tests: shell: Improve concurrent noflush restore test a bit

The described issue happens only if chain FOO does not exist at program
start so flush the ruleset after each iteration to make sure this is the
case. Sadly the bug is still not 100% reproducible on my testing VM.

While being at it, add a paragraph describing what exact situation the
test is trying to provoke.

Fixes: dac904bdcd9a1 ("nft: Fix for concurrent noflush restore calls")
Signed-off-by: Phil Sutter <phil@nwl.cc>

diff --git a/iptables/tests/shell/testcases/ipt-restore/0016-concurrent-restores_0 b/iptables/tests/shell/testcases/ipt-restore/0016-concurrent-restores_0
index 53ec12fa368af0a5cd78d973a8b180c8cd90a6a3..aa746ab458a3c4bb22912608a6bd5781f1549bc3 100755 (executable)
--- a/iptables/tests/shell/testcases/ipt-restore/0016-concurrent-restores_0
+++ b/iptables/tests/shell/testcases/ipt-restore/0016-concurrent-restores_0
@@ -1,5 +1,14 @@
 #!/bin/bash
 
+# test for iptables-restore --noflush skipping an explicitly requested chain
+# flush because the chain did not exist when cache was fetched. In order to
+# expect for that chain to appear when refreshing the transaction (due to a
+# concurrent ruleset change), the chain flush job has to be present in batch
+# job list (although disabled at first).
+# The input line requesting chain flush is ':FOO - [0:0]'. RS1 and RS2 contents
+# are crafted to cause EBUSY when deleting the BAR* chains if FOO is not
+# flushed in the same transaction.
+
 set -e
 
 RS="*filter
@@ -45,7 +54,12 @@ RS2="$RS
 COMMIT
 "
 
+NORS="*filter
+COMMIT
+"
+
 for n in $(seq 1 10); do
+       $XT_MULTI iptables-restore <<< "$NORS"
        $XT_MULTI iptables-restore --noflush -w <<< "$RS1" &
        $XT_MULTI iptables-restore --noflush -w <<< "$RS2" &
        wait -n