From: John Feeney <jfeeney@redhat.com>
Subject: [PATCH RHEL-5] BZ219288 /proc/bus/pci/devices speaks LIES
Date: Mon, 08 Jan 2007 15:52:30 -0500
Bugzilla: 219288
Message-Id: <45A2AF0E.5080901@redhat.com>
Changelog: edac: fix /proc/bus/pci/devices to allow X to start


RHBZ#: 219288 /proc/bus/pci/devices speaks LIES
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=219288

Description:
On specific Dell systems (e.g. PE 700), X fails to start with the following
errors in /var/log/Xorg.0.log:

(EE) No devices detected.

Fatal server error:
no screens found
XIO:  fatal IO error 104 (Connection reset by peer) on X server ":0.0"
    after 0 requests (0 known processed) with 0 events remaining.

After some investigation, by Ajax, it was determined that the
/proc/bus/pci/devices file and the /proc/bus/pci/* tree did not match,  
that is,
the devices file could list 15 devices while there would be 16 entries 
in the
tree. This discrepancy caused X to be unable to find its device, e.g. 
mach64.

Upon further review, it was determined that the device with the 8086:257e
pci id was missing from the devices file, hence the investigation lead to
drivers/edac/i82875p_edac.c where Dell determined that this driver was
adding an entry to the /proc tree with a call to pci_proc_attach_device()
but not to the global list of devices, the list that the devices file 
displays.

RHEL Version Found:
RHEL5-B2 (2.6.18-1.2767.el5)

The Proposed Fix:
The pci_bus_add_device function will replace pci_proc_attach_device()
because pci_bus_add_device() calls pci_proc_attach_device() while
populating the devices file as well.

In lieu of qualifying the call (with a "if(dev->procent==NULL)"), the
pci_bus_add_device() will be included in the if statement that detects a 
hidden
device. Thus, when the device previously hidden by BIOS is enabled and
detected, the /proc file system components will be fully implemented 
with the
pci_bus_add_device() call.

Note: In the drivers subdirectory, it appears as though i82875p_edac.c
is the only place outside of the pci directory where 
pci_proc_attach_device()
is called while pci_bus_add_device() is called by 17 other drivers. 
Perhaps,
this and the fact that pci_proc_attach_device() needed to be externed 
should
have been a red flag to the author of the original patch that added this
functionality.

Also note: even though the original code had pci_proc_attach_device() call
enveloped in "#ifdef CONFIG_PROC_FS", all other situations in drivers/*
where pci_bus_add_device() was called did not use this ifdef so it was not
carried forward.

Upstream Status:
It  does not  appear as though this fix is upstream at this point in
time. The offending function was added to 2.6.17-git17 in June of
2006. This patch will be submitted upstream.

--- linux-2.6.18.noarch/drivers/edac/i82875p_edac.c.dell
+++ linux-2.6.18.noarch/drivers/edac/i82875p_edac.c
@@ -261,10 +261,6 @@ static void i82875p_check(struct mem_ctl
 	i82875p_process_error_info(mci, &info, 1);
 }
 
-#ifdef CONFIG_PROC_FS
-extern int pci_proc_attach_device(struct pci_dev *);
-#endif
-
 /* Return 0 on success or 1 on failure. */
 static int i82875p_setup_overfl_dev(struct pci_dev *pdev,
 		struct pci_dev **ovrfl_pdev, void __iomem **ovrfl_window)
@@ -287,17 +283,12 @@ static int i82875p_setup_overfl_dev(stru
 
 		if (dev == NULL)
 			return 1;
+
+        	pci_bus_add_device(dev);
 	}
 
 	*ovrfl_pdev = dev;
 
-#ifdef CONFIG_PROC_FS
-	if ((dev->procent == NULL) && pci_proc_attach_device(dev)) {
-		i82875p_printk(KERN_ERR, "%s(): Failed to attach overflow "
-			       "device\n", __func__);
-		return 1;
-	}
-#endif  /* CONFIG_PROC_FS */
 	if (pci_enable_device(dev)) {
 		i82875p_printk(KERN_ERR, "%s(): Failed to enable overflow "
 			       "device\n", __func__);