T322: defer pagetable cap deletion to receive handler

On some machines, waiting until all messages were sent to delete the pagetable
cap in the sender was not enough, and the receiver would still get an
unexpected error while trying to identify the cap.

Signed-off-by: Simon Gerber <simon.gerber@inf.ethz.ch>

diff --git a/usr/tests/multihoptest/multihoptest.c b/usr/tests/multihoptest/multihoptest.c
index ea9f1cd..615dc9c 100644 (file)
--- a/usr/tests/multihoptest/multihoptest.c
+++ b/usr/tests/multihoptest/multihoptest.c
@@ -127,7 +127,10 @@ static void send_cont(void *arg)
         break;
 
     case 3:
-        // delete the caps, now they've been sent
+        // delete the Frame cap, now it's been sent, the page table cap is not
+        // movable, and if the sender (who's the owner) deletes it before the
+        // receiver is done with it's work, we run into odd situations.
+        // -SG 2016-11-08
         err = cap_destroy(myst->cap1);
         assert(err_is_ok(err));
 
@@ -139,11 +142,6 @@ static void send_cont(void *arg)
         // here is where we would deallocate the buffer, if it wasn't static
         printf("%s all done!\n", get_role_name());
 
-        // Clean up page table cap late as it is not moveable and deleting it
-        // to early will delete the receiver's copy as well! -SG,2016-11-07
-        err = cap_destroy(myst->cap2);
-        assert(err_is_ok(err));
-
         return;
 
     default:
@@ -213,6 +211,9 @@ static void rx_caps(struct test_binding *b, uint32_t arg, struct capref cap1,
     buf1[sizeof(buf1) - 1] = '\0';
     buf2[sizeof(buf2) - 1] = '\0';
     cap_destroy(cap1);
+    // Need to revoke and delete pagetable cap here, as otherwise it might get
+    // deleted before we get our hands on it! -SG,2016-11-08
+    cap_revoke(cap2);
     cap_destroy(cap2);
     printf("%s rx_caps %"PRIu32" [%s] [%s]\n", get_role_name(), arg, buf1, buf2);
 }