[cpia] USB anyone?

[sbertin@mindspring.com]

On 20 Jan, Jochen Scharrlach wrote:
> Ok, let's get back to the array-problem:

It looks like you have a good plan, but I have a few suggestions:

Keep the pointer to the list head in the high level driver, otherwise
you end up with a list in each lowlevel driver.  This will require
keeping track of which lowlevel driver is responsible for the camera.

struct cam_lowlevel {
	int type; /* CPIA_TYPE_PP or CPIA_TYPE_USB */
};
struct cam_data {
  struct cam_data **previous;
  struct cam_data *next;
  struct cam_lowlevel *lowlevel;
[...]
};

Because we are using a pointer, there is no need for more detail.  Each
lowlevel driver will have a structure that starts with the type.  A
typecast from struct cam_lowlevel * to the actual type will be needed in
the lowlevel driver.  This is somewhat like the way struct sockaddr is
handled.

cam_list_add and cam_list_remove could be functions provided by the high
level driver.  cam_unregister should be handled in each lowlevel driver.
(Can you tell I don't like macros used as functions.)

[ code snipped ]

Nice cpia_close implementation.  I like it :)

> Yes, the kernel_(un)lock is missing - did you use it only to protect
> the array operations or are there other critical parts? 

That's what it was there for, but I just realized that we probably need
a lock around the if(--cam->open_count == 0) block to prevent an open of
the camera before the cleanup is complete.

Scott J. Bertin
sbertin@mindspring.com