Skip to contentSkip to conversation

PCI, ACPI, hotplug: Fix BUS_CHECK event handle on root bridge

PATCH In review Yinghai Lu 4 participants 4 messages Apr 26, 2013 – May 10, 2013 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): CAAh6nkn2WxQAcmtLM1bu66Ngn+BheJNfoAUMN=dO9eY2HXi7YQ@mail.gmail.com

Yinghai Lu

author
#1 lore raw

PATCH 1/1 PCI, ACPI, hotplug: Fix BUS_CHECK event handle on root bridge

Gavin found that acpiphp does not handle hotplug anymore even after now we have acpiphp built-in preparing for v3.10.

Bjorn analyzed bootlog, he found that
acpiphp: ACPI Hot Plug PCI Controller Driver version: 0.5
ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-3e]) \_SB_.PCI0:_OSC invalid UUID
acpiphp: Slot [1] registered
acpiphp: Slot [1-1] registered
acpi root: \_SB_.PCI0 notify handler is installed
_handle_hotplug_event_root: Bus check notify on \_SB_.PCI0 _handle_hotplug_event_root: Bus check notify on \_SB_.PCI0
And that means:
So we should be using acpiphp, which you do have built in statically, and it found a couple slots. And we did get two bus check notifies on \_SB_.PCI0, so we *should* be re-enumerating PCI bus 0000:00. But it looks like we're handling this as a host bridge hotplug event instead of a PCI device hotplug. My guess is that handle_root_bridge_insertion() does nothing because the PCI0 ACPI device already exists, though I would expect to see the "acpi device exists..." in your dmesg log if this were the case.

Also according to Rafael and Bjorn, it is perfect fine that we should enumerate bus by sending event to root bridge after hotadd device to slots under that root bridge or children bridges.

It turns out that it is regression caused by

| commit 668192b678201d2fff27c6cc76bb003c1ec4a52a
| Author: Yinghai Lu <[email protected]>
| Date:   Mon Jan 21 13:20:48 2013 -0800
|
|    PCI: acpiphp: Move host bridge hotplug to pci_root.c

We should check slots when BUS_CHECK is sent to root bridge acpi handle.

Restore the old behavoir by calling acpi_check_bridge and check_sub_bridge in acpiphp.

Jiang Liu acctually have simimar patch but it forgets calling acpi_check_bridge() for system that have slots on root bus directly. That is still valid, as in QEMU we still have that slots on bus 0 at least. But my first version patch wrongly check if root bridge exists before check_sub_bridge for children bridges.

-v2: Don't check bridge for acpi_walk_namespace with check_sub_bridges.
also put back bridge reference.
-v3: More changelog and etc.

---
 drivers/acpi/pci_root.c            |    2 ++
 drivers/pci/hotplug/acpiphp_glue.c |   14 ++++++++++++++
 include/linux/pci-acpi.h           |    2 ++
 3 files changed, 18 insertions(+)
drivers/acpi/pci_root.c +2 -0
Index: linux-2.6/drivers/acpi/pci_root.c
===================================================================
@@ -643,6 +643,8 @@ static void _handle_hotplug_event_root(s
 				 (char *)buffer.pointer);
 		if (!root)
 			handle_root_bridge_insertion(handle);
+		else
+			acpiphp_check_host_bridge(handle);
 
 		break;
 
drivers/pci/hotplug/acpiphp_glue.c +14 -0
Index: linux-2.6/drivers/pci/hotplug/acpiphp_glue.c
===================================================================
@@ -950,6 +950,20 @@ check_sub_bridges(acpi_handle handle, u3
 	return AE_OK ;
 }
 
+void acpiphp_check_host_bridge(acpi_handle handle)
+{
+	struct acpiphp_bridge *bridge;
+
+	bridge = acpiphp_handle_to_bridge(handle);
+	if (bridge) {
+		acpiphp_check_bridge(bridge);
+		put_bridge(bridge);
+	}
+
+	acpi_walk_namespace(ACPI_TYPE_DEVICE, handle,
+		ACPI_UINT32_MAX, check_sub_bridges, NULL, NULL, NULL);
+}
+
 static void _handle_hotplug_event_bridge(struct work_struct *work)
 {
 	struct acpiphp_bridge *bridge;
include/linux/pci-acpi.h +2 -0
Index: linux-2.6/include/linux/pci-acpi.h
===================================================================
@@ -60,11 +60,13 @@ static inline void acpi_pci_slot_remove(
 void acpiphp_init(void);
 void acpiphp_enumerate_slots(struct pci_bus *bus, acpi_handle handle);
 void acpiphp_remove_slots(struct pci_bus *bus);
+void acpiphp_check_host_bridge(acpi_handle handle);
 #else
 static inline void acpiphp_init(void) { }
 static inline void acpiphp_enumerate_slots(struct pci_bus *bus,
 					   acpi_handle handle) { }
 static inline void acpiphp_remove_slots(struct pci_bus *bus) { }
+static inline void acpiphp_check_host_bridge(acpi_handle handle) { }
 #endif
 
