Merge tag 'net-5.19-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski: "Including fixes from netfilter. Current release - new code bugs: - clear msg_get_inq in __sys_recvfrom() and __copy_msghdr_from_user() - mptcp: - invoke MP_FAIL response only when needed - fix shutdown vs fallback race - consistent map handling on failure - octeon_ep: use bitwise AND Previous releases - regressions: - tipc: move bc link creation back to tipc_node_create, fix NPD Previous releases - always broken: - tcp: add a missing nf_reset_ct() in 3WHS handling to prevent socket buffered skbs from keeping refcount on the conntrack module - ipv6: take care of disable_policy when restoring routes - tun: make sure to always disable and unlink NAPI instances - phy: don't trigger state machine while in suspend - netfilter: nf_tables: avoid skb access on nf_stolen - asix: fix "can't send until first packet is send" issue - usb: asix: do not force pause frames support - nxp-nci: don't issue a zero length i2c_master_read() Misc: - ncsi: allow use of proper "mellanox" DT vendor prefix - act_api: add a message for user space if any actions were already flushed before the error was hit" * tag 'net-5.19-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (55 commits) net: dsa: felix: fix race between reading PSFP stats and port stats selftest: tun: add test for NAPI dismantle net: tun: avoid disabling NAPI twice net: sparx5: mdb add/del handle non-sparx5 devices net: sfp: fix memory leak in sfp_probe() mlxsw: spectrum_router: Fix rollback in tunnel next hop init net: rose: fix UAF bugs caused by timer handler net: usb: ax88179_178a: Fix packet receiving net: bonding: fix use-after-free after 802.3ad slave unbind ipv6: fix lockdep splat in in6_dump_addrs() net: phy: ax88772a: fix lost pause advertisement configuration net: phy: Don't trigger state machine while in suspend usbnet: fix memory allocation in helpers selftests net: fix kselftest net fatal error NFC: nxp-nci: don't print header length mismatch on i2c error NFC: nxp-nci: Don't issue a zero length i2c_master_read() net: tipc: fix possible refcount leak in tipc_sk_create() nfc: nfcmrvl: Fix irq_of_parse_and_map() return value net: ipv6: unexport __init-annotated seg6_hmac_net_init() ipv6/sit: fix ipip6_tunnel_get_prl return value ...

Linus Torvalds committed 4 years ago

Commit 5e837932 parents 868f9f2 58bf4db58 files · +905 −254Browse files

Unified · Split

58 files

MAINTAINERS +1 −2
drivers/net/bonding/bond_3ad.c +2 −1
drivers/net/bonding/bond_alb.c +1 −1
drivers/net/dsa/bcm_sf2.c +5 −0
drivers/net/dsa/hirschmann/hellcreek_ptp.c +1 −0
drivers/net/dsa/ocelot/felix_vsc9959.c +4 −0
drivers/net/ethernet/marvell/octeon_ep/octep_regs_cn9k_pf.h +1 −1
drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c +13 −1
drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c +6 −0
drivers/net/ethernet/smsc/epic100.c +2 −2
drivers/net/phy/ax88796b.c +4 −2
drivers/net/phy/dp83822.c +1 −3
drivers/net/phy/phy.c +23 −0
drivers/net/phy/phy_device.c +23 −0
drivers/net/phy/sfp.c +1 −1
drivers/net/tun.c +13 −2
drivers/net/usb/asix.h +1 −2
drivers/net/usb/asix_common.c +1 −0
drivers/net/usb/ax88179_178a.c +76 −25
drivers/net/usb/usbnet.c +2 −2
drivers/nfc/nfcmrvl/i2c.c +3 −3
drivers/nfc/nfcmrvl/spi.c +3 −3
drivers/nfc/nxp-nci/i2c.c +9 −2
include/linux/netdevice.h +1 −1
include/linux/phy.h +6 −0
include/net/netfilter/nf_tables.h +10 −6
include/uapi/linux/mptcp.h +5 −4
net/bridge/br_netfilter_hooks.c +18 −3
net/ipv4/ip_tunnel_core.c +1 −1
net/ipv4/tcp_ipv4.c +5 −1
net/ipv6/addrconf.c +2 −6
net/ipv6/route.c +8 −1
net/ipv6/seg6_hmac.c +0 −1
net/ipv6/sit.c +3 −5
net/mptcp/options.c +4 −3
net/mptcp/pm.c +4 −6
net/mptcp/protocol.c +53 −31
net/mptcp/protocol.h +20 −4
net/mptcp/subflow.c +98 −29
net/ncsi/ncsi-manage.c +2 −1
net/netfilter/nf_tables_core.c +21 −3
net/netfilter/nf_tables_trace.c +24 −20
net/netfilter/nft_set_hash.c +2 −0
net/rose/rose_timer.c +19 −15
net/sched/act_api.c +14 −8
net/socket.c +6 −10
net/tipc/node.c +22 −19
net/tipc/socket.c +1 −0
tools/testing/selftests/bpf/prog_tests/sockmap_ktls.c +75 −9
tools/testing/selftests/net/Makefile +1 −1
tools/testing/selftests/net/bpf/Makefile +1 −1
tools/testing/selftests/net/mptcp/diag.sh +40 −8
tools/testing/selftests/net/mptcp/mptcp_connect.c +1 −1
tools/testing/selftests/net/mptcp/mptcp_inq.c +1 −1
tools/testing/selftests/net/mptcp/mptcp_sockopt.c +1 −1
tools/testing/selftests/net/tun.c +162 −0
tools/testing/selftests/net/udpgso_bench.sh +1 −1
tools/testing/selftests/tc-testing/tc-tests/actions/gact.json +77 −0

MAINTAINERS+1−2

@@ -14397,9 +14397,8 @@ F: Documentation/devicetree/bindings/sound/nxp,tfa989x.yaml · 14396 unmodified lines
14397	14397	F: sound/soc/codecs/tfa989x.c
14398	14398
14399	14399	NXP-NCI NFC DRIVER
14400		−R: Charles Gorand <charles.gorand@effinnov.com>
14401	14400	L: linux-nfc@lists.01.org (subscribers-only)
14402		−S: Supported
	14401	+S: Orphan
14403	14402	F: Documentation/devicetree/bindings/net/nfc/nxp,nci.yaml
14404	14403	F: drivers/nfc/nxp-nci
14405	14404

drivers/net/bonding/bond_3ad.c+2−1

@@ -2228,7 +2228,8 @@ void bond_3ad_unbind_slave(struct slave *slave) · 2227 unmodified lines
2228	2228	temp_aggregator->num_of_ports--;
2229	2229	if (__agg_active_ports(temp_aggregator) == 0) {
2230	2230	select_new_active_agg = temp_aggregator->is_active;
2231		− ad_clear_agg(temp_aggregator);
	2231	+ if (temp_aggregator->num_of_ports == 0)
	2232	+ ad_clear_agg(temp_aggregator);
2232	2233	if (select_new_active_agg) {
2233	2234	slave_info(bond->dev, slave->dev, "Removing an active aggregator\n");
2234	2235	/* select new active aggregator */

drivers/net/bonding/bond_alb.c+1−1

@@ -1302,12 +1302,12 @@ int bond_alb_initialize(struct bonding *bond, int rlb_enabled) · 1301 unmodified lines
1302	1302	return res;
1303	1303
1304	1304	if (rlb_enabled) {
1305		− bond->alb_info.rlb_enabled = 1;
1306	1305	res = rlb_initialize(bond);
1307	1306	if (res) {
1308	1307	tlb_deinitialize(bond);
1309	1308	return res;
1310	1309	}
	1310	+ bond->alb_info.rlb_enabled = 1;
1311	1311	} else {
1312	1312	bond->alb_info.rlb_enabled = 0;
1313	1313	}

drivers/net/dsa/bcm_sf2.c+5−0

@@ -878,6 +878,11 @@ static void bcm_sf2_sw_mac_link_up(struct dsa_switch *ds, int port, · 877 unmodified lines
878	878	if (duplex == DUPLEX_FULL)
879	879	reg \|= DUPLX_MODE;
880	880
	881	+ if (tx_pause)
	882	+ reg \|= TXFLOW_CNTL;
	883	+ if (rx_pause)
	884	+ reg \|= RXFLOW_CNTL;
	885	+
881	886	core_writel(priv, reg, offset);
882	887	}
883	888

drivers/net/dsa/hirschmann/hellcreek_ptp.c+1−0

@@ -300,6 +300,7 @@ static int hellcreek_led_setup(struct hellcreek *hellcreek) · 299 unmodified lines
300	300	const char label, state;
301	301	int ret = -EINVAL;
302	302
	303	+ of_node_get(hellcreek->dev->of_node);
303	304	leds = of_find_node_by_name(hellcreek->dev->of_node, "leds");
304	305	if (!leds) {
305	306	dev_err(hellcreek->dev, "No LEDs specified in device tree!\n");

drivers/net/dsa/ocelot/felix_vsc9959.c+4−0

@@ -1886,6 +1886,8 @@ static void vsc9959_psfp_sgi_table_del(struct ocelot *ocelot, · 1885 unmodified lines
1886	1886	static void vsc9959_psfp_counters_get(struct ocelot *ocelot, u32 index,
1887	1887	struct felix_stream_filter_counters *counters)
1888	1888	{
	1889	+ mutex_lock(&ocelot->stats_lock);
	1890	+
1889	1891	ocelot_rmw(ocelot, SYS_STAT_CFG_STAT_VIEW(index),
1890	1892	SYS_STAT_CFG_STAT_VIEW_M,
1891	1893	SYS_STAT_CFG);
@@ -1900,6 +1902,8 @@ static void vsc9959_psfp_counters_get(struct ocelot *ocelot, u32 index, · 8 unmodified lines
1900	1902	SYS_STAT_CFG_STAT_VIEW(index) \|
1901	1903	SYS_STAT_CFG_STAT_CLEAR_SHOT(0x10),
1902	1904	SYS_STAT_CFG);
	1905	+
	1906	+ mutex_unlock(&ocelot->stats_lock);
1903	1907	}
1904	1908
1905	1909	static int vsc9959_psfp_filter_add(struct ocelot *ocelot, int port,

drivers/net/ethernet/marvell/octeon_ep/octep_regs_cn9k_pf.h+1−1

@@ -52,7 +52,7 @@ · 51 unmodified lines
52	52
53	53	#define CN93_SDP_EPF_RINFO_SRN(val) ((val) & 0xFF)
54	54	#define CN93_SDP_EPF_RINFO_RPVF(val) (((val) >> 32) & 0xF)
55		−#define CN93_SDP_EPF_RINFO_NVFS(val) (((val) >> 48) && 0xFF)
	55	+#define CN93_SDP_EPF_RINFO_NVFS(val) (((val) >> 48) & 0xFF)
56	56
57	57	/* SDP Function select */
58	58	#define CN93_SDP_FUNC_SEL_EPF_BIT_POS 8

drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c+13−1

@@ -4415,6 +4415,8 @@ static int mlxsw_sp_nexthop4_init(struct mlxsw_sp *mlxsw_sp, · 4414 unmodified lines
4415	4415	return 0;
4416	4416
4417	4417	err_nexthop_neigh_init:
	4418	+ list_del(&nh->router_list_node);
	4419	+ mlxsw_sp_nexthop_counter_free(mlxsw_sp, nh);
4418	4420	mlxsw_sp_nexthop_remove(mlxsw_sp, nh);
4419	4421	return err;
4420	4422	}
@@ -6740,6 +6742,7 @@ static int mlxsw_sp_nexthop6_init(struct mlxsw_sp *mlxsw_sp, · 2319 unmodified lines
6740	6742	const struct fib6_info *rt)
6741	6743	{
6742	6744	struct net_device *dev = rt->fib6_nh->fib_nh_dev;
	6745	+ int err;
6743	6746
6744	6747	nh->nhgi = nh_grp->nhgi;
6745	6748	nh->nh_weight = rt->fib6_nh->fib_nh_weight;
@@ -6755,7 +6758,16 @@ static int mlxsw_sp_nexthop6_init(struct mlxsw_sp *mlxsw_sp, · 9 unmodified lines
6755	6758	return 0;
6756	6759	nh->ifindex = dev->ifindex;
6757	6760
6758		− return mlxsw_sp_nexthop_type_init(mlxsw_sp, nh, dev);
	6761	+ err = mlxsw_sp_nexthop_type_init(mlxsw_sp, nh, dev);
	6762	+ if (err)
	6763	+ goto err_nexthop_type_init;
	6764	+
	6765	+ return 0;
	6766	+
	6767	+err_nexthop_type_init:
	6768	+ list_del(&nh->router_list_node);
	6769	+ mlxsw_sp_nexthop_counter_free(mlxsw_sp, nh);
	6770	+ return err;
6759	6771	}
6760	6772
6761	6773	static void mlxsw_sp_nexthop6_fini(struct mlxsw_sp *mlxsw_sp,

