[PATCH] wifi: ath11k_pci: add a soft dependency on qrtr-mhi

[Baochen Qiang]

On 7/22/2025 10:37 PM, Jeff Johnson wrote:
> On 7/22/2025 1:35 AM, Baochen Qiang wrote:
>>
>>
>> On 12/2/2022 9:06 PM, Christoph Hellwig wrote:
>>> While ath11k_pci can load without qrtr-mhi, probing the actual hardware
>>> will fail when qrtr and qrtr-mhi aren't loaded with
>>>
>>>    failed to initialize qmi handle: -517
>>>
>>> Add a MODULE_SOFTDEP statement to bring the module in (and as a hint
>>> for kernel packaging) for those cases where it isn't autoloaded already
>>> for some reason.
>>>
>>> Signed-off-by: Christoph Hellwig <hch at lst.de>
>>> ---
>>>  drivers/net/wireless/ath/ath11k/pci.c | 2 ++
>>>  1 file changed, 2 insertions(+)
>>>
>>> diff --git a/drivers/net/wireless/ath/ath11k/pci.c b/drivers/net/wireless/ath/ath11k/pci.c
>>> index 99cf3357c66e16..9d58856cbf8a94 100644
>>> --- a/drivers/net/wireless/ath/ath11k/pci.c
>>> +++ b/drivers/net/wireless/ath/ath11k/pci.c
>>> @@ -1037,6 +1037,8 @@ module_exit(ath11k_pci_exit);
>>>  MODULE_DESCRIPTION("Driver support for Qualcomm Technologies 802.11ax WLAN PCIe devices");
>>>  MODULE_LICENSE("Dual BSD/GPL");
>>>  
>>> +MODULE_SOFTDEP("pre: qrtr-mhi");
>>> +
>>>  /* QCA639x 2.0 firmware files */
>>>  MODULE_FIRMWARE(ATH11K_FW_DIR "/QCA6390/hw2.0/" ATH11K_BOARD_API2_FILE);
>>>  MODULE_FIRMWARE(ATH11K_FW_DIR "/QCA6390/hw2.0/" ATH11K_AMSS_FILE);
>>
>> Do we know why this patch is rejected?
> 
> This predates me becoming a maintainer, and I don't see any comment from Kalle
> to the patch, however he did have the following comment in a separate thread
> where the issue was being discussed:
> https://lore.kernel.org/all/87cz8v2xb2.fsf@kernel.org/
> 
>> Though I am happy to take your MODULE_SOFTDEP() patch, just wondering if
>> there is a better way to solve this. For example net/mac80211 (the
>> 802.11 stack) has a lot of crypto dependencies:
>>
>> 	select CRYPTO
>> 	select CRYPTO_LIB_ARC4
>> 	select CRYPTO_AES
>> 	select CRYPTO_CCM
>> 	select CRYPTO_GCM
>> 	select CRYPTO_CMAC
>> 	select CRC32
>>
>> And it's not using MODULE_SOFTDEP() at all.
> 
> So I'm guessing he considered this to be an issue that should be solved
> external to the individual drivers.
> 

Thanks for finding the thread, it really helps.

Hmm, I agree that we should not blame this on the driver as all dependencies have been
clearly documented in the Kconfig file. However seems the userspace tool (e.g. depmod on
Ubuntu) does not care about it but just looks into ko files under /lib/modules. So if we
currently does not have a better solution, maybe we can take the MODULE_SOFTDEP way?

> /jeff