RE: [PATCH] Fix SEGVs in xenconsoled

RE: [PATCH] Fix SEGVs in xenconsoled

[Ian Pratt]

> Under the right circumstances, xenconsoled will corrupt its 
> internal list of domains causing a SEGV.  This is usually 
> characterized by a rapid number of creations/destructions.  
> The attached patch fixes this.

I'm seeing frequent short pauses and even ocassional hangs from console
clients. Any chance this will fix that?

Thanks,
Ian

Re: [PATCH] Fix SEGVs in xenconsoled

[Anthony Liguori]

Ian Pratt wrote:

>>Under the right circumstances, xenconsoled will corrupt its 
>>internal list of domains causing a SEGV.  This is usually 
>>characterized by a rapid number of creations/destructions.  
>>The attached patch fixes this.
>>    
>>
>
>I'm seeing frequent short pauses and even ocassional hangs from console
>clients. Any chance this will fix that?
>  
>
I think this should fix the hangs from clients.  The code base is small 
so once I can reproduce a problem I can solve it pretty quickly.

The short pauses are part of the throttling to combat corruption.  We 
could get rid of the throttling and just deal with the corruption for 
now until we get the new console drivers.

Regards,

Anthony Liguori

>Thanks,
>Ian
>
>  
>

[PATCH] Fix SEGVs in xenconsoled

[Anthony Liguori]

[-- Attachment #1: Type: text/plain, Size: 240 bytes --]

Under the right circumstances, xenconsoled will corrupt its internal 
list of domains causing a SEGV.  This is usually characterized by a 
rapid number of creations/destructions.  The attached patch fixes this.

Regards,

Anthony Liguori



[-- Attachment #2: consoled_segv.diff --]
[-- Type: text/x-patch, Size: 2288 bytes --]

# HG changeset patch
# User Anthony Liguori <aliguori@us.ibm.com>
# Node ID f8e21ceb6820b6bdb48a161c5f401f3577a14fde
# Parent  a5e3caa6efbf3a46515c8b4f39219eaed75b2b6c
1) Fix uninitialized next pointer.  This could sometimes cause xenconsoled to
   SEGV on an invalid domain pointer
2) Fix race condition in iterating domain list where removing a domain in a
   callback could lead to the iterators becoming invalid.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

diff -r a5e3caa6efbf -r f8e21ceb6820 tools/console/daemon/io.c
--- a/tools/console/daemon/io.c	Sat Aug 13 19:31:13 2005
+++ b/tools/console/daemon/io.c	Mon Aug 15 20:11:29 2005
@@ -87,6 +87,7 @@
 {
 	int domid;
 	int tty_fd;
+	bool is_dead;
 	struct buffer buffer;
 	struct domain *next;
 };
@@ -156,10 +157,12 @@
 
 	dom->domid = domid;
 	dom->tty_fd = domain_create_tty(dom);
+	dom->is_dead = false;
 	dom->buffer.data = 0;
 	dom->buffer.size = 0;
 	dom->buffer.capacity = 0;
 	dom->buffer.max_capacity = 0;
+	dom->next = 0;
 
 	dolog(LOG_DEBUG, "New domain %d", domid);
 
@@ -206,6 +209,16 @@
 	}
 }
 
+static void remove_dead_domains(struct domain *dom)
+{
+	if (dom == NULL) return;
+	remove_dead_domains(dom->next);
+
+	if (dom->is_dead) {
+		remove_domain(dom);
+	}
+}
+
 static void handle_tty_read(struct domain *dom)
 {
 	ssize_t len;
@@ -224,7 +237,7 @@
 		if (domain_is_valid(dom->domid)) {
 			dom->tty_fd = domain_create_tty(dom);
 		} else {
-			remove_domain(dom);
+			dom->is_dead = true;
 		}
 	} else if (domain_is_valid(dom->domid)) {
 		msg.u.control.msg.length = len;
@@ -235,7 +248,7 @@
 		}
 	} else {
 		close(dom->tty_fd);
-		remove_domain(dom);
+		dom->is_dead = true;
 	}
 }
 
@@ -250,7 +263,7 @@
 		if (domain_is_valid(dom->domid)) {
 			dom->tty_fd = domain_create_tty(dom);
 		} else {
-			remove_domain(dom);
+			dom->is_dead = true;
 		}
 	} else {
 		buffer_advance(&dom->buffer, len);
@@ -333,13 +346,15 @@
 		}
 
 		for (d = dom_head; d; d = d->next) {
-			if (FD_ISSET(d->tty_fd, &readfds)) {
+			if (!d->is_dead && FD_ISSET(d->tty_fd, &readfds)) {
 				handle_tty_read(d);
 			}
 
-			if (FD_ISSET(d->tty_fd, &writefds)) {
+			if (!d->is_dead && FD_ISSET(d->tty_fd, &writefds)) {
 				handle_tty_write(d);
 			}
 		}
+
+		remove_dead_domains(dom_head);
 	} while (ret > -1);
 }

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

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