From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-oo1-f52.google.com (mail-oo1-f52.google.com [209.85.161.52]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 17C552BEC2B for ; Thu, 23 Jul 2026 23:49:01 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.161.52 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784850543; cv=none; b=A1tZVDRmk2IDdCOXfj/GZRhw1NVYBKMrHeJgC02eZFsw80n9SKa1TCZ/Tii70DgnpdDtRdmRdN/Nm44B0OBPNojxC0PzmNo41ibhXEYFss/BaqAS8JMqggHKn5TJTmLHhlQeUwGwIkHNRVa7kDIBe1MoJBSqCIS40ZQgRTZy2FA= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784850543; c=relaxed/simple; bh=JkKgwfbwSWE/W2vX/kkIONFlxUMWhEmRwOzG1inAE8M=; h=From:To:Cc:Subject:Date:Message-ID:MIME-Version:Content-Type; b=J8K7dzqyD83J5hiCr3JATXfcVDFLgkou/Zw4jH+nj5wnWXeENYzbVCuIxCMPsiQ0pmDqmHCTNOx4ocHW/Dpk1CoOlcY35EtCt2bCbyJorw4OxQ5hrmTx0WNTsqFYWQwDTZdsy17fjSc8NkGhWtqddTxBbZGuoyJjs0ezvktUxaM= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=r0nP4WlM; arc=none smtp.client-ip=209.85.161.52 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="r0nP4WlM" Received: by mail-oo1-f52.google.com with SMTP id 006d021491bc7-6a31c05e092so522585eaf.0 for ; Thu, 23 Jul 2026 16:49:01 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1784850541; x=1785455341; darn=vger.kernel.org; h=content-transfer-encoding:content-type:mime-version:message-id:date :subject:cc:to:from:from:to:cc:subject:date:message-id:reply-to :content-type; bh=8SM4Olva3UxHUMc0jHFquvfoKH+36uv2FSDslSJ2+uw=; b=r0nP4WlMSrLXySNH1N0znE9yrdne0mhGmO5ZrFLzo8rdV0F2bP+Wx5RldOu+ANri2V p33IrzxGR6vrJTELntNS01g1YE+WS29sZFaFWNx3mDUUBMDO6B/abAqtAI2CYOgWRAh6 +dPxJEGN7qsSjAIr+oL8GRUdOSm33n+wvNRRRefOJxz6TsdQA3cyClNY0ybbha1bwpB0 Q1nBAm4+9t8Usl/dHijqeAvZ6hPH+Y8BfwUgABvk4grd0o44UAzFlXZl30MFVaEypw01 fsBOTgF9iUNLuPmIznQy3OL0MTsSJVy6YQgRQsaFXWcSaIjBFqfm7lNtM4BEMav913n9 9jng== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1784850541; x=1785455341; h=content-transfer-encoding:content-type:mime-version:message-id:date :subject:cc:to:from:x-gm-gg:x-gm-message-state:from:to:cc:subject :date:message-id:reply-to:content-type; bh=8SM4Olva3UxHUMc0jHFquvfoKH+36uv2FSDslSJ2+uw=; b=SUGxCi/S7aZd6uN+FPEKikFDPQKow7zRWVU4MAFfyY3AqQP4NzcQFscsxaUiy50ael Pvx4k3bJlGZs8LKdlT6MRkVANwuEJ5him0u9/GCECd+YpJbxvz1tLC4GSUAPmhFLX4cb h10Zu05q+YDZVdW/zvByJ5LGrTaQgcU4HHmkf7She89yEAlObPseUGMiHnzSg7xUdvca BriUQrVxrsM879Drwds3g/SDqC9y2Zl3j4aOBQFkXHixTNILvYT68SBMvJxXMMP/T04k a6ThncDx97PQB7OGip8aLe2ipTC0T4EP3TYXPn/p0rHD1qkYmIFSzBsd7fwdCCkcyySX ZR2A== X-Forwarded-Encrypted: i=1; AHgh+RryDozIWP+1RIaunTvI7fSd9gyVOZ4oij0e8NzeIwTaGeWhaT7RK0rsz4YJwnbrP0+JRAAoUXpYxOTDqaJHmKo=@vger.kernel.org X-Gm-Message-State: AOJu0Yw2pSo9JUeYwNOJaHVv0tJTUgbYjiFapenMjiN+ZuwfQIjdYw47 pFyHgEG+RZAfA7dytqMFhjvP+N0TBhdvGbNqTe3c3ij3z9emZwUQCjhd X-Gm-Gg: AR+sD129beef6E5kAD4EIF36IvQA2EBOM5yDJZKL7LdPlHhnlv+UwJLioWhE4MVsC9K qc0RUCzTbO/Sgg1Utk1+6Sqlac4XYcWzoxAlD+7KLN0B30AbRGIkcDSEiGIL0bz3YOGa1kcE5MO 59J1t+YF8leL0bKyDK5almjfFNLR4p/xRd6g8qqpIJC5tciZBKNSwtWye+dykeQ1isEZClhzpQZ jD6U/d91zL/ItQGuDBPVyMX+qo/LcFw6vJ6iBUkBAa2Ey61IhdF/2umyViooRQ0sVT8PPaomQYU wB2CA+X96rmg61UI6QuFx0dxH/D4QvMKMzQVznnLBYY794SwL/k/N/A8O9uvVrGN5Wn2xRsh2FD ApzyNxrhq62Qw1/L9PEYn1yJsCVz+qXEBf2jzla/mUsqcElEt/16lhrvjkwRB7U0HiW7BBXpOK8 I= X-Received: by 2002:a05:6820:4c14:b0:696:836e:ba2b with SMTP id 