[LinuxBIOS] [PATCH] v3 drop malloc, clean up device model

[Stefan Reinauer]

* start using arch/foo.h again instead of archfoo.h (trivial)
* make constructor an initializer.
* fix memory leak/code flow error in current code
* add spinlocking
* drop malloc and use new_device for device allocation instead.
* ad CONFIG_SMP as it is needed by spinlocks and soon other stuff.

tested in qemu and works.

Signed-off-by: Stefan Reinauer

[ron minnich]

vice/device.h
===================================================================
--- include/device/device.h	(revision 417)
+++ include/device/device.h	(working copy)
@@ -112,18 +112,22 @@
    void (*set_link)(struct device * dev, unsigned int link);
    void (*reset_bus)(struct bus *bus);

-	/* a constructor. The constructor for a given device is defined in
the device source file.
-	 * When is this called? Not for the static tree. When the scan bus
code finds a new device, it must
-	 * create it and insert it into the device tree. To do this, it
calls a device constructor.
-	 * The set of all device constructors is concatenated into the
constructors array of structures via the usual
-	 * gcc hack of naming a segment.

Let's drop the "gcc hack" comment, it no longer applies.

Index: device/device.c
===================================================================
--- device/device.c	(revision 417)
+++ device/device.c	(working copy)

-//      spin_lock(&dev_lock);
+	spin_lock(&dev_lock);

/* Find the last child of our parent. */
    for (child = parent->children; child && child->sibling; /* */) {
    	child = child->sibling;
    }

-	dev = constructor(devid);
+	dev = new_device();
    if (!dev)
-		printk(BIOS_DEBUG, "%s: No constructor, going with empty dev",
-		       dev_id_string(devid));
+		return NULL;

are you missing a spin_unlock here? If we are doing lock/unlock then
the failure case should probably be a goto Error in the kernel style,
so we don' t leave a dangling lock.

thanks,

ron

[Stefan Reinauer]

* ron minnich  [070629 18:36]:
> Let's drop the "gcc hack" comment, it no longer applies.

dropped.
> if (!dev)
> printk(BIOS_DEBUG, "%s: No constructor, going with empty dev",
>        dev_id_string(devid));
> return NULL;
>
> are you missing a spin_unlock here? If we are doing lock/unlock then
>  the failure case should probably be a goto Error in the kernel style,
>  so we don' t leave a dangling lock.

Oh yeah! Even though this case does usually mean we can not really
continue.

new patch ...
Signed-off-by: Stefan Reinauer

[ron minnich]

Acked-by: Ronald G. Minnich

[Uwe Hermann]

On Fri, Jun 29, 2007 at 09:53:49AM -0700, ron minnich wrote:
> Acked-by: Ronald G. Minnich

r418.

Uwe.