and start device drivers for
+ * virtual function devices
+ *
+ */
+void pci_bus_add_virtfn_device(struct pci_dev *pdev)
+{
+	pci_bus_add_sysfs_entries(pdev);
+	pci_bus_match_virtfn_driver(pdev);
+}
+EXPORT_SYMBOL_GPL(pci_bus_add_virtfn_device);

Is there any way we can avoid adding this new interface?  I don't
really want a new global exported symbol just to support this highly
platform-specific functionality.

If we could figure out a way to set dev->match_driver to false instead
of true in pci_bus_add_device(), it *looks* like that would be enough
to accomplish what you need.  Maybe that could be done via a pcibios
hook or a new function pointer in struct pci_host_bridge, similar to
the swizzle_irq pointer?

Yes! I will remove the first two patches in the series and only utilize the third
patch but adding pdev->match_driver = -1; under pseries_pcibios_bus_add_device, which
is a new function I added for machine dependent bus add of a device. 

The patch will then require https://patchwork.kernel.org/patch/9882915/
which basically allows for match_driver state to not get changed later under the
pci_bus_match due to allowing for the match_driver to be a int vs a bool. 

I will push up another patch removing the first two patches and adding that extra line.

-Bryant

      

        
+#endif
+
+/**
+ * pci_bus_add_device - start driver for a single device
+ * @dev: device to add
+ *
+ * This adds add sysfs entries and start device drivers
+ */
+void pci_bus_add_device(struct pci_dev *dev)
+{
+	pci_bus_add_sysfs_entries(dev);
+	pci_bus_match_device_driver(dev);
+}
+
 EXPORT_SYMBOL_GPL(pci_bus_add_device);
 
 /**
diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index ac41c8be9200..16cc72545847 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -162,7 +162,7 @@ int pci_iov_add_virtfn(struct pci_dev *dev, int id, int reset)
 
 	pci_device_add(virtfn, virtfn->bus);
 
-	pci_bus_add_device(virtfn);
+	pci_bus_add_virtfn_device(virtfn);
 	sprintf(buf, "virtfn%u", id);
 	rc = sysfs_create_link(&dev->dev.kobj, &virtfn->dev.kobj, buf);
 	if (rc)
diff --git a/include/linux/pci.h b/include/linux/pci.h
index f68c58a93dd0..39f5c0b4bf23 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -911,6 +911,9 @@ struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn);
 void pci_device_add(struct pci_dev *dev, struct pci_bus *bus);
 unsigned int pci_scan_child_bus(struct pci_bus *bus);
 void pci_bus_add_device(struct pci_dev *dev);
+#ifdef CONFIG_PCI_IOV
+void pci_bus_add_virtfn_device(struct pci_dev *dev);
+#endif
 void pci_read_bridge_bases(struct pci_bus *child);
 struct resource *pci_find_parent_resource(const struct pci_dev *dev,
 					  struct resource *res);
-- 
2.11.0 (Apple Git-81)



      

      

    

    

  


--------------0B9EC22EE5D642BE41404790--