[PATCH] Unset msi and msix flags on pci_device_disable

[PATCH] Unset msi and msix flags on pci_device_disable

[Thomas Meyer]

The commit f5f2b13129a6541debf8851bae843cbbf48298b7 broke suspend/resume
to disk two or more  times in a row. This patches fixes the problem:

This patch should be included in 2.6.21

Signed-off-by: Thomas Meyer <thomas@m3y3r.de>

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index d3eab05..d7f8f96 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -886,7 +886,9 @@ pci_disable_device(struct pci_dev *dev)
         pci_write_config_word(dev, PCI_COMMAND, pci_command);
     }
     dev->is_busmaster = 0;
-
+    dev->msi_enabled=0;
+    dev->msix_enabled=0;
+   
     pcibios_disable_device(dev);
 }
 

Re: [PATCH] Unset msi and msix flags on pci_device_disable

[Eric W. Biederman]

Thomas Meyer <thomas@m3y3r.de> writes:

> The commit f5f2b13129a6541debf8851bae843cbbf48298b7 broke suspend/resume
> to disk two or more  times in a row. This patches fixes the problem:

Please clue me in, on what the problem is, I see why my patch would
have changed things I don't see how it is breaking things.  Clearing
the device bits has no effect on the hardware and it leaks our
msi handling data structures.

What is the user doing that this code broke?

I will be happy to work through a proper fix but this isn't it.

If this is actually helps something I am tempted to make the
added lines say WARN_ON.....

But I would really like to understand the nature of the problem
so that we can do something to fix it properly.

Eric