[PATCH] PCI vendor and device strings in sysfs

[PATCH] PCI vendor and device strings in sysfs

[Ananth N Mavinakayanahalli]

[-- Attachment #1: Type: text/plain, Size: 814 bytes --]

Hi,

Here is a patch against 2.6.0-test1 to display PCI vendor and
device strings in sysfs.

At present, the PCI "name" attribute has a length restriction
(DEVICE_NAME_SIZE) within which it tries to accomodate the vendor
and device strings, leading to, in most cases, truncation of one
or both strings.

This patch alleviates the issue by creating the vendor_name and
device_name attributes for PCI devices.

Here is an example:

ananth@llm06:/sys/devices/pci0000:00/0000:00:01.0> cat name
Intel Corp. 82810 DC-100 CGC [Ch
ananth@llm06:/sys/devices/pci0000:00/0000:00:01.0> cat vendor_name
Intel Corp.
ananth@llm06:/sys/devices/pci0000:00/0000:00:01.0> cat device_name
82810 DC-100 CGC [Chipset Graphics Controller]

Thanks,
-- 
Ananth Narayan <ananth@in.ibm.com>
Linux Technology Center,
IBM Software Lab, INDIA


[-- Attachment #2: km_pci.patch --]
[-- Type: text/plain, Size: 4920 bytes --]

diff -Naur linux-2.6.0-test1/include/linux/pci.h linux-2.6.0-test1-fix/include/linux/pci.h
--- linux-2.6.0-test1/include/linux/pci.h	2003-07-14 09:04:02.000000000 +0530
+++ linux-2.6.0-test1-fix/include/linux/pci.h	2003-07-17 14:23:05.000000000 +0530
@@ -416,6 +416,10 @@
 
 	char *		slot_name;	/* pointer to dev.bus_id */
 
+	/* pointer to export vendor and device strings to sysfs */
+	char *		vendor_name;
+	char *		device_name;
+
 	/* These fields are used by common fixups */
 	unsigned int	transparent:1;	/* Transparent PCI bridge */
 	unsigned int	multifunction:1;/* Part of multi-function device */

diff -Naur linux-2.6.0-test1/drivers/pci/names.c linux-2.6.0-test1-fix/drivers/pci/names.c
--- linux-2.6.0-test1/drivers/pci/names.c	2003-07-14 09:00:37.000000000 +0530
+++ linux-2.6.0-test1-fix/drivers/pci/names.c	2003-07-17 14:35:10.000000000 +0530
@@ -57,6 +57,8 @@
 	const struct pci_vendor_info *vendor_p = pci_vendor_list;
 	int i = VENDORS;
 	char *name = dev->dev.name;
+	char *vendor_unknown = "Unknown vendor";
+	char *device_unknown = "Unknown device";
 
 	do {
 		if (vendor_p->vendor == dev->vendor)
@@ -66,7 +68,15 @@
 
 	/* Couldn't find either the vendor nor the device */
 	sprintf(name, "PCI device %04x:%04x", dev->vendor, dev->device);
-	return;
+	dev->vendor_name = kmalloc(strlen(vendor_unknown)+1, GFP_KERNEL);
+	if (!dev->vendor_name)
+		goto out;
+	strcpy(dev->vendor_name, vendor_unknown);
+	dev->device_name = kmalloc(strlen(device_unknown)+1, GFP_KERNEL);
+	if (!dev->device_name)
+		goto out;
+	strcpy(dev->device_name, device_unknown);
+	goto out;
 
 	match_vendor: {
 		struct pci_device_info *device_p = vendor_p->devices;
@@ -82,19 +92,39 @@
 		/* Ok, found the vendor, but unknown device */
 		sprintf(name, "PCI device %04x:%04x (%." DEVICE_NAME_HALF "s)",
 				dev->vendor, dev->device, vendor_p->name);
-		return;
+		dev->vendor_name = kmalloc(strlen(vendor_p->name)+1, GFP_KERNEL);
+		if (!dev->vendor_name)
+			goto out;
+		strcpy(dev->vendor_name, vendor_p->name);
+		dev->device_name = kmalloc(strlen(device_unknown)+1, GFP_KERNEL);
+		if (!dev->device_name)
+			goto out;
+		strcpy(dev->device_name, device_unknown);
+		goto out;
 
 		/* Full match */
 		match_device: {
 			char *n = name + sprintf(name, "%." DEVICE_NAME_HALF
 					"s %." DEVICE_NAME_HALF "s",
 					vendor_p->name, device_p->name);
-			int nr = device_p->seen + 1;
-			device_p->seen = nr;
-			if (nr > 1)
-				sprintf(n, " (#%d)", nr);
+
+			dev->vendor_name = kmalloc(strlen(vendor_p->name)+1, GFP_KERNEL);
+			if (!dev->vendor_name)
+				goto update_count;
+			strcpy(dev->vendor_name, vendor_p->name);
+			dev->device_name = kmalloc(strlen(device_p->name)+1, GFP_KERNEL);
+			if (dev->device_name)
+				strcpy(dev->device_name, device_p->name);
+			update_count: {
+				int nr = device_p->seen + 1;
+				device_p->seen = nr;
+				if (nr > 1)
+					sprintf(n, " (#%d)", nr);
+			}
 		}
 	}
+out:
+	return;
 }
 
 /*

diff -Naur linux-2.6.0-test1/drivers/pci/pci-sysfs.c linux-2.6.0-test1-fix/drivers/pci/pci-sysfs.c
--- linux-2.6.0-test1/drivers/pci/pci-sysfs.c	2003-07-14 09:04:40.000000000 +0530
+++ linux-2.6.0-test1-fix/drivers/pci/pci-sysfs.c	2003-07-16 12:03:16.000000000 +0530
@@ -39,6 +39,42 @@
 pci_config_attr(class, "0x%06x\n");
 pci_config_attr(irq, "%u\n");
 
+/* show vendor name */
+static ssize_t
+pci_show_vendor_name(struct device *dev, char *buf)
+{
+	struct pci_dev * pci_dev = to_pci_dev(dev);
+	int len;
+	
+	if (!pci_dev->vendor_name)
+		return 0;
+	strcpy(buf, pci_dev->vendor_name);
+	len = strlen(pci_dev->vendor_name);
+	buf[len] = '\n';
+	buf[len+1] = 0;
+	return len+1;
+}
+
+static DEVICE_ATTR(vendor_name,S_IRUGO,pci_show_vendor_name,NULL);
+	
+/* show device name */
+static ssize_t
+pci_show_device_name(struct device *dev, char *buf)
+{
+	struct pci_dev * pci_dev = to_pci_dev(dev);
+	int len;
+	
+	if (!pci_dev->device_name)
+		return 0;
+	strcpy(buf, pci_dev->device_name);
+	len = strlen(pci_dev->device_name);
+	buf[len] = '\n';
+	buf[len+1] = 0;
+	return len+1;
+}
+
+static DEVICE_ATTR(device_name,S_IRUGO,pci_show_device_name,NULL);
+
 /* show resources */
 static ssize_t
 pci_show_resources(struct device * dev, char * buf)
@@ -176,5 +212,7 @@
 	device_create_file (dev, &dev_attr_class);
 	device_create_file (dev, &dev_attr_irq);
 	device_create_file (dev, &dev_attr_resource);
+	device_create_file (dev, &dev_attr_vendor_name);
+	device_create_file (dev, &dev_attr_device_name);
 	sysfs_create_bin_file(&dev->kobj, &pci_config_attr);
 }

diff -Naur linux-2.6.0-test1/drivers/pci/probe.c linux-2.6.0-test1-fix/drivers/pci/probe.c
--- linux-2.6.0-test1/drivers/pci/probe.c	2003-07-14 09:06:33.000000000 +0530
+++ linux-2.6.0-test1-fix/drivers/pci/probe.c	2003-07-15 17:25:42.000000000 +0530
@@ -478,6 +478,8 @@
 	struct pci_dev *pci_dev;
 
 	pci_dev = to_pci_dev(dev);
+	kfree(pci_dev->vendor_name);
+	kfree(pci_dev->device_name);
 	kfree(pci_dev);
 }
 

Re: [PATCH] PCI vendor and device strings in sysfs

[Patrick Mochel]

> Here is a patch against 2.6.0-test1 to display PCI vendor and
> device strings in sysfs.
> 
> At present, the PCI "name" attribute has a length restriction
> (DEVICE_NAME_SIZE) within which it tries to accomodate the vendor
> and device strings, leading to, in most cases, truncation of one
> or both strings.
> 
> This patch alleviates the issue by creating the vendor_name and
> device_name attributes for PCI devices.

We don't necessarily need to keep the ASCII strings around at all, and in 
the case in which CONFIG_PCI_NAMES=n, they are completely irrelevant. They 
are pretty, but we could just export the vendor/device IDs and have a 
userspace tool (e.g. sysutils from IBM) look up the name in a userspace 
database. 


	-pat

Re: [PATCH] PCI vendor and device strings in sysfs

[Greg KH]

On Thu, Jul 17, 2003 at 03:41:24PM +0530, Ananth N Mavinakayanahalli wrote:
> Hi,
> 
> Here is a patch against 2.6.0-test1 to display PCI vendor and
> device strings in sysfs.
> 
> At present, the PCI "name" attribute has a length restriction
> (DEVICE_NAME_SIZE) within which it tries to accomodate the vendor
> and device strings, leading to, in most cases, truncation of one
> or both strings.
> 
> This patch alleviates the issue by creating the vendor_name and
> device_name attributes for PCI devices.

I agree with Pat, this should be done in userspace.

We should really just get rid of the dev->name file all together to keep
people from relying on it :)

thanks,

greg k-h