Message-ID: <52EB818C.1070007@redpinesignals.com> (sfid-20140131_115718_198234_BD1876EA)
Date: Fri, 31 Jan 2014 16:27:16 +0530
From: Jahnavi Meher <jahnavi.meher@redpinesignals.com>
MIME-Version: 1.0
To: Dan Williams <dcbw@redhat.com>
CC: linux-wireless@vger.kernel.org
Subject: Re: [PATCH 3.13.1 4/9] rsi: USB functionality
References: <52EA75C9.5050005@redpinesignals.com> <1391111311.698.60.camel@dcbw.local>
In-Reply-To: <1391111311.698.60.camel@dcbw.local>
Content-Type: text/plain; charset=UTF-8
Sender: linux-wireless-owner@vger.kernel.org

Hi Dan,
Thank you for your review comments. Patches 5 and 8 did
not appear on patchwork.kernel.org, so I have sent them
again. We will make relevant changes going by your comments.
The driver maintains 5 queues, 4 for data and 1 for 
mgmt. Depending on the type of packet, data or mgmt, the 
packet is sent to the endpoint.
Will make the reset card changes and read the f/w version
from the device. The firmware used is the same for both 
SDIO and USB, is not bus dependent.

Regards,
Jahnavi

On 01/31/2014 01:18 AM, Dan Williams wrote:
> On Thu, 2014-01-30 at 21:24 +0530, Jahnavi wrote:
>> From: Jahnavi Meher <jahnavi.meher@redpinesignals.com>
>>
>> This file adds the rsi_usb module, enabling the USB interface for
>> the 91x chipsets.
>>
>> Signed-off-by: Jahnavi Meher <jahnavi.meher@redpinesignals.com>
>> ---
> In general the code is very clean and mostly follows kernel style.  It's
> also fairly easy to read.  That's great.
>
> However, I've got some architecture comments below...
>
>> rsi_usb.c |  642 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 642 insertions(+)
>>
>> diff -uprN a/drivers/net/wireless/rsi/common/rsi_usb.c b/drivers/net/wireless/rsi/common/rsi_usb.c
>> --- a/drivers/net/wireless/rsi/common/rsi_usb.c	1970-01-01 05:30:00.000000000 +0530
>> +++ b/drivers/net/wireless/rsi/common/rsi_usb.c	2014-01-30 16:15:14.126308228 +0530
>> @@ -0,0 +1,642 @@
>> +/**
>> + * @file rsi_usb.c
>> + * @author
>> + * @version 1.0
>> + *
>> + * @section LICENSE
>> + * Copyright (c) 2013 Redpine Signals Inc.
>> + *
>> + * Permission to use, copy, modify, and/or distribute this software for any
>> + * purpose with or without fee is hereby granted, provided that the above
>> + * copyright notice and this permission notice appear in all copies.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
>> + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
>> + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
>> + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
>> + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
>> + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
>> + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
>> + *
>> + * @section DESCRIPTION
>> + *
>> + * The file contains Generic HAL layer for USB.
>> + */
>> +
>> +#include "../include/rsi_main.h"
>> +#include "../include/rsi_hw_intf.h"
>> +#include "../include/rsi_device_ops.h"
>> +
>> +static unsigned short fw;
> Why does 'fw' need to be static?  It's only used in one function, and
> it's read from the device every time.
>
>> +static unsigned char reset_card = 1;
> See my notes below about reset_card; why is this a static global for the
> file?  First, it prevents using multiple USB devices with the driver.
> Second, what does it actually prevent and is that actually the best way
> to do that?
>
>> +/**
>> + * This function writes to the USB Card.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @param  buf Pointer to the buffer from where the data has to be taken.
>> + * @param  len Length to be written.
>> + * @param  endpoint Type of endpoint.
>> + * @return status: 0 on success, -1 on failure.
>> + */
>> +static int rsi_usb_card_write(struct rsi_hw *adapter,
>> +			      void *buf,
>> +			      unsigned short len,
>> +			      unsigned char endpoint)
> Use types like u16, u8, and u32 where the types are not
> endian-dependent.  Use types like __le32 or __be32 or __le16 or __be16
> where you interface with firmware, which is endian-dependent.  That way
> you can run 'sparse' and find out where your driver may have mismatching
> endian problems between host and the device.
