Skip to contentSkip to conversation

xc: deal with xen/evtchn and xen/gntdev device names

PATCH In review Jeremy Fitzhardinge 5 participants 13 messages May 29, 2010 – Jun 4, 2010

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xc: deal with xen/evtchn and xen/gntdev device names

This patch makes xc_linux properly deal with:
1. discovering and creating device nodes if necessary
2. the new form of xen/<dev> device names soon to be used by the kernel

This changes the logic slightly:
- If a device node already exists with the proper name, then it uses it as-is, assuming it has already been correctly created.
- If the path doesn't exist, or it exists but isn't a device node, and it has successfully found the major/minor for the device, then (re)create the device node.

Since this logic is identical for gntdev and evtchn, make a common function to handle both.

tools/libxc/xc_linux.c +65 -51
diff -r f146aa75120c tools/libxc/xc_linux.c
@@ -378,34 +378,73 @@
     return makedev(major, minor);
 }
 
-#define EVTCHN_DEV_NAME  "/dev/xen/evtchn"
+#define DEVXEN	"/dev/xen"
+
+static int make_dev_xen(void)
+{
+	if (mkdir(DEVXEN, 0755) != 0) {
+		struct stat st;
+
+		if (stat(DEVXEN, &st) != 0 || !S_ISDIR(st.st_mode))
+			return -1;
+	}
+
+	return 0;
+}
+
+static int xendev_open(const char *dev)
+{
+	int fd, devnum;
+	struct stat st;
+	char *devname, *devpath;
+
+	devname = devpath = NULL;
+	fd = -1;
+
+	if (asprintf(&devname, "xen!%s", dev) == 0)
+		goto fail;
+
+	if (asprintf(&devpath, "%s/%s", DEVXEN, dev) == 0)
+		goto fail;
+
+	devnum = xc_find_device_number(dev);
+	if (devnum == -1)
+		devnum = xc_find_device_number(devname);
+
+	/*
+	 * If we know what the correct device is and the path doesn't
+	 * exist or isn't a device, then remove it so we can create the
+	 * device.
+	 */
+	if (devnum != -1 &&
+		(stat(devpath, &st) != 0 || !S_ISCHR(st.st_mode))) {
+		unlink(devpath);
+
+		if (make_dev_xen() == -1)
+			goto fail;
+
+		if (mknod(devpath, S_IFCHR|0600, devnum) != 0) {
+			PERROR("Couldn't make %s",  devpath);
+			goto fail;
+		}
+	}
+
+    fd = open(devpath, O_RDWR);
+
+	if (fd == -1)
+        PERROR("Could not open %s", devpath);
+
+fail:
+	free(devname);
+	free(devpath);
+
+    return fd;
+}
+
 
 int xc_evtchn_open(void)
 {
-    struct stat st;
-    int fd;
-    int devnum;
-
-    devnum = xc_find_device_number("evtchn");
-
-    /* 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 != devnum) )
-        (void)unlink(EVTCHN_DEV_NAME);
-
- reopen:
-    if ( (fd = open(EVTCHN_DEV_NAME, O_RDWR)) == -1 )
-    {
-        if ( (errno == ENOENT) &&
-             ((mkdir("/dev/xen", 0755) == 0) || (errno == EEXIST)) &&
-             (mknod(EVTCHN_DEV_NAME, S_IFCHR|0600, devnum) == 0) )
-            goto reopen;
-
-        PERROR("Could not open event channel interface");
-        return -1;
-    }
-
-    return fd;
+	return xendev_open("evtchn");
 }
 
 int xc_evtchn_close(int xce_handle)
@@ -519,34 +558,9 @@
     errno = saved_errno;
 }
 
-#define GNTTAB_DEV_NAME "/dev/xen/gntdev"
-
 int xc_gnttab_open(void)
 {
-    struct stat st;
-    int fd;
-    int devnum;
-
-    devnum = xc_find_device_number("gntdev");
-
-    /* Make sure any existing device file links to correct device. */
-    if ( (lstat(GNTTAB_DEV_NAME, &st) != 0) || !S_ISCHR(st.st_mode) ||
-         (st.st_rdev != devnum) )
-        (void)unlink(GNTTAB_DEV_NAME);
-
-reopen:
-    if ( (fd = open(GNTTAB_DEV_NAME, O_RDWR)) == -1 )
-    {
-        if ( (errno == ENOENT) &&
-             ((mkdir("/dev/xen", 0755) == 0) || (errno == EEXIST)) &&
-             (mknod(GNTTAB_DEV_NAME, S_IFCHR|0600, devnum) == 0) )
-            goto reopen;
-
-        PERROR("Could not open grant table interface");
-        return -1;
-    }
-
-    return fd;
+	return xendev_open("gntdev");
 }
 