 #else	/* CONFIG_ACPI */
Acked-by on v1 1/1
#2 lore raw
Quoted 48 lines · Yinghai Lu

Gavin found that acpiphp does not handle hotplug anymore even after now we have acpiphp built-in preparing for v3.10.

Bjorn analyzed bootlog, he found that
acpiphp: ACPI Hot Plug PCI Controller Driver version: 0.5
ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-3e]) \_SB_.PCI0:_OSC invalid UUID
acpiphp: Slot [1] registered
acpiphp: Slot [1-1] registered
acpi root: \_SB_.PCI0 notify handler is installed
_handle_hotplug_event_root: Bus check notify on \_SB_.PCI0 _handle_hotplug_event_root: Bus check notify on \_SB_.PCI0
And that means:
So we should be using acpiphp, which you do have built in statically, and it found a couple slots. And we did get two bus check notifies on \_SB_.PCI0, so we *should* be re-enumerating PCI bus 0000:00. But it looks like we're handling this as a host bridge hotplug event instead of a PCI device hotplug. My guess is that handle_root_bridge_insertion() does nothing because the PCI0 ACPI device already exists, though I would expect to see the "acpi device exists..." in your dmesg log if this were the case.

Also according to Rafael and Bjorn, it is perfect fine that we should enumerate bus by sending event to root bridge after hotadd device to slots under that root bridge or children bridges.

It turns out that it is regression caused by

| commit 668192b678201d2fff27c6cc76bb003c1ec4a52a
| Author: Yinghai Lu <[email protected]>
| Date:   Mon Jan 21 13:20:48 2013 -0800
|
|    PCI: acpiphp: Move host bridge hotplug to pci_root.c

We should check slots when BUS_CHECK is sent to root bridge acpi handle.

Restore the old behavoir by calling acpi_check_bridge and check_sub_bridge in acpiphp.

Jiang Liu acctually have simimar patch but it forgets calling acpi_check_bridge() for system that have slots on root bus directly. That is still valid, as in QEMU we still have that slots on bus 0 at least. But my first version patch wrongly check if root bridge exists before check_sub_bridge for children bridges.

(quoted diff) +0 -2
-v2: Don't check bridge for acpi_walk_namespace with check_sub_bridges.
     also put back bridge reference.
-v3: More changelog and etc.
Quoted 3 lines
Quoted 62 lines
---
 drivers/acpi/pci_root.c            |    2 ++
 drivers/pci/hotplug/acpiphp_glue.c |   14 ++++++++++++++
 include/linux/pci-acpi.h           |    2 ++
 3 files changed, 18 insertions(+)

Index: linux-2.6/drivers/acpi/pci_root.c
===================================================================

drivers/acpi/pci_root.c +2 -0
@@ -643,6 +643,8 @@ static void _handle_hotplug_event_root(s
 				 (char *)buffer.pointer);
 		if (!root)
 			handle_root_bridge_insertion(handle);
+		else
+			acpiphp_check_host_bridge(handle);
 
 		break;
 

Index: linux-2.6/drivers/pci/hotplug/acpiphp_glue.c
===================================================================

drivers/pci/hotplug/acpiphp_glue.c +14 -0
@@ -950,6 +950,20 @@ check_sub_bridges(acpi_handle handle, u3
 	return AE_OK ;
 }
 
+void acpiphp_check_host_bridge(acpi_handle handle)
+{
+	struct acpiphp_bridge *bridge;
+
+	bridge = acpiphp_handle_to_bridge(handle);
+	if (bridge) {
+		acpiphp_check_bridge(bridge);
+		put_bridge(bridge);
+	}
+
+	acpi_walk_namespace(ACPI_TYPE_DEVICE, handle,
+		ACPI_UINT32_MAX, check_sub_bridges, NULL, NULL, NULL);
+}
+
 static void _handle_hotplug_event_bridge(struct work_struct *work)
 {
 	struct acpiphp_bridge *bridge;

