On 12/15/2016 05:40 PM, Valo, Kalle wrote:

Erik Stromdahl [off-list ref] writes:

quoted

Initial HIF sdio/mailbox implementation.

Signed-off-by: Erik Stromdahl <redacted>
---
 drivers/net/wireless/ath/ath10k/Kconfig  |    6 +
 drivers/net/wireless/ath/ath10k/Makefile |    3 +
 drivers/net/wireless/ath/ath10k/sdio.c   | 1860 ++++++++++++++++++++++++++++++
 drivers/net/wireless/ath/ath10k/sdio.h   |  276 +++++
 4 files changed, 2145 insertions(+)
 create mode 100644 drivers/net/wireless/ath/ath10k/sdio.c
 create mode 100644 drivers/net/wireless/ath/ath10k/sdio.h

AFAIK the sdio firmware binary is different from pci and usb. And as all
the firmware images need to coexist in the same repository I suspect we
can't continue to use firmware-N.bin for both pcie and sdio (and in
future usb). So should we somehow rename the sdio firmware filename to
something else, like firmware-sdio-N.bin?

I suppose you are right. I would be surprised if the firmware images
were the same for sdio, usb and pci. Hopefully all sdio chipsets sharing
the same device id and BMI version will use the same firmware image.
Currently, these two params are the only two that are used when
selecting firmware.

struct bmi_target_info also has a type parameter, but I can't see it
being used anywhere in the code.

If it turns out that several sdio chipsets uses the same BMI version and
device id, this parameter might be used (together with the other two) to
select hw params.