Re: [Bisected] [-next-20130204] usb/hcd: irq 18: nobody cared

[Yinghai Lu <yinghai@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 3476 bytes --]

On Mon, Feb 11, 2013 at 5:02 AM, Peter Hurley <peter@hurleysoftware.com> wrote:
> On Sun, 2013-02-10 at 22:40 -0800, Yinghai Lu wrote:
>> On Sun, Feb 10, 2013 at 12:33 PM, Yinghai Lu <yinghai@kernel.org> wrote:
>> > On Sun, Feb 10, 2013 at 6:23 AM, Peter Hurley <peter@hurleysoftware.com> wrote:
>> >> On Sat, 2013-02-09 at 22:14 -0500, Peter Hurley wrote:
>> >>> On Tue, 2013-02-05 at 15:26 -0500, Alan Stern wrote:
>> >>> > On Tue, 5 Feb 2013, Peter Hurley wrote:
>> >>> >
>> >>> > > With -next-20130204:
>> >>> > >
>> >>> > > [   33.855570] irq 18: nobody cared (try booting with the "irqpoll" option)
>> >>> > > [   33.855580] Pid: 0, comm: swapper/4 Not tainted 3.8.0-next-20130204-xeon #20130204
>> >>> > > [   33.855582] Call Trace:
>> >>> > > [   33.855585]  <IRQ>  [<ffffffff810f1076>] __report_bad_irq+0x36/0xe0
>> >>> > > [   33.855600]  [<ffffffff810f152a>] note_interrupt+0x1aa/0x200
>> >>> > > [   33.855606]  [<ffffffff8101edf2>] ? mwait_idle+0x82/0x1b0
>> >>> > > [   33.855610]  [<ffffffff810eed89>] handle_irq_event_percpu+0xc9/0x260
>> >>> > > [   33.855614]  [<ffffffff810eef68>] handle_irq_event+0x48/0x70
>> >>> > > [   33.855618]  [<ffffffff810f20ba>] handle_fasteoi_irq+0x5a/0x100
>> >>> > > [   33.855624]  [<ffffffff810182f2>] handle_irq+0x22/0x40
>> >>> > > [   33.855630]  [<ffffffff816e2a9a>] do_IRQ+0x5a/0xd0
>> >>> > > [   33.855636]  [<ffffffff816d97ad>] common_interrupt+0x6d/0x6d
>> >>> > > [   33.855638]  <EOI>  [<ffffffff810f8e2a>] ? rcu_eqs_enter_common+0x4a/0x320
>> >>> > > [   33.855646]  [<ffffffff8101edf2>] ? mwait_idle+0x82/0x1b0
>> >>> > > [   33.855649]  [<ffffffff8101ed99>] ? mwait_idle+0x29/0x1b0
>> >>> > > [   33.855653]  [<ffffffff8101f8a6>] cpu_idle+0x116/0x130
>> >>> > > [   33.855658]  [<ffffffff816c0c0f>] start_secondary+0x251/0x258
>> >>> > > [   33.855660] handlers:
>> >>> > > [   33.855664] [<ffffffff814f98a0>] usb_hcd_irq
>> >>> > > [   33.855667] Disabling IRQ #18
>> >>> > >
>> >> https://bugzilla.kernel.org/show_bug.cgi?id=53561
>> >>
>> >> Maybe this is some interaction with all the new ACPI code and fixes
>> >> written in those 8 days.
>> >
>> > interrupt routing seems get changed:
>> > next:
>> >    5:          0          0          0          0          0
>> > 0          0          0   IO-APIC-fasteoi   snd_ctxfi
>> >   18:      99970         13         16         20      99940
>> > 13         13         16   IO-APIC-fasteoi   uhci_hcd:usb4
>> > v3.8-rc7:
>> >   18:        424         15         11        112        420
>> > 16         18        105   IO-APIC-fasteoi   uhci_hcd:usb4, snd_ctxfi
>> >
>> > These messages in the bad dmesg log are interesting since PCI INT A is routed
>> > on
>> > irq 18 with the kernels that work.
>> > [    8.983246] pci 0000:00:1e.0: can't derive routing for PCI INT A
>> > [    8.983600] snd_ctxfi 0000:09:02.0: PCI INT A: no GSI - using ISA IRQ 5
>> > ...
>> >
>> > acpi_pci_irq_add_prt() add wrong bus for that bridge, because that
>> > that bridge is not scanned.
>> >
>> > Will check if I can produce one patch for it.
>>
>> Hi Peter,
>>
>> Can you try attached debug patch?
>
> Fixed, thanks.

Bjorn, Rafael,

acpi_pci_irq_add_prt need to be called after pci bridge get scanned,
so we can not call it from pci_acpi_setup, after we move dev_register
for pci_dev early.

The attached debug patch move down that calling into
pci_bus_add_devices and that will make prt works again.

Can acpi provide another hook after bridge get scanned?

Thanks

Yinghai

[-- Attachment #2: move_add_irq_prt_down.patch --]
[-- Type: application/octet-stream, Size: 2009 bytes --]

---
 drivers/pci/bus.c      |    5 +++++
 drivers/pci/pci-acpi.c |   13 +++++++++++--
 2 files changed, 16 insertions(+), 2 deletions(-)

Index: linux-2.6/drivers/pci/bus.c
===================================================================
--- linux-2.6.orig/drivers/pci/bus.c
+++ linux-2.6/drivers/pci/bus.c
@@ -160,6 +160,8 @@ pci_bus_alloc_resource(struct pci_bus *b
 
 void __weak pcibios_resource_survey_bus(struct pci_bus *bus) { }
 
+void __weak pcibios_irq_setup(struct device *dev) { }
+
 /**
  * pci_bus_add_device - start driver for a single device
  * @dev: device to add
@@ -170,6 +172,9 @@ int pci_bus_add_device(struct pci_dev *d
 {
 	int retval;
 
+	/* need to after bridge is scanned */
+	pcibios_irq_setup(&dev->dev);
+
 	/*
 	 * Can not put in pci_device_add yet because resources
 	 * are not assigned yet for some devices.
Index: linux-2.6/drivers/pci/pci-acpi.c
===================================================================
--- linux-2.6.orig/drivers/pci/pci-acpi.c
+++ linux-2.6/drivers/pci/pci-acpi.c
@@ -302,14 +302,16 @@ static int acpi_pci_find_device(struct d
 	return 0;
 }
 
-static void pci_acpi_setup(struct device *dev)
+void pcibios_irq_setup(struct device *dev)
 {
 	struct pci_dev *pci_dev = to_pci_dev(dev);
 	acpi_handle handle = ACPI_HANDLE(dev);
-	struct acpi_device *adev;
 	acpi_status status;
 	acpi_handle dummy;
 
+	if (acpi_disabled || acpi_pci_disabled)
+		return;
+
 	/*
 	 * Evaluate and parse _PRT, if exists.  This code allows parsing of
 	 * _PRT objects within the scope of non-bridge devices.  Note that
@@ -326,6 +328,13 @@ static void pci_acpi_setup(struct device
 			pci_dev->subordinate->number : pci_dev->bus->number;
 		acpi_pci_irq_add_prt(handle, pci_domain_nr(pci_dev->bus), bus);
 	}
+}
+
+static void pci_acpi_setup(struct device *dev)
+{
+	struct pci_dev *pci_dev = to_pci_dev(dev);
+	acpi_handle handle = ACPI_HANDLE(dev);
+	struct acpi_device *adev;
 
 	if (acpi_bus_get_device(handle, &adev) || !adev->wakeup.flags.valid)
 		return;