Our bus should allow defining children nodes as we may want to specify
devices attached to the bus. This is required e.g. to specify NAND or
ChipCommon cores and use bus's address and IRQ mappings.
Signed-off-by: Rafał Miłecki <[email protected]>
---
drivers/bcma/main.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
index 9635f10..5912847 100644
--- a/drivers/bcma/main.c
+++ b/drivers/bcma/main.c
@@ -12,6 +12,7 @@
#include <linux/slab.h>
#include <linux/of_address.h>
#include <linux/of_irq.h>
+#include <linux/of_platform.h>
MODULE_DESCRIPTION("Broadcom's specific AMBA driver");
MODULE_LICENSE("GPL");
@@ -409,6 +410,13 @@ int bcma_bus_register(struct bcma_bus *bus)
bcma_core_pci_early_init(&bus->drv_pci[0]);
}
+ if (bus->host_pdev) {
+ struct device *dev = &bus->host_pdev->dev;
+
+ of_platform_populate(dev->of_node, of_default_bus_match_table,
+ NULL, dev);
+ }
+
/* Cores providing flash access go before SPROM init */
list_for_each_entry(core, &bus->cores, list) {
if (bcma_is_core_needed_early(core->id.id))
--
1.8.4.5
On 06/28/2015 05:17 PM, Rafał Miłecki wrote:
> Our bus should allow defining children nodes as we may want to specify
> devices attached to the bus. This is required e.g. to specify NAND or
> ChipCommon cores and use bus's address and IRQ mappings.
>
> Signed-off-by: Rafał Miłecki <[email protected]>
> ---
> drivers/bcma/main.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
Rafał have you solved the IRQ problem? Is it possible to assign all the
sub devices the same IRQ number as the chipcommon core?
On 25 July 2015 at 20:55, Hauke Mehrtens <[email protected]> wrote:
> On 07/24/2015 06:11 PM, Joachim Eastwood wrote:
>> On 23 July 2015 at 08:06, Kalle Valo <[email protected]> wrote:
>>> Rafał Miłecki <[email protected]> writes:
>>>
>>>>> + if (bus->host_pdev) {
>>>>> + struct device *dev = &bus->host_pdev->dev;
>>>>> +
>>>>> + of_platform_populate(dev->of_node, of_default_bus_match_table,
>>>>> + NULL, dev);
>>>>> + }
>>>>> +
>>>>
>>>> This caused a compile error when using bcma as module:
>>>> ERROR: "of_default_bus_match_table" [drivers/bcma/bcma.ko] undefined!
>>
>> I encountered the same problem with my modular memory controller driver.
>>
>>>> There are two options I guess:
>>>> 1) Export of_default_bus_match_table
>>
>> See link below for Rob Herring's take on that.
>> http://marc.info/?l=linaro-kernel&m=143766694414121&w=2
>>
>>
>> regards,
>> Joachim Eastwood
>>
> Hi Joachim,
>
> are you working on a patch which exports "of_default_bus_match_table"
> or provides a wrapper function? If not then I will look into this.
Not really, so if you could take it that would be great. Thanks.
regards,
Joachim Eastwood
Rafał Miłecki <[email protected]> writes:
> Our bus should allow defining children nodes as we may want to specify
> devices attached to the bus. This is required e.g. to specify NAND or
> ChipCommon cores and use bus's address and IRQ mappings.
>
> Signed-off-by: Rafał Miłecki <[email protected]>
Thanks, manually applied to wireless-drivers-next.git.
--
Kalle Valo
Rafał Miłecki <[email protected]> writes:
>> + if (bus->host_pdev) {
>> + struct device *dev = &bus->host_pdev->dev;
>> +
>> + of_platform_populate(dev->of_node, of_default_bus_match_table,
>> + NULL, dev);
>> + }
>> +
>
> This caused a compile error when using bcma as module:
> ERROR: "of_default_bus_match_table" [drivers/bcma/bcma.ko] undefined!
>
> There are two options I guess:
> 1) Export of_default_bus_match_table
> 2) Use some better condition like
> if (IS_BUILTIN(CONFIG_BCMA) && bus->host_pdev)
>
> Unfortunately I'm on my long holidays right now.
>
> Hauke: do you think you can find a moment to handle this?
I can cook up a patch which uses IS_BUILTIN() so that we get the build
working again. You can fix this better once you come back.
> Sorry for the problem :|
No worries, enjoy your vacation :)
--
Kalle Valo
On 23 July 2015 at 08:06, Kalle Valo <[email protected]> wrote:
> Rafał Miłecki <[email protected]> writes:
>
>>> + if (bus->host_pdev) {
>>> + struct device *dev = &bus->host_pdev->dev;
>>> +
>>> + of_platform_populate(dev->of_node, of_default_bus_match_table,
>>> + NULL, dev);
>>> + }
>>> +
>>
>> This caused a compile error when using bcma as module:
>> ERROR: "of_default_bus_match_table" [drivers/bcma/bcma.ko] undefined!
I encountered the same problem with my modular memory controller driver.
>> There are two options I guess:
>> 1) Export of_default_bus_match_table
See link below for Rob Herring's take on that.
http://marc.info/?l=linaro-kernel&m=143766694414121&w=2
regards,
Joachim Eastwood
On 28 June 2015 at 17:17, Rafał Miłecki <[email protected]> wrote:
> Our bus should allow defining children nodes as we may want to specify
> devices attached to the bus. This is required e.g. to specify NAND or
> ChipCommon cores and use bus's address and IRQ mappings.
>
> Signed-off-by: Rafał Miłecki <[email protected]>
> ---
> drivers/bcma/main.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
> index 9635f10..5912847 100644
> --- a/drivers/bcma/main.c
> +++ b/drivers/bcma/main.c
> @@ -12,6 +12,7 @@
> #include <linux/slab.h>
> #include <linux/of_address.h>
> #include <linux/of_irq.h>
> +#include <linux/of_platform.h>
>
> MODULE_DESCRIPTION("Broadcom's specific AMBA driver");
> MODULE_LICENSE("GPL");
> @@ -409,6 +410,13 @@ int bcma_bus_register(struct bcma_bus *bus)
> bcma_core_pci_early_init(&bus->drv_pci[0]);
> }
>
> + if (bus->host_pdev) {
> + struct device *dev = &bus->host_pdev->dev;
> +
> + of_platform_populate(dev->of_node, of_default_bus_match_table,
> + NULL, dev);
> + }
> +
This caused a compile error when using bcma as module:
ERROR: "of_default_bus_match_table" [drivers/bcma/bcma.ko] undefined!
There are two options I guess:
1) Export of_default_bus_match_table
2) Use some better condition like
if (IS_BUILTIN(CONFIG_BCMA) && bus->host_pdev)
Unfortunately I'm on my long holidays right now.
Hauke: do you think you can find a moment to handle this?
Sorry for the problem :|
--
Rafał
On 07/24/2015 06:11 PM, Joachim Eastwood wrote:
> On 23 July 2015 at 08:06, Kalle Valo <[email protected]> wrote:
>> Rafał Miłecki <[email protected]> writes:
>>
>>>> + if (bus->host_pdev) {
>>>> + struct device *dev = &bus->host_pdev->dev;
>>>> +
>>>> + of_platform_populate(dev->of_node, of_default_bus_match_table,
>>>> + NULL, dev);
>>>> + }
>>>> +
>>>
>>> This caused a compile error when using bcma as module:
>>> ERROR: "of_default_bus_match_table" [drivers/bcma/bcma.ko] undefined!
>
> I encountered the same problem with my modular memory controller driver.
>
>>> There are two options I guess:
>>> 1) Export of_default_bus_match_table
>
> See link below for Rob Herring's take on that.
> http://marc.info/?l=linaro-kernel&m=143766694414121&w=2
>
>
> regards,
> Joachim Eastwood
>
Hi Joachim,
are you working on a patch which exports "of_default_bus_match_table"
or provides a wrapper function? If not then I will look into this. My
current plan for bcma is just to deactivate this when it is build as a
module as a temporary workaround
Hauke
On 30 June 2015 at 23:45, Hauke Mehrtens <[email protected]> wrote:
> On 06/28/2015 05:17 PM, Rafał Miłecki wrote:
>> Our bus should allow defining children nodes as we may want to specify
>> devices attached to the bus. This is required e.g. to specify NAND or
>> ChipCommon cores and use bus's address and IRQ mappings.
>>
>> Signed-off-by: Rafał Miłecki <[email protected]>
>> ---
>> drivers/bcma/main.c | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>
> Rafał have you solved the IRQ problem? Is it possible to assign all the
> sub devices the same IRQ number as the chipcommon core?
Not yet. The only working solution I have so far is:
serial0: serial@0300 {
compatible = "ns16550";
reg = <0x0300 0x100>;
interrupt-parent = <&gic>;
interrupts = <GIC_SPI 85 IRQ_TYPE_LEVEL_HIGH>;
clock-frequency = <100000000>;
};
(plus adding compatible = "simple-bus" to the chipcommon).
--
Rafał
On 06/28/2015 05:17 PM, Rafał Miłecki wrote:
> Our bus should allow defining children nodes as we may want to specify
> devices attached to the bus. This is required e.g. to specify NAND or
> ChipCommon cores and use bus's address and IRQ mappings.
>
> Signed-off-by: Rafał Miłecki <[email protected]>
> ---
> drivers/bcma/main.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/bcma/main.c b/drivers/bcma/main.c
> index 9635f10..5912847 100644
> --- a/drivers/bcma/main.c
> +++ b/drivers/bcma/main.c
> @@ -12,6 +12,7 @@
> #include <linux/slab.h>
> #include <linux/of_address.h>
> #include <linux/of_irq.h>
> +#include <linux/of_platform.h>
>
> MODULE_DESCRIPTION("Broadcom's specific AMBA driver");
> MODULE_LICENSE("GPL");
> @@ -409,6 +410,13 @@ int bcma_bus_register(struct bcma_bus *bus)
> bcma_core_pci_early_init(&bus->drv_pci[0]);
> }
>
> + if (bus->host_pdev) {
Here is an other problem. So many bugs in this line. ;-)
host_pdev is in a union so this is also true for PCIe devices, but then
the offset of the dev structure probably is wrong. You should check for
the hosttype. I will send a patch.
> + struct device *dev = &bus->host_pdev->dev;
> +
> + of_platform_populate(dev->of_node, of_default_bus_match_table,
> + NULL, dev);
> + }
> +
> /* Cores providing flash access go before SPROM init */
> list_for_each_entry(core, &bus->cores, list) {
> if (bcma_is_core_needed_early(core->id.id))
>