drivers/net/ethernet/microchip/sparx5/sparx5_switchdev.c+6−0

@@ -396,6 +396,9 @@ static int sparx5_handle_port_mdb_add(struct net_device *dev, · 395 unmodified lines
396	396	u32 mact_entry;
397	397	int res, err;
398	398
	399	+ if (!sparx5_netdevice_check(dev))
	400	+ return -EOPNOTSUPP;
	401	+
399	402	if (netif_is_bridge_master(v->obj.orig_dev)) {
400	403	sparx5_mact_learn(spx5, PGID_CPU, v->addr, v->vid);
401	404	return 0;
@@ -466,6 +469,9 @@ static int sparx5_handle_port_mdb_del(struct net_device *dev, · 64 unmodified lines
466	469	u32 mact_entry, res, pgid_entry[3];
467	470	int err;
468	471
	472	+ if (!sparx5_netdevice_check(dev))
	473	+ return -EOPNOTSUPP;
	474	+
469	475	if (netif_is_bridge_master(v->obj.orig_dev)) {
470	476	sparx5_mact_forget(spx5, v->addr, v->vid);
471	477	return 0;

drivers/net/ethernet/smsc/epic100.c+2−2

@@ -1515,14 +1515,14 @@ static void epic_remove_one(struct pci_dev *pdev) · 1514 unmodified lines
1515	1515	struct net_device *dev = pci_get_drvdata(pdev);
1516	1516	struct epic_private *ep = netdev_priv(dev);
1517	1517
	1518	+ unregister_netdev(dev);
1518	1519	dma_free_coherent(&pdev->dev, TX_TOTAL_SIZE, ep->tx_ring,
1519	1520	ep->tx_ring_dma);
1520	1521	dma_free_coherent(&pdev->dev, RX_TOTAL_SIZE, ep->rx_ring,
1521	1522	ep->rx_ring_dma);
1522		− unregister_netdev(dev);
1523	1523	pci_iounmap(pdev, ep->ioaddr);
1524		− pci_release_regions(pdev);
1525	1524	free_netdev(dev);
	1525	+ pci_release_regions(pdev);
1526	1526	pci_disable_device(pdev);
1527	1527	/* pci_power_off(pdev, -1); */
1528	1528	}

drivers/net/phy/ax88796b.c+4−2

@@ -88,8 +88,10 @@ static void asix_ax88772a_link_change_notify(struct phy_device *phydev) · 87 unmodified lines
88	88	/* Reset PHY, otherwise MII_LPA will provide outdated information.
89	89	* This issue is reproducible only with some link partner PHYs
90	90	*/
91		− if (phydev->state == PHY_NOLINK && phydev->drv->soft_reset)
92		− phydev->drv->soft_reset(phydev);
	91	+ if (phydev->state == PHY_NOLINK) {
	92	+ phy_init_hw(phydev);
	93	+ phy_start_aneg(phydev);
	94	+ }
93	95	}
94	96
95	97	static struct phy_driver asix_driver[] = {

drivers/net/phy/dp83822.c+1−3

@@ -229,9 +229,7 @@ static int dp83822_config_intr(struct phy_device *phydev) · 228 unmodified lines
229	229	if (misr_status < 0)
230	230	return misr_status;
231	231
232		− misr_status \|= (DP83822_RX_ERR_HF_INT_EN \|
233		− DP83822_FALSE_CARRIER_HF_INT_EN \|
234		− DP83822_LINK_STAT_INT_EN \|
	232	+ misr_status \|= (DP83822_LINK_STAT_INT_EN \|
235	233	DP83822_ENERGY_DET_INT_EN \|
236	234	DP83822_LINK_QUAL_INT_EN);
237	235

drivers/net/phy/phy.c+23−0

@@ -31,6 +31,7 @@ · 30 unmodified lines
31	31	#include <linux/io.h>
32	32	#include <linux/uaccess.h>
33	33	#include <linux/atomic.h>
	34	+#include <linux/suspend.h>
34	35	#include <net/netlink.h>
35	36	#include <net/genetlink.h>
36	37	#include <net/sock.h>
@@ -976,6 +977,28 @@ static irqreturn_t phy_interrupt(int irq, void *phy_dat) · 939 unmodified lines
976	977	struct phy_driver *drv = phydev->drv;
977	978	irqreturn_t ret;
978	979
	980	+ /* Wakeup interrupts may occur during a system sleep transition.
	981	+ * Postpone handling until the PHY has resumed.
	982	+ */
	983	+ if (IS_ENABLED(CONFIG_PM_SLEEP) && phydev->irq_suspended) {
	984	+ struct net_device *netdev = phydev->attached_dev;
	985	+
	986	+ if (netdev) {
	987	+ struct device *parent = netdev->dev.parent;
	988	+
	989	+ if (netdev->wol_enabled)
	990	+ pm_system_wakeup();
	991	+ else if (device_may_wakeup(&netdev->dev))
	992	+ pm_wakeup_dev_event(&netdev->dev, 0, true);
	993	+ else if (parent && device_may_wakeup(parent))
	994	+ pm_wakeup_dev_event(parent, 0, true);
	995	+ }
	996	+
	997	+ phydev->irq_rerun = 1;
	998	+ disable_irq_nosync(irq);
	999	+ return IRQ_HANDLED;
	1000	+ }
	1001	+
979	1002	mutex_lock(&phydev->lock);
980	1003	ret = drv->handle_interrupt(phydev);
981	1004	mutex_unlock(&phydev->lock);

drivers/net/phy/phy_device.c+23−0

@@ -278,6 +278,15 @@ static __maybe_unused int mdio_bus_phy_suspend(struct device *dev) · 277 unmodified lines
278	278	if (phydev->mac_managed_pm)
279	279	return 0;
280	280
	281	+ /* Wakeup interrupts may occur during the system sleep transition when
	282	+ * the PHY is inaccessible. Set flag to postpone handling until the PHY
	283	+ * has resumed. Wait for concurrent interrupt handler to complete.
	284	+ */
	285	+ if (phy_interrupt_is_valid(phydev)) {
	286	+ phydev->irq_suspended = 1;
	287	+ synchronize_irq(phydev->irq);
	288	+ }
	289	+
281	290	/* We must stop the state machine manually, otherwise it stops out of
282	291	* control, possibly with the phydev->lock held. Upon resume, netdev
283	292	* may call phy routines that try to grab the same lock, and that may
@@ -315,6 +324,20 @@ static __maybe_unused int mdio_bus_phy_resume(struct device *dev) · 31 unmodified lines
315	324	if (ret < 0)
316	325	return ret;
317	326	no_resume:
	327	+ if (phy_interrupt_is_valid(phydev)) {
	328	+ phydev->irq_suspended = 0;
	329	+ synchronize_irq(phydev->irq);
	330	+
	331	+ /* Rerun interrupts which were postponed by phy_interrupt()
	332	+ * because they occurred during the system sleep transition.
	333	+ */
	334	+ if (phydev->irq_rerun) {
	335	+ phydev->irq_rerun = 0;
	336	+ enable_irq(phydev->irq);
	337	+ irq_wake_thread(phydev->irq, phydev);
	338	+ }
	339	+ }
	340	+
318	341	if (phydev->attached_dev && phydev->adjust_link)
319	342	phy_start_machine(phydev);
320	343

drivers/net/phy/sfp.c+1−1

@@ -2516,7 +2516,7 @@ static int sfp_probe(struct platform_device *pdev) · 2515 unmodified lines
2516	2516
2517	2517	platform_set_drvdata(pdev, sfp);
2518	2518
2519		− err = devm_add_action(sfp->dev, sfp_cleanup, sfp);
	2519	+ err = devm_add_action_or_reset(sfp->dev, sfp_cleanup, sfp);
2520	2520	if (err < 0)
2521	2521	return err;
2522	2522

drivers/net/tun.c+13−2

@@ -273,6 +273,12 @@ static void tun_napi_init(struct tun_struct tun, struct tun_file tfile, · 272 unmodified lines
273	273	}
274	274	}
275	275
	276	+static void tun_napi_enable(struct tun_file *tfile)
	277	+{
	278	+ if (tfile->napi_enabled)
	279	+ napi_enable(&tfile->napi);
	280	+}
	281	+
276	282	static void tun_napi_disable(struct tun_file *tfile)
277	283	{
278	284	if (tfile->napi_enabled)
@@ -634,7 +640,8 @@ static void __tun_detach(struct tun_file *tfile, bool clean) · 355 unmodified lines
634	640	tun = rtnl_dereference(tfile->tun);
635	641
636	642	if (tun && clean) {
637		− tun_napi_disable(tfile);
	643	+ if (!tfile->detached)
	644	+ tun_napi_disable(tfile);
638	645	tun_napi_del(tfile);
639	646	}
640	647
@@ -653,8 +660,10 @@ static void __tun_detach(struct tun_file *tfile, bool clean) · 12 unmodified lines
653	660	if (clean) {
654	661	RCU_INIT_POINTER(tfile->tun, NULL);
655	662	sock_put(&tfile->sk);
656		− } else
	663	+ } else {
657	664	tun_disable_queue(tun, tfile);
	665	+ tun_napi_disable(tfile);
	666	+ }
658	667
659	668	synchronize_net();
660	669	tun_flow_delete_by_queue(tun, tun->numqueues + 1);
@@ -727,6 +736,7 @@ static void tun_detach_all(struct net_device *dev) · 66 unmodified lines
727	736	sock_put(&tfile->sk);
728	737	}
729	738	list_for_each_entry_safe(tfile, tmp, &tun->disabled, next) {
	739	+ tun_napi_del(tfile);
730	740	tun_enable_queue(tfile);
731	741	tun_queue_purge(tfile);
732	742	xdp_rxq_info_unreg(&tfile->xdp_rxq);
@@ -807,6 +817,7 @@ static int tun_attach(struct tun_struct tun, struct file file, · 74 unmodified lines
807	817
808	818	if (tfile->detached) {
809	819	tun_enable_queue(tfile);
	820	+ tun_napi_enable(tfile);
810	821	} else {
811	822	sock_hold(&tfile->sk);
812	823	tun_napi_init(tun, tfile, napi, napi_frags);

drivers/net/usb/asix.h+1−2

@@ -126,8 +126,7 @@ · 125 unmodified lines
126	126	AX_MEDIUM_RE)
127	127
128	128	#define AX88772_MEDIUM_DEFAULT \
129		− (AX_MEDIUM_FD \| AX_MEDIUM_RFC \| \
130		− AX_MEDIUM_TFC \| AX_MEDIUM_PS \| \
	129	+ (AX_MEDIUM_FD \| AX_MEDIUM_PS \| \
131	130	AX_MEDIUM_AC \| AX_MEDIUM_RE)
132	131
133	132	/* AX88772 & AX88178 RX_CTL values */

drivers/net/usb/asix_common.c+1−0

@@ -431,6 +431,7 @@ void asix_adjust_link(struct net_device *netdev) · 430 unmodified lines
431	431
432	432	asix_write_medium_mode(dev, mode, 0);
433	433	phy_print_status(phydev);
	434	+ usbnet_link_change(dev, phydev->link, 0);
434	435	}
435	436
436	437	int asix_write_gpio(struct usbnet *dev, u16 value, int sleep, int in_pm)