 int xc_gnttab_close(int xcg_handle)
#1 lore raw

tools/libxc/xc_linux.c line 407

Quoted 2 lines · Jeremy Fitzhardinge
tools/libxc/xc_linux.c +2 -0
+	if (asprintf(&devname, "xen!%s", dev) == 0)
+	if (asprintf(&devpath, "%s/%s", DEVXEN, dev) == 0)

asprintf is a glibc extension. Is it really necessary to support long device names for fun or would some 64 byte on the stack be enough?

tools/libxc/xc_linux.c line 412

Quoted 3 lines
tools/libxc/xc_linux.c +3 -0
+	devnum = xc_find_device_number(dev);
+	if (devnum == -1)
+		devnum = xc_find_device_number(devname);

Checks should always go from the special ("xen!...") to the gerneral. Also the device is dev_t, not integer.

tools/libxc/xc_linux.c line 420

Quoted 7 lines
tools/libxc/xc_linux.c +7 -0
+	/*
+	 * If we know what the correct device is and the path doesn't
+	 * exist or isn't a device, then remove it so we can create the
+	 * device.
+	 */
+	if (devnum != -1 &&
+		(stat(devpath, &st) != 0 || !S_ISCHR(st.st_mode))) {

Why should it exist and be something else then a character device? The admin is free to break the system as she likes.

Quoted 1 line

+ fd = open(devpath, O_RDWR);

Whitespace.

Bastian

Signature · 3 lines
... bacteriological warfare ... hard to believe we were once foolish
enough to play around with that.
		-- McCoy, "The Omega Glory", stardate unknown
#2 lore raw
In reply to Jeremy Fitzhardinge
Quoted 11 lines · Jeremy Fitzhardinge

This patch makes xc_linux properly deal with:
1. discovering and creating device nodes if necessary
2. the new form of xen/<dev> device names soon to be used by the kernel

This changes the logic slightly:
- If a device node already exists with the proper name, then it uses it as-is, assuming it has already been correctly created.
- If the path doesn't exist, or it exists but isn't a device node, and it has successfully found the major/minor for the device, then (re)create the device node.

Quoted 2 lines

Since this logic is identical for gntdev and evtchn, make a common function to handle both.

Does any of this logic really belong in libxc in the first place?

Can we not just rip it all out and make it the distro/platform's responsibility to ensure these devices exist and are correct? Perhaps that might involve shipping some default/example udev rules instead.

Ian.

#3 lore raw
Quoted 1 line · Ian Campbell

Does any of this logic really belong in libxc in the first place?

No.

Quoted 2 lines

Can we not just rip it all out and make it the distro/platform's responsibility to ensure these devices exist and are correct?

Thats what I forced in Debian right now.

Quoted 2 lines

Perhaps
that might involve shipping some default/example udev rules instead.

With the fixed kernel this needs no rules. Even with devtmpfs (some people call it devfs version 2) and no udev it will work then.

Bastian

Signature · 2 lines
Vulcans believe peace should not depend on force.
		-- Amanda, "Journey to Babel", stardate 3842.3
#4 lore raw
In reply to Ian Campbell
Quoted 2 lines · Ian Campbell

Does any of this logic really belong in libxc in the first place?

No. It's a relic of a simpler time.

Quoted 4 lines

Can we not just rip it all out and make it the distro/platform's responsibility to ensure these devices exist and are correct? Perhaps that might involve shipping some default/example udev rules instead.

I only kept it because I didn't want to break any existing installs, but updating the kernel's name is going to do that anyway. Unfortunately the current libxc code is broken in the worst possible way - it will unlink an existing correct device and then fail to replace it - so even if the system has set it up correctly it will still screw things up.

I think we're just going to have to have a flag day and fix kernel, libxc and udev (assuming it needs it) all at once...

J

#5 lore raw
Quoted 11 lines · Jeremy Fitzhardinge
Quoted 2 lines · Ian Campbell

Does any of this logic really belong in libxc in the first place?

