Re: PCMCIA driver resource allocation

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Fri, Oct 19, 2007 at 10:51:51AM -0600, Bjorn Helgaas wrote:
> Question 1: Does the linux-pcmcia list still exist?  It's in MAINTAINERS:
> 
>   PCMCIA SUBSYSTEM
>   P:      Linux PCMCIA Team
>   L:      linux-pcmcia@lists.infradead.org
>   L:      http://lists.infradead.org/mailman/listinfo/linux-pcmcia
>   T:      git kernel.org:/pub/scm/linux/kernel/git/brodo/pcmcia-2.6.git
>   S:      Maintained
> 
> but the archive: http://lists.infradead.org/mailman/listinfo/linux-pcmcia
> seems dead.

The list is still around, but Dominik seems to have vanished.

> Question 2: Documentation/pcmcia/driver-changes.txt says drivers should
> now claim their own resources:
> 
>   * Resource management. (as of 2.6.8)
>      Although the PCMCIA subsystem will allocate resources for cards,
>      it no longer marks these resources busy. This means that driver
>      authors are now responsible for claiming your resources as per
>      other drivers in Linux.
> 
> But I don't see any drivers that do that.  It looks like there should
> be a bunch of changes like the one below.  Is there a reason these
> changes didn't happen, other than just lack of interest?

That's from around the time that I handed PCMCIA over to Dominik, and was
a to-do item.  I had some drivers converted over - mainly the few that I
was using, those being serial and pcnet_cs (serial is converted over but
the patch I had for pcnet_cs is below.)

However, in spite of me pointing Dominik at my remaining patch sets several
times, as far as I could tell they got ignored.  So essentially I all lost
interest in helping out with PCMCIA.

> Index: work3/drivers/net/wireless/orinoco_cs.c
> ===================================================================
> --- work3.orig/drivers/net/wireless/orinoco_cs.c	2007-10-18 10:56:34.000000000 -0600
> +++ work3/drivers/net/wireless/orinoco_cs.c	2007-10-18 13:22:44.000000000 -0600
> @@ -296,6 +296,10 @@
>  	/* We initialize the hermes structure before completing PCMCIA
>  	 * configuration just in case the interrupt handler gets
>  	 * called. */
> +	priv->io_resource = request_region(link->io.BasePort1,
> +					   link->io.NumPorts1, DRIVER_NAME);
> +	if (!priv->io_resource)
> +		goto cs_failed;
>  	mem = ioport_map(link->io.BasePort1, link->io.NumPorts1);
>  	if (!mem)
>  		goto cs_failed;
> @@ -366,6 +370,10 @@
>  	pcmcia_disable_device(link);
>  	if (priv->hw.iobase)
>  		ioport_unmap(priv->hw.iobase);
> +	if (priv->io_resource) {
> +		release_resource(priv->io_resource);
> +		priv->io_resource = NULL;

Wrong function.  release_resource() doesn't pair with request_region().
request_region() allocates memory for the struct resource.
release_resource() merely removes the struct resource from the tree.
release_region() on the other hand removes the struct resource and
frees it.

---

Convert pcnet_cs and serial_cs to request their IO regions, thereby
marking them busy.  These are only two of many drivers which need this
update.

diff -u -x BitKeeper -x ChangeSet -x SCCS -x _xlk -x '*.orig' -x '*.rej' ref/drivers/net/pcmcia/pcnet_cs.c linux/drivers/net/pcmcia/pcnet_cs.c
--- ref/drivers/net/pcmcia/pcnet_cs.c	Sun Nov 16 19:12:33 2003
+++ linux/drivers/net/pcmcia/pcnet_cs.c	Tue Dec 23 09:43:22 2003
@@ -687,8 +687,15 @@
     dev->poll_controller = ei_poll;
 #endif
 
+    if (!request_region(dev->base_addr, link->io.NumPorts1, "pcnet_cs")) {
+	printk(KERN_NOTICE "pcnet_cs: request_region() failed\n");
+	link->dev = NULL;
+	goto failed;
+    }
+
     if (register_netdev(dev) != 0) {
 	printk(KERN_NOTICE "pcnet_cs: register_netdev() failed\n");
+	release_region(dev->base_addr, link->io.NumPorts1);
 	link->dev = NULL;
 	goto failed;
     }
@@ -736,6 +743,9 @@
 
     DEBUG(0, "pcnet_release(0x%p)\n", link);
 
+    if (link->dev)
+	release_region(link->io.BasePort1, link->io.NumPorts1);
+
     if (info->flags & USE_SHMEM) {
 	iounmap(info->base);
 	pcmcia_release_window(link->win);


-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of: