[BUG] WARN_ON(!context) in drivers/pci/hotplug/acpiphp_glue.c

Commit Message

Rafael J. Wysocki Oct. 11, 2013, 1:45 a.m. UTC

On Friday, October 11, 2013 01:09:33 AM Rafael J. Wysocki wrote:
> On Thursday, October 10, 2013 02:37:15 PM Linus Torvalds wrote:
> > On Thu, Oct 10, 2013 at 2:35 PM, Rafael J. Wysocki <[e-mail]> wrote:
> > >
> > > Well, I must have overlooked the original report.  Is it available anywhere
> > > I can find it?
> > 
> > I think Steven has some buggered email system, he has other emails
> > being eaten by lkml too, and apparently other mail gateways (because
> > you were direct-cc'd on the original).
> 
> Mailer issues aside, I've just seen the original (Bjorn forwarded it to me,
> thanks!) and I'm wondering if the message added by the debug patch below is
> triggered along with the WARN_ON().  If it is, I think it's better to drop
> the WARN_ON(), at least for now (until we sort out the acpiphp/pciehp
> coexistence).
>
> ---
>  drivers/pci/hotplug/acpiphp_glue.c |    5 +++++
>  1 file changed, 5 insertions(+)
> 
> Index: linux-pm/drivers/pci/hotplug/acpiphp_glue.c
> ===================================================================
> --- linux-pm.orig/drivers/pci/hotplug/acpiphp_glue.c
> +++ linux-pm/drivers/pci/hotplug/acpiphp_glue.c
> @@ -991,6 +991,11 @@ void acpiphp_enumerate_slots(struct pci_
>  
>  	if (!pci_is_root_bus(bridge->pci_bus)) {
>  		struct acpiphp_context *context;
> +		struct pci_dev *parent = bridge->pci_bus->parent->self;
> +
> +		if (parent && device_is_managed_by_native_pciehp(parent))
> +			dev_warn(&bridge->pci_dev->dev,
> +				 "Parent is managed by pciehp!\n");
>  
>  		/*
>  		 * This bridge should have been registered as a hotplug function
> 

Steve told me over IRC that the message added by the above triggered along
with the WARN_ON(), so this really was the issue I had in mind.  So, I asked
Steve to test the appended patch and it worked for him.

Well, I admit this is a gray area, because we've never tried it, but I think
we'll need to try it at one point anyway and see how it goes, so why don't we
actually do that now?

If it turns out to cause problems to happen for people, we can simply revert
it and remove the WARN_ON() instead.  And then we'll know that this is
problematic.

What do you think?

Rafael


---
From: Rafael J. Wysocki <[e-mail]>
Subject: ACPI / hotplug / PCI: Accept coexistence with native PCIe hotplug

Allow ACPIPHP (ACPI-based PCI hotplug) to handle event signaling for
devices that have already been claimed by the native PCIe hotplug
(pciehp).

This change prevents the WARN_ON() in acpiphp_enumerate_slots() from
triggering unnecessarily for bridges whose parents are managed by
pciehp.

Signed-off-by: Rafael J. Wysocki <[e-mail]>
---
 drivers/pci/hotplug/acpiphp_glue.c |    6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)


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

Comments

Linus Torvalds Oct. 11, 2013, 1:42 a.m. UTC | #1

On Thu, Oct 10, 2013 at 6:45 PM, Rafael J. Wysocki <[e-mail]> wrote:
>
>         /* Register slots for ejectable funtions only. */
> -       if (acpi_pci_check_ejectable(pbus, handle)  || is_dock_device(handle)) {
> +       if ((acpi_pci_check_ejectable(pbus, handle) || is_dock_device(handle))
> +           && !(pdev && device_is_managed_by_native_pciehp(pdev))) {
>                 unsigned long long sun;
>                 int retval;

I can't even begin to say whether this is a good solution or not,
because that if-conditional makes me want to go out and kill some
homeless people to let my aggressions out.

Can we please agree to *never* write code like this? Ever?

Use a well-named inline helper function where the name describes what
the f*ck the code is trying to do, and then comment the separate
issues. Because none of the above line noise makes me go "Ahh, it's
the test for an ejectable function".

What the heck _is_ an "ejectable function" anyway? The only comment
there just makes the code even less sensible.

Please?

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

Patch

Index: linux-pm/drivers/pci/hotplug/acpiphp_glue.c
===================================================================
--- linux-pm.orig/drivers/pci/hotplug/acpiphp_glue.c
+++ linux-pm/drivers/pci/hotplug/acpiphp_glue.c
@@ -274,9 +274,6 @@  static acpi_status register_slot(acpi_ha
 	struct pci_dev *pdev = bridge->pci_dev;
 	u32 val;
 
-	if (pdev && device_is_managed_by_native_pciehp(pdev))
-		return AE_OK;
-
 	status = acpi_evaluate_integer(handle, "_ADR", NULL, &adr);
 	if (ACPI_FAILURE(status)) {
 		acpi_handle_warn(handle, "can't evaluate _ADR (%#x)\n", status);
@@ -326,7 +323,8 @@  static acpi_status register_slot(acpi_ha
 	list_add_tail(&slot->node, &bridge->slots);
 
 	/* Register slots for ejectable funtions only. */
-	if (acpi_pci_check_ejectable(pbus, handle)  || is_dock_device(handle)) {
+	if ((acpi_pci_check_ejectable(pbus, handle) || is_dock_device(handle))
+	    && !(pdev && device_is_managed_by_native_pciehp(pdev))) {
 		unsigned long long sun;
 		int retval;