Hi Hillf,

On 5/22/22 07:15, Hillf Danton wrote:
> 
> In the call chain below
> 
> 	ath9k_hif_usb_firmware_cb()
> 	  ath9k_htc_hw_alloc()
> 	  ath9k_hif_usb_dev_init()
> 	  ret = ath9k_htc_hw_init()
> 	          ath9k_htc_probe_device()
> 		    htc_handle->drv_priv = priv;
> 		    ret = ath9k_htc_wait_for_target(priv);
> 		    if (ret)
> 			goto err_free;
> 	  if (ret)
> 		goto err_htc_hw_init;
> 
> err_free:
> 	ieee80211_free_hw(hw);
> 
> 
> err_htc_hw_init:
> 	ath9k_hif_usb_dev_deinit(hif_dev);
> 	  ath9k_hif_usb_dealloc_urbs()
> err_dev_init:
> 	ath9k_htc_hw_free(hif_dev->htc_handle);
> err_dev_alloc:
> 	release_firmware(fw);
> err_fw:
> 	ath9k_hif_usb_firmware_fail(hif_dev);
> 
> 
> hw should survive deallocating urbs, and changes should be added instead to
> the rollback in ath9k_htc_probe_device() by deferring cleanup of hw to its
> callsite in addition to urbs.
> 

Don't get it, sorry. I am not changing the life time of `hw`, I am just 
deferring htc_handle->drv_priv initialization.




With regards,
Pavel Skripkin