re: usbtouchscreen: version 0.4

re: usbtouchscreen: version 0.4

[Dan Carpenter]

Hi Daniel,

I had a question about patch 5d8926658ce4: "usbtouchscreen: version 0.4"
from Jul 31, 2006.

drivers/input/touchscreen/usbtouchscreen.c
  1326  static void usbtouch_process_multi(struct usbtouch_usb *usbtouch,
  1327                                     unsigned char *pkt, int len)
  1328  {
  1329          unsigned char *buffer;
  1330          int pkt_len, pos, buf_len, tmp;
  1331  
  1332          /* process buffer */
  1333          if (unlikely(usbtouch->buf_len)) {
  1334                  /* try to get size */
  1335                  pkt_len = usbtouch->type->get_pkt_len(
  1336                                  usbtouch->buffer, usbtouch->buf_len);

->get_pkt_len() returns -1 on error.

  1337  
  1338                  /* drop? */
  1339                  if (unlikely(!pkt_len))
  1340                          goto out_flush_buf;
  1341  
  1342                  /* need to append -pkt_len bytes before able to get size */
  1343                  if (unlikely(pkt_len < 0)) {
  1344                          int append = -pkt_len;

But then we treat the -1 as a length of 1.  Is that really what we
intended?

  1345                          if (unlikely(append > len))
  1346                                 append = len;
  1347                          if (usbtouch->buf_len + append >= usbtouch->type->rept_size)
  1348                                  goto out_flush_buf;
  1349                          memcpy(usbtouch->buffer + usbtouch->buf_len, pkt, append);
  1350                          usbtouch->buf_len += append;
  1351  
  1352                          pkt_len = usbtouch->type->get_pkt_len(
  1353                                          usbtouch->buffer, usbtouch->buf_len);
  1354                          if (pkt_len < 0)
  1355                                  return;
  1356                  }
  1357  
  1358                  /* append */
  1359                  tmp = pkt_len - usbtouch->buf_len;

regards,
dan carpenter

Re: usbtouchscreen: version 0.4

[Daniel Ritz]

Hi Dan

On 2015-06-06 18:52 , Dan Carpenter wrote:
> Hi Daniel,
>
> I had a question about patch 5d8926658ce4: "usbtouchscreen: version 0.4"
> from Jul 31, 2006.
That's ancient :)
>
> drivers/input/touchscreen/usbtouchscreen.c
>   1326  static void usbtouch_process_multi(struct usbtouch_usb *usbtouch,
>   1327                                     unsigned char *pkt, int len)
>   1328  {
>   1329          unsigned char *buffer;
>   1330          int pkt_len, pos, buf_len, tmp;
>   1331  
>   1332          /* process buffer */
>   1333          if (unlikely(usbtouch->buf_len)) {
>   1334                  /* try to get size */
>   1335                  pkt_len = usbtouch->type->get_pkt_len(
>   1336                                  usbtouch->buffer, usbtouch->buf_len);
>
> ->get_pkt_len() returns -1 on error.

No, it doesn't and never has. In 62aa366d9b015 I added a bit of documentation:

/*
 * used to get the packet len. possible return values:
 * > 0: packet len
 * = 0: skip one byte
 * < 0: -return value more bytes needed
 */
int (*get_pkt_len) (unsigned char *pkt, int len);

So -1 means "need one more byte for a complete data packet". That byte will be
in delivered in the next invocation of usbtouch_irq(). One "touchscreen data packet"
can be split across multiple "usb packets" for some devices.

regards
-Daniel

Re: usbtouchscreen: version 0.4

[Dan Carpenter]

Ah, I'm amazed your still responding to emails about this ancient
driver.  :)  Thanks for the explanation.

regards,
dan carpenter