Re: [9front] Kernel memory leak

[Alex Musolino <alex@musolino.id.au>]

> i think the cunique() would *ALWAYS* be required when we are going
> to stuff something on the chan's umh, as otherwise we'd have a race
> condition because the returned chan from domount() -> findmount()
> is shared.

Yeah, it makes more sense to do it at the same time/place as the umh
update.  I guess we do the same thing for the Aopen/Acreate case where
we might also update the umh?  We also change umh in devshr.c but I
guess this is fine since it's only temporary.

> can you check if this works?

Works fine as far as I can tell.  How about the following patch then?

diff -r ec3da4e6c943 sys/src/9/port/chan.c
--- a/sys/src/9/port/chan.c	Fri Aug 21 10:06:22 2020 +0930
+++ b/sys/src/9/port/chan.c	Sun Aug 23 12:01:46 2020 +0930
@@ -1410,7 +1410,11 @@
 		m = nil;
 		if(!nomount)
 			domount(&c, &m, nil);
-		putmhead(c->umh);
+		c = cunique(c);
+		if(c->umh != nil){	//BUG
+			print("bind umh\n");
+			putmhead(c->umh);
+		}
 		c->umh = m;
 		break;
 
@@ -1454,9 +1458,14 @@
 				c->umh = nil;
 			}
 			/* only save the mount head if it's a multiple element union */
-			if(m != nil && m->mount != nil && m->mount->next != nil)
+			if(m != nil && m->mount != nil && m->mount->next != nil){
+				c = cunique(c);
+				if(c->umh != nil){	//BUG
+					print("Aopen/Acreate umh\n");
+					putmhead(c->umh);
+				}
 				c->umh = m;
-			else
+			}else
 				putmhead(m);
 
 			/* save registers else error() in open has wrong value of c saved */

--
Cheers,
Alex Musolino