No. It's a relic of a simpler time.

Quoted 4 lines

Can we not just rip it all out and make it the distro/platform's responsibility to ensure these devices exist and are correct? Perhaps that might involve shipping some default/example udev rules instead.

Quoted 8 lines

I only kept it because I didn't want to break any existing installs, but updating the kernel's name is going to do that anyway. Unfortunately the current libxc code is broken in the worst possible way - it will unlink an existing correct device and then fail to replace it - so even if the system has set it up correctly it will still screw things up.

I think we're just going to have to have a flag day and fix kernel, libxc and udev (assuming it needs it) all at once...

I don't think we need a flag day. It seems like we already ship a udev rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is apparently unnecessary (but harmless) with the proposed kernel change.

I removed all the mknod stuff from my tools and things kept working as expected.

So lets try the following. I think once it has settled down in unstable we could consider it for inclusion in the stable branch.

---
tools: assume that special Xen devices have been created by the platform

Remove all the magic surrounding the special Xen devices in Linux specific code whereby we attempt to figure out what the correct major:minor number is and check the the existing device has these numbers etc. In 2010 we really should be able to trust that the platform has created the devices correctly or provide correct configuration settings such that they are without resorting to tearing down the platform configured state and rebuilding it.

tools/hotplug/Linux/xen-backend.rules already contains the necessary udev rules to create /dev/xen/evtchn and friends in the correct place.

tools/blktap/drivers/blktapctrl_linux.c +2 -22
diff -r 2f765c9825b2 -r 2962c9c4d26f tools/blktap/drivers/blktapctrl_linux.c
@@ -79,31 +79,11 @@
 
 int blktap_interface_open(void)
 {
-	char *devname;
-	int ret;
 	int ctlfd;
 
-	/* Attach to blktap0 */
-	if (asprintf(&devname,"%s/%s0", BLKTAP_DEV_DIR, BLKTAP_DEV_NAME) == -1)
-		goto open_failed;
-
-	ret = xc_find_device_number("blktap0");
-	if (ret < 0) {
-		DPRINTF("couldn't find device number for 'blktap0'\n");
-		goto open_failed;
-	}
-
-	blktap_major = major(ret);
-	make_blktap_dev(devname,blktap_major, 0);
-
-	ctlfd = open(devname, O_RDWR);
-	if (ctlfd == -1) {
+	ctlfd = open(BLKTAP_DEV_DIR "/" BLKTAP_DEV_NAME "0", O_RDWR);
+	if (ctlfd == -1)
 		DPRINTF("blktap0 open failed\n");
-		goto open_failed;
-	}
 
 	return ctlfd;
-
-open_failed:
-	return -1;
 }
tools/libxc/xc_linux.c +3 -118

136 lines not on this page · raw message

#6 lore raw
Quoted 21 lines · Ian Campbell
Quoted 19 lines · Jeremy Fitzhardinge
Quoted 3 lines · Ian Campbell

Does any of this logic really belong in libxc in the first place?

No. It's a relic of a simpler time.

Quoted 5 lines

Can we not just rip it all out and make it the distro/platform's responsibility to ensure these devices exist and are correct? Perhaps that might involve shipping some default/example udev rules instead.

I only kept it because I didn't want to break any existing installs, but updating the kernel's name is going to do that anyway. Unfortunately the current libxc code is broken in the worst possible way - it will unlink an existing correct device and then fail to replace it - so even if the system has set it up correctly it will still screw things up.

Quoted 8 lines
Quoted 3 lines

I think we're just going to have to have a flag day and fix kernel, libxc and udev (assuming it needs it) all at once...

I don't think we need a flag day. It seems like we already ship a udev rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is apparently unnecessary (but harmless) with the proposed kernel change.

My main concern is that an old libxc will screw anyone with new kernel and udev.

Quoted 17 lines

I removed all the mknod stuff from my tools and things kept working as expected.

So lets try the following. I think once it has settled down in unstable we could consider it for inclusion in the stable branch.

---
tools: assume that special Xen devices have been created by the platform

Remove all the magic surrounding the special Xen devices in Linux specific code whereby we attempt to figure out what the correct major:minor number is and check the the existing device has these numbers etc. In 2010 we really should be able to trust that the platform has created the devices correctly or provide correct configuration settings such that they are without resorting to tearing down the platform configured state and rebuilding it.