>
>> +{
>> +	int status;
>> +	int transfer;
>> +
>> +	status = usb_bulk_msg(adapter->usbdev,
>> +			      usb_sndbulkpipe(adapter->usbdev,
>> +			      adapter->bulkout_endpoint_addr[endpoint - 1]),
>> +			      buf,
>> +			      len,
>> +			      &transfer,
>> +			      HZ * 5);
>> +
>> +	if (status < 0) {
>> +		rsi_dbg(ERR_ZONE,
>> +			"Card write failed with error code :%10d\n", status);
>> +		adapter->write_fail = 1;
>> +	}
>> +	return status;
>> +}
>> +
>> +/**
>> + * This function writes multiple bytes of information to the USB card.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @param  addr Address of the register.
>> + * @param  data Pointer to the data that has to be written.
>> + * @param  count Number of multiple bytes to be written.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +static int rsi_write_multiple(struct rsi_hw *adapter,
>> +			      unsigned int addr,
>> +			      unsigned char *data,
>> +			      unsigned int count)
> Use u32, u8, etc.
>> +{
>> +	if ((adapter == NULL) || (adapter->write_fail) || (addr == 0)) {
>> +		return 0;
> Since this is returning, no need to put the rest of the code below into
> an indented block.
>
>> +	} else {
>> +		unsigned char *seg = adapter->tx_buffer;
>> +
>> +		if (addr == 1) {
> Why is addr = 1 special?  I see later on that it's determined from the
> queue that the packet is supposed to be in.  But rsi_write_multiple()
> actually calls rsi_usb_card_write() and passes 'addr' as the USB
> endpoint?
>
> So correct me if I'm wrong, but what really *is* happening here is that
> the driver is doing WMM internally and keeps a management queue and a
> data queue.  Then the USB bus layer inspects the host->device packet and
> send that packet to a specific USB endpoint depending on whether it's a
> mgmt frame or a data frame?
>
> This should really be more explicit.  Or maybe just add a new parameter
> to host_intf_write_pkt like 'bool mgmt' or something?  Or, if Johannes
> has a better suggestion for cleaning up the queue stuff, I'd defer to
> that.
>
>> +			memset(seg, 0, RSI_USB_TX_HEAD_ROOM);
>> +			memcpy(seg + RSI_USB_TX_HEAD_ROOM, data, count);
>> +		} else {
>> +			seg = ((unsigned char *)data - RSI_USB_TX_HEAD_ROOM);
>> +		}
>> +		return rsi_usb_card_write(adapter,
>> +					  seg,
>> +					  count + RSI_USB_TX_HEAD_ROOM,
>> +					  addr);
>> +	}
>> +}
>> +
>> +/**
>> + * This function initializes the bulk endpoints to the device.
>> + *
>> + * @param  interface Pointer to the USB interface structure.
>> + * @param  adapter Pointer to the adapter structure.
>> + * @return ret_val: 0 on success, -ENOMEM on failure.
>> + */
>> +static int rsi_find_bulk_in_and_out_endpoints(struct usb_interface *interface,
>> +					      struct rsi_hw *adapter)
>> +{
>> +	struct usb_host_interface *iface_desc;
>> +	struct usb_endpoint_descriptor *endpoint;
>> +	int buffer_size;
>> +	int ii, bep_found = 0;
>> +
>> +	if ((interface == NULL) || (adapter == NULL))
>> +		return -1;
>> +
>> +	iface_desc = &(interface->altsetting[0]);
>> +
>> +	for (ii = 0; ii < iface_desc->desc.bNumEndpoints; ++ii) {
>> +		endpoint = &(iface_desc->endpoint[ii].desc);
>> +
>> +		if ((!(adapter->bulkin_endpoint_addr)) &&
>> +		    (endpoint->bEndpointAddress & USB_DIR_IN) &&
>> +		    ((endpoint->bmAttributes &
>> +		    USB_ENDPOINT_XFERTYPE_MASK) ==
>> +		    USB_ENDPOINT_XFER_BULK)) {
>> +			buffer_size = endpoint->wMaxPacketSize;
>> +			adapter->bulkin_size = buffer_size;
>> +			adapter->bulkin_endpoint_addr =
>> +				endpoint->bEndpointAddress;
>> +		}
>> +
>> +		if (!adapter->bulkout_endpoint_addr[bep_found] &&
>> +		    !(endpoint->bEndpointAddress & USB_DIR_IN) &&
>> +		    ((endpoint->bmAttributes & USB_ENDPOINT_XFERTYPE_MASK) ==
>> +		      USB_ENDPOINT_XFER_BULK)) {
>> +			adapter->bulkout_endpoint_addr[bep_found] =
>> +				endpoint->bEndpointAddress;
>> +			buffer_size = endpoint->wMaxPacketSize;
>> +			adapter->bulkout_size[bep_found] = buffer_size;
>> +			bep_found++;
>> +		}
>> +
>> +		if (bep_found >= MAX_BULK_EP)
>> +			break;
>> +	}
>> +
>> +	if (!(adapter->bulkin_endpoint_addr) &&
>> +	    (adapter->bulkout_endpoint_addr[0]))
>> +		return -1;
>> +
>> +	return 0;
>> +}
>> +
>> +/* This function reads the data from given register address.