drivers/net/usb/ax88179_178a.c+76−25

@@ -1472,6 +1472,42 @@ static int ax88179_rx_fixup(struct usbnet dev, struct sk_buff skb) · 1471 unmodified lines
1472	1472	* are bundled into this buffer and where we can find an array of
1473	1473	* per-packet metadata (which contains elements encoded into u16).
1474	1474	*/
	1475	+
	1476	+ /* SKB contents for current firmware:
	1477	+ * <packet 1> <padding>
	1478	+ * ...
	1479	+ * <packet N> <padding>
	1480	+ * <per-packet metadata entry 1> <dummy header>
	1481	+ * ...
	1482	+ * <per-packet metadata entry N> <dummy header>
	1483	+ * <padding2> <rx_hdr>
	1484	+ *
	1485	+ * where:
	1486	+ * <packet N> contains pkt_len bytes:
	1487	+ * 2 bytes of IP alignment pseudo header
	1488	+ * packet received
	1489	+ * <per-packet metadata entry N> contains 4 bytes:
	1490	+ * pkt_len and fields AX_RXHDR_*
	1491	+ * <padding> 0-7 bytes to terminate at
	1492	+ * 8 bytes boundary (64-bit).
	1493	+ * <padding2> 4 bytes to make rx_hdr terminate at
	1494	+ * 8 bytes boundary (64-bit)
	1495	+ * <dummy-header> contains 4 bytes:
	1496	+ * pkt_len=0 and AX_RXHDR_DROP_ERR
	1497	+ * <rx-hdr> contains 4 bytes:
	1498	+ * pkt_cnt and hdr_off (offset of
	1499	+ * <per-packet metadata entry 1>)
	1500	+ *
	1501	+ * pkt_cnt is number of entrys in the per-packet metadata.
	1502	+ * In current firmware there is 2 entrys per packet.
	1503	+ * The first points to the packet and the
	1504	+ * second is a dummy header.
	1505	+ * This was done probably to align fields in 64-bit and
	1506	+ * maintain compatibility with old firmware.
	1507	+ * This code assumes that <dummy header> and <padding2> are
	1508	+ * optional.
	1509	+ */
	1510	+
1475	1511	if (skb->len < 4)
1476	1512	return 0;
1477	1513	skb_trim(skb, skb->len - 4);
@@ -1485,51 +1521,66 @@ static int ax88179_rx_fixup(struct usbnet dev, struct sk_buff skb) · 7 unmodified lines
1485	1521	/* Make sure that the bounds of the metadata array are inside the SKB
1486	1522	* (and in front of the counter at the end).
1487	1523	*/
1488		− if (pkt_cnt * 2 + hdr_off > skb->len)
	1524	+ if (pkt_cnt * 4 + hdr_off > skb->len)
1489	1525	return 0;
1490	1526	pkt_hdr = (u32 *)(skb->data + hdr_off);
1491	1527
1492	1528	/* Packets must not overlap the metadata array */
1493	1529	skb_trim(skb, hdr_off);
1494	1530
1495		− for (; ; pkt_cnt--, pkt_hdr++) {
	1531	+ for (; pkt_cnt > 0; pkt_cnt--, pkt_hdr++) {
	1532	+ u16 pkt_len_plus_padd;
1496	1533	u16 pkt_len;
1497	1534
1498	1535	le32_to_cpus(pkt_hdr);
1499	1536	pkt_len = (*pkt_hdr >> 16) & 0x1fff;
	1537	+ pkt_len_plus_padd = (pkt_len + 7) & 0xfff8;
1500	1538
1501		− if (pkt_len > skb->len)
	1539	+ /* Skip dummy header used for alignment
	1540	+ */
	1541	+ if (pkt_len == 0)
	1542	+ continue;
	1543	+
	1544	+ if (pkt_len_plus_padd > skb->len)
1502	1545	return 0;
1503	1546
1504	1547	/* Check CRC or runt packet */
1505		− if (((*pkt_hdr & (AX_RXHDR_CRC_ERR \| AX_RXHDR_DROP_ERR)) == 0) &&
1506		− pkt_len >= 2 + ETH_HLEN) {
1507		− bool last = (pkt_cnt == 0);
1508		−
1509		− if (last) {
1510		− ax_skb = skb;
1511		− } else {
1512		− ax_skb = skb_clone(skb, GFP_ATOMIC);
1513		− if (!ax_skb)
1514		− return 0;
1515		− }
1516		− ax_skb->len = pkt_len;
1517		− /* Skip IP alignment pseudo header */
1518		− skb_pull(ax_skb, 2);
1519		− skb_set_tail_pointer(ax_skb, ax_skb->len);
1520		− ax_skb->truesize = pkt_len + sizeof(struct sk_buff);
1521		− ax88179_rx_checksum(ax_skb, pkt_hdr);
	1548	+ if ((*pkt_hdr & (AX_RXHDR_CRC_ERR \| AX_RXHDR_DROP_ERR)) \|\|
	1549	+ pkt_len < 2 + ETH_HLEN) {
	1550	+ dev->net->stats.rx_errors++;
	1551	+ skb_pull(skb, pkt_len_plus_padd);
	1552	+ continue;
	1553	+ }
1522	1554
1523		− if (last)
1524		− return 1;
	1555	+ /* last packet */
	1556	+ if (pkt_len_plus_padd == skb->len) {
	1557	+ skb_trim(skb, pkt_len);
1525	1558
1526		− usbnet_skb_return(dev, ax_skb);
	1559	+ /* Skip IP alignment pseudo header */
	1560	+ skb_pull(skb, 2);
	1561	+
	1562	+ skb->truesize = SKB_TRUESIZE(pkt_len_plus_padd);
	1563	+ ax88179_rx_checksum(skb, pkt_hdr);
	1564	+ return 1;
1527	1565	}
1528	1566
1529		− /* Trim this packet away from the SKB */
1530		− if (!skb_pull(skb, (pkt_len + 7) & 0xFFF8))
	1567	+ ax_skb = skb_clone(skb, GFP_ATOMIC);
	1568	+ if (!ax_skb)
1531	1569	return 0;
	1570	+ skb_trim(ax_skb, pkt_len);
	1571	+
	1572	+ /* Skip IP alignment pseudo header */
	1573	+ skb_pull(ax_skb, 2);
	1574	+
	1575	+ skb->truesize = pkt_len_plus_padd +
	1576	+ SKB_DATA_ALIGN(sizeof(struct sk_buff));
	1577	+ ax88179_rx_checksum(ax_skb, pkt_hdr);
	1578	+ usbnet_skb_return(dev, ax_skb);
	1579	+
	1580	+ skb_pull(skb, pkt_len_plus_padd);
1532	1581	}
	1582	+
	1583	+ return 0;
1533	1584	}
1534	1585
1535	1586	static struct sk_buff *

drivers/net/usb/usbnet.c+2−2

@@ -2004,7 +2004,7 @@ static int __usbnet_read_cmd(struct usbnet *dev, u8 cmd, u8 reqtype, · 2003 unmodified lines
2004	2004	cmd, reqtype, value, index, size);
2005	2005
2006	2006	if (size) {
2007		− buf = kmalloc(size, GFP_KERNEL);
	2007	+ buf = kmalloc(size, GFP_NOIO);
2008	2008	if (!buf)
2009	2009	goto out;
2010	2010	}
@@ -2036,7 +2036,7 @@ static int __usbnet_write_cmd(struct usbnet *dev, u8 cmd, u8 reqtype, · 25 unmodified lines
2036	2036	cmd, reqtype, value, index, size);
2037	2037
2038	2038	if (data) {
2039		− buf = kmemdup(data, size, GFP_KERNEL);
	2039	+ buf = kmemdup(data, size, GFP_NOIO);
2040	2040	if (!buf)
2041	2041	goto out;
2042	2042	} else {

drivers/nfc/nfcmrvl/i2c.c+3−3

@@ -167,9 +167,9 @@ static int nfcmrvl_i2c_parse_dt(struct device_node *node, · 166 unmodified lines
167	167	pdata->irq_polarity = IRQF_TRIGGER_RISING;
168	168
169	169	ret = irq_of_parse_and_map(node, 0);
170		− if (ret < 0) {
171		− pr_err("Unable to get irq, error: %d\n", ret);
172		− return ret;
	170	+ if (!ret) {
	171	+ pr_err("Unable to get irq\n");
	172	+ return -EINVAL;
173	173	}
174	174	pdata->irq = ret;
175	175

drivers/nfc/nfcmrvl/spi.c+3−3

@@ -115,9 +115,9 @@ static int nfcmrvl_spi_parse_dt(struct device_node *node, · 114 unmodified lines
115	115	}
116	116
117	117	ret = irq_of_parse_and_map(node, 0);
118		− if (ret < 0) {
119		− pr_err("Unable to get irq, error: %d\n", ret);
120		− return ret;
	118	+ if (!ret) {
	119	+ pr_err("Unable to get irq\n");
	120	+ return -EINVAL;
121	121	}
122	122	pdata->irq = ret;
123	123

drivers/nfc/nxp-nci/i2c.c+9−2

@@ -122,7 +122,9 @@ static int nxp_nci_i2c_fw_read(struct nxp_nci_i2c_phy *phy, · 121 unmodified lines
122	122	skb_put_data(*skb, &header, NXP_NCI_FW_HDR_LEN);
123	123
124	124	r = i2c_master_recv(client, skb_put(*skb, frame_len), frame_len);
125		− if (r != frame_len) {
	125	+ if (r < 0) {
	126	+ goto fw_read_exit_free_skb;
	127	+ } else if (r != frame_len) {
126	128	nfc_err(&client->dev,
127	129	"Invalid frame length: %u (expected %zu)\n",
128	130	r, frame_len);
@@ -162,8 +164,13 @@ static int nxp_nci_i2c_nci_read(struct nxp_nci_i2c_phy *phy, · 33 unmodified lines
162	164
163	165	skb_put_data(skb, (void )&header, NCI_CTRL_HDR_SIZE);
164	166
	167	+ if (!header.plen)
	168	+ return 0;
	169	+
165	170	r = i2c_master_recv(client, skb_put(*skb, header.plen), header.plen);
166		− if (r != header.plen) {
	171	+ if (r < 0) {
	172	+ goto nci_read_exit_free_skb;
	173	+ } else if (r != header.plen) {
167	174	nfc_err(&client->dev,
168	175	"Invalid frame payload length: %u (expected %u)\n",
169	176	r, header.plen);

include/linux/netdevice.h+1−1

@@ -1671,7 +1671,7 @@ enum netdev_priv_flags { · 1670 unmodified lines
1671	1671	IFF_FAILOVER_SLAVE = 1<<28,
1672	1672	IFF_L3MDEV_RX_HANDLER = 1<<29,
1673	1673	IFF_LIVE_RENAME_OK = 1<<30,
1674		− IFF_TX_SKB_NO_LINEAR = 1<<31,
	1674	+ IFF_TX_SKB_NO_LINEAR = BIT_ULL(31),
1675	1675	IFF_CHANGE_PROTO_DOWN = BIT_ULL(32),
1676	1676	};
1677	1677

include/linux/phy.h+6−0

@@ -572,6 +572,10 @@ struct macsec_ops; · 571 unmodified lines
572	572	* @mdix_ctrl: User setting of crossover
573	573	* @pma_extable: Cached value of PMA/PMD Extended Abilities Register
574	574	* @interrupts: Flag interrupts have been enabled
	575	+ * @irq_suspended: Flag indicating PHY is suspended and therefore interrupt
	576	+ * handling shall be postponed until PHY has resumed
	577	+ * @irq_rerun: Flag indicating interrupts occurred while PHY was suspended,
	578	+ * requiring a rerun of the interrupt handler after resume
575	579	* @interface: enum phy_interface_t value
576	580	* @skb: Netlink message for cable diagnostics
577	581	* @nest: Netlink nest used for cable diagnostics
@@ -626,6 +630,8 @@ struct phy_device { · 48 unmodified lines
626	630
627	631	/* Interrupts are enabled */
628	632	unsigned interrupts:1;
	633	+ unsigned irq_suspended:1;
	634	+ unsigned irq_rerun:1;
629	635
630	636	enum phy_state state;
631	637

include/net/netfilter/nf_tables.h+10−6

@@ -1338,24 +1338,28 @@ void nft_unregister_flowtable_type(struct nf_flowtable_type *type); · 1337 unmodified lines
1338	1338	/**
1339	1339	* struct nft_traceinfo - nft tracing information and state
1340	1340	*
	1341	+ * @trace: other struct members are initialised
	1342	+ * @nf_trace: copy of skb->nf_trace before rule evaluation
	1343	+ * @type: event type (enum nft_trace_types)
	1344	+ * @skbid: hash of skb to be used as trace id
	1345	+ * @packet_dumped: packet headers sent in a previous traceinfo message
1341	1346	* @pkt: pktinfo currently processed
1342	1347	* @basechain: base chain currently processed
1343	1348	* @chain: chain currently processed
1344	1349	* @rule: rule that was evaluated
1345	1350	* @verdict: verdict given by rule
1346		− * @type: event type (enum nft_trace_types)
1347		− * @packet_dumped: packet headers sent in a previous traceinfo message
1348		− * @trace: other struct members are initialised
1349	1351	*/
1350	1352	struct nft_traceinfo {
	1353	+ bool trace;
	1354	+ bool nf_trace;
	1355	+ bool packet_dumped;
	1356	+ enum nft_trace_types type:8;
	1357	+ u32 skbid;
1351	1358	const struct nft_pktinfo *pkt;
1352	1359	const struct nft_base_chain *basechain;
1353	1360	const struct nft_chain *chain;
1354	1361	const struct nft_rule_dp *rule;
1355	1362	const struct nft_verdict *verdict;
1356		− enum nft_trace_types type;
1357		− bool packet_dumped;
1358		− bool trace;
1359	1363	};
1360	1364
1361	1365	void nft_trace_init(struct nft_traceinfo info, const struct nft_pktinfo pkt,

include/uapi/linux/mptcp.h+5−4

@@ -2,16 +2,17 @@ · 1 unmodified line
2	2	#ifndef _UAPI_MPTCP_H
3	3	#define _UAPI_MPTCP_H
4	4
	5	+#ifndef __KERNEL__
	6	+#include <netinet/in.h> /* for sockaddr_in and sockaddr_in6 */
	7	+#include <sys/socket.h> /* for struct sockaddr */
	8	+#endif
	9	+
5	10	#include <linux/const.h>
6	11	#include <linux/types.h>
7	12	#include <linux/in.h> /* for sockaddr_in */
8	13	#include <linux/in6.h> /* for sockaddr_in6 */
9	14	#include <linux/socket.h> /* for sockaddr_storage and sa_family */
10	15
11		−#ifndef __KERNEL__
12		−#include <sys/socket.h> /* for struct sockaddr */
13		−#endif
14		−
15	16	#define MPTCP_SUBFLOW_FLAG_MCAP_REM _BITUL(0)
16	17	#define MPTCP_SUBFLOW_FLAG_MCAP_LOC _BITUL(1)
17	18	#define MPTCP_SUBFLOW_FLAG_JOIN_REM _BITUL(2)

net/bridge/br_netfilter_hooks.c+18−3

@@ -1012,9 +1012,24 @@ int br_nf_hook_thresh(unsigned int hook, struct net *net, · 1011 unmodified lines
1012	1012	return okfn(net, sk, skb);
1013	1013
1014	1014	ops = nf_hook_entries_get_hook_ops(e);
1015		− for (i = 0; i < e->num_hook_entries &&
1016		− ops[i]->priority <= NF_BR_PRI_BRNF; i++)
1017		− ;
	1015	+ for (i = 0; i < e->num_hook_entries; i++) {
	1016	+ /* These hooks have already been called */
	1017	+ if (ops[i]->priority < NF_BR_PRI_BRNF)
	1018	+ continue;
	1019	+
	1020	+ /* These hooks have not been called yet, run them. */
	1021	+ if (ops[i]->priority > NF_BR_PRI_BRNF)
	1022	+ break;
	1023	+
	1024	+ /* take a closer look at NF_BR_PRI_BRNF. */
	1025	+ if (ops[i]->hook == br_nf_pre_routing) {
	1026	+ /* This hook diverted the skb to this function,
	1027	+ * hooks after this have not been run yet.
	1028	+ */
	1029	+ i++;
	1030	+ break;
	1031	+ }
	1032	+ }
1018	1033
1019	1034	nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1020	1035	sk, net, okfn);

net/ipv4/ip_tunnel_core.c+1−1

@@ -410,7 +410,7 @@ int skb_tunnel_check_pmtu(struct sk_buff skb, struct dst_entry encap_dst, · 409 unmodified lines
410	410	u32 mtu = dst_mtu(encap_dst) - headroom;
411	411
412	412	if ((skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu)) \|\|
413		− (!skb_is_gso(skb) && (skb->len - skb_mac_header_len(skb)) <= mtu))
	413	+ (!skb_is_gso(skb) && (skb->len - skb_network_offset(skb)) <= mtu))
414	414	return 0;
415	415
416	416	skb_dst_update_pmtu_no_confirm(skb, mtu);

net/ipv4/tcp_ipv4.c+5−1

@@ -1964,7 +1964,10 @@ process: · 1963 unmodified lines
1964	1964	struct sock *nsk;
1965	1965
1966	1966	sk = req->rsk_listener;
1967		− drop_reason = tcp_inbound_md5_hash(sk, skb,
	1967	+ if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
	1968	+ drop_reason = SKB_DROP_REASON_XFRM_POLICY;
	1969	+ else
	1970	+ drop_reason = tcp_inbound_md5_hash(sk, skb,
1968	1971	&iph->saddr, &iph->daddr,
1969	1972	AF_INET, dif, sdif);
1970	1973	if (unlikely(drop_reason)) {
@@ -2016,6 +2019,7 @@ process: · 45 unmodified lines
2016	2019	}
2017	2020	goto discard_and_relse;
2018	2021	}
	2022	+ nf_reset_ct(skb);
2019	2023	if (nsk == sk) {
2020	2024	reqsk_put(req);
2021	2025	tcp_v4_restore_cb(skb);

net/ipv6/addrconf.c+2−6

@@ -1109,10 +1109,6 @@ ipv6_add_addr(struct inet6_dev idev, struct ifa6_config cfg, · 1108 unmodified lines
1109	1109	goto out;
1110	1110	}
1111	1111
1112		− if (net->ipv6.devconf_all->disable_policy \|\|
1113		− idev->cnf.disable_policy)
1114		− f6i->dst_nopolicy = true;
1115		−
1116	1112	neigh_parms_data_state_setall(idev->nd_parms);
1117	1113
1118	1114	ifa->addr = *cfg->pfx;
@@ -5172,9 +5168,9 @@ next: · 4053 unmodified lines
5172	5168	fillargs->event = RTM_GETMULTICAST;
5173	5169
5174	5170	/* multicast address */
5175		− for (ifmca = rcu_dereference(idev->mc_list);
	5171	+ for (ifmca = rtnl_dereference(idev->mc_list);
5176	5172	ifmca;
5177		− ifmca = rcu_dereference(ifmca->next), ip_idx++) {
	5173	+ ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
5178	5174	if (ip_idx < s_ip_idx)
5179	5175	continue;
5180	5176	err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);

net/ipv6/route.c+8−1

@@ -4569,8 +4569,15 @@ struct fib6_info addrconf_f6i_alloc(struct net net, · 4568 unmodified lines
4569	4569	}
4570	4570
4571	4571	f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4572		− if (!IS_ERR(f6i))
	4572	+ if (!IS_ERR(f6i)) {
4573	4573	f6i->dst_nocount = true;
	4574	+
	4575	+ if (!anycast &&
	4576	+ (net->ipv6.devconf_all->disable_policy \|\|
	4577	+ idev->cnf.disable_policy))
	4578	+ f6i->dst_nopolicy = true;
	4579	+ }
	4580	+
4574	4581	return f6i;
4575	4582	}
4576	4583

net/ipv6/seg6_hmac.c+0−1

@@ -406,7 +406,6 @@ int __net_init seg6_hmac_net_init(struct net *net) · 405 unmodified lines
406	406
407	407	return rhashtable_init(&sdata->hmac_infos, &rht_params);
408	408	}
409		−EXPORT_SYMBOL(seg6_hmac_net_init);
410	409
411	410	void seg6_hmac_exit(void)
412	411	{

net/ipv6/sit.c+3−5

@@ -323,8 +323,6 @@ static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __u · 322 unmodified lines
323	323	kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT \| __GFP_NOWARN) :
324	324	NULL;
325	325
326		− rcu_read_lock();
327		−
328	326	ca = min(t->prl_count, cmax);
329	327
330	328	if (!kp) {
@@ -341,7 +339,7 @@ static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __u · 10 unmodified lines
341	339	}
342	340	}
343	341
344		− c = 0;
	342	+ rcu_read_lock();
345	343	for_each_prl_rcu(t->prl) {
346	344	if (c >= cmax)
347	345	break;
@@ -353,7 +351,7 @@ static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __u · 5 unmodified lines
353	351	if (kprl.addr != htonl(INADDR_ANY))
354	352	break;
355	353	}
356		−out:
	354	+
357	355	rcu_read_unlock();
358	356
359	357	len = sizeof(kp) c;
@@ -362,7 +360,7 @@ out: · 2 unmodified lines
362	360	ret = -EFAULT;
363	361
364	362	kfree(kp);
365		−
	363	+out:
366	364	return ret;
367	365	}
368	366

net/mptcp/options.c+4−3

@@ -765,6 +765,7 @@ static noinline bool mptcp_established_options_rst(struct sock *sk, struct sk_bu · 764 unmodified lines
765	765	opts->suboptions \|= OPTION_MPTCP_RST;
766	766	opts->reset_transient = subflow->reset_transient;
767	767	opts->reset_reason = subflow->reset_reason;
	768	+ MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPRSTTX);
768	769
769	770	return true;
770	771	}
@@ -788,6 +789,7 @@ static bool mptcp_established_options_fastclose(struct sock *sk, · 17 unmodified lines
788	789	opts->rcvr_key = msk->remote_key;
789	790
790	791	pr_debug("FASTCLOSE key=%llu", opts->rcvr_key);
	792	+ MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFASTCLOSETX);
791	793	return true;
792	794	}
793	795
@@ -809,6 +811,7 @@ static bool mptcp_established_options_mp_fail(struct sock *sk, · 15 unmodified lines
809	811	opts->fail_seq = subflow->map_seq;
810	812
811	813	pr_debug("MP_FAIL fail_seq=%llu", opts->fail_seq);
	814	+ MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILTX);
812	815
813	816	return true;
814	817	}
@@ -833,13 +836,11 @@ bool mptcp_established_options(struct sock sk, struct sk_buff skb, · 18 unmodified lines
833	836	mptcp_established_options_mp_fail(sk, &opt_size, remaining, opts)) {
834	837	*size += opt_size;
835	838	remaining -= opt_size;
836		− MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFASTCLOSETX);
837	839	}
838	840	/* MP_RST can be used with MP_FASTCLOSE and MP_FAIL if there is room */
839	841	if (mptcp_established_options_rst(sk, skb, &opt_size, remaining, opts)) {
840	842	*size += opt_size;
841	843	remaining -= opt_size;
842		− MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPRSTTX);
843	844	}
844	845	return true;
845	846	}
@@ -966,7 +967,7 @@ static bool check_fully_established(struct mptcp_sock msk, struct sock ssk, · 120 unmodified lines
966	967	goto reset;
967	968	subflow->mp_capable = 0;
968	969	pr_fallback(msk);
969		− __mptcp_do_fallback(msk);
	970	+ mptcp_do_fallback(ssk);
970	971	return false;
971	972	}
972	973

net/mptcp/pm.c+4−6

@@ -299,23 +299,21 @@ void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq) · 298 unmodified lines
299	299	{
300	300	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
301	301	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
302		− struct sock s = (struct sock )msk;
303	302
304	303	pr_debug("fail_seq=%llu", fail_seq);
305	304
306	305	if (!READ_ONCE(msk->allow_infinite_fallback))
307	306	return;
308	307
309		− if (!READ_ONCE(subflow->mp_fail_response_expect)) {
	308	+ if (!subflow->fail_tout) {
310	309	pr_debug("send MP_FAIL response and infinite map");
311	310
312	311	subflow->send_mp_fail = 1;
313		− MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILTX);
314	312	subflow->send_infinite_map = 1;
315		− } else if (!sock_flag(sk, SOCK_DEAD)) {
	313	+ tcp_send_ack(sk);
	314	+ } else {
316	315	pr_debug("MP_FAIL response received");
317		−
318		− sk_stop_timer(s, &s->sk_timer);
	316	+ WRITE_ONCE(subflow->fail_tout, 0);
319	317	}
320	318	}
321	319

net/mptcp/protocol.c+53−31

@@ -500,7 +500,7 @@ static void mptcp_set_timeout(struct sock *sk) · 499 unmodified lines
500	500	__mptcp_set_timeout(sk, tout);
501	501	}
502	502
503		−static bool tcp_can_send_ack(const struct sock *ssk)
	503	+static inline bool tcp_can_send_ack(const struct sock *ssk)
504	504	{
505	505	return !((1 << inet_sk_state_load(ssk)) &
506	506	(TCPF_SYN_SENT \| TCPF_SYN_RECV \| TCPF_TIME_WAIT \| TCPF_CLOSE \| TCPF_LISTEN));
@@ -1245,7 +1245,7 @@ static void mptcp_update_infinite_map(struct mptcp_sock *msk, · 738 unmodified lines
1245	1245	MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPTX);
1246	1246	mptcp_subflow_ctx(ssk)->send_infinite_map = 0;
1247	1247	pr_fallback(msk);
1248		− __mptcp_do_fallback(msk);
	1248	+ mptcp_do_fallback(ssk);
1249	1249	}
1250	1250
1251	1251	static int mptcp_sendmsg_frag(struct sock sk, struct sock ssk,
@@ -2175,21 +2175,6 @@ static void mptcp_retransmit_timer(struct timer_list *t) · 923 unmodified lines
2175	2175	sock_put(sk);
2176	2176	}
2177	2177
2178		−static struct mptcp_subflow_context *
2179		−mp_fail_response_expect_subflow(struct mptcp_sock *msk)
2180		−{
2181		− struct mptcp_subflow_context subflow, ret = NULL;
2182		−
2183		− mptcp_for_each_subflow(msk, subflow) {
2184		− if (READ_ONCE(subflow->mp_fail_response_expect)) {
2185		− ret = subflow;
2186		− break;
2187		− }
2188		− }
2189		−
2190		− return ret;
2191		−}
2192		−
2193	2178	static void mptcp_timeout_timer(struct timer_list *t)
2194	2179	{
2195	2180	struct sock *sk = from_timer(sk, t, sk_timer);
@@ -2346,6 +2331,11 @@ static void __mptcp_close_ssk(struct sock sk, struct sock ssk, · 150 unmodified lines
2346	2331	kfree_rcu(subflow, rcu);
2347	2332	} else {
2348	2333	/* otherwise tcp will dispose of the ssk and subflow ctx */
	2334	+ if (ssk->sk_state == TCP_LISTEN) {
	2335	+ tcp_set_state(ssk, TCP_CLOSE);
	2336	+ mptcp_subflow_queue_clean(ssk);
	2337	+ inet_csk_listen_stop(ssk);
	2338	+ }
2349	2339	__tcp_close(ssk, 0);
2350	2340
2351	2341	/* close acquired an extra ref */
@@ -2518,27 +2508,50 @@ reset_timer: · 166 unmodified lines
2518	2508	mptcp_reset_timer(sk);
2519	2509	}
2520	2510
	2511	+/* schedule the timeout timer for the relevant event: either close timeout
	2512	+ * or mp_fail timeout. The close timeout takes precedence on the mp_fail one
	2513	+ */
	2514	+void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout)
	2515	+{
	2516	+ struct sock sk = (struct sock )msk;
	2517	+ unsigned long timeout, close_timeout;
	2518	+
	2519	+ if (!fail_tout && !sock_flag(sk, SOCK_DEAD))
	2520	+ return;
	2521	+
	2522	+ close_timeout = inet_csk(sk)->icsk_mtup.probe_timestamp - tcp_jiffies32 + jiffies + TCP_TIMEWAIT_LEN;
	2523	+
	2524	+ /* the close timeout takes precedence on the fail one, and here at least one of
	2525	+ * them is active
	2526	+ */
	2527	+ timeout = sock_flag(sk, SOCK_DEAD) ? close_timeout : fail_tout;
	2528	+
	2529	+ sk_reset_timer(sk, &sk->sk_timer, timeout);
	2530	+}
	2531	+
2521	2532	static void mptcp_mp_fail_no_response(struct mptcp_sock *msk)
2522	2533	{
2523		− struct mptcp_subflow_context *subflow;
2524		− struct sock *ssk;
	2534	+ struct sock *ssk = msk->first;
2525	2535	bool slow;
2526	2536
2527		− subflow = mp_fail_response_expect_subflow(msk);
2528		− if (subflow) {
2529		− pr_debug("MP_FAIL doesn't respond, reset the subflow");
	2537	+ if (!ssk)
	2538	+ return;
2530	2539
2531		− ssk = mptcp_subflow_tcp_sock(subflow);
2532		− slow = lock_sock_fast(ssk);
2533		− mptcp_subflow_reset(ssk);
2534		− unlock_sock_fast(ssk, slow);
2535		− }
	2540	+ pr_debug("MP_FAIL doesn't respond, reset the subflow");
	2541	+
	2542	+ slow = lock_sock_fast(ssk);
	2543	+ mptcp_subflow_reset(ssk);
	2544	+ WRITE_ONCE(mptcp_subflow_ctx(ssk)->fail_tout, 0);
	2545	+ unlock_sock_fast(ssk, slow);
	2546	+
	2547	+ mptcp_reset_timeout(msk, 0);
2536	2548	}
2537	2549
2538	2550	static void mptcp_worker(struct work_struct *work)
2539	2551	{
2540	2552	struct mptcp_sock *msk = container_of(work, struct mptcp_sock, work);
2541	2553	struct sock *sk = &msk->sk.icsk_inet.sk;
	2554	+ unsigned long fail_tout;
2542	2555	int state;
2543	2556
2544	2557	lock_sock(sk);
@@ -2575,7 +2588,9 @@ static void mptcp_worker(struct work_struct *work) · 30 unmodified lines
2575	2588	if (test_and_clear_bit(MPTCP_WORK_RTX, &msk->flags))
2576	2589	__mptcp_retrans(sk);
2577	2590
2578		− mptcp_mp_fail_no_response(msk);
	2591	+ fail_tout = msk->first ? READ_ONCE(mptcp_subflow_ctx(msk->first)->fail_tout) : 0;
	2592	+ if (fail_tout && time_after(jiffies, fail_tout))
	2593	+ mptcp_mp_fail_no_response(msk);
2579	2594
2580	2595	unlock:
2581	2596	release_sock(sk);
@@ -2822,6 +2837,7 @@ static void __mptcp_destroy_sock(struct sock *sk) · 240 unmodified lines
2822	2837	static void mptcp_close(struct sock *sk, long timeout)
2823	2838	{
2824	2839	struct mptcp_subflow_context *subflow;
	2840	+ struct mptcp_sock *msk = mptcp_sk(sk);
2825	2841	bool do_cancel_work = false;
2826	2842
2827	2843	lock_sock(sk);
@@ -2840,10 +2856,16 @@ static void mptcp_close(struct sock *sk, long timeout) · 12 unmodified lines
2840	2856	cleanup:
2841	2857	/* orphan all the subflows */
2842	2858	inet_csk(sk)->icsk_mtup.probe_timestamp = tcp_jiffies32;
2843		− mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
	2859	+ mptcp_for_each_subflow(msk, subflow) {
2844	2860	struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
2845	2861	bool slow = lock_sock_fast_nested(ssk);
2846	2862
	2863	+ /* since the close timeout takes precedence on the fail one,
	2864	+ * cancel the latter
	2865	+ */
	2866	+ if (ssk == msk->first)
	2867	+ subflow->fail_tout = 0;
	2868	+
2847	2869	sock_orphan(ssk);
2848	2870	unlock_sock_fast(ssk, slow);
2849	2871	}
@@ -2852,13 +2874,13 @@ cleanup: · 2 unmodified lines
2852	2874	sock_hold(sk);
2853	2875	pr_debug("msk=%p state=%d", sk, sk->sk_state);
2854	2876	if (mptcp_sk(sk)->token)
2855		− mptcp_event(MPTCP_EVENT_CLOSED, mptcp_sk(sk), NULL, GFP_KERNEL);
	2877	+ mptcp_event(MPTCP_EVENT_CLOSED, msk, NULL, GFP_KERNEL);
2856	2878
2857	2879	if (sk->sk_state == TCP_CLOSE) {
2858	2880	__mptcp_destroy_sock(sk);
2859	2881	do_cancel_work = true;
2860	2882	} else {
2861		− sk_reset_timer(sk, &sk->sk_timer, jiffies + TCP_TIMEWAIT_LEN);
	2883	+ mptcp_reset_timeout(msk, 0);
2862	2884	}
2863	2885	release_sock(sk);
2864	2886	if (do_cancel_work)

net/mptcp/protocol.h+20−4

@@ -306,6 +306,7 @@ struct mptcp_sock { · 305 unmodified lines
306	306
307	307	u32 setsockopt_seq;
308	308	char ca_name[TCP_CA_NAME_MAX];
	309	+ struct mptcp_sock *dl_next;
309	310	};
310	311
311	312	#define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
@@ -468,7 +469,6 @@ struct mptcp_subflow_context { · 156 unmodified lines
468	469	local_id_valid : 1, /* local_id is correctly initialized */
469	470	valid_csum_seen : 1; /* at least one csum validated */
470	471	enum mptcp_data_avail data_avail;
471		− bool mp_fail_response_expect;
472	472	u32 remote_nonce;
473	473	u64 thmac;
474	474	u32 local_nonce;
@@ -482,6 +482,7 @@ struct mptcp_subflow_context { · 7 unmodified lines
482	482	u8 stale_count;
483	483
484	484	long delegated_status;
	485	+ unsigned long fail_tout;
485	486
486	487	);
487	488
@@ -608,6 +609,7 @@ void mptcp_close_ssk(struct sock sk, struct sock ssk, · 120 unmodified lines
608	609	struct mptcp_subflow_context *subflow);
609	610	void mptcp_subflow_send_ack(struct sock *ssk);
610	611	void mptcp_subflow_reset(struct sock *ssk);
	612	+void mptcp_subflow_queue_clean(struct sock *ssk);
611	613	void mptcp_sock_graft(struct sock sk, struct socket parent);
612	614	struct socket __mptcp_nmpc_socket(const struct mptcp_sock msk);
613	615
@@ -662,6 +664,7 @@ void mptcp_get_options(const struct sk_buff *skb, · 48 unmodified lines
662	664
663	665	void mptcp_finish_connect(struct sock *sk);
664	666	void __mptcp_set_connected(struct sock *sk);
	667	+void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
665	668	static inline bool mptcp_is_fully_established(struct sock *sk)
666	669	{
667	670	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
@@ -926,12 +929,25 @@ static inline void __mptcp_do_fallback(struct mptcp_sock *msk) · 258 unmodified lines
926	929	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
927	930	}
928	931
929		−static inline void mptcp_do_fallback(struct sock *sk)
	932	+static inline void mptcp_do_fallback(struct sock *ssk)
930	933	{
931		− struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
932		− struct mptcp_sock *msk = mptcp_sk(subflow->conn);
	934	+ struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
	935	+ struct sock *sk = subflow->conn;
	936	+ struct mptcp_sock *msk;
933	937
	938	+ msk = mptcp_sk(sk);
934	939	__mptcp_do_fallback(msk);
	940	+ if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
	941	+ gfp_t saved_allocation = ssk->sk_allocation;
	942	+
	943	+ /* we are in a atomic (BH) scope, override ssk default for data
	944	+ * fin allocation
	945	+ */
	946	+ ssk->sk_allocation = GFP_ATOMIC;
	947	+ ssk->sk_shutdown \|= SEND_SHUTDOWN;
	948	+ tcp_shutdown(ssk, SEND_SHUTDOWN);
	949	+ ssk->sk_allocation = saved_allocation;
	950	+ }
935	951	}
936	952
937	953	#define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)

