[PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[linas]

[-- Attachment #1: Type: text/plain, Size: 495 bytes --]



Greg,

This patch implements the catching of EEH events by the hotplug subsystem.
Its preliminary in that it doesn't do much with these events; right now, 
my goal is to stub-in the required interfaces so that development can occur 
on both the arch/ppc64 tree and the drivers/pci/hotplug tree without 
excessive co-dependency of patches.   I'm hoping to have these routines 
do a whole lot more once some firmware issues get resolved.

Signed-off-by: Linas Vepstas <linas@linas.org>

--linas

[-- Attachment #2: eeh-notifier-rpaphp.patch --]
[-- Type: text/plain, Size: 4038 bytes --]

===== drivers/pci/hotplug/rpaphp.h 1.9 vs edited =====
--- 1.9/drivers/pci/hotplug/rpaphp.h	Fri Jul  2 11:14:11 2004
+++ edited/drivers/pci/hotplug/rpaphp.h	Wed Jul  7 15:44:35 2004
@@ -124,7 +124,8 @@
 extern int register_pci_slot(struct slot *slot);
 extern int rpaphp_unconfig_pci_adapter(struct slot *slot);
 extern int rpaphp_get_pci_adapter_status(struct slot *slot, int is_init, u8 * value);
-extern struct hotplug_slot *rpaphp_find_hotplug_slot(struct pci_dev *dev);
+extern void init_eeh_handler (void);
+extern void exit_eeh_handler (void);
 
 /* rpaphp_core.c */
 extern int rpaphp_add_slot(struct device_node *dn);
===== drivers/pci/hotplug/rpaphp_core.c 1.14 vs edited =====
--- 1.14/drivers/pci/hotplug/rpaphp_core.c	Tue Jun  8 17:53:59 2004
+++ edited/drivers/pci/hotplug/rpaphp_core.c	Wed Jul  7 13:50:20 2004
@@ -54,8 +54,6 @@
 MODULE_DESCRIPTION(DRIVER_DESC);
 MODULE_LICENSE("GPL");
 
-void eeh_register_disable_func(int (*)(struct pci_dev *));
-
 module_param(debug, bool, 0644);
 
 static int enable_slot(struct hotplug_slot *slot);
@@ -65,7 +63,6 @@
 static int get_attention_status(struct hotplug_slot *slot, u8 * value);
 static int get_adapter_status(struct hotplug_slot *slot, u8 * value);
 static int get_max_bus_speed(struct hotplug_slot *hotplug_slot, enum pci_bus_speed *value);
-static int rpaphp_disable_slot(struct pci_dev *dev);
 
 struct hotplug_slot_ops rpaphp_hotplug_slot_ops = {
 	.owner = THIS_MODULE,
@@ -407,8 +404,8 @@
 {
 	info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
 
-	/* let EEH know they can use hotplug */
-	eeh_register_disable_func(&rpaphp_disable_slot);
+	/* Get set to handle EEH events */
+	init_eeh_handler();
 
 	/* read all the PRA info from the system */
 	return init_rpa();
@@ -416,8 +413,8 @@
 
 static void __exit rpaphp_exit(void)
 {
-	/* let EEH know we are going away */
-	eeh_register_disable_func(NULL);
+	/* Stop handling EEH events */
+	exit_eeh_handler();
 
 	cleanup_slots();
 }
@@ -448,11 +445,6 @@
 exit:
 	dbg("%s - Exit: rc[%d]\n", __FUNCTION__, retval);
 	return retval;
-}
-
-static int rpaphp_disable_slot(struct pci_dev *dev)
-{
-	return disable_slot(rpaphp_find_hotplug_slot(dev));
 }
 
 static int disable_slot(struct hotplug_slot *hotplug_slot)
===== drivers/pci/hotplug/rpaphp_pci.c 1.9 vs edited =====
--- 1.9/drivers/pci/hotplug/rpaphp_pci.c	Thu Jul  1 18:31:49 2004
+++ edited/drivers/pci/hotplug/rpaphp_pci.c	Wed Jul  7 13:56:25 2004
@@ -22,7 +22,9 @@
  * Send feedback to <lxie@us.ibm.com>
  *
  */
+#include <linux/notifier.h>
 #include <linux/pci.h>
+#include <asm/eeh.h>
 #include <asm/pci-bridge.h>
 #include "../pci.h"		/* for pci_add_new_bus */
 
@@ -227,7 +229,7 @@
 	}
 	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
 	/* do pci_scan_child_bus */
-	pci_scan_child_bus(child_bus);
+	// pci_scan_child_bus(child_bus);
 
 	list_for_each_entry(child_dev, &child_bus->devices, bus_list) {
 		eeh_add_device_late(child_dev);
@@ -503,7 +505,7 @@
 	return retval;
 }
 
-struct hotplug_slot *rpaphp_find_hotplug_slot(struct pci_dev *dev)
+static struct slot *rpaphp_find_slot(struct pci_dev *dev)
 {
 	struct list_head	*tmp, *n;
 	struct slot		*slot;
@@ -528,11 +530,34 @@
 		for (ln = bus->devices.next; ln != &bus->devices; ln = ln->next) {
                                 struct pci_dev *pdev = pci_dev_b(ln);
 				if (pdev == dev)
-					return slot->hotplug_slot;
+					return slot;
 		}
 	}
 
 	return NULL;
 }
 
-EXPORT_SYMBOL_GPL(rpaphp_find_hotplug_slot);
+int handle_eeh_events (struct notifier_block *self, 
+                       unsigned long reason, void *ev)
+{
+	struct eeh_event *event = ev;
+
+	/* Just turn it off for now */
+	rpaphp_unconfig_pci_adapter (rpaphp_find_slot(event->dev));
+	return 0;
+}
+
+static struct notifier_block eeh_block;
+
+void __init init_eeh_handler (void)
+{
+	eeh_block.notifier_call = handle_eeh_events;
+	eeh_register_notifier (&eeh_block);
+}
+
+void __exit exit_eeh_handler (void)
+{
+	eeh_block.notifier_call = handle_eeh_events;
+	eeh_register_notifier (&eeh_block);
+}
+

Re: [Pcihpd-discuss] [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Greg KH]

On Wed, Jul 07, 2004 at 03:59:07PM -0500, linas@austin.ibm.com wrote:
> +static struct notifier_block eeh_block;
> +
> +void __init init_eeh_handler (void)
> +{
> +	eeh_block.notifier_call = handle_eeh_events;
> +	eeh_register_notifier (&eeh_block);
> +}
> +
> +void __exit exit_eeh_handler (void)
> +{
> +	eeh_block.notifier_call = handle_eeh_events;
> +	eeh_register_notifier (&eeh_block);
> +}
> +

Um, I don't think you want your exit_* function to look identical to
your init_* function :)

greg k-h

Re: [Pcihpd-discuss] [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[linas]

[-- Attachment #1: Type: text/plain, Size: 1174 bytes --]

On Wed, Jul 07, 2004 at 02:16:57PM -0700, Greg KH wrote:
> On Wed, Jul 07, 2004 at 03:59:07PM -0500, linas@austin.ibm.com wrote:
> > +void __exit exit_eeh_handler (void)
> 
> Um, I don't think you want your exit_* function to look identical to
> your init_* function :)

Ooops .... it helps to hit 'save' every now and then efter editing 
a file.  Attached below is the corrected patch, and a repeat copy of 
the original mail...


---------
Greg,
                                                                                
This patch implements the catching of EEH events by the hotplug subsystem.
Its preliminary in that it doesn't do much with these events; right now,
my goal is to stub-in the required interfaces so that development can occur
on both the arch/ppc64 tree and the drivers/pci/hotplug tree without
excessive co-dependency of patches.   I'm hoping to have these routines
do a whole lot more once some firmware issues get resolved.
                                                                                
Signed-off-by: Linas Vepstas <linas@linas.org>
                                                                                
--linas




[-- Attachment #2: eeh-notifier-rpaphp.patch --]
[-- Type: text/plain, Size: 3993 bytes --]

===== drivers/pci/hotplug/rpaphp.h 1.9 vs edited =====
--- 1.9/drivers/pci/hotplug/rpaphp.h	Fri Jul  2 11:14:11 2004
+++ edited/drivers/pci/hotplug/rpaphp.h	Wed Jul  7 15:44:35 2004
@@ -124,7 +124,8 @@
 extern int register_pci_slot(struct slot *slot);
 extern int rpaphp_unconfig_pci_adapter(struct slot *slot);
 extern int rpaphp_get_pci_adapter_status(struct slot *slot, int is_init, u8 * value);
-extern struct hotplug_slot *rpaphp_find_hotplug_slot(struct pci_dev *dev);
+extern void init_eeh_handler (void);
+extern void exit_eeh_handler (void);
 
 /* rpaphp_core.c */
 extern int rpaphp_add_slot(struct device_node *dn);
===== drivers/pci/hotplug/rpaphp_core.c 1.14 vs edited =====
--- 1.14/drivers/pci/hotplug/rpaphp_core.c	Tue Jun  8 17:53:59 2004
+++ edited/drivers/pci/hotplug/rpaphp_core.c	Wed Jul  7 13:50:20 2004
@@ -54,8 +54,6 @@
 MODULE_DESCRIPTION(DRIVER_DESC);
 MODULE_LICENSE("GPL");
 
-void eeh_register_disable_func(int (*)(struct pci_dev *));
-
 module_param(debug, bool, 0644);
 
 static int enable_slot(struct hotplug_slot *slot);
@@ -65,7 +63,6 @@
 static int get_attention_status(struct hotplug_slot *slot, u8 * value);
 static int get_adapter_status(struct hotplug_slot *slot, u8 * value);
 static int get_max_bus_speed(struct hotplug_slot *hotplug_slot, enum pci_bus_speed *value);
-static int rpaphp_disable_slot(struct pci_dev *dev);
 
 struct hotplug_slot_ops rpaphp_hotplug_slot_ops = {
 	.owner = THIS_MODULE,
@@ -407,8 +404,8 @@
 {
 	info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
 
-	/* let EEH know they can use hotplug */
-	eeh_register_disable_func(&rpaphp_disable_slot);
+	/* Get set to handle EEH events */
+	init_eeh_handler();
 
 	/* read all the PRA info from the system */
 	return init_rpa();
@@ -416,8 +413,8 @@
 
 static void __exit rpaphp_exit(void)
 {
-	/* let EEH know we are going away */
-	eeh_register_disable_func(NULL);
+	/* Stop handling EEH events */
+	exit_eeh_handler();
 
 	cleanup_slots();
 }
@@ -448,11 +445,6 @@
 exit:
 	dbg("%s - Exit: rc[%d]\n", __FUNCTION__, retval);
 	return retval;
-}
-
-static int rpaphp_disable_slot(struct pci_dev *dev)
-{
-	return disable_slot(rpaphp_find_hotplug_slot(dev));
 }
 
 static int disable_slot(struct hotplug_slot *hotplug_slot)
===== drivers/pci/hotplug/rpaphp_pci.c 1.9 vs edited =====
--- 1.9/drivers/pci/hotplug/rpaphp_pci.c	Thu Jul  1 18:31:49 2004
+++ edited/drivers/pci/hotplug/rpaphp_pci.c	Wed Jul  7 16:39:03 2004
@@ -22,7 +22,9 @@
  * Send feedback to <lxie@us.ibm.com>
  *
  */
+#include <linux/notifier.h>
 #include <linux/pci.h>
+#include <asm/eeh.h>
 #include <asm/pci-bridge.h>
 #include "../pci.h"		/* for pci_add_new_bus */
 
@@ -227,7 +229,7 @@
 	}
 	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
 	/* do pci_scan_child_bus */
-	pci_scan_child_bus(child_bus);
+	// pci_scan_child_bus(child_bus);
 
 	list_for_each_entry(child_dev, &child_bus->devices, bus_list) {
 		eeh_add_device_late(child_dev);
@@ -503,7 +505,7 @@
 	return retval;
 }
 
-struct hotplug_slot *rpaphp_find_hotplug_slot(struct pci_dev *dev)
+static struct slot *rpaphp_find_slot(struct pci_dev *dev)
 {
 	struct list_head	*tmp, *n;
 	struct slot		*slot;
@@ -528,11 +530,33 @@
 		for (ln = bus->devices.next; ln != &bus->devices; ln = ln->next) {
                                 struct pci_dev *pdev = pci_dev_b(ln);
 				if (pdev == dev)
-					return slot->hotplug_slot;
+					return slot;
 		}
 	}
 
 	return NULL;
 }
 
-EXPORT_SYMBOL_GPL(rpaphp_find_hotplug_slot);
+int handle_eeh_events (struct notifier_block *self, 
+                       unsigned long reason, void *ev)
+{
+	struct eeh_event *event = ev;
+
+	/* Just turn it off for now */
+	rpaphp_unconfig_pci_adapter (rpaphp_find_slot(event->dev));
+	return 0;
+}
+
+static struct notifier_block eeh_block;
+
+void __init init_eeh_handler (void)
+{
+	eeh_block.notifier_call = handle_eeh_events;
+	eeh_register_notifier (&eeh_block);
+}
+
+void __exit exit_eeh_handler (void)
+{
+	eeh_unregister_notifier (&eeh_block);
+}
+

Re: [Pcihpd-discuss] [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Greg KH]

On Wed, Jul 07, 2004 at 04:47:39PM -0500, linas@austin.ibm.com wrote:

> ===== drivers/pci/hotplug/rpaphp.h 1.9 vs edited =====
> --- 1.9/drivers/pci/hotplug/rpaphp.h	Fri Jul  2 11:14:11 2004
> +++ edited/drivers/pci/hotplug/rpaphp.h	Wed Jul  7 15:44:35 2004
> @@ -124,7 +124,8 @@
>  extern int register_pci_slot(struct slot *slot);
>  extern int rpaphp_unconfig_pci_adapter(struct slot *slot);
>  extern int rpaphp_get_pci_adapter_status(struct slot *slot, int is_init, u8 * value);
> -extern struct hotplug_slot *rpaphp_find_hotplug_slot(struct pci_dev *dev);
> +extern void init_eeh_handler (void);
> +extern void exit_eeh_handler (void);

This belongs in the eeh header file, not the rpaphp.h file.

> @@ -227,7 +229,7 @@
>  	}
>  	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
>  	/* do pci_scan_child_bus */
> -	pci_scan_child_bus(child_bus);
> +	// pci_scan_child_bus(child_bus);

And the reason you are commenting out this function is...

thanks,

greg k-h

Re: [Pcihpd-discuss] [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[linas]

On Wed, Jul 07, 2004 at 10:26:04PM -0700, Greg KH wrote:
> On Wed, Jul 07, 2004 at 04:47:39PM -0500, linas@austin.ibm.com wrote:
> 
> > ===== drivers/pci/hotplug/rpaphp.h 1.9 vs edited =====
> > +extern void init_eeh_handler (void);
> > +extern void exit_eeh_handler (void);
> 
> This belongs in the eeh header file, not the rpaphp.h file.

But these are implemented in rpaphp_pci.c and are called 
in rpaphp_core.c.  Perchance a different function name, such as
init_rpaphp_eeh_handler() be better?

> > +	// pci_scan_child_bus(child_bus);
> 
> And the reason you are commenting out this function is...

Due to a mistake.

--linas

Re: [Pcihpd-discuss] [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Greg KH]

On Thu, Jul 08, 2004 at 10:04:33AM -0500, linas@austin.ibm.com wrote:
> On Wed, Jul 07, 2004 at 10:26:04PM -0700, Greg KH wrote:
> > On Wed, Jul 07, 2004 at 04:47:39PM -0500, linas@austin.ibm.com wrote:
> > 
> > > ===== drivers/pci/hotplug/rpaphp.h 1.9 vs edited =====
> > > +extern void init_eeh_handler (void);
> > > +extern void exit_eeh_handler (void);
> > 
> > This belongs in the eeh header file, not the rpaphp.h file.
> 
> But these are implemented in rpaphp_pci.c and are called 
> in rpaphp_core.c.  Perchance a different function name, such as
> init_rpaphp_eeh_handler() be better?

Ah, yes, see I didn't even catch that :)

They should start with "rpaphp" as they are global symbols.

thanks,

greg k-h

Re: [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Linda Xie]

linas@austin.ibm.com wrote:

>  
>
>
> 	}
> 	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
> 	/* do pci_scan_child_bus */
>-	pci_scan_child_bus(child_bus);
>+	// pci_scan_child_bus(child_bus);
> 
>
>  
>
Why remove pci_scan_child_bus call?

Linda

Re: [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[linas]

On Wed, Jul 07, 2004 at 07:49:06PM -0500, Linda Xie wrote:
> linas@austin.ibm.com wrote:
> 
> > 	}
> > 	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
> > 	/* do pci_scan_child_bus */
> >-	pci_scan_child_bus(child_bus);
> >+	// pci_scan_child_bus(child_bus);
> > 
> Why remove pci_scan_child_bus call?

Because it won't compile otherwise.  
(Actually, I didn't mean to leave that in the patch, 
it was a work-around to get my tree to compile).

pci_scan_child_bus() is currently defined only as a static fuction
in drivers/pci/probe.c and thus cannot be called outside of that 
file.  Maybe there's a patch to drivers/pci/probe.c that hasn't 
been applied yet?

--linas

Re: [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Greg KH]

On Wed, Jul 07, 2004 at 07:06:42PM -0500, linas@austin.ibm.com wrote:
> On Wed, Jul 07, 2004 at 07:49:06PM -0500, Linda Xie wrote:
> > linas@austin.ibm.com wrote:
> > 
> > > 	}
> > > 	sprintf(child_bus->name, "PCI Bus #%02x", child_bus->number);
> > > 	/* do pci_scan_child_bus */
> > >-	pci_scan_child_bus(child_bus);
> > >+	// pci_scan_child_bus(child_bus);
> > > 
> > Why remove pci_scan_child_bus call?
> 
> Because it won't compile otherwise.  
> (Actually, I didn't mean to leave that in the patch, 
> it was a work-around to get my tree to compile).
> 
> pci_scan_child_bus() is currently defined only as a static fuction
> in drivers/pci/probe.c and thus cannot be called outside of that 
> file.  Maybe there's a patch to drivers/pci/probe.c that hasn't 
> been applied yet?

It's in the latest -bk tree, right?

greg k-h

Re: [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[linas]

On Wed, Jul 07, 2004 at 11:09:33PM -0700, Greg KH wrote:
> > pci_scan_child_bus() is currently ...
>
> It's in the latest -bk tree, right?

yes, it looks correct in the current tree (it wasn't when 
I'd previously cloned).    It's possible I'm not using bk correctly;
once I've modified a file, 'bk pull' never merges in newer changes 
made upstream.  So I have to 'bk clone' all the time, which is a 
pain in the neck. What am I doing wrong?

--linas

Re: [PATCH] 2.6 PCI Hotplug: receive PPC64 EEH events

[Greg KH]

On Thu, Jul 08, 2004 at 10:24:25AM -0500, linas@austin.ibm.com wrote:
> On Wed, Jul 07, 2004 at 11:09:33PM -0700, Greg KH wrote:
> > > pci_scan_child_bus() is currently ...
> >
> > It's in the latest -bk tree, right?
> 
> yes, it looks correct in the current tree (it wasn't when 
> I'd previously cloned).    It's possible I'm not using bk correctly;
> once I've modified a file, 'bk pull' never merges in newer changes 
> made upstream.  So I have to 'bk clone' all the time, which is a 
> pain in the neck. What am I doing wrong?

Not checking in your file before pulling.

greg k-h