006d021491bc7-6aad3f4c0demr2535127eaf.11.1784850540853; Thu, 23 Jul 2026 16:49:00 -0700 (PDT) Received: from localhost ([2a03:2880:31ff:43::]) by smtp.gmail.com with ESMTPSA id 586e51a60fabf-45766f16dfcsm5904191fac.7.2026.07.23.16.48.59 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 23 Jul 2026 16:49:00 -0700 (PDT) From: Mohsin Bashir To: netdev@vger.kernel.org Cc: andrew+netdev@lunn.ch, davem@davemloft.net, edumazet@google.com, kuba@kernel.org, pabeni@redhat.com, shuah@kernel.org, linux-kselftest@vger.kernel.org, mohsin.bashr@gmail.com Subject: [PATCH net-next V3] selftests: drv-net: Test queue stall upon reconfig Date: Thu, 23 Jul 2026 16:48:55 -0700 Message-ID: <20260723234857.2880270-1-mohsin.bashr@gmail.com> X-Mailer: git-send-email 2.53.0 Precedence: bulk X-Mailing-List: linux-kselftest@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From: Mohsin Bashir Add a reconfig_tx_stall test that detects the possibility of a TX stall after ring reconfiguration. The key observation is that drivers using netif_tx_start_all_queues() are prone to experiencing a stall when reconfiguration completes compared to drivers using netif_tx_wake_all_queues(). start_all_queues only clears DRV_XOFF, while wake_all_queues also calls __netif_schedule() to kick the qdisc. Without the kick, qdisc backlog present at reconfig time can stay stuck until a new trigger is issued. The test caps the TX ring at 64 entries so it fills quickly, then installs FQ on a target TX queue and sends UDP packets with SO_TXTIME scheduled in the future. With napi_defer_hard_irqs slowing completions, the small ring can fill when FQ releases the burst, leaving requeued qdisc backlog with no FQ timer to rescue it. A subsequent ring reconfig must wake the queues to drain the backlog. Simply starting the queues can leave it stuck. On host with problematic driver: Sent 128 SO_TXTIME packets (+100ms) Sent 128 SO_TXTIME packets (+200ms) Backlog before reconfig: 52632 bytes Check| At /root/ksft-net-drv/./drivers/net/ring_reconfig.py, ... Check| ksft_eq(0, backlog, Check failed 0 != 52632 qdisc backlog stuck on queue 1 after ring reconfig not ok 3 ring_reconfig.reconfig_tx_stall On host with fixed driver: Sent 128 SO_TXTIME packets (+100ms) Sent 128 SO_TXTIME packets (+200ms) Backlog before reconfig: 76024 bytes ok 3 ring_reconfig.reconfig_tx_stall Signed-off-by: Mohsin Bashir Signed-off-by: Jakub Kicinski --- Changelog: V3: - Preserve the device's existing TX/RX qdisc policy. On a device with a previously-configured (addressable) mq, only the target queue's leaf is swapped and restored. The default mq (unaddressable handle 0:) or a noqueue root is temporarily given a real handle and restored to the default mq on exit. Skip when the root is a non-mq qdisc rather than disturb it. - Request 2+ queues from NetDrvEpEnv via a new queue_count parameter so the test runs under netdevsim. - Fail instead of skip when the qdisc backlog cannot be built. - Skip on devices without 2+ combined channels; bump combined-count to 2 (restored on exit) when supported but not currently configured. V2: https://lore.kernel.org/netdev/20260613014855.1717712-1-mohsin.bashr@gmail.com V1: https://lore.kernel.org/netdev/20260612001754.2489868-1-mohsin.bashr@gmail.com --- tools/testing/selftests/drivers/net/config | 3 + .../selftests/drivers/net/lib/py/env.py | 11 +- .../selftests/drivers/net/ring_reconfig.py | 207 +++++++++++++++++- 3 files changed, 215 insertions(+), 6 deletions(-) diff --git a/tools/testing/selftests/drivers/net/config b/tools/testing/selftests/drivers/net/config index 2070e890e064..f3933cf3e6be 100644 --- a/tools/testing/selftests/drivers/net/config +++ b/tools/testing/selftests/drivers/net/config @@ -4,8 +4,11 @@ CONFIG_DEBUG_INFO_BTF_MODULES=n CONFIG_INET_PSP=y CONFIG_IPV6=y CONFIG_MACSEC=m +CONFIG_NET_ACT_SKBEDIT=m CONFIG_NET_CLS_ACT=y CONFIG_NET_CLS_BPF=y +CONFIG_NET_CLS_FLOWER=m +CONFIG_NET_CLS_MATCHALL=m CONFIG_NETCONSOLE=m CONFIG_NETCONSOLE_DYNAMIC=y CONFIG_NETCONSOLE_EXTENDED_LOG=y diff --git a/tools/testing/selftests/drivers/net/lib/py/env.py b/tools/testing/selftests/drivers/net/lib/py/env.py index e4acf3d8333f..b1c6f6cef7dd 100644 --- a/tools/testing/selftests/drivers/net/lib/py/env.py +++ b/tools/testing/selftests/drivers/net/lib/py/env.py @@ -114,10 +114,11 @@ class NetDrvEpEnv(NetDrvEnvBase): nsim_v4_pfx = "192.0.2." nsim_v6_pfx = "2001:db8::" - def __init__(self, src_path, nsim_test=None): + def __init__(self, src_path, nsim_test=None, queue_count=None): super().__init__(src_path) self._stats_settle_time = None + self._queue_count = queue_count # Things we try to destroy self.remote = None @@ -173,9 +174,13 @@ class NetDrvEpEnv(NetDrvEnvBase): self._required_cmd = {} def create_local(self): + nsim_kwargs = {} + if self._queue_count: + nsim_kwargs["queue_count"] = self._queue_count + self._netns = NetNS() - self._ns = NetdevSimDev() - self._ns_peer = NetdevSimDev(ns=self._netns) + self._ns = NetdevSimDev(**nsim_kwargs) + self._ns_peer = NetdevSimDev(ns=self._netns, **nsim_kwargs) with open("/proc/self/ns/net") as nsfd0, \ open("/var/run/netns/" + self._netns.name) as nsfd1: diff --git a/tools/testing/selftests/drivers/net/ring_reconfig.py b/tools/testing/selftests/drivers/net/ring_reconfig.py index f9530a8b0856..a5a2defc4fb1 100755 --- a/tools/testing/selftests/drivers/net/ring_reconfig.py +++ b/tools/testing/selftests/drivers/net/ring_reconfig.py @@ -5,10 +5,21 @@ Test channel and ring size configuration via ethtool (-L / -G). """ +import socket +import struct +import time + from lib.py import ksft_run, ksft_exit, ksft_pr from lib.py import ksft_eq +from lib.py import KsftSkipEx, KsftFailEx from lib.py import NetDrvEpEnv, EthtoolFamily, GenerateTraffic -from lib.py import defer, NlError +from lib.py import cmd, defer, rand_port, tc, NlError + +# Added in Python 3.13; fallback to 61 for x86/ARM/MIPS +SO_TXTIME = getattr(socket, "SO_TXTIME", 61) + +# TX ring size the test shrinks to so the ring fills quickly. +MIN_TX_RING = 64 def channels(cfg) -> None: @@ -151,14 +162,204 @@ def ringparam(cfg) -> None: GenerateTraffic(cfg).wait_pkts_and_stop(10000) +def _write_file(path, val): + """Write val to a file.""" + with open(path, "w", encoding="utf-8") as fp: + fp.write(str(val)) + + +def _write_sysfs(path, val): + """Write val to a sysfs file, restoring the original value on exit.""" + with open(path, "r", encoding="utf-8") as fp: + orig_val = fp.read().strip() + if str(val) == orig_val: + return + _write_file(path, val) + defer(_write_file, path, orig_val) + + +def _get_qdisc_backlog(cfg, mq_handle, queue): + """Return