Index: linux-2.6/include/linux/pci-acpi.h
===================================================================

include/linux/pci-acpi.h +2 -0
@@ -60,11 +60,13 @@ static inline void acpi_pci_slot_remove(
 void acpiphp_init(void);
 void acpiphp_enumerate_slots(struct pci_bus *bus, acpi_handle handle);
 void acpiphp_remove_slots(struct pci_bus *bus);
+void acpiphp_check_host_bridge(acpi_handle handle);
 #else
 static inline void acpiphp_init(void) { }
 static inline void acpiphp_enumerate_slots(struct pci_bus *bus,
 					   acpi_handle handle) { }
 static inline void acpiphp_remove_slots(struct pci_bus *bus) { }
+static inline void acpiphp_check_host_bridge(acpi_handle handle) { }
 #endif
 
 #else	/* CONFIG_ACPI */
Signature · 2 lines
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
Reviewed-by on v1 1/1
#3 lore raw
In reply to Yinghai Lu

Hi Yinghai,

Thanks!

Quoted 119 lines · Yinghai Lu

Gavin found that acpiphp does not handle hotplug anymore even after now we have acpiphp built-in preparing for v3.10.

Bjorn analyzed bootlog, he found that
acpiphp: ACPI Hot Plug PCI Controller Driver version: 0.5
ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-3e]) \_SB_.PCI0:_OSC invalid UUID
acpiphp: Slot [1] registered
acpiphp: Slot [1-1] registered
acpi root: \_SB_.PCI0 notify handler is installed
_handle_hotplug_event_root: Bus check notify on \_SB_.PCI0 _handle_hotplug_event_root: Bus check notify on \_SB_.PCI0
And that means:
So we should be using acpiphp, which you do have built in statically, and it found a couple slots. And we did get two bus check notifies on \_SB_.PCI0, so we *should* be re-enumerating PCI bus 0000:00. But it looks like we're handling this as a host bridge hotplug event instead of a PCI device hotplug. My guess is that handle_root_bridge_insertion() does nothing because the PCI0 ACPI device already exists, though I would expect to see the "acpi device exists..." in your dmesg log if this were the case.

Also according to Rafael and Bjorn, it is perfect fine that we should enumerate bus by sending event to root bridge after hotadd device to slots under that root bridge or children bridges.

It turns out that it is regression caused by

| commit 668192b678201d2fff27c6cc76bb003c1ec4a52a
| Author: Yinghai Lu <[email protected]>
| Date:   Mon Jan 21 13:20:48 2013 -0800
|
|    PCI: acpiphp: Move host bridge hotplug to pci_root.c

We should check slots when BUS_CHECK is sent to root bridge acpi handle.

Restore the old behavoir by calling acpi_check_bridge and check_sub_bridge in acpiphp.

Jiang Liu acctually have simimar patch but it forgets calling acpi_check_bridge() for system that have slots on root bus directly. That is still valid, as in QEMU we still have that slots on bus 0 at least. But my first version patch wrongly check if root bridge exists before check_sub_bridge for children bridges.

(quoted diff) +0 -2
-v2: Don't check bridge for acpi_walk_namespace with check_sub_bridges.
     also put back bridge reference.
-v3: More changelog and etc.
---
 drivers/acpi/pci_root.c            |    2 ++
 drivers/pci/hotplug/acpiphp_glue.c |   14 ++++++++++++++
 include/linux/pci-acpi.h           |    2 ++
 3 files changed, 18 insertions(+)

Index: linux-2.6/drivers/acpi/pci_root.c
===================================================================

drivers/acpi/pci_root.c +2 -0
@@ -643,6 +643,8 @@ static void _handle_hotplug_event_root(s
 				 (char *)buffer.pointer);
 		if (!root)
 			handle_root_bridge_insertion(handle);
+		else
+			acpiphp_check_host_bridge(handle);
 
 		break;
 

Index: linux-2.6/drivers/pci/hotplug/acpiphp_glue.c
===================================================================

drivers/pci/hotplug/acpiphp_glue.c +14 -0
@@ -950,6 +950,20 @@ check_sub_bridges(acpi_handle handle, u3
 	return AE_OK ;
 }
 
