Re: [PATCH v3 2/3] staging: r8188eu: Shorten calls chain of rtw_read8/16/32()

[Pavel Skripkin <paskripkin@gmail.com>]

On 9/6/21 5:07 PM, Greg Kroah-Hartman wrote:
> On Sun, Sep 05, 2021 at 12:00:47AM +0200, Fabio M. De Francesco wrote:
>> Shorten the calls chain of rtw_read8/16/32() down to the actual reads.
>> For this purpose unify the three usb_read8/16/32 into the new
>> usb_read(); make the latter parameterizable with 'size'; embed most of
>> the code of usbctrl_vendorreq() into usb_read() and use in it the new
>> usb_control_msg_recv() API of USB Core.
>> 
>> Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
>> Co-developed-by: Pavel Skripkin <paskripkin@gmail.com>
>> Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
>> Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
>> ---
>> 
>> v2->v3: No changes.
>> 
>> v1->v2: No changes.
>> 
>>  drivers/staging/r8188eu/hal/usb_ops_linux.c | 92 +++++++++++++++++----
>>  1 file changed, 78 insertions(+), 14 deletions(-)
>> 
>> diff --git a/drivers/staging/r8188eu/hal/usb_ops_linux.c b/drivers/staging/r8188eu/hal/usb_ops_linux.c
>> index a87b0d2e87d0..f9c4fd5a2c53 100644
>> --- a/drivers/staging/r8188eu/hal/usb_ops_linux.c
>> +++ b/drivers/staging/r8188eu/hal/usb_ops_linux.c
>> @@ -97,38 +97,102 @@ static int usbctrl_vendorreq(struct intf_hdl *pintfhdl, u16 value, void *pdata,
>>  	return status;
>>  }
>>  
>> +static int usb_read(struct intf_hdl *intfhdl, u32 addr, void *data, u8 size)
>> +{
>> +	u16 value = (u16)(addr & 0x0000ffff);
> 
> Why not just pass in the address as a 16bit value?
> 
> 
>> +	struct adapter *adapt = intfhdl->padapter;
>> +	struct dvobj_priv *dvobjpriv = adapter_to_dvobj(adapt);
>> +	struct usb_device *udev = dvobjpriv->pusbdev;
>> +	int status;
>> +	u8 *io_buf;
>> +	int vendorreq_times = 0;
>> +
>> +	if (adapt->bSurpriseRemoved || adapt->pwrctrlpriv.pnp_bstop_trx) {
>> +		status = -EPERM;
>> +		goto exit;
> 
> This is "interesting" to see if it's really even working as they think
> it does, but let's leave it alone for now...
> 
>> +	}
>> +
>> +	mutex_lock(&dvobjpriv->usb_vendor_req_mutex);
>> +
>> +	/*  Acquire IO memory for vendorreq */
>> +	io_buf = dvobjpriv->usb_vendor_req_buf;
>> +
>> +	if (!io_buf) {
>> +		DBG_88E("[%s] io_buf == NULL\n", __func__);
> 
> How can this buffer ever be NULL?
> 
>> +		status = -ENOMEM;
>> +		goto release_mutex;
>> +	}
> 
> Why share a buffer at all anyway?
> 
>> +	while (++vendorreq_times <= MAX_USBCTRL_VENDORREQ_TIMES) {
>> +		status = usb_control_msg_recv(udev, 0, REALTEK_USB_VENQT_CMD_REQ,
>> +					      REALTEK_USB_VENQT_READ, value,
>> +					      REALTEK_USB_VENQT_CMD_IDX, io_buf,
>> +					      size, RTW_USB_CONTROL_MSG_TIMEOUT,
>> +					      GFP_KERNEL);
>> +		if (!status) {   /*  Success this control transfer. */
> 
> Comments go on the next line.
> 
>> +			rtw_reset_continual_urb_error(dvobjpriv);
>> +			memcpy(data, io_buf, size);
>> +		} else { /*  error cases */
> 
> Again, next line for the comment.
> 
>> +			DBG_88E("reg 0x%x, usb %s %u fail, status:%d vendorreq_times:%d\n",
>> +				value, "read", size, status, vendorreq_times);
> 
> These should be removed eventually...
> 
>> +
>> +			if (status == (-ESHUTDOWN) || status == -ENODEV) {
>> +				adapt->bSurpriseRemoved = true;
> 
> Odd, but ok...
> 
>> +			} else {
>> +				struct hal_data_8188e *haldata = GET_HAL_DATA(adapt);
>> +
>> +				haldata->srestpriv.wifi_error_status = USB_VEN_REQ_CMD_FAIL;
> 
> Why are we not saying the command failed even if the device was removed?
> 
> But if we do say an error happened, why are we trying to send this out
> again?  What would happen to make it work the second time?
> 
>> +			}
>> +
>> +			if (rtw_inc_and_chk_continual_urb_error(dvobjpriv)) {
>> +				adapt->bSurpriseRemoved = true;
> 
> So we try to see if the device was removed again?
> 
>> +				break;
>> +			}
>> +		}
>> +
>> +		/*  firmware download is checksummed, don't retry */
>> +		if ((value >= FW_8188E_START_ADDRESS && value <= FW_8188E_END_ADDRESS) || !status)
>> +			break;
> 
> Nothing like a special case for firmware magic.
> 
> Those calls should just use a different write function entirely,
> eventually, to remove this...
> 
> Ok, I know you are just moving code around, this is fine, just pointing
> out things that should be fixed up eventually...
> 

I agree with all statements, and I've asked maintainers about some of 
them before. It's 100% should be fixed, but we (me and Fabio) want to 
fix them step by step to not rebase every time new clean up goes in.

Our plan is: (Fabio, please, correct me, if I am wrong here)

1. ops removal + shorten call chain (this patch series)
2. my error handling series [1]
3. Clean ups for rtw_* functions (old usb_*)
4. Remove dead proc code + introduce new sysfs interface, based on old 
proc code.


We can prepare them all at once, but, IMO:

1. It will be really hard to review
2. A lot of rebasing through new versions, since there are a lot
clean ups appear every day.


I believe, our step-by-step plan should be more comfortable for all of us :)


Thank you!


[1] 
https://lore.kernel.org/linux-staging/cover.1629479152.git.paskripkin@gmail.com/



With regards,
Pavel Skripkin