Quoted 4 lines

tools/hotplug/Linux/xen-backend.rules already contains the necessary udev rules to create /dev/xen/evtchn and friends in the correct place.

Looks fine to me.

J

Quoted 183 lines

diff -r 2f765c9825b2 -r 2962c9c4d26f tools/blktap/drivers/blktapctrl_linux.c

tools/blktap/drivers/blktapctrl_linux.c +2 -22
@@ -79,31 +79,11 @@
 
 int blktap_interface_open(void)
 {
-	char *devname;
-	int ret;
 	int ctlfd;
 
-	/* Attach to blktap0 */
-	if (asprintf(&devname,"%s/%s0", BLKTAP_DEV_DIR, BLKTAP_DEV_NAME) == -1)
-		goto open_failed;
-
-	ret = xc_find_device_number("blktap0");
-	if (ret < 0) {
-		DPRINTF("couldn't find device number for 'blktap0'\n");
-		goto open_failed;
-	}
-
-	blktap_major = major(ret);
-	make_blktap_dev(devname,blktap_major, 0);
-
-	ctlfd = open(devname, O_RDWR);
-	if (ctlfd == -1) {
+	ctlfd = open(BLKTAP_DEV_DIR "/" BLKTAP_DEV_NAME "0", O_RDWR);
+	if (ctlfd == -1)
 		DPRINTF("blktap0 open failed\n");
-		goto open_failed;
-	}
 
 	return ctlfd;
-
-open_failed:
-	return -1;
 }

diff -r 2f765c9825b2 -r 2962c9c4d26f tools/libxc/xc_linux.c

tools/libxc/xc_linux.c +3 -118
@@ -316,126 +316,11 @@
                       (unsigned long)hypercall);
 }
 
-#define MTAB "/proc/mounts"
-#define MAX_PATH 255
-#define _STR(x) #x
-#define STR(x) _STR(x)
-
-static int find_sysfsdir(char *sysfsdir)
-{
-    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);
-}
-
-int xc_find_device_number(const char *name)
-{
-    FILE *fp;
-    int i, major, minor;
-    char sysfsdir[MAX_PATH + 1];
-    static char *classlist[] = { "xen", "misc" };
-
-    for ( i = 0; i < (sizeof(classlist) / sizeof(classlist[0])); i++ )
-    {
-        if ( find_sysfsdir(sysfsdir) < 0 )
-            goto not_found;
-
-        /* <base>/class/<classname>/<devname>/dev */
-        strncat(sysfsdir, "/class/", MAX_PATH);
-        strncat(sysfsdir, classlist[i], MAX_PATH);
-        strncat(sysfsdir, "/", MAX_PATH);
-        strncat(sysfsdir, name, MAX_PATH);
-        strncat(sysfsdir, "/dev", MAX_PATH);
-
-        if ( (fp = fopen(sysfsdir, "r")) != NULL )
-            goto found;
-    }
-
- not_found:
-    errno = -ENOENT;
-    return -1;
-
- found:
-    if ( fscanf(fp, "%d:%d", &major, &minor) != 2 )
-    {
-        fclose(fp);
-        goto not_found;
-    }
-
-    fclose(fp);
-
-    return makedev(major, minor);
-}
-
-#define DEVXEN "/dev/xen"
-
-static int make_dev_xen(void)
-{
-    if ( mkdir(DEVXEN, 0755) != 0 )
-    {
-        struct stat st;
-        if ( (stat(DEVXEN, &st) != 0) || !S_ISDIR(st.st_mode) )
-            return -1;
-    }
-
-    return 0;
-}
-
-static int xendev_open(const char *dev)
-{
-    int fd, devnum;
-    struct stat st;
-    char *devname, *devpath;
-
-    devname = devpath = NULL;
-    fd = -1;
-
-    if ( asprintf(&devname, "xen!%s", dev) == 0 )
-        goto fail;
-
-    if ( asprintf(&devpath, "%s/%s", DEVXEN, dev) == 0 )
-        goto fail;
-
-    devnum = xc_find_device_number(dev);
-    if ( devnum == -1 )
-        devnum = xc_find_device_number(devname);
-
-    /*
-     * If we know what the correct device is and the path doesn't exist or 
-     * isn't a device, then remove it so we can create the device.
-     */
-    if ( (devnum != -1) &&
-         ((stat(devpath, &st) != 0) || !S_ISCHR(st.st_mode)) )
-    {
-        unlink(devpath);
-        if ( (make_dev_xen() == -1) ||
-             (mknod(devpath, S_IFCHR|0600, devnum) != 0) )
-            goto fail;
-    }
-
-    fd = open(devpath, O_RDWR);
-
- fail:
-    free(devname);
-    free(devpath);
-
-    return fd;
-}
+#define DEVXEN "/dev/xen/"
 
