Re: [patch] PCI Express Enhanced Config Patch - 2.6.0-test11

[Vladimir Kondratiev <vladimir.kondratiev@intel.com>]

My inputs:

- I do not like pci_express_read implemented as inline function. It is 
called only in one place. It is more appropriate, on my opinion, to 
merge all stuff added to include/asm-i386/pci.h , into 
arch/i386/pci/direct.c.

- if you will present 4k config space for all devices, it will save lots 
of work: you do not need to modify struct pci_dev, do not need almost 
all stuff in drivers/pci/proc.c. By presenting 4k config for PCI device 
you should not broke anything.

- Here and in _write function:
+static int pci_express_conf_read(int seg, int bus,
+        int devfn, int reg, int len, u32 *value)
+{
+    if (!value || (bus > 255) || (devfn > 255) || (reg > 4095)) {
+        printk(KERN_ERR "pci_express_conf_read: "
+                    "Invalid Parameter\n");
Worth to use
         printk(KERN_ERR "%s: Invalid Parameter\n",__FUNCTION__);

Durairaj, Sundarapandian wrote:

>Hi All, 
>
>Thanks for your comments. I am posting this patch after incorporating
>the review comments.
>
>Please find the attached patch file. Please review this and send your
>comments.
>
>Thanks,
>Sundar
>
>Note:
>This is the patch on PCI Express Enhanced configuration for 2.6.0 test11
>kernel following up to the Vladimir (Vladimir.Kondratiev@intel.com) and
>Harinarayanan (Harinarayanan.Seshadri@intel.com)  and my previous
>patches .
>I tested it on our i386 platform. 
>
>This patch also implements a mechanism for the kernel to find the
>chipset specific mmcfg base address. The kernel will detect the base
>address of the chipset through the ACPI table entry and based on that
>the PCI subsystem will be initialized.  
>  
>