+void acpiphp_check_host_bridge(acpi_handle handle)
+{
+	struct acpiphp_bridge *bridge;
+
+	bridge = acpiphp_handle_to_bridge(handle);
+	if (bridge) {
+		acpiphp_check_bridge(bridge);
+		put_bridge(bridge);
+	}
+
+	acpi_walk_namespace(ACPI_TYPE_DEVICE, handle,
+		ACPI_UINT32_MAX, check_sub_bridges, NULL, NULL, NULL);
+}
+
 static void _handle_hotplug_event_bridge(struct work_struct *work)
 {
 	struct acpiphp_bridge *bridge;

Index: linux-2.6/include/linux/pci-acpi.h
===================================================================

include/linux/pci-acpi.h +2 -1
@@ -60,11 +60,13 @@ static inline void acpi_pci_slot_remove(
 void acpiphp_init(void);
 void acpiphp_enumerate_slots(struct pci_bus *bus, acpi_handle handle);
 void acpiphp_remove_slots(struct pci_bus *bus);
+void acpiphp_check_host_bridge(acpi_handle handle);
 #else
 static inline void acpiphp_init(void) { }
 static inline void acpiphp_enumerate_slots(struct pci_bus *bus,
 					   acpi_handle handle) { }
 static inline void acpiphp_remove_slots(struct pci_bus *bus) { }
+static inline void acpiphp_check_host_bridge(acpi_handle handle) { }
 #endif
 
 #else	/* CONFIG_ACPI */
--

To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

maintainer
#4 lore raw
Quoted 49 lines · Jiang Liu

Hi Yinghai,

Thanks!

Quoted 44 lines · Yinghai Lu

Gavin found that acpiphp does not handle hotplug anymore even after now we have acpiphp built-in preparing for v3.10.

Bjorn analyzed bootlog, he found that
acpiphp: ACPI Hot Plug PCI Controller Driver version: 0.5
ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-3e]) \_SB_.PCI0:_OSC invalid UUID
acpiphp: Slot [1] registered
acpiphp: Slot [1-1] registered
acpi root: \_SB_.PCI0 notify handler is installed
_handle_hotplug_event_root: Bus check notify on \_SB_.PCI0 _handle_hotplug_event_root: Bus check notify on \_SB_.PCI0
And that means:
So we should be using acpiphp, which you do have built in statically, and it found a couple slots. And we did get two bus check notifies on \_SB_.PCI0, so we *should* be re-enumerating PCI bus 0000:00. But it looks like we're handling this as a host bridge hotplug event instead of a PCI device hotplug. My guess is that handle_root_bridge_insertion() does nothing because the PCI0 ACPI device already exists, though I would expect to see the "acpi device exists..." in your dmesg log if this were the case.

Also according to Rafael and Bjorn, it is perfect fine that we should enumerate bus by sending event to root bridge after hotadd device to slots under that root bridge or children bridges.

It turns out that it is regression caused by

| commit 668192b678201d2fff27c6cc76bb003c1ec4a52a
| Author: Yinghai Lu <[email protected]>
| Date:   Mon Jan 21 13:20:48 2013 -0800
|
|    PCI: acpiphp: Move host bridge hotplug to pci_root.c

We should check slots when BUS_CHECK is sent to root bridge acpi handle.

Restore the old behavoir by calling acpi_check_bridge and check_sub_bridge in acpiphp.

Jiang Liu acctually have simimar patch but it forgets calling acpi_check_bridge() for system that have slots on root bus directly. That is still valid, as in QEMU we still have that slots on bus 0 at least. But my first version patch wrongly check if root bridge exists before check_sub_bridge for children bridges.

Quoted 7 lines
Quoted 7 lines
(quoted diff) +0 -2
-v2: Don't check bridge for acpi_walk_namespace with check_sub_bridges.
     also put back bridge reference.
-v3: More changelog and etc.

I added the Acked-by and Reviewed-by and opened
https://bugzilla.kernel.org/show_bug.cgi?id=57961 just as a place to store the dmesg logs. I also backported it to v3.9 and added a CC:
stable tag because I think v3.9 has the same issue. The backport is also attached to the bugzilla.

I'll ask Linus to put this in v3.10.

Thanks,
Bjorn