 int xc_evtchn_open(void)
 {
-    return xendev_open("evtchn");
+    return open(DEVXEN "evtchn", O_RDWR);
 }
 
 int xc_evtchn_close(int xce_handle)
@@ -551,7 +436,7 @@
 
 int xc_gnttab_open(xc_interface *xch)
 {
-    return xendev_open("gntdev");
+    return open(DEVXEN "gntdev", O_RDWR);
 }
 
 int xc_gnttab_close(xc_interface *xch, int xcg_handle)
#7 lore raw
Quoted 10 lines · Jeremy Fitzhardinge
Quoted 6 lines · Ian Campbell

On Tue, 2010-06-01 at 17:35 +0100, Jeremy Fitzhardinge wrote:
I don't think we need a flag day. It seems like we already ship a udev rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is apparently unnecessary (but harmless) with the proposed kernel change.

My main concern is that an old libxc will screw anyone with new kernel and udev.

Is it any more likely to screw them with a new kernel than with an old one?

If so I think that's an argument for propagating the removal of this functionality into stable trees sooner rather than later rather than papering over the craziness for even longer.

Ian.

#8 lore raw
Quoted 11 lines · Ian Campbell
Quoted 9 lines · Jeremy Fitzhardinge
Quoted 7 lines · Ian Campbell

On Tue, 2010-06-01 at 17:35 +0100, Jeremy Fitzhardinge wrote:
I don't think we need a flag day. It seems like we already ship a udev rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is apparently unnecessary (but harmless) with the proposed kernel change.

Quoted 6 lines
Quoted 3 lines

My main concern is that an old libxc will screw anyone with new kernel and udev.

Is it any more likely to screw them with a new kernel than with an old one?

Yeah, because libxc's rummage around in sysfs will actually work. If we rename the devices to be correct, it won't find them and it just ends up deleting the old device and either failing to create a new one, or creating it with a bogus major/minor.

Quoted 4 lines

If so I think that's an argument for propagating the removal of this functionality into stable trees sooner rather than later rather than papering over the craziness for even longer.

Yep.

J

#9 lore raw
Quoted 15 lines · Jeremy Fitzhardinge
Quoted 14 lines · Ian Campbell
Quoted 12 lines · Jeremy Fitzhardinge
Quoted 7 lines · Ian Campbell

On Tue, 2010-06-01 at 17:35 +0100, Jeremy Fitzhardinge wrote:
I don't think we need a flag day. It seems like we already ship a udev rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is apparently unnecessary (but harmless) with the proposed kernel change.

My main concern is that an old libxc will screw anyone with new kernel and udev.

Quoted 8 lines
Quoted 3 lines

Is it any more likely to screw them with a new kernel than with an old one?

Yeah, because libxc's rummage around in sysfs will actually work. If we rename the devices to be correct, it won't find them and it just ends up deleting the old device and either failing to create a new one, or creating it with a bogus major/minor.

That sucks ;-)

Quoted 6 lines
Quoted 4 lines

If so I think that's an argument for propagating the removal of this functionality into stable trees sooner rather than later rather than papering over the craziness for even longer.

Yep.

Keir has applied my patch so I'll keep an eye out for fallout and suggest it for backporting to the stable branch when it looks like we've gotten away with it.

#10 lore raw
Quoted 7 lines · Ian Campbell
Quoted 3 lines
Quoted 3 lines

If so I think that's an argument for propagating the removal of this functionality into stable trees sooner rather than later rather than papering over the craziness for even longer.

Keir has applied my patch so I'll keep an eye out for fallout and suggest it for backporting to the stable branch when it looks like we've gotten away with it.

