Re: [PATCH v4 4/8] PCI: replace pci_dev::driver usage that gets the driver name

Hello,

On Tue, Sep 28, 2021 at 12:17:59PM -0500, Bjorn Helgaas wrote:

[+to Oliver, Russell for eeh_driver_name() question below]

On Mon, Sep 27, 2021 at 10:43:22PM +0200, Uwe Kleine-König wrote:

quoted

From: Uwe Kleine-König <redacted>

struct pci_dev::driver holds (apart from a constant offset) the same
data as struct pci_dev::dev->driver. With the goal to remove struct
pci_dev::driver to get rid of data duplication replace getting the
driver name by dev_driver_string() which implicitly makes use of struct
pci_dev::dev->driver.

When you repost to fix the build issue, can you capitalize the subject
line to match the other?

Yes, sure.

Also, would you mind using "pci_dev.driver" instead of
"pci_dev::driver"?  AFAIK, the "::" operator is not actually part of
C, so I think it's more confusing than useful.

pci_dev.driver doesn't work either in C because pci_dev is a type and
not a variable. This is probably subjective, but for me pci_dev.driver
looks definitively stranger than pci_dev::driver. And :: is at least not
unseen in the kernel commit logs. (git log --grep=::)
But if you insist I can change to .

quoted

diff --git a/arch/powerpc/include/asm/ppc-pci.h b/arch/powerpc/include/asm/ppc-pci.h
index 2b9edbf6e929..e8f1795a2acf 100644
--- a/arch/powerpc/include/asm/ppc-pci.h
+++ b/arch/powerpc/include/asm/ppc-pci.h

@@ -57,7 +57,14 @@ void eeh_sysfs_remove_device(struct pci_dev *pdev);
 
 static inline const char *eeh_driver_name(struct pci_dev *pdev)
 {
-	return (pdev && pdev->driver) ? pdev->driver->name : "<null>";
+	if (pdev) {
+		const char *drvstr = dev_driver_string(&pdev->dev);
+
+		if (strcmp(drvstr, ""))
+			return drvstr;
+	}
+
+	return "<null>";

Can we just do this?

  if (pdev)
    return dev_driver_string(&pdev->dev);

  return "<null>";

Works for me, too. It behaves a bit differerently than my suggestion
(which nearly behaves identical to the status quo), but only in some
degenerated cases.

I think it's more complicated than it's worth to include a strcmp().
It's possible this will change those error messages about "Might be
infinite loop in %s driver", but that doesn't seem like a huge deal.

I moved Oliver to "to:" and added Russell in case they object.

quoted

 }
 
 #endif /* CONFIG_EEH */

diff --git a/drivers/bcma/host_pci.c b/drivers/bcma/host_pci.c
index 69c10a7b7c61..0973022d4b13 100644
--- a/drivers/bcma/host_pci.c
+++ b/drivers/bcma/host_pci.c

@@ -175,9 +175,10 @@ static int bcma_host_pci_probe(struct pci_dev *dev,
 	if (err)
 		goto err_kfree_bus;
 
-	name = dev_name(&dev->dev);
-	if (dev->driver && dev->driver->name)
-		name = dev->driver->name;
+	name = dev_driver_string(&dev->dev);
+	if (!strcmp(name, ""))
+		name = dev_name(&dev->dev);
 	err = pci_request_regions(dev, name);

Again seems more complicated than it's worth to me.  This is in the
driver's .probe() method, so really_probe() has already set
"dev->driver = drv", which means dev->driver is always set to
&bcma_pci_bridge_driver here, and bcma_pci_bridge_driver.name is
always "bcma-pci-bridge".

Almost all callers of pci_request_regions() just hardcode the driver
name or use a DRV_NAME #define

So I think we should just do:

  err = pci_request_regions(dev, "bcma-pci-bridge");

Yes, looks right. I'd put this in a separate patch.

quoted

 	if (err)
 		goto err_pci_disable;
[...]

diff --git a/drivers/ssb/pcihost_wrapper.c b/drivers/ssb/pcihost_wrapper.c
index 410215c16920..4938ed5cfae5 100644
--- a/drivers/ssb/pcihost_wrapper.c
+++ b/drivers/ssb/pcihost_wrapper.c

@@ -78,9 +78,11 @@ static int ssb_pcihost_probe(struct pci_dev *dev,
 	err = pci_enable_device(dev);
 	if (err)
 		goto err_kfree_ssb;
-	name = dev_name(&dev->dev);
-	if (dev->driver && dev->driver->name)
-		name = dev->driver->name;
+
+	name = dev_driver_string(&dev->dev);
+	if (*name == '\0')
+		name = dev_name(&dev->dev);
+
 	err = pci_request_regions(dev, name);

Also seems like more trouble than it's worth.  This one is a little
strange but is always called for either b43_pci_bridge_driver or
b44_pci_driver, both of which have .name set, so I think we should
simply do:

  err = pci_request_regions(dev, dev_driver_string(&dev->dev));

yes, agreed, too.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |