public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] ipmi: Make sure drivers were registered before unregistering them
@ 2010-06-10 17:53 Matthew Garrett
  2010-06-10 18:11 ` Corey Minyard
  2010-06-14 20:36 ` Andrew Morton
  0 siblings, 2 replies; 4+ messages in thread
From: Matthew Garrett @ 2010-06-10 17:53 UTC (permalink / raw)
  To: minyard; +Cc: linux-kernel, Matthew Garrett

The ipmi code will never register a PCI or Open Firmware driver if
a hardcoded device is provided by the user. This can cause us to attempt
to unregister a driver that was never registered, resulting in an oops.
Keep track of registration in order to avoid this.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
---
 drivers/char/ipmi/ipmi_si_intf.c |   21 +++++++++++++++++----
 1 files changed, 17 insertions(+), 4 deletions(-)

diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
index 35603dd..311f85b 100644
--- a/drivers/char/ipmi/ipmi_si_intf.c
+++ b/drivers/char/ipmi/ipmi_si_intf.c
@@ -302,6 +302,12 @@ struct smi_info {
 
 static int force_kipmid[SI_MAX_PARMS];
 static int num_force_kipmid;
+#ifdef CONFIG_PCI
+static int pci_registered;
+#endif
+#ifdef CONFIG_PPC_OF
+static int of_registered;
+#endif
 
 static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
 static int num_max_busy_us;
@@ -3314,6 +3320,8 @@ static __devinit int init_ipmi_si(void)
 	rv = pci_register_driver(&ipmi_pci_driver);
 	if (rv)
 		printk(KERN_ERR PFX "Unable to register PCI driver: %d\n", rv);
+	else
+		pci_registered = 1;
 #endif
 
 #ifdef CONFIG_ACPI
@@ -3330,6 +3338,7 @@ static __devinit int init_ipmi_si(void)
 
 #ifdef CONFIG_PPC_OF
 	of_register_platform_driver(&ipmi_of_platform_driver);
+	of_registered = 1;
 #endif
 
 	/* We prefer devices with interrupts, but in the case of a machine
@@ -3383,11 +3392,13 @@ static __devinit int init_ipmi_si(void)
 	if (unload_when_empty && list_empty(&smi_infos)) {
 		mutex_unlock(&smi_infos_lock);
 #ifdef CONFIG_PCI
-		pci_unregister_driver(&ipmi_pci_driver);
+		if (pci_registered)
+			pci_unregister_driver(&ipmi_pci_driver);
 #endif
 
 #ifdef CONFIG_PPC_OF
-		of_unregister_platform_driver(&ipmi_of_platform_driver);
+		if (of_registered)
+			of_unregister_platform_driver(&ipmi_of_platform_driver);
 #endif
 		driver_unregister(&ipmi_driver.driver);
 		printk(KERN_WARNING PFX
@@ -3478,14 +3489,16 @@ static __exit void cleanup_ipmi_si(void)
 		return;
 
 #ifdef CONFIG_PCI
-	pci_unregister_driver(&ipmi_pci_driver);
+	if (pci_registered)
+		pci_unregister_driver(&ipmi_pci_driver);
 #endif
 #ifdef CONFIG_ACPI
 	pnp_unregister_driver(&ipmi_pnp_driver);
 #endif
 
 #ifdef CONFIG_PPC_OF
-	of_unregister_platform_driver(&ipmi_of_platform_driver);
+	if (of_registered)
+		of_unregister_platform_driver(&ipmi_of_platform_driver);
 #endif
 
 	mutex_lock(&smi_infos_lock);
-- 
1.7.0.1


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] ipmi: Make sure drivers were registered before unregistering them
  2010-06-10 17:53 [PATCH] ipmi: Make sure drivers were registered before unregistering them Matthew Garrett
@ 2010-06-10 18:11 ` Corey Minyard
  2010-06-14 20:36 ` Andrew Morton
  1 sibling, 0 replies; 4+ messages in thread
From: Corey Minyard @ 2010-06-10 18:11 UTC (permalink / raw)
  To: Matthew Garrett; +Cc: linux-kernel, Andrew Morton

Yes, thanks for this.