>> + *
>> + * @param  usbdev Pointer to the usb_device structure.
>> + * @param  reg Address of the register to be read.
>> + * @param  value Value to be read.
>> + * @param  len length of data to be read.
>> + * @return status: 0 on success, -1 on failure.
>> + */
>> +static int rsi_usb_reg_read(struct usb_device *usbdev,
>> +			    unsigned int reg,
> u32 perhaps?
>
>> +			    unsigned short *value,
>> +			    unsigned short len)
> u16 instead of unsigned short.
>
>> +{
>> +	unsigned char temp_buf[4];
> Use 'u8' instead of unsigned char.
>
>> +	int status = 0;
>> +
>> +	len = 2;/* FIXME */
> Every caller of rsi_usb_reg_read() already passed len = 2, so why is
> this function overriding the length the callers want?
>
>> +
>> +	status = usb_control_msg(usbdev,
>> +				 usb_rcvctrlpipe(usbdev, 0),
>> +				 USB_VENDOR_REGISTER_READ,
>> +				 USB_TYPE_VENDOR,
>> +				 ((reg & 0xffff0000) >> 16), (reg & 0xffff),
>> +				 (void *)temp_buf,
>> +				 len,
>> +				 HZ * 5);
>> +
>> +	*value = (temp_buf[0] | (temp_buf[1] << 8));
>> +	if (status < 0) {
>> +		rsi_dbg(ERR_ZONE,
>> +			"%s: Reg read failed with error code :%d\n",
>> +			__func__, status);
>> +	}
>> +	return status;
>> +}
>> +
>> +/**
>> + * This function writes the given data into the given register address.
>> + *
>> + * @param  usbdev Pointer to the usb_device structure.
>> + * @param  reg Address of the register.
>> + * @param  value Value to write.
>> + * @param  len Length of data to be written.
>> + * @return status: 0 on success, -1 on failure.
>> + */
>> +static int rsi_usb_reg_write(struct usb_device *usbdev,
>> +			     unsigned int reg,
>> +			     unsigned short value,
>> +			     unsigned short len)
>> +{
>> +	unsigned char usb_reg_buf[4];
> u8
>
>> +	int status = 0;
>> +
>> +	usb_reg_buf[0] = (value & 0x00ff);
>> +	usb_reg_buf[1] = (value & 0xff00) >> 8;
>> +	usb_reg_buf[2] = 0x0;
>> +	usb_reg_buf[3] = 0x0;
>> +
>> +	status = usb_control_msg(usbdev,
>> +				 usb_sndctrlpipe(usbdev, 0),
>> +				 USB_VENDOR_REGISTER_WRITE,
>> +				 USB_TYPE_VENDOR,
>> +				 ((reg & 0xffff0000) >> 16),
>> +				 (reg & 0xffff),
>> +				 (void *)usb_reg_buf,
>> +				 len,
>> +				 HZ * 5);
>> +	if (status < 0) {
>> +		rsi_dbg(ERR_ZONE,
>> +			"%s: Reg write failed with error code :%d\n",
>> +			__func__, status);
>> +	}
>> +	return status;
>> +}
>> +
>> +/**
>> + * This function is called when a packet is received from USB
>> + * the stack. This is a callback to recieve done.
>> + *
>> + * @param  urb Received URB.
>> + * @return None.
