diff mbox

pci/hotplug: work-around for missing _RMV on HP ZBook G2

Message ID 1948383.UKjRPomMUp@vostro.rjw.lan (mailing list archive)
State New, archived
Delegated to: Bjorn Helgaas
Headers show

Commit Message

Rafael J. Wysocki May 18, 2015, 11:06 p.m. UTC
On Monday, May 18, 2015 04:45:28 PM Jarod Wilson wrote:
> On 5/18/2015 12:17 PM, Jarod Wilson wrote:
> > On 5/18/2015 10:33 AM, Jarod Wilson wrote:
> >> On 5/17/2015 8:26 PM, Rafael J. Wysocki wrote:
> >>> On Saturday, May 16, 2015 09:41:55 AM Bjorn Helgaas wrote:
> >>>> On Sat, May 16, 2015 at 09:37:50AM -0500, Bjorn Helgaas wrote:
> >>>>> Hi Jarod,
> >>>>>
> >>>>> On Thu, May 14, 2015 at 03:33:58PM -0400, Jarod Wilson wrote:
> >>>>>> The HP ZBook 15 and 17 Mobile Workstations, generation 2, up to and
> >>>>>> including at least BIOS revision 01.07, do not have an ACPI _RMV
> >>>>>> object
> >>>>>> associated with their expresscard slots, so acpi-based
> >>>>>> hotplug-capable
> >>>>>> slot detection fails. If we fall back to pcie-based detection, the
> >>>>>> systems
> >>>>>> work just fine, so this uses dmi matching to do that. With luck, a
> >>>>>> future
> >>>>>> BIOS will remedy this (I've let someone at HP know about the
> >>>>>> problem),
> >>>>>> but for now, just use this for all existing versions.
> >> ...
> >>>>> Oh, my goodness.  I forgot how terrible this path is.  Can anyone
> >>>>> write a
> >>>>> simple explanation of how we choose to use acpiphp or pciehp?
> >>>
> >>> In theory, that should depend on the _OSC handshake in
> >>> acpi_pci_root_add().
> >>>
> >>> If the firmware doesn't give us control of the PCIe features, we'll
> >>> not use
> >>> pciehp (or at least that's the idea).
> >>>
> >>> acpiphp is used if pciehp doesn't claim the device, AFAICS.
> >>
> >> [    4.013326] acpi PNP0A08:00: _OSC: OS supports [ExtendedConfig ASPM
> >> ClockPM Segments MSI]
> >> [    4.015860] acpi PNP0A08:00: _OSC: OS now controls [PCIeHotplug PME
> >> AER PCIeCapability]
> >>
> >> So at a glance, it would appear that pciehp *should* be claiming it,
> >> right? Something I noted in the bug I filed is that the device ID
> >> reported there is PNP0A08, and the root_device_id table that associates
> >> with acpi_pci_root_add() only includes PNP0A03 in it. Is that correct,
> >> or should 08 also be in there, which might remedy this? (I can test this
> >> out easily enough).
> >
> > Nope, makes no difference, seems those are just two different references
> > to the same bus, based on a peek at the extracted dsdt:
> >
> > Name (_HID, EisaId ("PNP0A08") /* PCI Express Bus */)  // _HID: Hardware ID
> > Name (_CID, EisaId ("PNP0A03") /* PCI Bus */)  // _CID: Compatible ID
> 
> Ah, I forgot some additional details. pciehp_probe() in 
> drivers/pci/hotplug/pciehp_core.c fails on the 
> pciehp_acpi_slot_detection_check() call for the expresscard slot, which 
> is why the base pciehp doesn't bind. DEVICE_ACPI_HANDLE(&dev->dev) in 
> the slot detection check is winding up with a NULL acpi device.

So IMO the bug is that select_detection_mode() assumes that ACPI should be
used as the PCIe hotplug detection method if it has found at least one
device that looks like an "ACPI hotplug slot" (Thuderbolt breaks that "logic").

To be honest, I'm not sure why we need the pciehp_acpi_slot_detection_check()
in pciehp_probe() at all.  It doesn't add any value as far as I can say.

If pciehp_probe() is called at all, we have registered a PCIe port service
and if this is a "hotplug" service, we wouldn't have registered it if the
_OSC handshake had not given us contol over native hotplug.

So I wonder if the patch below makes any difference.

---
 drivers/pci/hotplug/pciehp_core.c |    9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)


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

Comments

Jarod Wilson May 19, 2015, 3:06 a.m. UTC | #1
On 5/18/2015 7:06 PM, Rafael J. Wysocki wrote:
> On Monday, May 18, 2015 04:45:28 PM Jarod Wilson wrote:
...
>>>>>>> On Thu, May 14, 2015 at 03:33:58PM -0400, Jarod Wilson wrote:
>>>>>>>> The HP ZBook 15 and 17 Mobile Workstations, generation 2, up to and
>>>>>>>> including at least BIOS revision 01.07, do not have an ACPI _RMV
>>>>>>>> object
>>>>>>>> associated with their expresscard slots, so acpi-based
>>>>>>>> hotplug-capable
>>>>>>>> slot detection fails. If we fall back to pcie-based detection, the
>>>>>>>> systems
>>>>>>>> work just fine
...
>> Ah, I forgot some additional details. pciehp_probe() in
>> drivers/pci/hotplug/pciehp_core.c fails on the
>> pciehp_acpi_slot_detection_check() call for the expresscard slot, which
>> is why the base pciehp doesn't bind. DEVICE_ACPI_HANDLE(&dev->dev) in
>> the slot detection check is winding up with a NULL acpi device.
>
> So IMO the bug is that select_detection_mode() assumes that ACPI should be
> used as the PCIe hotplug detection method if it has found at least one
> device that looks like an "ACPI hotplug slot" (Thuderbolt breaks that "logic").
>
> To be honest, I'm not sure why we need the pciehp_acpi_slot_detection_check()
> in pciehp_probe() at all.  It doesn't add any value as far as I can say.
>
> If pciehp_probe() is called at all, we have registered a PCIe port service
> and if this is a "hotplug" service, we wouldn't have registered it if the
> _OSC handshake had not given us contol over native hotplug.
>
> So I wonder if the patch below makes any difference.

Yeah, that also works, for the most part. You still get spew from 
pciehp_acpi_slot_detection_init() saying "Using ACPI for slot 
detection.", but in re-reading pciehp_acpi.c in its entirety... I can't 
see anything productive that it actually does. I'm of the mind that the 
entire file should just be nuked, the path from pciehp_core.c that your 
patch alters was the only one that called 
pciehp_acpi_slot_detection_check(), and everything else is basically the 
dummy probe and fluff, nothing meaningful actually happens. I can whip 
up a follow-up patch that neuters that file entirely in the morning. At 
the very least, the "Using ACPI" bit needs to be beaten into submission, 
since its not going to be accurate.
Rafael J. Wysocki May 19, 2015, 11:36 a.m. UTC | #2
On Monday, May 18, 2015 11:06:53 PM Jarod Wilson wrote:
> On 5/18/2015 7:06 PM, Rafael J. Wysocki wrote:
> > On Monday, May 18, 2015 04:45:28 PM Jarod Wilson wrote:
> ...
> >>>>>>> On Thu, May 14, 2015 at 03:33:58PM -0400, Jarod Wilson wrote:
> >>>>>>>> The HP ZBook 15 and 17 Mobile Workstations, generation 2, up to and
> >>>>>>>> including at least BIOS revision 01.07, do not have an ACPI _RMV
> >>>>>>>> object
> >>>>>>>> associated with their expresscard slots, so acpi-based
> >>>>>>>> hotplug-capable
> >>>>>>>> slot detection fails. If we fall back to pcie-based detection, the
> >>>>>>>> systems
> >>>>>>>> work just fine
> ...
> >> Ah, I forgot some additional details. pciehp_probe() in
> >> drivers/pci/hotplug/pciehp_core.c fails on the
> >> pciehp_acpi_slot_detection_check() call for the expresscard slot, which
> >> is why the base pciehp doesn't bind. DEVICE_ACPI_HANDLE(&dev->dev) in
> >> the slot detection check is winding up with a NULL acpi device.
> >
> > So IMO the bug is that select_detection_mode() assumes that ACPI should be
> > used as the PCIe hotplug detection method if it has found at least one
> > device that looks like an "ACPI hotplug slot" (Thuderbolt breaks that "logic").
> >
> > To be honest, I'm not sure why we need the pciehp_acpi_slot_detection_check()
> > in pciehp_probe() at all.  It doesn't add any value as far as I can say.
> >
> > If pciehp_probe() is called at all, we have registered a PCIe port service
> > and if this is a "hotplug" service, we wouldn't have registered it if the
> > _OSC handshake had not given us contol over native hotplug.
> >
> > So I wonder if the patch below makes any difference.
> 
> Yeah, that also works, for the most part. You still get spew from 
> pciehp_acpi_slot_detection_init() saying "Using ACPI for slot 
> detection.", but in re-reading pciehp_acpi.c in its entirety... I can't 
> see anything productive that it actually does. I'm of the mind that the 
> entire file should just be nuked, the path from pciehp_core.c that your 
> patch alters was the only one that called 
> pciehp_acpi_slot_detection_check(), and everything else is basically the 
> dummy probe and fluff, nothing meaningful actually happens.

Right, all that is horrible horrible garbage.

> I can whip up a follow-up patch that neuters that file entirely in the morning.

Well, let me have that pleasure. :-)

Anyway, code that is only used in one place should be dropped along with its
only user.

> At the very least, the "Using ACPI" bit needs to be beaten into submission, 
> since its not going to be accurate.

An updated patch will follow this message.
diff mbox

Patch

Index: linux-pm/drivers/pci/hotplug/pciehp_core.c
===================================================================
--- linux-pm.orig/drivers/pci/hotplug/pciehp_core.c
+++ linux-pm/drivers/pci/hotplug/pciehp_core.c
@@ -248,12 +248,9 @@  static int pciehp_probe(struct pcie_devi
 	struct slot *slot;
 	u8 occupied, poweron;
 
-	if (pciehp_force)
-		dev_info(&dev->device,
-			 "Bypassing BIOS check for pciehp use on %s\n",
-			 pci_name(dev->port));
-	else if (pciehp_acpi_slot_detection_check(dev->port))
-		goto err_out_none;
+	/* If this is not a "hotplug" service, we have no business here. */
+	if (dev->service != PCIE_PORT_SERVICE_HP)
+		return -ENODEV;
 
 	if (!dev->port->subordinate) {
 		/* Can happen if we run out of bus numbers during probe */