On 2014-04-15 18:37, H Hartley Sweeten wrote:
Step 5 of the (*do_cmdtest) validates that the chanlist, chan/range/aref/etc,
is compatible with the actual hardware. At this point in the (*do_cmdtest)
the chanlist is valid, due to checks in the core, so the sanity check is not
needed.

For aesthetics, factor the step 5 code into a helper function. Tidy up the
factored out code.

Signed-off-by: H Hartley Sweeten <hswee...@visionengravers.com>
Cc: Ian Abbott <abbo...@mev.co.uk>
Cc: Greg Kroah-Hartman <gre...@linuxfoundation.org>
---
  drivers/staging/comedi/drivers/dmm32at.c | 51 +++++++++++++++++++-------------
  1 file changed, 30 insertions(+), 21 deletions(-)

diff --git a/drivers/staging/comedi/drivers/dmm32at.c 
b/drivers/staging/comedi/drivers/dmm32at.c
index c8a36eb..c6a6930 100644
--- a/drivers/staging/comedi/drivers/dmm32at.c
+++ b/drivers/staging/comedi/drivers/dmm32at.c
@@ -243,13 +243,39 @@ static int dmm32at_ns_to_timer(unsigned int *ns, int 
round)
        return *ns;
  }

+static int dmm32at_ai_check_chanlist(struct comedi_device *dev,
+                                    struct comedi_subdevice *s,
+                                    struct comedi_cmd *cmd)
+{
+       unsigned int chan0 = CR_CHAN(cmd->chanlist[0]);
+       unsigned int range0 = CR_RANGE(cmd->chanlist[0]);
+       int i;
+
+       for (i = 1; i < cmd->chanlist_len; i++) {
+               unsigned int chan = CR_CHAN(cmd->chanlist[i]);
+               unsigned int range = CR_RANGE(cmd->chanlist[i]);
+
+               if (chan != (chan0 + i) % s->n_chan) {
+                       dev_err(dev->class_dev,
+                               "chanlist must be consecutive channels, counting 
upwards\n");
+                       return -EINVAL;
+               }
+
+               if (range != range0) {
+                       dev_err(dev->class_dev,
+                               "chanlist must all have the same range\n");
+                       return -EINVAL;
+               }
+       }
+       return 0;
+}
+
  static int dmm32at_ai_cmdtest(struct comedi_device *dev,
                              struct comedi_subdevice *s,
                              struct comedi_cmd *cmd)
  {
        int err = 0;
        int tmp;
-       int start_chan, gain, i;

        /* Step 1 : check if triggers are trivially valid */

@@ -349,26 +375,9 @@ static int dmm32at_ai_cmdtest(struct comedi_device *dev,
        if (err)
                return 4;

-       /* step 5 check the channel list, the channel list for this
-          board must be consecutive and gains must be the same */
-
-       if (cmd->chanlist) {
-               gain = CR_RANGE(cmd->chanlist[0]);
-               start_chan = CR_CHAN(cmd->chanlist[0]);
-               for (i = 1; i < cmd->chanlist_len; i++) {
-                       if (CR_CHAN(cmd->chanlist[i]) !=
-                           (start_chan + i) % s->n_chan) {
-                               comedi_error(dev,
-                                            "entries in chanlist must be 
consecutive channels, counting upwards\n");
-                               err++;
-                       }
-                       if (CR_RANGE(cmd->chanlist[i]) != gain) {
-                               comedi_error(dev,
-                                            "entries in chanlist must all have the 
same gain\n");
-                               err++;
-                       }
-               }
-       }
+       /* Step 5: check channel list */
+
+       err |= dmm32at_ai_check_chanlist(dev, s, cmd);

        if (err)
                return 5;


As for the remarks in patch 02, it's possible for chanlist to be NULL or for chanlist_len to be 0, so dmm32at_ai_check_chanlist() should only be called if chanlist is non-NULL and chanlist_len is non-0.

But if chanlist is forced to be NULL by __comedi_get_user_chanlist() when chanlist_len is 0, it would be sufficient to check that chanlist is non-NULL before calling dmm32at_ai_check_chanlist().

The original code is buggy as well because it accesses cmd->chanlist[0] without checking cmd->chanlist_len.

--
-=( Ian Abbott @ MEV Ltd.    E-mail: <abbo...@mev.co.uk>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-
_______________________________________________
devel mailing list
de...@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel

Reply via email to