linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] PCI: warn on driver probe return value greater than zero
@ 2013-11-01 19:34 Stephen M. Cameron
  2013-11-06 23:30 ` Bjorn Helgaas
  0 siblings, 1 reply; 2+ messages in thread
From: Stephen M. Cameron @ 2013-11-01 19:34 UTC (permalink / raw)
  To: bhelgaas; +Cc: axboe, linux-pci, linux-kernel, stephenmcameron, thenzl, akpm

From: Stephen M. Cameron <scameron@beardog.cce.hp.com>

Ages ago, drivers could return values greater than zero from
their probe function and this would be regarded as success.
Commit f3ec4f87d607f40497 "PCI: change device runtime PM
settings for probe and remove" slightly altered this in 2010,
and commit 967577b062417b4e4b8e27b "PCI/PM: Keep runtime PM
enabled for unbound PCI devices" in late 2012 altered it more
signficantly, setting pci_dev->driver to NULL if the driver's
probe function returned a value greater than zero, which would
for example prevent the driver's remove function from being
called on rmmod.

Neither of those changes would necessarily make the driver fail
in an obvious way though, and so at least a couple drivers (cciss,
hpsa) fell into this hole since they were returning 1, and this
situation went unnoticed for quite some time.

If a driver's probe function returns a value greater than zero,
issue a warning, but otherwise treat this as success.

Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
---
 drivers/pci/pci-driver.c |   12 ++++++++++--
 1 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
index 98f7b9b..7fbe343 100644
--- a/drivers/pci/pci-driver.c
+++ b/drivers/pci/pci-driver.c
@@ -264,11 +264,19 @@ static long local_pci_probe(void *_ddi)
 	pm_runtime_get_sync(dev);
 	pci_dev->driver = pci_drv;
 	rc = pci_drv->probe(pci_dev, ddi->id);
-	if (rc) {
+	if (!rc)
+		return rc;
+	if (rc < 0) {
 		pci_dev->driver = NULL;
 		pm_runtime_put_sync(dev);
+		return rc;
 	}
-	return rc;
+	/*
+	 * Probe function should return < 0 for failure, 0 for success
+	 * Treat values > 0 as success, but warn.
+	 */
+	dev_warn(dev, "Driver probe function unexpectedly returned %d\n", rc);
+	return 0;
 }
 
 static int pci_call_probe(struct pci_driver *drv, struct pci_dev *dev,


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] PCI: warn on driver probe return value greater than zero
  2013-11-01 19:34 [PATCH] PCI: warn on driver probe return value greater than zero Stephen M. Cameron
@ 2013-11-06 23:30 ` Bjorn Helgaas
  0 siblings, 0 replies; 2+ messages in thread
From: Bjorn Helgaas @ 2013-11-06 23:30 UTC (permalink / raw)
  To: Stephen M. Cameron
  Cc: axboe, linux-pci, linux-kernel, stephenmcameron, thenzl, akpm

On Fri, Nov 01, 2013 at 02:34:55PM -0500, Stephen M. Cameron wrote:
> From: Stephen M. Cameron <scameron@beardog.cce.hp.com>
> 
> Ages ago, drivers could return values greater than zero from
> their probe function and this would be regarded as success.
> Commit f3ec4f87d607f40497 "PCI: change device runtime PM
> settings for probe and remove" slightly altered this in 2010,
> and commit 967577b062417b4e4b8e27b "PCI/PM: Keep runtime PM
> enabled for unbound PCI devices" in late 2012 altered it more
> signficantly, setting pci_dev->driver to NULL if the driver's
> probe function returned a value greater than zero, which would
> for example prevent the driver's remove function from being
> called on rmmod.
> 
> Neither of those changes would necessarily make the driver fail
> in an obvious way though, and so at least a couple drivers (cciss,
> hpsa) fell into this hole since they were returning 1, and this
> situation went unnoticed for quite some time.
> 
> If a driver's probe function returns a value greater than zero,
> issue a warning, but otherwise treat this as success.
> 
> Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>

Applied to my pci/misc branch for v3.13, thanks.

Bjorn

> ---
>  drivers/pci/pci-driver.c |   12 ++++++++++--
>  1 files changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/pci-driver.c b/drivers/pci/pci-driver.c
> index 98f7b9b..7fbe343 100644
> --- a/drivers/pci/pci-driver.c
> +++ b/drivers/pci/pci-driver.c
> @@ -264,11 +264,19 @@ static long local_pci_probe(void *_ddi)
>  	pm_runtime_get_sync(dev);
>  	pci_dev->driver = pci_drv;
>  	rc = pci_drv->probe(pci_dev, ddi->id);
> -	if (rc) {
> +	if (!rc)
> +		return rc;
> +	if (rc < 0) {
>  		pci_dev->driver = NULL;
>  		pm_runtime_put_sync(dev);
> +		return rc;
>  	}
> -	return rc;
> +	/*
> +	 * Probe function should return < 0 for failure, 0 for success
> +	 * Treat values > 0 as success, but warn.
> +	 */
> +	dev_warn(dev, "Driver probe function unexpectedly returned %d\n", rc);
> +	return 0;
>  }
>  
>  static int pci_call_probe(struct pci_driver *drv, struct pci_dev *dev,
> 

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2013-11-06 23:30 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-11-01 19:34 [PATCH] PCI: warn on driver probe return value greater than zero Stephen M. Cameron
2013-11-06 23:30 ` Bjorn Helgaas

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).