Any fallout should be dealt with by getting users to set up proper udev rules, I'm sure. I backported to 4.0 for 4.0.1, so I have confidence. ;-)

-- Keir

#11 lore raw
In reply to Ian Campbell

Since udev magic is something that few developers should need to deal with, would it be too much to ask for a good wiki page to list likely failure conditions and how to solve them BEFORE clueless developers (including me) run into them?

Quoted 60 lines

-----Original Message-----
From: Ian Campbell [mailto:[email protected]]
Sent: Friday, June 04, 2010 3:55 AM
To: Jeremy Fitzhardinge
Cc: Bastian Blank; Xen-devel
Subject: Re: [Xen-devel] [PATCH] xc: deal with xen/evtchn and xen/gntdev device names

Quoted 7 lines · Jeremy Fitzhardinge
Quoted 6 lines · Ian Campbell
Quoted 4 lines · Jeremy Fitzhardinge
Quoted 2 lines · Ian Campbell

On Tue, 2010-06-01 at 17:35 +0100, Jeremy Fitzhardinge wrote:
I don't think we need a flag day. It seems like we already ship a

udev

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is

apparently

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

unnecessary (but harmless) with the proposed kernel change.

My main concern is that an old libxc will screw anyone with new

kernel

Quoted 3 lines
Quoted 3 lines
Quoted 2 lines

and udev.

Is it any more likely to screw them with a new kernel than with an

old

Quoted 4 lines
Quoted 2 lines

one?

Yeah, because libxc's rummage around in sysfs will actually work. If

we

Quoted 1 line

rename the devices to be correct, it won't find them and it just ends

up

Quoted 2 lines

deleting the old device and either failing to create a new one, or creating it with a bogus major/minor.

That sucks ;-)

Quoted 1 line
Quoted 1 line

If so I think that's an argument for propagating the removal of

this

Quoted 1 line
Quoted 1 line

functionality into stable trees sooner rather than later rather

than

Quoted 4 lines
Quoted 2 lines

papering over the craziness for even longer.

Yep.

Keir has applied my patch so I'll keep an eye out for fallout and suggest it for backporting to the stable branch when it looks like we've
gotten away with it.

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

#12 lore raw
Quoted 5 lines · Dan Magenheimer

Since udev magic is something that few developers should need to deal with, would it be too much to ask for a good wiki page to list likely failure conditions and how to solve them BEFORE clueless developers (including me) run into them?

The expectation is that things will mostly Just Work. If this turns out not to be the case then the wiki would be an appropriate place to document the issues and solutions as we determine what they are.

Ian.

Quoted 61 lines
Quoted 61 lines

-----Original Message-----
From: Ian Campbell [mailto:[email protected]]
Sent: Friday, June 04, 2010 3:55 AM
To: Jeremy Fitzhardinge
Cc: Bastian Blank; Xen-devel
Subject: Re: [Xen-devel] [PATCH] xc: deal with xen/evtchn and xen/gntdev device names

Quoted 7 lines · Jeremy Fitzhardinge
Quoted 6 lines · Ian Campbell
Quoted 4 lines · Jeremy Fitzhardinge
Quoted 2 lines · Ian Campbell

On Tue, 2010-06-01 at 17:35 +0100, Jeremy Fitzhardinge wrote:
I don't think we need a flag day. It seems like we already ship a

udev

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

rule (in tools/hotplug/Linux/xen-backend.rules) which correctly created /dev/xen/evtchn with the current kernel and which is

apparently

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 3 lines

unnecessary (but harmless) with the proposed kernel change.

My main concern is that an old libxc will screw anyone with new

kernel

Quoted 3 lines
Quoted 3 lines
Quoted 2 lines

and udev.

Is it any more likely to screw them with a new kernel than with an

old

Quoted 4 lines
Quoted 2 lines

one?

Yeah, because libxc's rummage around in sysfs will actually work. If

we

Quoted 1 line

rename the devices to be correct, it won't find them and it just ends

up

Quoted 2 lines

deleting the old device and either failing to create a new one, or creating it with a bogus major/minor.

That sucks ;-)

Quoted 1 line
Quoted 1 line

If so I think that's an argument for propagating the removal of

this

Quoted 1 line
Quoted 1 line

functionality into stable trees sooner rather than later rather

than

Quoted 4 lines
Quoted 2 lines

papering over the craziness for even longer.

Yep.

Keir has applied my patch so I'll keep an eye out for fallout and suggest it for backporting to the stable branch when it looks like we've
gotten away with it.

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel