All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH nft v2] tests: 0044interval_overlap_0: Repeat insertion tests with timeout
@ 2020-08-03 14:06 Stefano Brivio
  2020-08-04 11:43 ` Pablo Neira Ayuso
  0 siblings, 1 reply; 2+ messages in thread
From: Stefano Brivio @ 2020-08-03 14:06 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: Mike Dillinger, Phil Sutter, netfilter-devel

Mike Dillinger reported issues with insertion of entries into sets
supporting intervals that were denied because of false conflicts with
elements that were already expired. Partial failures would occur to,
leading to the generation of new intervals the user didn't specify,
as only the opening or the closing elements wouldn't be inserted.

The reproducer provided by Mike looks like this:

  #!/bin/bash
  nft list set ip filter blacklist4-ip-1m
  for ((i=1;i<=10;i++)); do
        nft add element filter blacklist4-ip-1m {$i.$i.$i.$i}
        sleep 1
  done
  nft list set ip filter blacklist4-ip-1m

which, run in a loop at different intervals, show the different kind
of failures.

Extend the existing test case for overlapping and non-overlapping
intervals to systematically cover sets with a configured timeout.

As reported by Pablo, the test would fail if we keep a one-second
timeout if it runs on a "slow" kernel (e.g. with KASan), using the
libtool wrapper in src/nft as $NFT, because we can't issue 218
commands within one second. To avoid that, introduce an adaptive
timeout based on how many times we can list a single entry with a
fixed one-second timeout.

On a single 2.9GHz AMD Epyc 7351 thread:
                                     test run   nft commands/s  timeout
- src/nft libtool wrapper, KASan:       68.4s          10         32s
- nft binary, KASan:                     5.1s         168          2s
- src/nft libtool wrapper, w/o KASan:   18.3s          37          8s
- nft binary, w/o KASan:                 2.4s         719          1s

While at it, fix expectation for insertion of '15-20 . 50-60' (it's
expected to succeed, given the list), and the reason why I didn't
notice: a simple command preceded by ! won't actually result in
the shell exiting, even if it fails. Add some clearer failure reports
too.

v2:
  - adjust set timeouts to nft commands/s
  - fix checks on expected outcome of insertions and reports

Reported-by: Mike Dillinger <miked@softtalker.com>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
---
 .../testcases/sets/0044interval_overlap_0     | 136 +++++++++++++++---
 1 file changed, 117 insertions(+), 19 deletions(-)

diff --git a/tests/shell/testcases/sets/0044interval_overlap_0 b/tests/shell/testcases/sets/0044interval_overlap_0
index fad92ddcf356..face90f2e9ae 100755
--- a/tests/shell/testcases/sets/0044interval_overlap_0
+++ b/tests/shell/testcases/sets/0044interval_overlap_0
@@ -7,6 +7,20 @@
 #   existing one
 # - for concatenated ranges, the new element is less specific than any existing
 #   overlapping element, as elements are evaluated in order of insertion
+#
+# Then, repeat the test with a set configured with a timeout, checking that:
+# - we can insert all the elements as described above
+# - once the timeout has expired, we can insert all the elements again, and old
+#   elements are not present
+# - before the timeout expires again, we can re-add elements that are not
+#   expected to fail, but old elements might be present
+#
+# If $NFT points to a libtool wrapper, and we're running on a slow machine or
+# kernel (e.g. KASan enabled), it might not be possible to execute hundreds of
+# commands within an otherwise reasonable 1 second timeout. Estimate a usable
+# timeout first, by counting commands and measuring against one nft rule timeout
+# itself, so that we can keep this fast for a binary $NFT on a reasonably fast
+# kernel.
 
 #	Accept	Interval	List
 intervals_simple="
@@ -32,35 +46,119 @@ intervals_concat="
 	y	0-2 . 0-3	0-2 . 0-3
 	n	0-1 . 0-2	0-2 . 0-3
 	y	10-20 . 30-40	0-2 . 0-3, 10-20 . 30-40
-	n	15-20 . 50-60	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60
+	y	15-20 . 50-60	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60
 	y	3-9 . 4-29	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60, 3-9 . 4-29
 	y	3-9 . 4-29	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60, 3-9 . 4-29
 	n	11-19 . 30-40	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60, 3-9 . 4-29
 	y	15-20 . 49-61	0-2 . 0-3, 10-20 . 30-40, 15-20 . 50-60, 3-9 . 4-29, 15-20 . 49-61
 "
 
-$NFT add table t
-$NFT add set t s '{ type inet_service ; flags interval ; }'
-$NFT add set t c '{ type inet_service . inet_service ; flags interval ; }'
+count_elements() {
+	pass=
+	interval=
+	elements=0
+	for t in ${intervals_simple} ${intervals_concat}; do
+		[ -z "${pass}" ]      && pass="${t}"     && continue
+		[ -z "${interval}" ]  && interval="${t}" && continue
+		unset IFS
+
+		elements=$((elements + 1))
 
-IFS='	
+		IFS='	
 '
-set="s"
-for t in ${intervals_simple} switch ${intervals_concat}; do
-	[ "${t}" = "switch" ] && set="c"         && continue
-	[ -z "${pass}" ]      && pass="${t}"     && continue
-	[ -z "${interval}" ]  && interval="${t}" && continue
-
-	if [ "${pass}" = "y" ]; then
-		$NFT add element t ${set} "{ ${interval} }"
-	else
-		! $NFT add element t ${set} "{ ${interval} }" 2>/dev/null
+	done
+	unset IFS
+}
+
+match_elements() {
+	skip=0
+	n=0
+	out=
+	for a in $($NFT list set t ${1})}; do
+		[ ${n} -eq 0 ] && { [ "${a}" = "elements" ] && n=1; continue; }
+		[ ${n} -eq 1 ] && { [ "${a}" = "=" ] 	    && n=2; continue; }
+		[ ${n} -eq 2 ] && { [ "${a}" = "{" ]	    && n=3; continue; }
+
+		[ "${a}" = "}" ]				 && break
+
+		[ ${skip} -eq 1 ] && skip=0 && out="${out},"	 && continue
+		[ "${a}" = "expires" ] && skip=1		 && continue
+
+		[ -n "${out}" ] && out="${out} ${a}" || out="${a}"
+
+	done
+
+	if [ "${out%,}" != "${2}" ]; then
+		echo "Expected: ${2}, got: ${out%,}"
+		return 1
 	fi
-	$NFT list set t ${set} | tr -d '\n\t' | tr -s ' ' | \
-		grep -q "elements = { ${t} }"
+}
+
+estimate_timeout() {
+	count_elements
 
+	$NFT add table t
+	$NFT add set t s '{ type inet_service ; flags timeout; timeout 1s; gc-interval 1s; }'
+	execs_1s=1
+	$NFT add element t s "{ 0 }"
+	while match_elements s "0" >/dev/null; do
+		execs_1s=$((execs_1s + 1))
+	done
+
+	timeout="$((elements / execs_1s * 3 / 2 + 1))"
+}
+
+add_elements() {
+	set="s"
 	pass=
 	interval=
-done
+	IFS='	
+'
+	for t in ${intervals_simple} switch ${intervals_concat}; do
+		[ "${t}" = "switch" ] && set="c"         && continue
+		[ -z "${pass}" ]      && pass="${t}"     && continue
+		[ -z "${interval}" ]  && interval="${t}" && continue
+		unset IFS
+
+		if [ "${pass}" = "y" ]; then
+			if ! $NFT add element t ${set} "{ ${interval} }"; then
+				echo "Failed to insert ${interval} given:"
+				$NFT list ruleset
+				exit 1
+			fi
+		else
+			if $NFT add element t ${set} "{ ${interval} }" 2>/dev/null; then
+				echo "Could insert ${interval} given:"
+				$NFT list ruleset
+				exit 1
+			fi
+		fi
+
+		[ "${1}" != "nomatch" ] && match_elements "${set}" "${t}"
+
+		pass=
+		interval=
+		IFS='	
+'
+	done
+	unset IFS
+}
+
+$NFT add table t
+$NFT add set t s '{ type inet_service ; flags interval ; }'
+$NFT add set t c '{ type inet_service . inet_service ; flags interval ; }'
+add_elements
+
+$NFT flush ruleset
+estimate_timeout
+
+$NFT flush ruleset
+$NFT add table t
+$NFT add set t s "{ type inet_service ; flags interval,timeout; timeout ${timeout}s; gc-interval ${timeout}s; }"
+$NFT add set t c "{ type inet_service . inet_service ; flags interval,timeout ; timeout ${timeout}s; gc-interval ${timeout}s; }"
+add_elements
+
+sleep $((timeout * 3 / 2))
+add_elements
 
-unset IFS
+add_elements nomatch
-- 
2.27.0


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH nft v2] tests: 0044interval_overlap_0: Repeat insertion tests with timeout
  2020-08-03 14:06 [PATCH nft v2] tests: 0044interval_overlap_0: Repeat insertion tests with timeout Stefano Brivio
@ 2020-08-04 11:43 ` Pablo Neira Ayuso
  0 siblings, 0 replies; 2+ messages in thread
From: Pablo Neira Ayuso @ 2020-08-04 11:43 UTC (permalink / raw)
  To: Stefano Brivio; +Cc: Mike Dillinger, Phil Sutter, netfilter-devel

On Mon, Aug 03, 2020 at 04:06:21PM +0200, Stefano Brivio wrote:
> Mike Dillinger reported issues with insertion of entries into sets
> supporting intervals that were denied because of false conflicts with
> elements that were already expired. Partial failures would occur to,
> leading to the generation of new intervals the user didn't specify,
> as only the opening or the closing elements wouldn't be inserted.

Applied, thanks.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2020-08-04 11:44 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-03 14:06 [PATCH nft v2] tests: 0044interval_overlap_0: Repeat insertion tests with timeout Stefano Brivio
2020-08-04 11:43 ` Pablo Neira Ayuso

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.