Acked-by: Corey Minyard <cminyard@mvista.com>

Matthew Garrett wrote:
> The ipmi code will never register a PCI or Open Firmware driver if
> a hardcoded device is provided by the user. This can cause us to attempt
> to unregister a driver that was never registered, resulting in an oops.
> Keep track of registration in order to avoid this.
>
> Signed-off-by: Matthew Garrett <mjg@redhat.com>
> ---
>  drivers/char/ipmi/ipmi_si_intf.c |   21 +++++++++++++++++----
>  1 files changed, 17 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/char/ipmi/ipmi_si_intf.c b/drivers/char/ipmi/ipmi_si_intf.c
> index 35603dd..311f85b 100644
> --- a/drivers/char/ipmi/ipmi_si_intf.c
> +++ b/drivers/char/ipmi/ipmi_si_intf.c
> @@ -302,6 +302,12 @@ struct smi_info {
>  
>  static int force_kipmid[SI_MAX_PARMS];
>  static int num_force_kipmid;
> +#ifdef CONFIG_PCI
> +static int pci_registered;
> +#endif
> +#ifdef CONFIG_PPC_OF
> +static int of_registered;
> +#endif
>  
>  static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
>  static int num_max_busy_us;
> @@ -3314,6 +3320,8 @@ static __devinit int init_ipmi_si(void)
>  	rv = pci_register_driver(&ipmi_pci_driver);
>  	if (rv)
>  		printk(KERN_ERR PFX "Unable to register PCI driver: %d\n", rv);
> +	else
> +		pci_registered = 1;
>  #endif
>  
>  #ifdef CONFIG_ACPI
> @@ -3330,6 +3338,7 @@ static __devinit int init_ipmi_si(void)
>  
>  #ifdef CONFIG_PPC_OF
>  	of_register_platform_driver(&ipmi_of_platform_driver);
> +	of_registered = 1;
>  #endif
>  
>  	/* We prefer devices with interrupts, but in the case of a machine
> @@ -3383,11 +3392,13 @@ static __devinit int init_ipmi_si(void)
>  	if (unload_when_empty && list_empty(&smi_infos)) {
>  		mutex_unlock(&smi_infos_lock);
>  #ifdef CONFIG_PCI
> -		pci_unregister_driver(&ipmi_pci_driver);
> +		if (pci_registered)
> +			pci_unregister_driver(&ipmi_pci_driver);
>  #endif
>  
>  #ifdef CONFIG_PPC_OF
> -		of_unregister_platform_driver(&ipmi_of_platform_driver);
> +		if (of_registered)
> +			of_unregister_platform_driver(&ipmi_of_platform_driver);
>  #endif
>  		driver_unregister(&ipmi_driver.driver);
>  		printk(KERN_WARNING PFX
> @@ -3478,14 +3489,16 @@ static __exit void cleanup_ipmi_si(void)
>  		return;
>  
>  #ifdef CONFIG_PCI
> -	pci_unregister_driver(&ipmi_pci_driver);
> +	if (pci_registered)
> +		pci_unregister_driver(&ipmi_pci_driver);
>  #endif
>  #ifdef CONFIG_ACPI
>  	pnp_unregister_driver(&ipmi_pnp_driver);
>  #endif
>  
>  #ifdef CONFIG_PPC_OF
> -	of_unregister_platform_driver(&ipmi_of_platform_driver);
> +	if (of_registered)
> +		of_unregister_platform_driver(&ipmi_of_platform_driver);
>  #endif
>  
>  	mutex_lock(&smi_infos_lock);
>   


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] ipmi: Make sure drivers were registered before unregistering them
  2010-06-10 17:53 [PATCH] ipmi: Make sure drivers were registered before unregistering them Matthew Garrett
  2010-06-10 18:11 ` Corey Minyard
@ 2010-06-14 20:36 ` Andrew Morton
  2010-06-14 20:47   ` Matthew Garrett
  1 sibling, 1 reply; 4+ messages in thread
From: Andrew Morton @ 2010-06-14 20:36 UTC (permalink / raw)
  To: Matthew Garrett; +Cc: minyard, linux-kernel

On Thu, 10 Jun 2010 13:53:26 -0400
Matthew Garrett <mjg@redhat.com> wrote:

> The ipmi code will never register a PCI or Open Firmware driver if
> a hardcoded device is provided by the user.

How does a user "provide a hardcoded device"?

> This can cause us to attempt
> to unregister a driver that was never registered, resulting in an oops.
> Keep track of registration in order to avoid this.
> 

I tried to work out from the above whether we want to backport this fix
into -stable and failed.  And I reckon that if I can't work this out
from a changelog, the changelog is inadequate.

> --- a/drivers/char/ipmi/ipmi_si_intf.c
> +++ b/drivers/char/ipmi/ipmi_si_intf.c
> @@ -302,6 +302,12 @@ struct smi_info {
>  
>  static int force_kipmid[SI_MAX_PARMS];
>  static int num_force_kipmid;
> +#ifdef CONFIG_PCI
> +static int pci_registered;
> +#endif
> +#ifdef CONFIG_PPC_OF
> +static int of_registered;
> +#endif
>  
>  static unsigned int kipmid_max_busy_us[SI_MAX_PARMS];
>  static int num_max_busy_us;
> @@ -3314,6 +3320,8 @@ static __devinit int init_ipmi_si(void)
>  	rv = pci_register_driver(&ipmi_pci_driver);
>  	if (rv)
>  		printk(KERN_ERR PFX "Unable to register PCI driver: %d\n", rv);
> +	else
> +		pci_registered = 1;
>  #endif
>  
>  #ifdef CONFIG_ACPI
> @@ -3330,6 +3338,7 @@ static __devinit int init_ipmi_si(void)
>  
>  #ifdef CONFIG_PPC_OF
>  	of_register_platform_driver(&ipmi_of_platform_driver);
> +	of_registered = 1;

I assume the code will still oops if of_register_platform_driver() failed.

>  #endif
>  
>  	/* We prefer devices with interrupts, but in the case of a machine
> @@ -3383,11 +3392,13 @@ static __devinit int init_ipmi_si(void)
>  	if (unload_when_empty && list_empty(&smi_infos)) {
>  		mutex_unlock(&smi_infos_lock);
>  #ifdef CONFIG_PCI
> -		pci_unregister_driver(&ipmi_pci_driver);
> +		if (pci_registered)
> +			pci_unregister_driver(&ipmi_pci_driver);
>  #endif
>  
>  #ifdef CONFIG_PPC_OF
> -		of_unregister_platform_driver(&ipmi_of_platform_driver);
> +		if (of_registered)
> +			of_unregister_platform_driver(&ipmi_of_platform_driver);
>  #endif
>  		driver_unregister(&ipmi_driver.driver);
>  		printk(KERN_WARNING PFX
> @@ -3478,14 +3489,16 @@ static __exit void cleanup_ipmi_si(void)
>  		return;
>  
>  #ifdef CONFIG_PCI
> -	pci_unregister_driver(&ipmi_pci_driver);
> +	if (pci_registered)
> +		pci_unregister_driver(&ipmi_pci_driver);
>  #endif
>  #ifdef CONFIG_ACPI
>  	pnp_unregister_driver(&ipmi_pnp_driver);
>  #endif
>  
>  #ifdef CONFIG_PPC_OF
> -	of_unregister_platform_driver(&ipmi_of_platform_driver);
> +	if (of_registered)
> +		of_unregister_platform_driver(&ipmi_of_platform_driver);
>  #endif
>  
>  	mutex_lock(&smi_infos_lock);


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] ipmi: Make sure drivers were registered before unregistering them
  2010-06-14 20:36 ` Andrew Morton
@ 2010-06-14 20:47   ` Matthew Garrett
  0 siblings, 0 replies; 4+ messages in thread
From: Matthew Garrett @ 2010-06-14 20:47 UTC (permalink / raw)
  To: Andrew Morton; +Cc: minyard, linux-kernel

On Mon, Jun 14, 2010 at 01:36:08PM -0700, Andrew Morton wrote:
> On Thu, 10 Jun 2010 13:53:26 -0400
> Matthew Garrett <mjg@redhat.com> wrote:
> 
> > The ipmi code will never register a PCI or Open Firmware driver if
> > a hardcoded device is provided by the user.
> 
> How does a user "provide a hardcoded device"?

By providing device addresses via the module parameters. This is the 
terminology used in the driver.

> I tried to work out from the above whether we want to backport this fix
> into -stable and failed.  And I reckon that if I can't work this out
> from a changelog, the changelog is inadequate.

We don't. The change that caused this bug was merged post-.34.

> >  #ifdef CONFIG_ACPI
> > @@ -3330,6 +3338,7 @@ static __devinit int init_ipmi_si(void)
> >  
> >  #ifdef CONFIG_PPC_OF
> >  	of_register_platform_driver(&ipmi_of_platform_driver);
> > +	of_registered = 1;
> 
> I assume the code will still oops if of_register_platform_driver() failed.

Indeed. I should fix that.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2010-06-14 20:47 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-06-10 17:53 [PATCH] ipmi: Make sure drivers were registered before unregistering them Matthew Garrett
2010-06-10 18:11 ` Corey Minyard
2010-06-14 20:36 ` Andrew Morton
2010-06-14 20:47   ` Matthew Garrett

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox