Some Broadcom based wireless devices contain dangling ethernet cores.
This triggers the ssb probing mechanism and tries to load the b44 driver
on this core.
Ignore the dangling core in the ssb core scanning code to avoid
access to the core and failure of b44 probing.
Signed-off-by: Michael Buesch <[email protected]>
Tested-by: Larry Finger <[email protected]>
---
Does not need to go into stable, because probing of that core
doesn't hurt except for failure messages in the logs.
Index: linux-2.6.37/drivers/ssb/scan.c
===================================================================
--- linux-2.6.37.orig/drivers/ssb/scan.c 2011-01-07 15:35:10.518000002 +0100
+++ linux-2.6.37/drivers/ssb/scan.c 2011-01-07 15:45:54.231998930 +0100
@@ -420,6 +420,16 @@
bus->pcicore.dev = dev;
#endif /* CONFIG_SSB_DRIVER_PCICORE */
break;
+ case SSB_DEV_ETHERNET:
+ if (bus->bustype == SSB_BUSTYPE_PCI) {
+ if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM &&
+ (bus->host_pci->device & 0xFF00) == 0x4300) {
+ /* This is a dangling ethernet core on a
+ * wireless device. Ignore it. */
+ continue;
+ }
+ }
+ break;
default:
break;
}
--
Greetings Michael.
On Wed, Nov 9, 2011 at 12:14 PM, David Woodhouse <[email protected]> wrote:
> On Fri, 2011-01-07 at 19:48 +0100, Michael B?sch wrote:
>>
>> + ? ? ? ? ? ? ? case SSB_DEV_ETHERNET:
>> + ? ? ? ? ? ? ? ? ? ? ? if (bus->bustype == SSB_BUSTYPE_PCI) {
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM &&
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? (bus->host_pci->device & 0xFF00) == 0x4300) {
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? /* This is a dangling ethernet core on a
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?* wireless device. Ignore it. */
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? continue;
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? }
>> + ? ? ? ? ? ? ? ? ? ? ? }
>> + ? ? ? ? ? ? ? ? ? ? ? break;
>
> Do you also need to check for (bus->host_pci->device / 1000) == 43?
> Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
>
> Would it be better to invert the test and check for != 0x4400?
I do not know of any Broadcom wireless device with a decimal PCI ID
(as opposed to a decimal Chip ID).
>
> --
> dwmw2
>
--
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)
On 9 November 2011 12:51, Gábor Stefanik <[email protected]> wrote:
> On Wed, Nov 9, 2011 at 12:50 PM, Gábor Stefanik <[email protected]> wrote:
>> On Wed, Nov 9, 2011 at 12:14 PM, David Woodhouse <[email protected]> wrote:
>>> On Fri, 2011-01-07 at 19:48 +0100, Michael Büsch wrote:
>>>>
>>>> + case SSB_DEV_ETHERNET:
>>>> + if (bus->bustype == SSB_BUSTYPE_PCI) {
>>>> + if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM &&
>>>> + (bus->host_pci->device & 0xFF00) == 0x4300) {
>>>> + /* This is a dangling ethernet core on a
>>>> + * wireless device. Ignore it. */
>>>> + continue;
>>>> + }
>>>> + }
>>>> + break;
>>>
>>> Do you also need to check for (bus->host_pci->device / 1000) == 43?
>>> Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
>>>
>>> Would it be better to invert the test and check for != 0x4400?
>>
>> I do not know of any Broadcom wireless device with a decimal PCI ID
>> (as opposed to a decimal Chip ID).
>
> Edit: However, 0x4700 should also be checked, as some BCM43xx chips
> use 0x47xx PCI IDs.
As far as I can tell from this snippet (I'm missing the original
message), this code is SSB, and the only 0x47xx I know of is the
BCM4313, and that's a BCMA card. So this doesn't apply here.
Same for the five digit Chip IDs (which might leak into the PCI ID, if
the card has no SPROM), AFAIK these are also BCMA exclusive.
A check for != 0x44xx would probably still the safest way.
Regards
Jonas
On Wed, Nov 9, 2011 at 12:50 PM, G?bor Stefanik <[email protected]> wrote:
> On Wed, Nov 9, 2011 at 12:14 PM, David Woodhouse <[email protected]> wrote:
>> On Fri, 2011-01-07 at 19:48 +0100, Michael B?sch wrote:
>>>
>>> + ? ? ? ? ? ? ? case SSB_DEV_ETHERNET:
>>> + ? ? ? ? ? ? ? ? ? ? ? if (bus->bustype == SSB_BUSTYPE_PCI) {
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM &&
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? (bus->host_pci->device & 0xFF00) == 0x4300) {
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? /* This is a dangling ethernet core on a
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?* wireless device. Ignore it. */
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? continue;
>>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? }
>>> + ? ? ? ? ? ? ? ? ? ? ? }
>>> + ? ? ? ? ? ? ? ? ? ? ? break;
>>
>> Do you also need to check for (bus->host_pci->device / 1000) == 43?
>> Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
>>
>> Would it be better to invert the test and check for != 0x4400?
>
> I do not know of any Broadcom wireless device with a decimal PCI ID
> (as opposed to a decimal Chip ID).
Edit: However, 0x4700 should also be checked, as some BCM43xx chips
use 0x47xx PCI IDs.
>
>>
>> --
>> dwmw2
>>
>
>
>
> --
> Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)
>
--
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)
2011/11/9 Larry Finger <[email protected]>:
> On 11/09/2011 06:16 AM, Jonas Gorski wrote:
>>
>> On 9 November 2011 12:51, Gábor Stefanik<[email protected]> wrote:
>>>
>>> On Wed, Nov 9, 2011 at 12:50 PM, Gábor Stefanik<[email protected]>
>>> wrote:
>>>>
>>>> On Wed, Nov 9, 2011 at 12:14 PM, David Woodhouse<[email protected]>
>>>> wrote:
>>>>>
>>>>> On Fri, 2011-01-07 at 19:48 +0100, Michael Büsch wrote:
>>>>>>
>>>>>> + case SSB_DEV_ETHERNET:
>>>>>> + if (bus->bustype == SSB_BUSTYPE_PCI) {
>>>>>> + if (bus->host_pci->vendor ==
>>>>>> PCI_VENDOR_ID_BROADCOM&&
>>>>>> + (bus->host_pci->device& 0xFF00)
>>>>>> == 0x4300) {
>>>>>> + /* This is a dangling ethernet
>>>>>> core on a
>>>>>> + * wireless device. Ignore it.
>>>>>> */
>>>>>> + continue;
>>>>>> + }
>>>>>> + }
>>>>>> + break;
>>>>>
>>>>> Do you also need to check for (bus->host_pci->device / 1000) == 43?
>>>>> Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
>>>>>
>>>>> Would it be better to invert the test and check for != 0x4400?
>>>>
>>>> I do not know of any Broadcom wireless device with a decimal PCI ID
>>>> (as opposed to a decimal Chip ID).
>>>
>>> Edit: However, 0x4700 should also be checked, as some BCM43xx chips
>>> use 0x47xx PCI IDs.
>>
>> As far as I can tell from this snippet (I'm missing the original
>> message), this code is SSB, and the only 0x47xx I know of is the
>> BCM4313, and that's a BCMA card. So this doesn't apply here.
>>
>> Same for the five digit Chip IDs (which might leak into the PCI ID, if
>> the card has no SPROM), AFAIK these are also BCMA exclusive.
>
> The only known card with this problem is the BCM4303, with PCI IDs
> 14e4:4301. My suspicion is that Broadcom created a chip that could be used
> for wireless or wired depending on which core was connected. Thus, it is an
> artifact of the early days. One can clean up the code as much as you want,
> but I do not believe any other chips are involved.
It's BCM4301 btw. There was incorrect entry in lspci db for some time.
b43legacy has been always detecting it as BCM4301.
--
Rafał
On Fri, 2011-01-07 at 19:48 +0100, Michael Büsch wrote:
>
> + case SSB_DEV_ETHERNET:
> + if (bus->bustype == SSB_BUSTYPE_PCI) {
> + if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM &&
> + (bus->host_pci->device & 0xFF00) == 0x4300) {
> + /* This is a dangling ethernet core on a
> + * wireless device. Ignore it. */
> + continue;
> + }
> + }
> + break;
Do you also need to check for (bus->host_pci->device / 1000) == 43?
Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
Would it be better to invert the test and check for != 0x4400?
--
dwmw2
On 11/09/2011 06:16 AM, Jonas Gorski wrote:
> On 9 November 2011 12:51, Gábor Stefanik<[email protected]> wrote:
>> On Wed, Nov 9, 2011 at 12:50 PM, Gábor Stefanik<[email protected]> wrote:
>>> On Wed, Nov 9, 2011 at 12:14 PM, David Woodhouse<[email protected]> wrote:
>>>> On Fri, 2011-01-07 at 19:48 +0100, Michael Büsch wrote:
>>>>>
>>>>> + case SSB_DEV_ETHERNET:
>>>>> + if (bus->bustype == SSB_BUSTYPE_PCI) {
>>>>> + if (bus->host_pci->vendor == PCI_VENDOR_ID_BROADCOM&&
>>>>> + (bus->host_pci->device& 0xFF00) == 0x4300) {
>>>>> + /* This is a dangling ethernet core on a
>>>>> + * wireless device. Ignore it. */
>>>>> + continue;
>>>>> + }
>>>>> + }
>>>>> + break;
>>>>
>>>> Do you also need to check for (bus->host_pci->device / 1000) == 43?
>>>> Or do the chips with 5-digit 'decimal' IDs not have the Ethernet cores?
>>>>
>>>> Would it be better to invert the test and check for != 0x4400?
>>>
>>> I do not know of any Broadcom wireless device with a decimal PCI ID
>>> (as opposed to a decimal Chip ID).
>>
>> Edit: However, 0x4700 should also be checked, as some BCM43xx chips
>> use 0x47xx PCI IDs.
>
> As far as I can tell from this snippet (I'm missing the original
> message), this code is SSB, and the only 0x47xx I know of is the
> BCM4313, and that's a BCMA card. So this doesn't apply here.
>
> Same for the five digit Chip IDs (which might leak into the PCI ID, if
> the card has no SPROM), AFAIK these are also BCMA exclusive.
The only known card with this problem is the BCM4303, with PCI IDs 14e4:4301. My
suspicion is that Broadcom created a chip that could be used for wireless or
wired depending on which core was connected. Thus, it is an artifact of the
early days. One can clean up the code as much as you want, but I do not believe
any other chips are involved.
Larry