>> + */
>> +static void rsi_rx_done_handler(struct urb *urb)
>> +{
>> +	struct rsi_hw *adapter = urb->context;
>> +	struct rsi_common *common = adapter->priv;
>> +
>> +	if (urb->status)
>> +		return;
>> +
>> +	adapter->total_usb_rx_urbs_done++;
> I can't find where this is used, but for whatever reason I don't see
> patches 5, 6, or 8 in my mailbox.  If it's not used, it should be
> removed.
>
>> +	rsi_set_event(&common->rx_event);
>> +	return;
>> +}
>> +
>> +/**
>> + * This function submits the given URB to the USB stack.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +int rsi_rx_urb_submit(struct rsi_hw *adapter)
>> +{
>> +	struct urb *urb = adapter->rx_usb_urb[0];
>> +
>> +	adapter->total_usb_rx_urbs_submitted++;
>> +	usb_fill_bulk_urb(urb,
>> +			  adapter->usbdev,
>> +			  usb_rcvbulkpipe(adapter->usbdev,
>> +				adapter->bulkin_endpoint_addr),
>> +			  urb->transfer_buffer,
>> +			  3000,
>> +			  rsi_rx_done_handler,
>> +			  adapter);
>> +	if (usb_submit_urb(urb, GFP_KERNEL)) {
>> +		rsi_dbg(ERR_ZONE, "%s: Failed in urb submission\n", __func__);
>> +		return -1;
>> +	} else {
>> +		return 0;
>> +	}
> This would typically be written as:
>
> if (usb_submit_urb(..)) {
>    rsi_dbg();
>    return -1;
> }
>
> return 0;
>
>> +}
>> +
>> +/**
>> + * This function writes multiple bytes of information to multiple registers.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @param  addr Address of the register.
>> + * @param  data Pointer to the data that has to be written.
>> + * @param  count Number of multiple bytes to be written on to the registers.
>> + * @return status: 0 on success, -1 on failure.
>> + */
>> +int rsi_write_ta_register_multiple(struct rsi_hw *adapter,
>> +				   unsigned int addr,
>> +				   unsigned char *data,
>> +				   unsigned int count)
> u32, u8, etc, you get the point.
>
>> +{
>> +	unsigned char *buf;
>> +	unsigned char transfer;
>> +	int status = 0;
>> +
>> +	buf = kzalloc(4096, GFP_KERNEL);
>> +	if (!buf)
>> +		return -ENOMEM;
>> +
>> +	while (count) {
>> +		transfer = min_t(int, count, 4096);
>> +		memcpy(buf, data, transfer);
>> +		status = usb_control_msg(adapter->usbdev,
>> +					 usb_sndctrlpipe(adapter->usbdev, 0),
>> +					 USB_VENDOR_REGISTER_WRITE,
>> +					 USB_TYPE_VENDOR,
>> +					 ((addr & 0xffff0000) >> 16),
>> +					 (addr & 0xffff),
>> +					 (void *)buf,
>> +					 transfer,
>> +					 HZ * 5);
>> +		if (status < 0) {
>> +			rsi_dbg(ERR_ZONE,
>> +				"Reg write failed with error code :%d\n",
>> +				status);
>> +		} else {
>> +			count -= transfer;
>> +			data += transfer;
>> +			addr += transfer;
>> +		}
>> +	}
>> +
>> +	kfree(buf);
>> +	return 0;
>> +}
>> +
>> +/**
>> + * This function writes the packet to the USB card.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @param  pkt Pointer to the data to be written on to the card.
>> + * @param  len Length of the data to be written on to the card.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +int rsi_write_pkt(struct rsi_hw *adapter,
>> +		  unsigned char *pkt,
>> +		  unsigned int len)
>> +{
>> +	unsigned int block_size = adapter->tx_blk_size;
>> +	unsigned int num_blocks, address;
>> +	unsigned int queueno = ((pkt[1] >> 4) & 0xf);
>> +
>> +	if ((!len) && (queueno == RSI_WIFI_DATA_Q)) {
>> +		rsi_dbg(ERR_ZONE, "%s: Wrong length\n", __func__);
>> +		return -1;
>> +	}
> See comments about queues and endpoints above.  But really, if it's an
> error to send a zero-length data packet, then perhaps it should be
> BUG_ON() since the driver should never, ever do that?
>
>> +	num_blocks = (len / block_size);
>> +
>> +	if (len % block_size)
>> +		num_blocks++;
> num_blocks doesn't get used anywhere, should be removed.
