Re: patch for kernel 2.6.10-rc2

[Alexandru Fomin <yo3fxy at spymac dot com>]

On Mon Nov 15  8:53 , Alexandru Fomin yo3fxy at spymac dot com> sent:
> Hi,
>
> I'm sending you a second version of the patch, makeing use of pci_dev_put(). Also
> note that the pci_dev struct stored in agp_bridge.dev is "released" in function
> agp_backend_cleanup, but probably functions agp_cleanup is a better place; if we
> take look at function agp_init, the call to function agp_backend_cleanup() looks
> to be in the wrong if block to me.

ok this is my last version of the patch (i think); the only difference should be
 in functions agp_init() and _X(agp_init)(). I don't know if it's appropriate to
 ask, but maybe someone from the devel team at ATI could advise us which patch we
 should consider correct.

ps: please note that this version of the patch is untested, but is not verry
different from
 the previous one, so it should be safe.
 
budha at l0pht:/usr/src/patches$ interdiff fglrx-pci_find_class-v2.patch fglrx-pci_
find_class-v3.patch 
diff -u agpgart_be.c agpgart_be.c
--- agpgart_be.c        2004-11-11 14:39:05.000000000 +0200
+++ agpgart_be.c        2004-11-15 19:40:12.000000000 +0200
@@ -7449,6 +7449,9 @@
        ret_val = agp_backend_initialize();
        if (ret_val) {
                agp_bridge.type = NOT_SUPPORTED;
+#if LINUX_VERSION_CODE > 0x020609
+               pci_dev_put(agp_bridge.dev);
+#endif
                return ret_val;
        }
        ret_val = agp_frontend_initialize();
@@ -7496,6 +7499,9 @@
        ret_val = agp_backend_initialize();
        if (ret_val) {
                agp_bridge.type = NOT_SUPPORTED;
+#if LINUX_VERSION_CODE > 0x020609
+               pci_dev_put(agp_bridge.dev);
+#endif
                return ret_val;
        }