Skip to contentSkip to conversation

AGPGart / AMD K7

DISCUSSION Preston A. Elder 3 participants 18 messages Apr 20, 2007 – Apr 21, 2007 Driver core

lore
Share & tools
lore
#0 lore raw

Hi,

I have a Tyan Thunder K7x Pro (S2469) and the amd-k7-agp module does not seem to be probing my AGP device. I have even tried putting debugging code into the amd-k7-agp module, and sure enough I can see it being loaded, but the probe function is never called. This is with kernel 2.6.19.

Here is my lspci -vvv for the relevent devices:

00:00.0 Host bridge: Advanced Micro Devices [AMD] AMD-760 MP [IGD4-2P] System Co
ntroller (rev 20)
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Step
ping- SERR- FastB2B-
Status: Cap+ 66MHz+ UDF- FastB2B- ParErr- DEVSEL=medium >TAbort-
<TAbort
- <MAbort+ >SERR- <PERR-

        Latency: 32
        Region 0: Memory at d0000000 (32-bit, prefetchable) [size=256M]
        Region 1: Memory at c2200000 (32-bit, prefetchable) [size=4K]
        Region 2: I/O ports at 1050 [disabled] [size=4]
        Capabilities: [a0] AGP version 2.0
                Status: RQ=16 Iso- ArqSz=0 Cal=0 SBA+ ITACoh- GART64-
HTrans- 64
bit- FW+ AGP3- Rate=x1,x2,x4
                Command: RQ=1 ArqSz=0 Cal=0 SBA+ AGP+ GART64- 64bit- FW-

Rate=<n
one>

00:01.0 PCI bridge: Advanced Micro Devices [AMD] AMD-760 MP [IGD4-2P] AGP Bridge
(prog-if 00 [Normal decode])
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Step
ping- SERR- FastB2B-
Status: Cap- 66MHz+ UDF- FastB2B- ParErr- DEVSEL=medium >TAbort-
<TAbort
- <MAbort- >SERR- <PERR-

        Latency: 99
        Bus: primary=00, secondary=01, subordinate=01, sec-latency=64
        I/O behind bridge: 00002000-00002fff
        Memory behind bridge: c0100000-c01fffff
        Prefetchable memory behind bridge: e0000000-efffffff
        Secondary status: 66MHz+ FastB2B- ParErr- DEVSEL=medium >TAbort-

<TAbort
- <MAbort+ <SERR- <PERR-
BridgeCtl: Parity- SERR- NoISA+ VGA+ MAbort- >Reset- FastB2B-

As you can see, the first device is indeed showing up as AGP capable and such, its just never probed (at least the probe function in amd-k7-agp is never called once the module is loaded).

To simplefy things, here is a pcitweak -l of the top two devices above:
PCI: 00:00:0: chip 1022,700c card 0000,0000 rev 20 class 06,00,00 hdr 00
PCI: 00:01:0: chip 1022,700d card 0000,0000 rev 00 class 06,04,00 hdr 01

In the amd-k7-agp code, this is in the device list:

    {
    .class      = (PCI_CLASS_BRIDGE_HOST << 8),
    .class_mask = ~0,
    .vendor     = PCI_VENDOR_ID_AMD,
    .device     = PCI_DEVICE_ID_AMD_FE_GATE_700C,
    .subvendor  = PCI_ANY_ID,
    .subdevice  = PCI_ANY_ID,
    },

Which matches the first device. So I'm completely unsure as to why this device is never probed or how to fix it.

Any help appreciated,

PreZ :)

PS - please CC me specifically on replies. I will check the list myself manually, but just in case I miss something .. :)

#1 lore raw
Quoted 4 lines · Preston A. Elder

I have a Tyan Thunder K7x Pro (S2469) and the amd-k7-agp module does not seem to be probing my AGP device. I have even tried putting debugging code into the amd-k7-agp module, and sure enough I can see it being loaded, but the probe function is never called. This is with kernel 2.6.19.

This is the second report of this I've heard, and I really have no good explanation for it.

Quoted 18 lines

As you can see, the first device is indeed showing up as AGP capable and such, its just never probed (at least the probe function in amd-k7-agp is never called once the module is loaded).

To simplefy things, here is a pcitweak -l of the top two devices above:
PCI: 00:00:0: chip 1022,700c card 0000,0000 rev 20 class 06,00,00 hdr 00
PCI: 00:01:0: chip 1022,700d card 0000,0000 rev 00 class 06,04,00 hdr 01

In the amd-k7-agp code, this is in the device list:

    {
    .class      = (PCI_CLASS_BRIDGE_HOST << 8),
    .class_mask = ~0,
    .vendor     = PCI_VENDOR_ID_AMD,
    .device     = PCI_DEVICE_ID_AMD_FE_GATE_700C,
    .subvendor  = PCI_ANY_ID,
    .subdevice  = PCI_ANY_ID,
    },
Quoted 2 lines

Which matches the first device. So I'm completely unsure as to why this device is never probed or how to fix it.

try adding some instrumentation to __pci_register_driver and the functions it calls.

oh, one thought.. do you have CONFIG_PCI_MULTITHREAD_PROBE set? I'm wondering if the probing is racing with another driver which is claiming the same PCI ID. (Edac, or watchdog for example)

Dave

Signature · 1 line
http://www.codemonkey.org.uk
#2 lore raw
Quoted 7 lines · Dave Jones

try adding some instrumentation to __pci_register_driver and the functions it calls.

oh, one thought.. do you have CONFIG_PCI_MULTITHREAD_PROBE set? I'm wondering if the probing is racing with another driver which is claiming the same PCI ID. (Edac, or watchdog for example)

Dave,

Multithread is not set.

Here is the output of my instrumentation:

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register __pci_register_driver: Error = 0 __pci_register_driver: Returning 0

The DEBUG 0 and 1 are coming from agp_amdk7_init() There is a DEBUG 2 at the top of agp_amdk7_probe(), even before pci_find_capability, but the function never gets called.

I assume an error of 0 means no conflict. Any further steps to take for this? I would really like my AGP working :S
Especially since lspci obviously detects it, even tells me it supports 1x, 2x and 4x.

PreZ :S

#3 lore raw
Quoted 11 lines · Preston A. Elder

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register __pci_register_driver: Error = 0 __pci_register_driver: Returning 0

Quoted 3 lines

The DEBUG 0 and 1 are coming from agp_amdk7_init() There is a DEBUG 2 at the top of agp_amdk7_probe(), even before pci_find_capability, but the function never gets called.

bus_add_driver() returns 0 on error, but there's a few different cases it can fail, which isn't helpful. Add some printk's to the error cases there, and see if that gives any more clues.

Dave

Signature · 1 line
http://www.codemonkey.org.uk
#4 lore raw
Quoted 17 lines · Dave Jones
Quoted 11 lines · Preston A. Elder

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register __pci_register_driver: Error = 0 __pci_register_driver: Returning 0

Quoted 3 lines

The DEBUG 0 and 1 are coming from agp_amdk7_init() There is a DEBUG 2 at the top of agp_amdk7_probe(), even before pci_find_capability, but the function never gets called.

Quoted 7 lines

bus_add_driver() returns 0 on error, but there's a few different cases it can fail, which isn't helpful. Add some printk's to the error cases there, and see if that gives any more clues.

Dave

Here you go:

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register
bus_add_driver: In Function
bus_add_driver: Before kobject_set_name (agpgart-amdk7)
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail
bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: Error = 0 __pci_register_driver: Returning 0

PreZ :)

#5 lore raw
Quoted 14 lines · Preston A. Elder
Quoted 13 lines · Dave Jones
Quoted 11 lines · Preston A. Elder

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register __pci_register_driver: Error = 0 __pci_register_driver: Returning 0

Quoted 6 lines
Quoted 6 lines
Quoted 3 lines

The DEBUG 0 and 1 are coming from agp_amdk7_init() There is a DEBUG 2 at the top of agp_amdk7_probe(), even before pci_find_capability, but the function never gets called.

bus_add_driver() returns 0 on error, but there's a few different cases it can fail, which isn't helpful.

Actually I misparsed this function, see below..

Quoted 4 lines
Quoted 3 lines

Add some printk's to
the error cases there, and see if that gives any more clues.

Here you go:

This is odd..

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

__pci_register_driver: Before Driver Register
__pci_register_driver: Error = 0
__pci_register_driver: Returning 0

That __pci_register_driver code is (presumably with your printk's added..)

        error = driver_register(&drv->driver);
	printk("Error = %d\n", error);
        if (error) {
		printk("Returning %d\n" error);
                return error;
	}

Which doesn't make much sense. If 'error' is 0, we shouldn't be taking that second printk & return. What compiler version is this?

btw Greg, wtf does driver_register return a 0 as 'success' if it completes the function, and 0 as 'failure' if !bus ? That seems doomed to failure.

Quoted 16 lines

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register
bus_add_driver: In Function
bus_add_driver: Before kobject_set_name (agpgart-amdk7)
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail

Quoted 8 lines

bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: Error = 0
__pci_register_driver: Returning 0

So we completed bus_add_driver without failing, then popped back up to __pci_register_driver and were somehow treated as
if we failed. *scratches head*

Dave

Signature · 1 line
http://www.codemonkey.org.uk
#6 lore raw
Quoted 76 lines · Dave Jones
Quoted 14 lines · Preston A. Elder
Quoted 13 lines · Dave Jones
Quoted 11 lines · Preston A. Elder

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register __pci_register_driver: Error = 0 __pci_register_driver: Returning 0

Quoted 6 lines
Quoted 6 lines
Quoted 3 lines

The DEBUG 0 and 1 are coming from agp_amdk7_init() There is a DEBUG 2 at the top of agp_amdk7_probe(), even before pci_find_capability, but the function never gets called.

bus_add_driver() returns 0 on error, but there's a few different cases it can fail, which isn't helpful.

Actually I misparsed this function, see below..

Quoted 4 lines
Quoted 3 lines

Add some printk's to
the error cases there, and see if that gives any more clues.

Here you go:

This is odd..

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

__pci_register_driver: Before Driver Register
__pci_register_driver: Error = 0
__pci_register_driver: Returning 0

That __pci_register_driver code is (presumably with your printk's added..)

        error = driver_register(&drv->driver);
	printk("Error = %d\n", error);
        if (error) {
		printk("Returning %d\n" error);
                return error;
	}

Which doesn't make much sense. If 'error' is 0, we shouldn't be taking that second printk & return. What compiler version is this?

btw Greg, wtf does driver_register return a 0 as 'success' if it completes the function, and 0 as 'failure' if !bus ? That seems doomed to failure.

Quoted 16 lines

Linux agpgart interface v0.101 (c) Dave Jones
agpgart: DEBUG 0
agpgart: DEBUG 1
__pci_register_driver: In function __pci_register_driver: driver = agpgart-amdk7, multithread = 0 __pci_register_driver: Before Spinlock __pci_register_driver: Before List Init __pci_register_driver: Before Driver Register
bus_add_driver: In Function
bus_add_driver: Before kobject_set_name (agpgart-amdk7)
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail

Quoted 8 lines

bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: Error = 0
__pci_register_driver: Returning 0

Quoted 7 lines

So we completed bus_add_driver without failing, then popped back up to __pci_register_driver and were somehow treated as
if we failed. *scratches head*

Dave

Dave,

Here is the code for __pci_register_driver:

  int __pci_register_driver(struct pci_driver *drv, struct module *owner)
  {
      int error;

  printk(KERN_INFO "__pci_register_driver: In function\n");
      /* initialize common driver fields */
      drv->driver.name = drv->name;
      drv->driver.bus = &pci_bus_type;
      drv->driver.owner = owner;
      drv->driver.kobj.ktype = &pci_driver_kobj_type;

  printk(KERN_INFO "__pci_register_driver: driver = %s, multithread = %d\n",
            drv->name, pci_multithread_probe);
      if (pci_multithread_probe)
          drv->driver.multithread_probe = pci_multithread_probe;
      else
          drv->driver.multithread_probe = drv->multithread_probe;

  printk(KERN_INFO "__pci_register_driver: Before Spinlock\n");
      spin_lock_init(&drv->dynids.lock);
  printk(KERN_INFO "__pci_register_driver: Before List Init\n");
     INIT_LIST_HEAD(&drv->dynids.list);

  printk(KERN_INFO "__pci_register_driver: Before Driver Register\n");
      /* register with core */
      error = driver_register(&drv->driver);

  printk(KERN_INFO "__pci_register_driver: Error = %d\n", error);
      if (!error)
          error = pci_create_newid_file(drv);

  printk(KERN_INFO "__pci_register_driver: Returning %d\n", error);
      return error;
  }

So in the above case, we ARE saying if driver_register returns 0 then pci_create_newid_file.

Is it different to the code you have? As I said, this IS 2.6.19.

PreZ :)

#7 lore raw
Quoted 7 lines · Preston A. Elder

Here is the code for __pci_register_driver:
...

So in the above case, we ARE saying if driver_register returns 0 then pci_create_newid_file.

Is it different to the code you have? As I said, this IS 2.6.19.

Yes, .20 changed this in this way..

@@ -445,9 +442,12 @@ int __pci_register_driver(struct pci_driver *drv, struct module *owner)

        /* register with core */
        error = driver_register(&drv->driver);

+ if (error)
+ return error;

- if (!error)
- error = pci_create_newid_file(drv);
+ error = pci_create_newid_file(drv);
+ if (error)
+ driver_unregister(&drv->driver);

        return error;

}

Retry your tracing with .20 (or better yet, .21rc7/todays git)

Dave

Signature · 1 line
http://www.codemonkey.org.uk
#8 lore raw

Dave, Greg,

Here is the trace with 2.6.20.6

I added back in my trace code, as you see. As you can also see, agp_amdk7_probe is still not called.

Linux agpgart interface v0.101 (c) Dave Jones
agp_amdk7_init: In function
agp_amdk7_init: Before pci_register_driver __pci_register_driver: In Function (driver = agpgart-amdk7, multithread = 0) __pci_register_driver: Before Spinlock
__pci_register_driver: Before Init List Head __pci_register_driver: Before driver_register
bus_add_driver: In Function (c048e920)
bus_add_driver: Before kobject_set_name
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail
bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: error = 0
__pci_register_driver: Before pci_create_newid_file __pci_register_driver: error = 0
__pci_register_driver: Returning 0

Even when I start X (using the fglrx driver) I still do not see the probe function being called.

Everything looks successful, too :(

I will try with 2.6.21rc7, but I don't hold out too much hope.

PreZ

Quoted 34 lines · Dave Jones
Quoted 7 lines · Preston A. Elder

Here is the code for __pci_register_driver:
...

So in the above case, we ARE saying if driver_register returns 0 then pci_create_newid_file.

Is it different to the code you have? As I said, this IS 2.6.19.

Yes, .20 changed this in this way..

(quoted diff) +5 -2
@@ -445,9 +442,12 @@ int __pci_register_driver(struct pci_driver *drv, struct module *owner)
 
        /* register with core */
        error = driver_register(&drv->driver);
+       if (error)
+               return error;
 
-       if (!error)
-               error = pci_create_newid_file(drv);
+       error = pci_create_newid_file(drv);
+       if (error)
+               driver_unregister(&drv->driver);
 
        return error;
 }

Retry your tracing with .20 (or better yet, .21rc7/todays git)

Dave

#9 lore raw
Quoted 6 lines · Preston A. Elder

Dave, Greg,

Here is the trace with 2.6.20.6

I added back in my trace code, as you see. As you can also see, agp_amdk7_probe is still not called.

Try looking down in __driver_attach()
The fact that we're not calling the ->probe function is quite bizarre.

It could be this in __driver_attach

        if (!dev->driver)
                driver_probe_device(drv, dev);

Though that'd be odd.

Putting a #define DEBUG 1 in drivers/base/dd.c may also yield some clues.

Dave

Signature · 1 line
http://www.codemonkey.org.uk
#10 lore raw
Quoted 13 lines · Dave Jones
Quoted 6 lines · Preston A. Elder

Dave, Greg,

Here is the trace with 2.6.20.6

I added back in my trace code, as you see. As you can also see, agp_amdk7_probe is still not called.

Try looking down in __driver_attach()
The fact that we're not calling the ->probe function is quite bizarre.

It could be this in __driver_attach

Quoted 10 lines
        if (!dev->driver)
                driver_probe_device(drv, dev);

Though that'd be odd.

Putting a #define DEBUG 1 in drivers/base/dd.c may also yield some clues.

Dave

OK, I found it!

Here is more trace:
Linux agpgart interface v0.101 (c) Dave Jones
agp_amdk7_init: In function
agp_amdk7_init: Before pci_register_driver __pci_register_driver: In Function (driver = agpgart-amdk7, multithread = 0) __pci_register_driver: Before Spinlock
__pci_register_driver: Before Init List Head __pci_register_driver: Before driver_register
bus_add_driver: In Function (c0492920)
bus_add_driver: Before kobject_set_name
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
__driver_attach (0000:00:00.0,1): Before Down (parent) (c21c8600) __driver_attach (0000:00:00.0, 1): Before Down
__driver_attach (0000:00:00.0, 1): Before Probe Device (c049fe54) __driver_attach (0000:00:00.0, 1): alreay registered with driver amd76x_edac __driver_attach (0000:00:00.0, 1): Before Up
__driver_attach (0000:00:00.0, 1): Before Up (parent) (c21c8600) __driver_attach (0000:00:00.0, 1): Returning 0
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail
bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: error = 0
__pci_register_driver: Before pci_create_newid_file __pci_register_driver: error = 0
__pci_register_driver: Returning 0

I snipped some since __driver_attach is called many times.

But the long and short is that 00:00:00 is already associated with the 'amd76x_edac' driver, and as such will not call the agp probe call. What is this edac, btw?

PreZ :)

#11 lore raw

Final followup,

If I compile EDAC out of the kernel completely, everything works now.

This should be resolved though.
1) dd.c should produce some kind of warning when it wants to assign a driver to a device, but it can't because a driver is already assigned to a device

ie. change:

      if (!dev->driver)
          driver_probe_device(drv, dev);
to:
      if (!dev->driver)
          driver_probe_device(drv, dev);
      else
          printk(KERN_WARNING "__driver_attach (%s): alreay registered
with driver %s\n",
             dev->bus_id, dev->driver->name);

2) Possibly a device should be able to have more than one driver associated with it - so the AGP driver and EDAC could both use the device in question here (though this would probably be a sizable change).

Either way, at least I found the culprit :)

PreZ :)

Quoted 73 lines · Preston A. Elder
Quoted 13 lines · Dave Jones
Quoted 6 lines · Preston A. Elder

Dave, Greg,

Here is the trace with 2.6.20.6

I added back in my trace code, as you see. As you can also see, agp_amdk7_probe is still not called.

Try looking down in __driver_attach()
The fact that we're not calling the ->probe function is quite bizarre.

It could be this in __driver_attach

Quoted 11 lines
        if (!dev->driver)
                driver_probe_device(drv, dev);

Though that'd be odd.

Putting a #define DEBUG 1 in drivers/base/dd.c may also yield some clues.

Dave

OK, I found it!

Here is more trace:
Linux agpgart interface v0.101 (c) Dave Jones
agp_amdk7_init: In function
agp_amdk7_init: Before pci_register_driver __pci_register_driver: In Function (driver = agpgart-amdk7, multithread = 0) __pci_register_driver: Before Spinlock
__pci_register_driver: Before Init List Head __pci_register_driver: Before driver_register
bus_add_driver: In Function (c0492920)
bus_add_driver: Before kobject_set_name
bus_add_driver: error = 0
bus_add_driver: Before kobject_register
bus_add_driver: error = 0
bus_add_driver: Before driver_attach
__driver_attach (0000:00:00.0,1): Before Down (parent) (c21c8600) __driver_attach (0000:00:00.0, 1): Before Down
__driver_attach (0000:00:00.0, 1): Before Probe Device (c049fe54) __driver_attach (0000:00:00.0, 1): alreay registered with driver amd76x_edac __driver_attach (0000:00:00.0, 1): Before Up
__driver_attach (0000:00:00.0, 1): Before Up (parent) (c21c8600) __driver_attach (0000:00:00.0, 1): Returning 0
bus_add_driver: error = 0
bus_add_driver: Before klist_add_tail
bus_add_driver: Before module_add_driver
bus_add_driver: Before driver_add_attrs
bus_add_driver: error = 0
bus_add_driver: Before add_bind_files
bus_add_driver: error = 0
bus_add_driver: Returning 0
__pci_register_driver: error = 0
__pci_register_driver: Before pci_create_newid_file __pci_register_driver: error = 0
__pci_register_driver: Returning 0

I snipped some since __driver_attach is called many times.

But the long and short is that 00:00:00 is already associated with the 'amd76x_edac' driver, and as such will not call the agp probe call. What is this edac, btw?

PreZ :)

#12 lore raw
Quoted 20 lines · Preston A. Elder

Final followup,

If I compile EDAC out of the kernel completely, everything works now.

This should be resolved though.
1) dd.c should produce some kind of warning when it wants to assign a driver to a device, but it can't because a driver is already assigned to a device

ie. change:

      if (!dev->driver)
          driver_probe_device(drv, dev);
to:
      if (!dev->driver)
          driver_probe_device(drv, dev);
      else
          printk(KERN_WARNING "__driver_attach (%s): alreay registered
with driver %s\n",
             dev->bus_id, dev->driver->name);
Quoted 3 lines

2) Possibly a device should be able to have more than one driver associated with it - so the AGP driver and EDAC could both use the device in question here (though this would probably be a sizable change).

I'm working on this change for PCI devices right now, but it's slow going due to some other external things (OLS paper that I am woefully behind on, etc...)

thanks,

greg k-h

#13 lore raw
In reply to Dave Jones
Quoted 11 lines · Dave Jones
Quoted 6 lines · Preston A. Elder

Dave, Greg,

Here is the trace with 2.6.20.6

I added back in my trace code, as you see. As you can also see, agp_amdk7_probe is still not called.

Try looking down in __driver_attach()
The fact that we're not calling the ->probe function is quite bizarre.

Quoted 8 lines

It could be this in __driver_attach

        if (!dev->driver)
                driver_probe_device(drv, dev);

Though that'd be odd.

Putting a #define DEBUG 1 in drivers/base/dd.c may also yield some clues.

Setting CONFIG_DEBUG_DRIVER automatically enables this also and might provide some more hints.

thanks,

greg k-h

#14 lore raw
In reply to Dave Jones
Quoted 4 lines · Dave Jones

btw Greg, wtf does driver_register return a 0 as 'success' if it completes the function, and 0 as 'failure' if !bus ? That seems doomed to failure.

I don't know why the code does that, we should always have a bus assigned to a driver. I'll change that and watch to see what breaks :)

thanks,

greg k-h

#15 lore raw
Quoted 8 lines · Greg Kroah-Hartman
Quoted 4 lines · Dave Jones

btw Greg, wtf does driver_register return a 0 as 'success' if it completes the function, and 0 as 'failure' if !bus ? That seems doomed to failure.

I don't know why the code does that, we should always have a bus assigned to a driver. I'll change that and watch to see what breaks :)

Maybe this?

We should always have a bus in bus_add_driver() Instead of returning success when we don't, BUG().

drivers/base/bus.c +1 -2
index 253868e..3ba8f1f 100644
@@ -530,8 +530,7 @@ int bus_add_driver(struct device_driver *drv)
 	struct bus_type * bus = get_bus(drv->bus);
 	int error = 0;
 
-	if (!bus)
-		return 0;
+	BUG_ON(!bus);
 
 	pr_debug("bus %s: add driver %s\n", bus->name, drv->name);
 	error = kobject_set_name(&drv->kobj, "%s", drv->name);
Signature · 1 line
http://www.codemonkey.org.uk
#16 lore raw
Quoted 7 lines · Dave Jones
Quoted 6 lines · Greg Kroah-Hartman
Quoted 4 lines · Dave Jones

btw Greg, wtf does driver_register return a 0 as 'success' if it completes the function, and 0 as 'failure' if !bus ? That seems doomed to failure.

Quoted 7 lines
Quoted 2 lines

I don't know why the code does that, we should always have a bus assigned to a driver. I'll change that and watch to see what breaks :)

Maybe this?

We should always have a bus in bus_add_driver() Instead of returning success when we don't, BUG().

Nah, I don't like adding BUG() calls to the kernel if it can be helped, how about the version I copied you on a few hours ago, which is also below?

thanks,

greg k-h

------
From: Greg Kroah-Hartman <[email protected]>
Subject: driver core: bus_add_driver should return an error if no bus

As pointed out by Dave Jones.

---
 drivers/base/bus.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/base/bus.c +1 -1
@@ -601,7 +601,7 @@ int bus_add_driver(struct device_driver 
 	int error = 0;
 
 	if (!bus)
-		return 0;
+		return -EINVAL;
 
 	pr_debug("bus %s: add driver %s\n", bus->name, drv->name);
 	error = kobject_set_name(&drv->kobj, "%s", drv->name);
#17 lore raw
Quoted 6 lines · Greg Kroah-Hartman
Quoted 2 lines

We should always have a bus in bus_add_driver() Instead of returning success when we don't, BUG().

Nah, I don't like adding BUG() calls to the kernel if it can be helped, how about the version I copied you on a few hours ago, which is also below?

Either works for me..

Dave

Signature · 1 line
http://www.codemonkey.org.uk