>
>> +	if (((*(unsigned short *)pkt) < 14) && (queueno == RSI_WIFI_DATA_Q)) {
>> +		rsi_dbg(ERR_ZONE, "%s: Too small pkt\n", __func__);
>> +		return -1;
>> +	}
>> +
>> +	address = ((queueno == RSI_WIFI_MGMT_Q) ? 1 : 2);
>> +
>> +	return rsi_write_multiple(adapter,
>> +				  address,
>> +				  (unsigned char *)pkt,
>> +				  len);
>> +}
>> +
>> +/**
>> + * This function initializes the usb interface.
>> + *
>> + * @param  adapter Pointer to the adapter structure.
>> + * @param  pfunction Pointer to USB interface structure.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +static int rsi_init_usb_interface(struct rsi_hw *adapter,
>> +				  struct usb_interface *pfunction)
>> +{
>> +	adapter->usbdev = interface_to_usbdev(pfunction);
>> +
>> +	if (rsi_find_bulk_in_and_out_endpoints(pfunction, adapter))
>> +		return -1;
>> +
>> +	adapter->device = &pfunction->dev;
>> +	usb_set_intfdata(pfunction, adapter);
>> +
>> +	adapter->tx_buffer = kmalloc(2048, GFP_ATOMIC);
>> +	adapter->rx_usb_urb[0] = usb_alloc_urb(0, GFP_KERNEL);
>> +	adapter->rx_usb_urb[0]->transfer_buffer = adapter->priv->rx_data_pkt;
>> +	adapter->tx_blk_size = 252;
>> +
>> +	rsi_dbg(INIT_ZONE, "%s: Enabled the interface\n", __func__);
>> +	return 0;
>> +}
>> +
>> +/**
>> + * This function writes into various registers to do USB reset.
>> + *
>> + * @param  usbdev Pointer to the usb_device structure.
>> + * @param  addr Address to write to.
>> + * @param  data Data to be written onto the address.
>> + * @param  len_in_bits Length in bits.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +static int rsi_ulp_read_write(struct usb_device *usbdev,
>> +			      unsigned short addr,
>> +			      unsigned short *data,
>> +			      unsigned short len_in_bits)
>> +{
>> +	unsigned short buffer = 0;
>> +
>> +	if (rsi_usb_reg_write(usbdev,
>> +			      GSPI_DATA_REG2,
>> +			      *(unsigned short *)&data[2], 2) < 0)
>> +		return -1;
>> +
>> +	if (rsi_usb_reg_write(usbdev,
>> +			      GSPI_DATA_REG1,
>> +			      ((addr << 6) | (data[1] & 0x3f)),
>> +			      2) < 0)
>> +		return -1;
>> +
>> +	if (rsi_usb_reg_write(usbdev,
>> +			      GSPI_DATA_REG0,
>> +			      *(unsigned short *)&data[0],
>> +			      2) < 0)
>> +		return -1;
>> +
>> +	if (rsi_usb_reg_read(usbdev, GSPI_CTRL_REG0, &buffer, 2) < 0)
>> +		return -1;
>> +
>> +	buffer  &= ~GSPI_2_ULP;
>> +
>> +	if (rsi_usb_reg_write(usbdev, GSPI_CTRL_REG0, buffer, 2) < 0)
>> +		return -1;
>> +
>> +	if (rsi_usb_reg_write(usbdev,
>> +			      GSPI_CTRL_REG1,
>> +			      ((len_in_bits - 1) | GSPI_TRIG),
>> +			      2) < 0)
>> +		return -1;
>> +
>> +	if (rsi_usb_reg_read(usbdev, GSPI_CTRL_REG0, &buffer, 2) < 0)
>> +		return -1;
>> +
>> +	buffer |= GSPI_2_ULP;
>> +
>> +	if (rsi_usb_reg_write(usbdev, GSPI_CTRL_REG0, buffer, 2) < 0)
>> +		return -1;
>> +	return 0;
>> +}
>> +
>> +/**
>> + * This function resets the USB card.
>> + *
>> + * @param  pfunction Pointer to the USB interface structure.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +static int rsi_reset_card(struct usb_interface *pfunction)
>> +{
>> +	struct usb_device *usbdev = interface_to_usbdev(pfunction);
>> +	unsigned short temp[4];
>> +
>> +	memset(temp, 0, sizeof(temp));
>> +	*(unsigned int *)temp = 10;
>> +	if (rsi_ulp_read_write(usbdev, WATCH_DOG_TIMER_1, &temp[0], 32) < 0)
>> +		return -1;
>> +
>> +	*(unsigned int *)temp = 0;
>> +	if (rsi_ulp_read_write(usbdev, WATCH_DOG_TIMER_2, temp, 32) < 0)
>> +		return -1;
>> +
>> +	*(unsigned int *)temp = 1;
>> +	if (rsi_ulp_read_write(usbdev, WATCH_DOG_DELAY_TIMER_1, temp, 32) < 0)
>> +		return -1;
>> +	*(unsigned int *)temp = 0;
>> +	if (rsi_ulp_read_write(usbdev, WATCH_DOG_DELAY_TIMER_2, temp, 32) < 0)
>> +		return -1;
>> +
>> +	*(unsigned int *)temp = ((0xaa000) | RESTART_WDT | BYPASS_ULP_ON_WDT);
>> +	if (rsi_ulp_read_write(usbdev, WATCH_DOG_TIMER_ENABLE, temp, 32) < 0)
>> +		return -1;
>> +
>> +	return 0;
>> +}
>> +
>> +/**
>> + * This function is called by kernel when the driver provided
>> + * Vendor and device IDs are matched. All the initialization
>> + * work is done here.
>> + *
>> + * @param  pfunction Pointer to the USB interface structure.
>> + * @param  id Pointer to the usb_device_id structure.
>> + * @return 0 on success, -1 on failure.
>> + */
>> +static int rsi_probe(struct usb_interface *pfunction,
>> +		     const struct usb_device_id *id)
>> +{
>> +	struct rsi_hw *adapter;
>> +
>> +	rsi_dbg(INIT_ZONE, "%s: Init function called\n", __func__);
>> +
>> +	if (reset_card) {
>> +		rsi_dbg(INIT_ZONE, "%s: Resetting card here\n", __func__);
>> +		reset_card = 0;
>> +		if (rsi_reset_card(pfunction) < 0) {
>> +			rsi_dbg(ERR_ZONE, "%s: Card reset failed\n", __func__);
>> +			return 1;
>> +		} else {
>> +			return 0;
>> +		}
>> +	}
> Ok, reset_card... The first time a card is probed by the module, it
> resets the device.  Then, if on any further probe or if the card fails
> to initialize, reset_card is set to 1, and then the *next* time probe()
> is called the card is reset?
>
> That's quite hacktastic.  If the driver always wants to ensure clean
> firmware state when it loads, it should reset the device any time
> firmware is loaded and alive to clear that firmware and get some new
> firmware.
>
> If there are any errors, just reset the card.  Since any reset of the
> card (presumably!) causes the device to drop off the bus, probe() will
> simply get re-run when the firmware reboots.  No need to dance around
> with reset_card across probe()s.
>
>> +	adapter = rsi_init_os_intf_ops();
>> +	if (!adapter) {
>> +		rsi_dbg(ERR_ZONE, "%s: Failed to init os intf ops\n",
>> +			__func__);
>> +		return 1;
>> +	}
>> +
>> +	if (rsi_init_usb_interface(adapter, pfunction)) {
>> +		rsi_dbg(ERR_ZONE, "%s: Failed to init usb interface\n",
>> +			__func__);
>> +		goto fail;
>> +	}
>> +
>> +	rsi_dbg(ERR_ZONE, "%s: Initialized os intf ops\n", __func__);
>> +
>> +	if (rsi_usb_reg_read(adapter->usbdev, 0x41050012, &fw, 2) < 0)
> There should really be a #define for this magic number.
>
>> +		goto fail;
>> +	else
>> +		fw &= 1;
> There should be some comments about 'fw' here, what does the returned
> value from rsi_usb_reg_read() mean?  Does it mean the firmware is alive?
> If so, then it should be named something more descriptive.
>
>> +
>> +	if (rsi_device_init(adapter->priv, fw)) {
> And it was a bit unclear to track down this usage, where passing
> anything > 0 to rsi_device_init() really just causes
> rsi_load_ta_instructions() to fill in a common firmware version struct
> from the on-disk firmware file (but not running firmware?), and at least
> for USB, not to attempt to send more firmware to the device.
>
>> +		rsi_dbg(ERR_ZONE, "%s: Failed in device init\n",
>> +			__func__);
>> +		goto fail;
>> +	}
>> +
>> +	if (!fw)  {
>> +		if (rsi_usb_reg_write(adapter->usbdev, 0x25000, 0xab, 1) < 0)
> More magic numbers, these should really have defines.
>
>> +			goto fail;
>> +		rsi_dbg(INIT_ZONE, "%s: Performed device init\n", __func__);
>> +	} else {
>> +		reset_card = 1;
>> +	}
>> +
>> +	if (rsi_rx_urb_submit(adapter))
>> +		goto fail;
>> +
>> +	return 0;
>> +fail:
>> +	reset_card = 1;
>> +	rsi_deinit_os_intf_ops(adapter);
>> +	rsi_dbg(ERR_ZONE, "%s: Failed in probe...Exiting\n", __func__);
>> +	return 1;
>> +}
>> +
>> +/**
>> + * This function performs the reverse of the probe function,
>> + * it deintialize the driver structure.
>> + *
>> + * @param  pfunction Pointer to the USB interface structure.
>> + * @return None.
>> + */
>> +static void rsi_disconnect(struct usb_interface *pfunction)
>> +{
>> +	struct rsi_hw *adapter = usb_get_intfdata(pfunction);
>> +
>> +	if (!adapter)
>> +		return;
>> +
>> +	rsi_device_deinit(adapter);
>> +	rsi_deinit_os_intf_ops(adapter);
>> +
>> +	rsi_dbg(INFO_ZONE, "%s: Deinitialization completed\n", __func__);
>> +	return;
>> +}
>> +
>> +#ifdef CONFIG_PM
>> +static int rsi_suspend(struct usb_interface *intf, pm_message_t message)
>> +{
>> +	/* Not yet implemented */
>> +	return -ENOSYS;
>> +}
>> +
>> +static int rsi_resume(struct usb_interface *intf)
>> +{
>> +	/* Not yet implemented */
>> +	return -ENOSYS;
>> +}
>> +#endif
>> +
>> +static const struct usb_device_id rsi_dev_table[] = {
>> +	{ USB_DEVICE(0x0303, 0x0100) },
>> +	{ USB_DEVICE(0x041B, 0x0301) },
>> +	{ USB_DEVICE(0x041B, 0x0201) },
>> +	{ USB_DEVICE(0x041B, 0x9330) },
>> +	{ /* Blank */},
>> +};
>> +
>> +static struct usb_driver rsi_driver = {
>> +	.name       = "RSI-USB WLAN",
>> +	.probe      = rsi_probe,
>> +	.disconnect = rsi_disconnect,
>> +	.id_table   = rsi_dev_table,
>> +#ifdef CONFIG_PM
>> +	.suspend    = rsi_suspend,
>> +	.resume     = rsi_resume,
>> +#endif
>> +};
>> +
>> +/**
>> + * This function registers the client driver.
>> + *
>> + * @param  Void.
>> + * @return 0 on success.
>> + */
>> +static int rsi_module_init(void)
>> +{
>> +	usb_register(&rsi_driver);
>> +	rsi_dbg(INIT_ZONE, "%s: Registering driver\n", __func__);
>> +	return 0;
>> +}
>> +
>> +/**
>> + * This function unregisters the client driver.
>> + *
>> + * @param  Void.
>> + * @return None.
>> + */
>> +static void rsi_module_exit(void)
>> +{
>> +	usb_deregister(&rsi_driver);
>> +	rsi_dbg(INFO_ZONE, "%s: Unregistering driver\n", __func__);
>> +	return;
>> +}
>> +
>> +module_init(rsi_module_init);
>> +module_exit(rsi_module_exit);
>> +
>> +MODULE_AUTHOR("Redpine Signals Inc");
>> +MODULE_DESCRIPTION("Common layer for RSI drivers");
>> +MODULE_SUPPORTED_DEVICE("RSI-91x");
>> +MODULE_VERSION("0.1");
>> +MODULE_LICENSE("GPL");
> As mentioned in other patches, if the firmware that's being loaded by
> rsi_load_ta_instructions() is really bus-type dependent (eg, USB or
> SDIO) then each bus type module should have MODULE_FIRMWARE() tags too.
>
> Dan
>
>
>