[PATCH] make evtchn use a dynamic minor

[PATCH] make evtchn use a dynamic minor

[Steven Rostedt]

[-- Attachment #1: Type: text/plain, Size: 933 bytes --]

Along with the blktap hardcoding of a major number. It has come to my 
attention that evtchn also hardcodes a minor number.  Not so bad, but 
can later cause conflicts.  evtchn uses the misc device to get its major 
number.

This patch makes evtchn use a dynamic minor number instead.  It also 
updates the code in tools to create the device node if udev fails.  The 
tools now read the sysfs system to find the minor number needed.


One thing that is missing though, in this patch, is the update to 
./debugger/pdb/evtchn.ml, which still has the hard coded 201 as the 
minor.  This calls evtchn_open which is a C function that creates the 
node if it doesn't already exist.  I don't know much about Modula-2 but 
the major and minor probably don't need to be passed in, and the C code 
can do the same thing as the tools do now.  But I'll let others decide 
on that.

-- Steve

Signed-off-by: Steven Rostedt <srostedt@redhat.com>


[-- Attachment #2: xen-linux-evtchn-dynamic-minor.patch --]
[-- Type: text/x-patch, Size: 3187 bytes --]

diff -r 748e3c64d8f1 linux-2.6-xen-sparse/drivers/xen/evtchn/evtchn.c
--- a/linux-2.6-xen-sparse/drivers/xen/evtchn/evtchn.c	Thu Sep 28 10:04:25 2006 -0400
+++ b/linux-2.6-xen-sparse/drivers/xen/evtchn/evtchn.c	Thu Sep 28 20:55:10 2006 -0400
@@ -419,7 +419,7 @@ static struct file_operations evtchn_fop
 };
 
 static struct miscdevice evtchn_miscdev = {
-	.minor        = EVTCHN_MINOR,
+	.minor        = MISC_DYNAMIC_MINOR,
 	.name         = "evtchn",
 	.fops         = &evtchn_fops,
 	.devfs_name   = "misc/evtchn",
diff -r 748e3c64d8f1 linux-2.6-xen-sparse/include/xen/public/evtchn.h
--- a/linux-2.6-xen-sparse/include/xen/public/evtchn.h	Thu Sep 28 10:04:25 2006 -0400
+++ b/linux-2.6-xen-sparse/include/xen/public/evtchn.h	Thu Sep 28 20:55:10 2006 -0400
@@ -32,9 +32,6 @@
 
 #ifndef __LINUX_PUBLIC_EVTCHN_H__
 #define __LINUX_PUBLIC_EVTCHN_H__
-
-/* /dev/xen/evtchn resides at device number major=10, minor=201 */
-#define EVTCHN_MINOR 201
 
 /*
  * Bind a fresh port to VIRQ @virq.
diff -r 748e3c64d8f1 tools/libxc/xc_linux.c
--- a/tools/libxc/xc_linux.c	Thu Sep 28 10:04:25 2006 -0400
+++ b/tools/libxc/xc_linux.c	Thu Sep 28 20:55:10 2006 -0400
@@ -135,16 +135,71 @@ int do_xen_hypercall(int xc_handle, priv
 
 #define EVTCHN_DEV_NAME  "/dev/xen/evtchn"
 #define EVTCHN_DEV_MAJOR 10
-#define EVTCHN_DEV_MINOR 201
+
+#define MTAB "/proc/mounts"
+#define EVTCHN_SYSFS_PATH "/class/misc/evtchn/dev"
+
+#define MAX_PATH 255
+#define _STR(x) #x
+#define STR(x) _STR(x)
+
+static char sysfsdir[MAX_PATH + 1];
+
+static int find_sysfsdir(void)
+{
+    FILE *fp;
+    char type[MAX_PATH + 1];
+
+    if ((fp = fopen(MTAB, "r")) == NULL)
+        return -1;
+
+    while (fscanf(fp, "%*s %"
+                  STR(MAX_PATH)
+                  "s %"
+                  STR(MAX_PATH)
+                  "s %*s %*d %*d\n",
+                  sysfsdir, type) == 2) {
+        if (strncmp(type, "sysfs", 5) == 0)
+            break;
+    }
+    fclose(fp);
+
+    return strncmp(type, "sysfs", 5) == 0 ? 0 : -1;
+}
+
+static int find_evtchn_minor(void)
+{
+	FILE *fp;
+	int major;
+    int minor;
+
+    if (find_sysfsdir() < 0)
+        return -1;
+
+    strncat(sysfsdir, EVTCHN_SYSFS_PATH, MAX_PATH);
+
+	if ((fp = fopen(sysfsdir, "r")) == NULL)
+		return -1;
+
+	/* Skip title */
+	fscanf(fp,"%d:%d",&major, &minor);
+
+	fclose(fp);
+
+    return minor;
+}
 
 int xc_evtchn_open(void)
 {
     struct stat st;
     int fd;
+    int minor;
+
+    minor = find_evtchn_minor();
 
     /* Make sure any existing device file links to correct device. */
     if ((lstat(EVTCHN_DEV_NAME, &st) != 0) || !S_ISCHR(st.st_mode) ||
-        (st.st_rdev != makedev(EVTCHN_DEV_MAJOR, EVTCHN_DEV_MINOR)))
+        (st.st_rdev != makedev(EVTCHN_DEV_MAJOR, minor)))
         (void)unlink(EVTCHN_DEV_NAME);
 
 reopen:
@@ -153,7 +208,7 @@ reopen:
         if ( (errno == ENOENT) &&
             ((mkdir("/dev/xen", 0755) == 0) || (errno == EEXIST)) &&
             (mknod(EVTCHN_DEV_NAME, S_IFCHR|0600,
-            makedev(EVTCHN_DEV_MAJOR, EVTCHN_DEV_MINOR)) == 0) )
+            makedev(EVTCHN_DEV_MAJOR, minor)) == 0) )
             goto reopen;
 
         PERROR("Could not open event channel interface");

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: [PATCH] make evtchn use a dynamic minor

[John Levon]

On Thu, Sep 28, 2006 at 09:40:57PM -0400, Steven Rostedt wrote:

> One thing that is missing though, in this patch, is the update to 
> ./debugger/pdb/evtchn.ml, which still has the hard coded 201 as the 

As far as I can tell this is completely abandoned code, it has been
unable to compile for a long time. Can't we just remove the directory?
The false positives get annoying after a while.

regards
john

Re: [PATCH] make evtchn use a dynamic minor

[Keir Fraser]

On 29/9/06 2:40 am, "Steven Rostedt" <srostedt@redhat.com> wrote:

> This patch makes evtchn use a dynamic minor number instead.  It also
> updates the code in tools to create the device node if udev fails.  The
> tools now read the sysfs system to find the minor number needed.

Are there really systems out there that mount sysfs yet don't use udev?

It'd be nice to avoid the skanky search-and-create code in userspace. But
we'll take it if it really is likely to be useful.

 -- Keir

Re: [PATCH] make evtchn use a dynamic minor

[Steven Rostedt]

Keir Fraser wrote:
> On 29/9/06 2:40 am, "Steven Rostedt" <srostedt@redhat.com> wrote:
> 
>> This patch makes evtchn use a dynamic minor number instead.  It also
>> updates the code in tools to create the device node if udev fails.  The
>> tools now read the sysfs system to find the minor number needed.
> 
> Are there really systems out there that mount sysfs yet don't use udev?

I use to :-)   (and sometimes still do!)

> 
> It'd be nice to avoid the skanky search-and-create code in userspace. But
> we'll take it if it really is likely to be useful.
> 

Yeah, I didn't care much for it either, but my intent was to make it as 
robust as possible.  Either way is fine with me.

-- Steve