the qdisc backlog (bytes) for the given TX queue's leaf.""" + target_parent = f"{mq_handle}{queue + 1:x}" + for q in tc(f"-s qdisc show dev {cfg.ifname}", json=True): + if q.get("parent", "") == target_parent: + return q.get("backlog") or 0 + return 0 + + +def _setup_fq_qdisc(cfg, port, target_queue, other_queue): + """Put an fq qdisc on target_queue's leaf and return the mq handle in use. + + We must not disturb the device's existing TX/RX qdisc policy. On a real + NIC the root mq already has an addressable handle, so we leave the root + and every other queue alone and only swap this one leaf, restoring its + original qdisc afterwards. + """ + qdiscs = tc(f"qdisc show dev {cfg.ifname}", json=True) + root = next((q for q in qdiscs if q.get("root")), None) + + if root and root["kind"] == "mq" and root["handle"] != "0:": + # Addressable mq (previously-configured): touch only the target queue's + # leaf and restore its original qdisc afterwards. + mq_handle = root["handle"] + parent = f"{mq_handle}{target_queue + 1:x}" + orig = next((q for q in qdiscs if q.get("parent") == parent), None) + orig_kind = orig["kind"] if orig else \ + cmd("sysctl -n net.core.default_qdisc").stdout.strip() + defer(tc, f"qdisc replace dev {cfg.ifname} parent {parent} {orig_kind}") + elif root is None or root["kind"] in ("mq", "noqueue"): + # The auto-attached root mq has handle 0: on any device (real or sim), + # which the kernel rejects as a qdisc parent. A 0: handle means the mq + # is the untouched kernel default - no custom child qdiscs can hang off + # an unaddressable parent - so installing a real handle and restoring + # the default mq on exit preserves the device's effective policy. + mq_handle = "1:" + tc(f"qdisc replace dev {cfg.ifname} root handle {mq_handle} mq") + defer(tc, f"qdisc replace dev {cfg.ifname} root mq") + parent = f"{mq_handle}{target_queue + 1:x}" + else: + raise KsftSkipEx(f"root qdisc '{root['kind']}' is not mq; " + "refusing to disturb existing qdisc policy") + + try: + tc(f"qdisc replace dev {cfg.ifname} parent {parent} fq") + except Exception as exc: + raise KsftSkipEx( + f"fq not available (CONFIG_NET_SCH_FQ): {exc}") from exc + + qdisc_j = tc(f"qdisc show dev {cfg.ifname}", json=True) + has_clsact = any(q['kind'] == 'clsact' for q in qdisc_j) + if not has_clsact: + tc(f"qdisc add dev {cfg.ifname} clsact") + defer(tc, f"qdisc del dev {cfg.ifname} clsact") + + proto = "ipv6" if int(cfg.addr_ipver) == 6 else "ip" + try: + tc(f"filter add dev {cfg.ifname} egress protocol {proto} " + f"pref 1 flower ip_proto udp dst_port {port} " + f"action skbedit queue_mapping {target_queue}") + except Exception as exc: + raise KsftSkipEx("tc flower/act_skbedit not available") from exc + defer(tc, f"filter del dev {cfg.ifname} egress pref 1") + + tc(f"filter add dev {cfg.ifname} egress pref 100 " + f"matchall action skbedit queue_mapping {other_queue}") + defer(tc, f"filter del dev {cfg.ifname} egress pref 100") + + return mq_handle + + +def _create_sotxtime_socket(cfg): + """Create a UDP socket with SO_TXTIME enabled, bound to the test device.""" + sock = socket.socket(socket.AF_INET6 if cfg.addr_ipver == "6" + else socket.AF_INET, socket.SOCK_DGRAM) + try: + sock.setsockopt(socket.SOL_SOCKET, SO_TXTIME, struct.pack("Ii", 1, 0)) + except OSError as exc: + sock.close() + raise KsftSkipEx("SO_TXTIME not supported") from exc + sock.setsockopt(socket.SOL_SOCKET, socket.SO_BINDTODEVICE, + cfg.ifname.encode()) + return sock + + +def _send_sotxtime_burst(cfg, sock, port, count, delay_ns): + """Send count UDP packets scheduled delay_ns ahead using SO_TXTIME.""" + payload = b'\x00' * 1400 + txtime_ns = time.clock_gettime_ns(time.CLOCK_MONOTONIC) + delay_ns + + ancdata = [(socket.SOL_SOCKET, SO_TXTIME, struct.pack("Q", txtime_ns))] + if int(cfg.addr_ipver) == 6: + dest = (cfg.remote_addr, port, 0, 0) + else: + dest = (cfg.remote_addr, port) + for _ in range(count): + sock.sendmsg([payload], ancdata, 0, dest) + + +def reconfig_tx_stall(cfg) -> None: + """Test that qdisc backlog drains after ring reconfiguration.""" + target_queue = 1 + other_queue = 0 + + ehdr = {'header': {'dev-index': cfg.ifindex}} + chans = cfg.eth.channels_get(ehdr) + + if "combined-max" not in chans: + raise KsftSkipEx("device does not support combined channels") + if chans.get("combined-max", 0) < 2: + raise KsftSkipEx("device does not support 2+ combined channels") + if chans["combined-count"] < 2: + defer(cfg.eth.channels_set, + ehdr | {"combined-count": chans["combined-count"]}) + cfg.eth.channels_set(ehdr | {"combined-count": 2}) + + rings = cfg.eth.rings_get(ehdr) + if 'rx' not in rings or 'tx' not in rings: + raise KsftSkipEx("device does not expose rx/tx ring params") + tx_cur = rings['tx'] + if tx_cur <= MIN_TX_RING: + raise KsftSkipEx("tx ring size already at minimum") + defer(cfg.eth.rings_set, ehdr | {'tx': tx_cur}) + + cfg.eth.rings_set(ehdr | {'tx': MIN_TX_RING}) + + # Slow completions so the ring stays full after FQ releases packets + napi_defer = f"/sys/class/net/{cfg.ifname}/napi_defer_hard_irqs" + gro_timeout = f"/sys/class/net/{cfg.ifname}/gro_flush_timeout" + _write_sysfs(napi_defer, 100) + _write_sysfs(gro_timeout, 1000000000) + + port = rand_port() + mq_handle = _setup_fq_qdisc(cfg, port, target_queue, other_queue) + + sock = _create_sotxtime_socket(cfg) + defer(sock.close) + + pkt_count = MIN_TX_RING * 2 + + for delay_ms in [100, 200, 500]: + _send_sotxtime_burst(cfg, sock, port, pkt_count, + delay_ms * 1_000_000) + ksft_pr(f"Sent {pkt_count} SO_TXTIME packets (+{delay_ms}ms)") + time.sleep(delay_ms / 1000 + 0.3) + + backlog = _get_qdisc_backlog(cfg, mq_handle, target_queue) + if backlog: + break + else: + raise KsftFailEx("failed to build qdisc backlog") + + ksft_pr(f"Backlog before reconfig: {backlog} bytes") + + # Trigger ring reconfig — driver should call wake, not just start + cfg.eth.rings_set(ehdr | {'tx': tx_cur}) + + # Let completions proceed normally + _write_sysfs(napi_defer, 0) + _write_sysfs(gro_timeout, 0) + + # Poll for backlog to drain + for _ in range(100): + backlog = _get_qdisc_backlog(cfg, mq_handle, target_queue) + if not backlog: + break + time.sleep(0.1) + + ksft_eq(0, backlog, + comment=f"qdisc backlog stuck on queue {target_queue} " + f"after ring reconfig") + + def main() -> None: """ Ksft boiler plate main """ - with NetDrvEpEnv(__file__) as cfg: + with NetDrvEpEnv(__file__, queue_count=2) as cfg: cfg.eth = EthtoolFamily() ksft_run([channels, - ringparam], + ringparam, + reconfig_tx_stall], args=(cfg, )) ksft_exit() -- 2.53.0-Meta