net/mptcp/subflow.c+98−29

@@ -843,7 +843,8 @@ enum mapping_status { · 842 unmodified lines
843	843	MAPPING_INVALID,
844	844	MAPPING_EMPTY,
845	845	MAPPING_DATA_FIN,
846		− MAPPING_DUMMY
	846	+ MAPPING_DUMMY,
	847	+ MAPPING_BAD_CSUM
847	848	};
848	849
849	850	static void dbg_bad_map(struct mptcp_subflow_context *subflow, u32 ssn)
@@ -958,11 +959,7 @@ static enum mapping_status validate_data_csum(struct sock ssk, struct sk_buff · 108 unmodified lines
958	959	subflow->map_data_csum);
959	960	if (unlikely(csum)) {
960	961	MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DATACSUMERR);
961		− if (subflow->mp_join \|\| subflow->valid_csum_seen) {
962		− subflow->send_mp_fail = 1;
963		− MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_MPFAILTX);
964		− }
965		− return subflow->mp_join ? MAPPING_INVALID : MAPPING_DUMMY;
	962	+ return MAPPING_BAD_CSUM;
966	963	}
967	964
968	965	subflow->valid_csum_seen = 1;
@@ -974,7 +971,6 @@ static enum mapping_status get_mapping_status(struct sock *ssk, · 5 unmodified lines
974	971	{
975	972	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
976	973	bool csum_reqd = READ_ONCE(msk->csum_enabled);
977		− struct sock sk = (struct sock )msk;
978	974	struct mptcp_ext *mpext;
979	975	struct sk_buff *skb;
980	976	u16 data_len;
@@ -1016,9 +1012,6 @@ static enum mapping_status get_mapping_status(struct sock *ssk, · 35 unmodified lines
1016	1012	pr_debug("infinite mapping received");
1017	1013	MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPRX);
1018	1014	subflow->map_data_len = 0;
1019		− if (!sock_flag(ssk, SOCK_DEAD))
1020		− sk_stop_timer(sk, &sk->sk_timer);
1021		−
1022	1015	return MAPPING_INVALID;
1023	1016	}
1024	1017
@@ -1165,6 +1158,33 @@ static bool subflow_can_fallback(struct mptcp_subflow_context *subflow) · 140 unmodified lines
1165	1158	return !subflow->fully_established;
1166	1159	}
1167	1160
	1161	+static void mptcp_subflow_fail(struct mptcp_sock msk, struct sock ssk)
	1162	+{
	1163	+ struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
	1164	+ unsigned long fail_tout;
	1165	+
	1166	+ /* greceful failure can happen only on the MPC subflow */
	1167	+ if (WARN_ON_ONCE(ssk != READ_ONCE(msk->first)))
	1168	+ return;
	1169	+
	1170	+ /* since the close timeout take precedence on the fail one,
	1171	+ * no need to start the latter when the first is already set
	1172	+ */
	1173	+ if (sock_flag((struct sock *)msk, SOCK_DEAD))
	1174	+ return;
	1175	+
	1176	+ /* we don't need extreme accuracy here, use a zero fail_tout as special
	1177	+ * value meaning no fail timeout at all;
	1178	+ */
	1179	+ fail_tout = jiffies + TCP_RTO_MAX;
	1180	+ if (!fail_tout)
	1181	+ fail_tout = 1;
	1182	+ WRITE_ONCE(subflow->fail_tout, fail_tout);
	1183	+ tcp_send_ack(ssk);
	1184	+
	1185	+ mptcp_reset_timeout(msk, subflow->fail_tout);
	1186	+}
	1187	+
1168	1188	static bool subflow_check_data_avail(struct sock *ssk)
1169	1189	{
1170	1190	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
@@ -1184,10 +1204,8 @@ static bool subflow_check_data_avail(struct sock *ssk) · 13 unmodified lines
1184	1204
1185	1205	status = get_mapping_status(ssk, msk);
1186	1206	trace_subflow_check_data_avail(status, skb_peek(&ssk->sk_receive_queue));
1187		− if (unlikely(status == MAPPING_INVALID))
1188		− goto fallback;
1189		−
1190		− if (unlikely(status == MAPPING_DUMMY))
	1207	+ if (unlikely(status == MAPPING_INVALID \|\| status == MAPPING_DUMMY \|\|
	1208	+ status == MAPPING_BAD_CSUM))
1191	1209	goto fallback;
1192	1210
1193	1211	if (status != MAPPING_OK)
@@ -1229,22 +1247,17 @@ no_data: · 35 unmodified lines
1229	1247	fallback:
1230	1248	if (!__mptcp_check_fallback(msk)) {
1231	1249	/* RFC 8684 section 3.7. */
1232		− if (subflow->send_mp_fail) {
	1250	+ if (status == MAPPING_BAD_CSUM &&
	1251	+ (subflow->mp_join \|\| subflow->valid_csum_seen)) {
	1252	+ subflow->send_mp_fail = 1;
	1253	+
1233	1254	if (!READ_ONCE(msk->allow_infinite_fallback)) {
1234		− ssk->sk_err = EBADMSG;
1235		− tcp_set_state(ssk, TCP_CLOSE);
1236	1255	subflow->reset_transient = 0;
1237	1256	subflow->reset_reason = MPTCP_RST_EMIDDLEBOX;
1238		− tcp_send_active_reset(ssk, GFP_ATOMIC);
1239		− while ((skb = skb_peek(&ssk->sk_receive_queue)))
1240		− sk_eat_skb(ssk, skb);
1241		− } else if (!sock_flag(ssk, SOCK_DEAD)) {
1242		− WRITE_ONCE(subflow->mp_fail_response_expect, true);
1243		− sk_reset_timer((struct sock *)msk,
1244		− &((struct sock *)msk)->sk_timer,
1245		− jiffies + TCP_RTO_MAX);
	1257	+ goto reset;
1246	1258	}
1247		− WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA);
	1259	+ mptcp_subflow_fail(msk, ssk);
	1260	+ WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_DATA_AVAIL);
1248	1261	return true;
1249	1262	}
1250	1263
@@ -1252,16 +1265,20 @@ fallback: · 1 unmodified line
1252	1265	/* fatal protocol error, close the socket.
1253	1266	* subflow_error_report() will introduce the appropriate barriers
1254	1267	*/
1255		− ssk->sk_err = EBADMSG;
1256		− tcp_set_state(ssk, TCP_CLOSE);
1257	1268	subflow->reset_transient = 0;
1258	1269	subflow->reset_reason = MPTCP_RST_EMPTCP;
	1270	+
	1271	+reset:
	1272	+ ssk->sk_err = EBADMSG;
	1273	+ tcp_set_state(ssk, TCP_CLOSE);
	1274	+ while ((skb = skb_peek(&ssk->sk_receive_queue)))
	1275	+ sk_eat_skb(ssk, skb);
1259	1276	tcp_send_active_reset(ssk, GFP_ATOMIC);
1260	1277	WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA);
1261	1278	return false;
1262	1279	}
1263	1280
1264		− __mptcp_do_fallback(msk);
	1281	+ mptcp_do_fallback(ssk);
1265	1282	}
1266	1283
1267	1284	skb = skb_peek(&ssk->sk_receive_queue);
@@ -1706,6 +1723,58 @@ static void subflow_state_change(struct sock *sk) · 438 unmodified lines
1706	1723	}
1707	1724	}
1708	1725
	1726	+void mptcp_subflow_queue_clean(struct sock *listener_ssk)
	1727	+{
	1728	+ struct request_sock_queue *queue = &inet_csk(listener_ssk)->icsk_accept_queue;
	1729	+ struct mptcp_sock msk, next, *head = NULL;
	1730	+ struct request_sock *req;
	1731	+
	1732	+ /* build a list of all unaccepted mptcp sockets */
	1733	+ spin_lock_bh(&queue->rskq_lock);
	1734	+ for (req = queue->rskq_accept_head; req; req = req->dl_next) {
	1735	+ struct mptcp_subflow_context *subflow;
	1736	+ struct sock *ssk = req->sk;
	1737	+ struct mptcp_sock *msk;
	1738	+
	1739	+ if (!sk_is_mptcp(ssk))
	1740	+ continue;
	1741	+
	1742	+ subflow = mptcp_subflow_ctx(ssk);
	1743	+ if (!subflow \|\| !subflow->conn)
	1744	+ continue;
	1745	+
	1746	+ /* skip if already in list */
	1747	+ msk = mptcp_sk(subflow->conn);
	1748	+ if (msk->dl_next \|\| msk == head)
	1749	+ continue;
	1750	+
	1751	+ msk->dl_next = head;
	1752	+ head = msk;
	1753	+ }
	1754	+ spin_unlock_bh(&queue->rskq_lock);
	1755	+ if (!head)
	1756	+ return;
	1757	+
	1758	+ /* can't acquire the msk socket lock under the subflow one,
	1759	+ * or will cause ABBA deadlock
	1760	+ */
	1761	+ release_sock(listener_ssk);
	1762	+
	1763	+ for (msk = head; msk; msk = next) {
	1764	+ struct sock sk = (struct sock )msk;
	1765	+ bool slow;
	1766	+
	1767	+ slow = lock_sock_fast_nested(sk);
	1768	+ next = msk->dl_next;
	1769	+ msk->first = NULL;
	1770	+ msk->dl_next = NULL;
	1771	+ unlock_sock_fast(sk, slow);
	1772	+ }
	1773	+
	1774	+ /* we are still under the listener msk socket lock */
	1775	+ lock_sock_nested(listener_ssk, SINGLE_DEPTH_NESTING);
	1776	+}
	1777	+
1709	1778	static int subflow_ulp_init(struct sock *sk)
1710	1779	{
1711	1780	struct inet_connection_sock *icsk = inet_csk(sk);

net/ncsi/ncsi-manage.c+2−1

@@ -1803,7 +1803,8 @@ struct ncsi_dev ncsi_register_dev(struct net_device dev, · 1802 unmodified lines
1803	1803	pdev = to_platform_device(dev->dev.parent);
1804	1804	if (pdev) {
1805	1805	np = pdev->dev.of_node;
1806		− if (np && of_get_property(np, "mlx,multi-host", NULL))
	1806	+ if (np && (of_get_property(np, "mellanox,multi-host", NULL) \|\|
	1807	+ of_get_property(np, "mlx,multi-host", NULL)))
1807	1808	ndp->mlx_multi_host = true;
1808	1809	}
1809	1810

net/netfilter/nf_tables_core.c+21−3

@@ -25,9 +25,7 @@ static noinline void __nft_trace_packet(struct nft_traceinfo *info, · 24 unmodified lines
25	25	const struct nft_chain *chain,
26	26	enum nft_trace_types type)
27	27	{
28		− const struct nft_pktinfo *pkt = info->pkt;
29		−
30		− if (!info->trace \|\| !pkt->skb->nf_trace)
	28	+ if (!info->trace \|\| !info->nf_trace)
31	29	return;
32	30
33	31	info->chain = chain;
@@ -42,11 +40,24 @@ static inline void nft_trace_packet(struct nft_traceinfo *info, · 8 unmodified lines
42	40	enum nft_trace_types type)
43	41	{
44	42	if (static_branch_unlikely(&nft_trace_enabled)) {
	43	+ const struct nft_pktinfo *pkt = info->pkt;
	44	+
	45	+ info->nf_trace = pkt->skb->nf_trace;
45	46	info->rule = rule;
46	47	__nft_trace_packet(info, chain, type);
47	48	}
48	49	}
49	50
	51	+static inline void nft_trace_copy_nftrace(struct nft_traceinfo *info)
	52	+{
	53	+ if (static_branch_unlikely(&nft_trace_enabled)) {
	54	+ const struct nft_pktinfo *pkt = info->pkt;
	55	+
	56	+ if (info->trace)
	57	+ info->nf_trace = pkt->skb->nf_trace;
	58	+ }
	59	+}
	60	+
50	61	static void nft_bitwise_fast_eval(const struct nft_expr *expr,
51	62	struct nft_regs *regs)
52	63	{
@@ -85,6 +96,7 @@ static noinline void __nft_trace_verdict(struct nft_traceinfo *info, · 32 unmodified lines
85	96	const struct nft_chain *chain,
86	97	const struct nft_regs *regs)
87	98	{
	99	+ const struct nft_pktinfo *pkt = info->pkt;
88	100	enum nft_trace_types type;
89	101
90	102	switch (regs->verdict.code) {
@@ -92,8 +104,13 @@ static noinline void __nft_trace_verdict(struct nft_traceinfo *info, · 1 unmodified line
92	104	case NFT_RETURN:
93	105	type = NFT_TRACETYPE_RETURN;
94	106	break;
	107	+ case NF_STOLEN:
	108	+ type = NFT_TRACETYPE_RULE;
	109	+ /* can't access skb->nf_trace; use copy */
	110	+ break;
95	111	default:
96	112	type = NFT_TRACETYPE_RULE;
	113	+ info->nf_trace = pkt->skb->nf_trace;
97	114	break;
98	115	}
99	116
@@ -254,6 +271,7 @@ next_rule: · 154 unmodified lines
254	271	switch (regs.verdict.code) {
255	272	case NFT_BREAK:
256	273	regs.verdict.code = NFT_CONTINUE;
	274	+ nft_trace_copy_nftrace(&info);
257	275	continue;
258	276	case NFT_CONTINUE:
259	277	nft_trace_packet(&info, chain, rule,

net/netfilter/nf_tables_trace.c+24−20

@@ -7,7 +7,7 @@ · 6 unmodified lines
7	7	#include <linux/module.h>
8	8	#include <linux/static_key.h>
9	9	#include <linux/hash.h>
10		−#include <linux/jhash.h>
	10	+#include <linux/siphash.h>
11	11	#include <linux/if_vlan.h>
12	12	#include <linux/init.h>
13	13	#include <linux/skbuff.h>
@@ -25,22 +25,6 @@ · 11 unmodified lines
25	25	DEFINE_STATIC_KEY_FALSE(nft_trace_enabled);
26	26	EXPORT_SYMBOL_GPL(nft_trace_enabled);
27	27
28		−static int trace_fill_id(struct sk_buff nlskb, struct sk_buff skb)
29		−{
30		− __be32 id;
31		−
32		− /* using skb address as ID results in a limited number of
33		− * values (and quick reuse).
34		− *
35		− * So we attempt to use as many skb members that will not
36		− * change while skb is with netfilter.
37		− */
38		− id = (__be32)jhash_2words(hash32_ptr(skb), skb_get_hash(skb),
39		− skb->skb_iif);
40		−
41		− return nla_put_be32(nlskb, NFTA_TRACE_ID, id);
42		−}
43		−
44	28	static int trace_fill_header(struct sk_buff *nlskb, u16 type,
45	29	const struct sk_buff *skb,
46	30	int off, unsigned int len)
@@ -186,6 +170,7 @@ void nft_trace_notify(struct nft_traceinfo *info) · 139 unmodified lines
186	170	struct nlmsghdr *nlh;
187	171	struct sk_buff *skb;
188	172	unsigned int size;
	173	+ u32 mark = 0;
189	174	u16 event;
190	175
191	176	if (!nfnetlink_has_listeners(nft_net(pkt), NFNLGRP_NFTRACE))
@@ -229,7 +214,7 @@ void nft_trace_notify(struct nft_traceinfo *info) · 37 unmodified lines
229	214	if (nla_put_be32(skb, NFTA_TRACE_TYPE, htonl(info->type)))
230	215	goto nla_put_failure;
231	216
232		− if (trace_fill_id(skb, pkt->skb))
	217	+ if (nla_put_u32(skb, NFTA_TRACE_ID, info->skbid))
233	218	goto nla_put_failure;
234	219
235	220	if (nla_put_string(skb, NFTA_TRACE_CHAIN, info->chain->name))
@@ -249,16 +234,24 @@ void nft_trace_notify(struct nft_traceinfo *info) · 13 unmodified lines
249	234	case NFT_TRACETYPE_RULE:
250	235	if (nft_verdict_dump(skb, NFTA_TRACE_VERDICT, info->verdict))
251	236	goto nla_put_failure;
	237	+
	238	+ /* pkt->skb undefined iff NF_STOLEN, disable dump */
	239	+ if (info->verdict->code == NF_STOLEN)
	240	+ info->packet_dumped = true;
	241	+ else
	242	+ mark = pkt->skb->mark;
	243	+
252	244	break;
253	245	case NFT_TRACETYPE_POLICY:
	246	+ mark = pkt->skb->mark;
	247	+
254	248	if (nla_put_be32(skb, NFTA_TRACE_POLICY,
255	249	htonl(info->basechain->policy)))
256	250	goto nla_put_failure;
257	251	break;
258	252	}
259	253
260		− if (pkt->skb->mark &&
261		− nla_put_be32(skb, NFTA_TRACE_MARK, htonl(pkt->skb->mark)))
	254	+ if (mark && nla_put_be32(skb, NFTA_TRACE_MARK, htonl(mark)))
262	255	goto nla_put_failure;
263	256
264	257	if (!info->packet_dumped) {
@@ -283,9 +276,20 @@ void nft_trace_init(struct nft_traceinfo info, const struct nft_pktinfo pkt, · 18 unmodified lines
283	276	const struct nft_verdict *verdict,
284	277	const struct nft_chain *chain)
285	278	{
	279	+ static siphash_key_t trace_key __read_mostly;
	280	+ struct sk_buff *skb = pkt->skb;
	281	+
286	282	info->basechain = nft_base_chain(chain);
287	283	info->trace = true;
	284	+ info->nf_trace = pkt->skb->nf_trace;
288	285	info->packet_dumped = false;
289	286	info->pkt = pkt;
290	287	info->verdict = verdict;
	288	+
	289	+ net_get_random_once(&trace_key, sizeof(trace_key));
	290	+
	291	+ info->skbid = (u32)siphash_3u32(hash32_ptr(skb),
	292	+ skb_get_hash(skb),
	293	+ skb->skb_iif,
	294	+ &trace_key);
291	295	}

net/netfilter/nft_set_hash.c+2−0

@@ -143,6 +143,7 @@ static bool nft_rhash_update(struct nft_set set, const u32 key, · 142 unmodified lines
143	143	/* Another cpu may race to insert the element with the same key */
144	144	if (prev) {
145	145	nft_set_elem_destroy(set, he, true);
	146	+ atomic_dec(&set->nelems);
146	147	he = prev;
147	148	}
148	149
@@ -152,6 +153,7 @@ out: · 3 unmodified lines
152	153
153	154	err2:
154	155	nft_set_elem_destroy(set, he, true);
	156	+ atomic_dec(&set->nelems);
155	157	err1:
156	158	return false;
157	159	}

net/rose/rose_timer.c+19−15

@@ -31,89 +31,89 @@ static void rose_idletimer_expiry(struct timer_list *); · 30 unmodified lines
31	31
32	32	void rose_start_heartbeat(struct sock *sk)
33	33	{
34		− del_timer(&sk->sk_timer);
	34	+ sk_stop_timer(sk, &sk->sk_timer);
35	35
36	36	sk->sk_timer.function = rose_heartbeat_expiry;
37	37	sk->sk_timer.expires = jiffies + 5 * HZ;
38	38
39		− add_timer(&sk->sk_timer);
	39	+ sk_reset_timer(sk, &sk->sk_timer, sk->sk_timer.expires);
40	40	}
41	41
42	42	void rose_start_t1timer(struct sock *sk)
43	43	{
44	44	struct rose_sock *rose = rose_sk(sk);
45	45
46		− del_timer(&rose->timer);
	46	+ sk_stop_timer(sk, &rose->timer);
47	47
48	48	rose->timer.function = rose_timer_expiry;
49	49	rose->timer.expires = jiffies + rose->t1;
50	50
51		− add_timer(&rose->timer);
	51	+ sk_reset_timer(sk, &rose->timer, rose->timer.expires);
52	52	}
53	53
54	54	void rose_start_t2timer(struct sock *sk)
55	55	{
56	56	struct rose_sock *rose = rose_sk(sk);
57	57
58		− del_timer(&rose->timer);
	58	+ sk_stop_timer(sk, &rose->timer);
59	59
60	60	rose->timer.function = rose_timer_expiry;
61	61	rose->timer.expires = jiffies + rose->t2;
62	62
63		− add_timer(&rose->timer);
	63	+ sk_reset_timer(sk, &rose->timer, rose->timer.expires);
64	64	}
65	65
66	66	void rose_start_t3timer(struct sock *sk)
67	67	{
68	68	struct rose_sock *rose = rose_sk(sk);
69	69
70		− del_timer(&rose->timer);
	70	+ sk_stop_timer(sk, &rose->timer);
71	71
72	72	rose->timer.function = rose_timer_expiry;
73	73	rose->timer.expires = jiffies + rose->t3;
74	74
75		− add_timer(&rose->timer);
	75	+ sk_reset_timer(sk, &rose->timer, rose->timer.expires);
76	76	}
77	77
78	78	void rose_start_hbtimer(struct sock *sk)
79	79	{
80	80	struct rose_sock *rose = rose_sk(sk);
81	81
82		− del_timer(&rose->timer);
	82	+ sk_stop_timer(sk, &rose->timer);
83	83
84	84	rose->timer.function = rose_timer_expiry;
85	85	rose->timer.expires = jiffies + rose->hb;
86	86
87		− add_timer(&rose->timer);
	87	+ sk_reset_timer(sk, &rose->timer, rose->timer.expires);
88	88	}
89	89
90	90	void rose_start_idletimer(struct sock *sk)
91	91	{
92	92	struct rose_sock *rose = rose_sk(sk);
93	93
94		− del_timer(&rose->idletimer);
	94	+ sk_stop_timer(sk, &rose->idletimer);
95	95
96	96	if (rose->idle > 0) {
97	97	rose->idletimer.function = rose_idletimer_expiry;
98	98	rose->idletimer.expires = jiffies + rose->idle;
99	99
100		− add_timer(&rose->idletimer);
	100	+ sk_reset_timer(sk, &rose->idletimer, rose->idletimer.expires);
101	101	}
102	102	}
103	103
104	104	void rose_stop_heartbeat(struct sock *sk)
105	105	{
106		− del_timer(&sk->sk_timer);
	106	+ sk_stop_timer(sk, &sk->sk_timer);
107	107	}
108	108
109	109	void rose_stop_timer(struct sock *sk)
110	110	{
111		− del_timer(&rose_sk(sk)->timer);
	111	+ sk_stop_timer(sk, &rose_sk(sk)->timer);
112	112	}
113	113
114	114	void rose_stop_idletimer(struct sock *sk)
115	115	{
116		− del_timer(&rose_sk(sk)->idletimer);
	116	+ sk_stop_timer(sk, &rose_sk(sk)->idletimer);
117	117	}
118	118
119	119	static void rose_heartbeat_expiry(struct timer_list *t)
@@ -130,6 +130,7 @@ static void rose_heartbeat_expiry(struct timer_list *t) · 10 unmodified lines
130	130	(sk->sk_state == TCP_LISTEN && sock_flag(sk, SOCK_DEAD))) {
131	131	bh_unlock_sock(sk);
132	132	rose_destroy_socket(sk);
	133	+ sock_put(sk);
133	134	return;
134	135	}
135	136	break;
@@ -152,6 +153,7 @@ static void rose_heartbeat_expiry(struct timer_list *t) · 16 unmodified lines
152	153
153	154	rose_start_heartbeat(sk);
154	155	bh_unlock_sock(sk);
	156	+ sock_put(sk);
155	157	}
156	158
157	159	static void rose_timer_expiry(struct timer_list *t)
@@ -181,6 +183,7 @@ static void rose_timer_expiry(struct timer_list *t) · 23 unmodified lines
181	183	break;
182	184	}
183	185	bh_unlock_sock(sk);
	186	+ sock_put(sk);
184	187	}
185	188
186	189	static void rose_idletimer_expiry(struct timer_list *t)
@@ -205,4 +208,5 @@ static void rose_idletimer_expiry(struct timer_list *t) · 18 unmodified lines
205	208	sock_set_flag(sk, SOCK_DEAD);
206	209	}
207	210	bh_unlock_sock(sk);
	211	+ sock_put(sk);
208	212	}

net/sched/act_api.c+14−8

@@ -588,7 +588,8 @@ static int tcf_idr_release_unsafe(struct tc_action *p) · 587 unmodified lines
588	588	}
589	589
590	590	static int tcf_del_walker(struct tcf_idrinfo idrinfo, struct sk_buff skb,
591		− const struct tc_action_ops *ops)
	591	+ const struct tc_action_ops *ops,
	592	+ struct netlink_ext_ack *extack)
592	593	{
593	594	struct nlattr *nest;
594	595	int n_i = 0;
@@ -604,20 +605,25 @@ static int tcf_del_walker(struct tcf_idrinfo idrinfo, struct sk_buff skb, · 9 unmodified lines
604	605	if (nla_put_string(skb, TCA_KIND, ops->kind))
605	606	goto nla_put_failure;
606	607
	608	+ ret = 0;
607	609	mutex_lock(&idrinfo->lock);
608	610	idr_for_each_entry_ul(idr, p, tmp, id) {
609	611	if (IS_ERR(p))
610	612	continue;
611	613	ret = tcf_idr_release_unsafe(p);
612		− if (ret == ACT_P_DELETED) {
	614	+ if (ret == ACT_P_DELETED)
613	615	module_put(ops->owner);
614		− n_i++;
615		− } else if (ret < 0) {
616		− mutex_unlock(&idrinfo->lock);
617		− goto nla_put_failure;
618		− }
	616	+ else if (ret < 0)
	617	+ break;
	618	+ n_i++;
619	619	}
620	620	mutex_unlock(&idrinfo->lock);
	621	+ if (ret < 0) {
	622	+ if (n_i)
	623	+ NL_SET_ERR_MSG(extack, "Unable to flush all TC actions");
	624	+ else
	625	+ goto nla_put_failure;
	626	+ }
621	627
622	628	ret = nla_put_u32(skb, TCA_FCNT, n_i);
623	629	if (ret)
@@ -638,7 +644,7 @@ int tcf_generic_walker(struct tc_action_net tn, struct sk_buff skb, · 14 unmodified lines
638	644	struct tcf_idrinfo *idrinfo = tn->idrinfo;
639	645
640	646	if (type == RTM_DELACTION) {
641		− return tcf_del_walker(idrinfo, skb, ops);
	647	+ return tcf_del_walker(idrinfo, skb, ops, extack);
642	648	} else if (type == RTM_GETACTION) {
643	649	return tcf_dump_walker(idrinfo, skb, cb);
644	650	} else {

net/socket.c+6−10

@@ -2149,10 +2149,13 @@ SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len, · 2148 unmodified lines
2149	2149	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2150	2150	struct sockaddr __user addr, int __user addr_len)
2151	2151	{
	2152	+ struct sockaddr_storage address;
	2153	+ struct msghdr msg = {
	2154	+ /* Save some cycles and don't copy the address if not needed */
	2155	+ .msg_name = addr ? (struct sockaddr *)&address : NULL,
	2156	+ };
2152	2157	struct socket *sock;
2153	2158	struct iovec iov;
2154		− struct msghdr msg;
2155		− struct sockaddr_storage address;
2156	2159	int err, err2;
2157	2160	int fput_needed;
2158	2161
@@ -2163,14 +2166,6 @@ int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags, · 4 unmodified lines
2163	2166	if (!sock)
2164	2167	goto out;
2165	2168
2166		− msg.msg_control = NULL;
2167		− msg.msg_controllen = 0;
2168		− /* Save some cycles and don't copy the address if not needed */
2169		− msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2170		− /* We assume all kernel code knows the size of sockaddr_storage */
2171		− msg.msg_namelen = 0;
2172		− msg.msg_iocb = NULL;
2173		− msg.msg_flags = 0;
2174	2169	if (sock->file->f_flags & O_NONBLOCK)
2175	2170	flags \|= MSG_DONTWAIT;
2176	2171	err = sock_recvmsg(sock, &msg, flags);
@@ -2375,6 +2370,7 @@ int __copy_msghdr_from_user(struct msghdr *kmsg, · 198 unmodified lines
2375	2370	return -EFAULT;
2376	2371
2377	2372	kmsg->msg_control_is_user = true;
	2373	+ kmsg->msg_get_inq = 0;
2378	2374	kmsg->msg_control_user = msg.msg_control;
2379	2375	kmsg->msg_controllen = msg.msg_controllen;
2380	2376	kmsg->msg_flags = msg.msg_flags;

net/tipc/node.c+22−19

@@ -472,8 +472,8 @@ struct tipc_node tipc_node_create(struct net net, u32 addr, u8 *peer_id, · 471 unmodified lines
472	472	bool preliminary)
473	473	{
474	474	struct tipc_net *tn = net_generic(net, tipc_net_id);
	475	+ struct tipc_link l, snd_l = tipc_bc_sndlink(net);
475	476	struct tipc_node n, temp_node;
476		− struct tipc_link *l;
477	477	unsigned long intv;
478	478	int bearer_id;
479	479	int i;
@@ -488,6 +488,16 @@ struct tipc_node tipc_node_create(struct net net, u32 addr, u8 *peer_id, · 8 unmodified lines
488	488	goto exit;
489	489	/* A preliminary node becomes "real" now, refresh its data */
490	490	tipc_node_write_lock(n);
	491	+ if (!tipc_link_bc_create(net, tipc_own_addr(net), addr, peer_id, U16_MAX,
	492	+ tipc_link_min_win(snd_l), tipc_link_max_win(snd_l),
	493	+ n->capabilities, &n->bc_entry.inputq1,
	494	+ &n->bc_entry.namedq, snd_l, &n->bc_entry.link)) {
	495	+ pr_warn("Broadcast rcv link refresh failed, no memory\n");
	496	+ tipc_node_write_unlock_fast(n);
	497	+ tipc_node_put(n);
	498	+ n = NULL;
	499	+ goto exit;
	500	+ }
491	501	n->preliminary = false;
492	502	n->addr = addr;
493	503	hlist_del_rcu(&n->hash);
@@ -567,7 +577,16 @@ update: · 73 unmodified lines
567	577	n->signature = INVALID_NODE_SIG;
568	578	n->active_links[0] = INVALID_BEARER_ID;
569	579	n->active_links[1] = INVALID_BEARER_ID;
570		− n->bc_entry.link = NULL;
	580	+ if (!preliminary &&
	581	+ !tipc_link_bc_create(net, tipc_own_addr(net), addr, peer_id, U16_MAX,
	582	+ tipc_link_min_win(snd_l), tipc_link_max_win(snd_l),
	583	+ n->capabilities, &n->bc_entry.inputq1,
	584	+ &n->bc_entry.namedq, snd_l, &n->bc_entry.link)) {
	585	+ pr_warn("Broadcast rcv link creation failed, no memory\n");
	586	+ kfree(n);
	587	+ n = NULL;
	588	+ goto exit;
	589	+ }
571	590	tipc_node_get(n);
572	591	timer_setup(&n->timer, tipc_node_timeout, 0);
573	592	/* Start a slow timer anyway, crypto needs it */
@@ -1155,7 +1174,7 @@ void tipc_node_check_dest(struct net *net, u32 addr, · 581 unmodified lines
1155	1174	bool respond, bool dupl_addr)
1156	1175	{
1157	1176	struct tipc_node *n;
1158		− struct tipc_link l, snd_l;
	1177	+ struct tipc_link *l;
1159	1178	struct tipc_link_entry *le;
1160	1179	bool addr_match = false;
1161	1180	bool sign_match = false;
@@ -1175,22 +1194,6 @@ void tipc_node_check_dest(struct net *net, u32 addr, · 13 unmodified lines
1175	1194	return;
1176	1195
1177	1196	tipc_node_write_lock(n);
1178		− if (unlikely(!n->bc_entry.link)) {
1179		− snd_l = tipc_bc_sndlink(net);
1180		− if (!tipc_link_bc_create(net, tipc_own_addr(net),
1181		− addr, peer_id, U16_MAX,
1182		− tipc_link_min_win(snd_l),
1183		− tipc_link_max_win(snd_l),
1184		− n->capabilities,
1185		− &n->bc_entry.inputq1,
1186		− &n->bc_entry.namedq, snd_l,
1187		− &n->bc_entry.link)) {
1188		− pr_warn("Broadcast rcv link creation failed, no mem\n");
1189		− tipc_node_write_unlock_fast(n);
1190		− tipc_node_put(n);
1191		− return;
1192		− }
1193		− }
1194	1197
1195	1198	le = &n->links[b->identity];
1196	1199

net/tipc/socket.c+1−0

@@ -502,6 +502,7 @@ static int tipc_sk_create(struct net net, struct socket sock, · 501 unmodified lines
502	502	sock_init_data(sock, sk);
503	503	tipc_set_sk_state(sk, TIPC_OPEN);
504	504	if (tipc_sk_insert(tsk)) {
	505	+ sk_free(sk);
505	506	pr_warn("Socket create failed; port number exhausted\n");
506	507	return -EINVAL;
507	508	}

tools/testing/selftests/bpf/prog_tests/sockmap_ktls.c+75−9

@@ -4,6 +4,7 @@ · 3 unmodified lines
4	4	* Tests for sockmap/sockhash holding kTLS sockets.
5	5	*/
6	6
	7	+#include <netinet/tcp.h>
7	8	#include "test_progs.h"
8	9
9	10	#define MAX_TEST_NAME 80
@@ -92,9 +93,78 @@ close_srv: · 82 unmodified lines
92	93	close(srv);
93	94	}
94	95
	96	+static void test_sockmap_ktls_update_fails_when_sock_has_ulp(int family, int map)
	97	+{
	98	+ struct sockaddr_storage addr = {};
	99	+ socklen_t len = sizeof(addr);
	100	+ struct sockaddr_in6 *v6;
	101	+ struct sockaddr_in *v4;
	102	+ int err, s, zero = 0;
	103	+
	104	+ switch (family) {
	105	+ case AF_INET:
	106	+ v4 = (struct sockaddr_in *)&addr;
	107	+ v4->sin_family = AF_INET;
	108	+ break;
	109	+ case AF_INET6:
	110	+ v6 = (struct sockaddr_in6 *)&addr;
	111	+ v6->sin6_family = AF_INET6;
	112	+ break;
	113	+ default:
	114	+ PRINT_FAIL("unsupported socket family %d", family);
	115	+ return;
	116	+ }
	117	+
	118	+ s = socket(family, SOCK_STREAM, 0);
	119	+ if (!ASSERT_GE(s, 0, "socket"))
	120	+ return;
	121	+
	122	+ err = bind(s, (struct sockaddr *)&addr, len);
	123	+ if (!ASSERT_OK(err, "bind"))
	124	+ goto close;
	125	+
	126	+ err = getsockname(s, (struct sockaddr *)&addr, &len);
	127	+ if (!ASSERT_OK(err, "getsockname"))
	128	+ goto close;
	129	+
	130	+ err = connect(s, (struct sockaddr *)&addr, len);
	131	+ if (!ASSERT_OK(err, "connect"))
	132	+ goto close;
	133	+
	134	+ /* save sk->sk_prot and set it to tls_prots */
	135	+ err = setsockopt(s, IPPROTO_TCP, TCP_ULP, "tls", strlen("tls"));
	136	+ if (!ASSERT_OK(err, "setsockopt(TCP_ULP)"))
	137	+ goto close;
	138	+
	139	+ /* sockmap update should not affect saved sk_prot */
	140	+ err = bpf_map_update_elem(map, &zero, &s, BPF_ANY);
	141	+ if (!ASSERT_ERR(err, "sockmap update elem"))
	142	+ goto close;
	143	+
	144	+ /* call sk->sk_prot->setsockopt to dispatch to saved sk_prot */
	145	+ err = setsockopt(s, IPPROTO_TCP, TCP_NODELAY, &zero, sizeof(zero));
	146	+ ASSERT_OK(err, "setsockopt(TCP_NODELAY)");
	147	+
	148	+close:
	149	+ close(s);
	150	+}
	151	+
	152	+static const char fmt_test_name(const char subtest_name, int family,
	153	+ enum bpf_map_type map_type)
	154	+{
	155	+ const char *map_type_str = BPF_MAP_TYPE_SOCKMAP ? "SOCKMAP" : "SOCKHASH";
	156	+ const char *family_str = AF_INET ? "IPv4" : "IPv6";
	157	+ static char test_name[MAX_TEST_NAME];
	158	+
	159	+ snprintf(test_name, MAX_TEST_NAME,
	160	+ "sockmap_ktls %s %s %s",
	161	+ subtest_name, family_str, map_type_str);
	162	+
	163	+ return test_name;
	164	+}
	165	+
95	166	static void run_tests(int family, enum bpf_map_type map_type)
96	167	{
97		− char test_name[MAX_TEST_NAME];
98	168	int map;
99	169
100	170	map = bpf_map_create(map_type, NULL, sizeof(int), sizeof(int), 1, NULL);
@@ -103,14 +173,10 @@ static void run_tests(int family, enum bpf_map_type map_type) · 2 unmodified lines
103	173	return;
104	174	}
105	175
106		− snprintf(test_name, MAX_TEST_NAME,
107		− "sockmap_ktls disconnect_after_delete %s %s",
108		− family == AF_INET ? "IPv4" : "IPv6",
109		− map_type == BPF_MAP_TYPE_SOCKMAP ? "SOCKMAP" : "SOCKHASH");
110		− if (!test__start_subtest(test_name))
111		− return;
112		−
113		− test_sockmap_ktls_disconnect_after_delete(family, map);
	176	+ if (test__start_subtest(fmt_test_name("disconnect_after_delete", family, map_type)))
	177	+ test_sockmap_ktls_disconnect_after_delete(family, map);
	178	+ if (test__start_subtest(fmt_test_name("update_fails_when_sock_has_ulp", family, map_type)))
	179	+ test_sockmap_ktls_update_fails_when_sock_has_ulp(family, map);
114	180
115	181	close(map);
116	182	}

tools/testing/selftests/net/Makefile+1−1

@@ -54,7 +54,7 @@ TEST_GEN_FILES += ipsec · 53 unmodified lines
54	54	TEST_GEN_FILES += ioam6_parser
55	55	TEST_GEN_FILES += gro
56	56	TEST_GEN_PROGS = reuseport_bpf reuseport_bpf_cpu reuseport_bpf_numa
57		−TEST_GEN_PROGS += reuseport_dualstack reuseaddr_conflict tls
	57	+TEST_GEN_PROGS += reuseport_dualstack reuseaddr_conflict tls tun
58	58	TEST_GEN_FILES += toeplitz
59	59	TEST_GEN_FILES += cmsg_sender
60	60	TEST_GEN_FILES += stress_reuseport_listen

tools/testing/selftests/net/bpf/Makefile+1−1

@@ -2,7 +2,7 @@ · 1 unmodified line
2	2
3	3	CLANG ?= clang
4	4	CCINCLUDE += -I../../bpf
5		−CCINCLUDE += -I../../../lib
	5	+CCINCLUDE += -I../../../../lib
6	6	CCINCLUDE += -I../../../../../usr/include/
7	7
8	8	TEST_CUSTOM_PROGS = $(OUTPUT)/bpf/nat6to4.o

tools/testing/selftests/net/mptcp/diag.sh+40−8

@@ -61,6 +61,39 @@ chk_msk_nr() · 60 unmodified lines
61	61	__chk_nr "grep -c token:" $*
62	62	}
63	63
	64	+wait_msk_nr()
	65	+{
	66	+ local condition="grep -c token:"
	67	+ local expected=$1
	68	+ local timeout=20
	69	+ local msg nr
	70	+ local max=0
	71	+ local i=0
	72	+
	73	+ shift 1
	74	+ msg=$*
	75	+
	76	+ while [ $i -lt $timeout ]; do
	77	+ nr=$(ss -inmHMN $ns \| $condition)
	78	+ [ $nr == $expected ] && break;
	79	+ [ $nr -gt $max ] && max=$nr
	80	+ i=$((i + 1))
	81	+ sleep 1
	82	+ done
	83	+
	84	+ printf "%-50s" "$msg"
	85	+ if [ $i -ge $timeout ]; then
	86	+ echo "[ fail ] timeout while expecting $expected max $max last $nr"
	87	+ ret=$test_cnt
	88	+ elif [ $nr != $expected ]; then
	89	+ echo "[ fail ] expected $expected found $nr"
	90	+ ret=$test_cnt
	91	+ else
	92	+ echo "[ ok ]"
	93	+ fi
	94	+ test_cnt=$((test_cnt+1))
	95	+}
	96	+
64	97	chk_msk_fallback_nr()
65	98	{
66	99	__chk_nr "grep -c fallback" $*
@@ -146,7 +179,7 @@ ip -n $ns link set dev lo up · 79 unmodified lines
146	179	echo "a" \| \
147	180	timeout ${timeout_test} \
148	181	ip netns exec $ns \
149		− ./mptcp_connect -p 10000 -l -t ${timeout_poll} \
	182	+ ./mptcp_connect -p 10000 -l -t ${timeout_poll} -w 20 \
150	183	0.0.0.0 >/dev/null &
151	184	wait_local_port_listen $ns 10000
152	185	chk_msk_nr 0 "no msk on netns creation"
@@ -155,7 +188,7 @@ chk_msk_listen 10000 · 2 unmodified lines
155	188	echo "b" \| \
156	189	timeout ${timeout_test} \
157	190	ip netns exec $ns \
158		− ./mptcp_connect -p 10000 -r 0 -t ${timeout_poll} \
	191	+ ./mptcp_connect -p 10000 -r 0 -t ${timeout_poll} -w 20 \
159	192	127.0.0.1 >/dev/null &
160	193	wait_connected $ns 10000
161	194	chk_msk_nr 2 "after MPC handshake "
@@ -167,13 +200,13 @@ flush_pids · 5 unmodified lines
167	200	echo "a" \| \
168	201	timeout ${timeout_test} \
169	202	ip netns exec $ns \
170		− ./mptcp_connect -p 10001 -l -s TCP -t ${timeout_poll} \
	203	+ ./mptcp_connect -p 10001 -l -s TCP -t ${timeout_poll} -w 20 \
171	204	0.0.0.0 >/dev/null &
172	205	wait_local_port_listen $ns 10001
173	206	echo "b" \| \
174	207	timeout ${timeout_test} \
175	208	ip netns exec $ns \
176		− ./mptcp_connect -p 10001 -r 0 -t ${timeout_poll} \
	209	+ ./mptcp_connect -p 10001 -r 0 -t ${timeout_poll} -w 20 \
177	210	127.0.0.1 >/dev/null &
178	211	wait_connected $ns 10001
179	212	chk_msk_fallback_nr 1 "check fallback"
@@ -184,7 +217,7 @@ for I in `seq 1 $NR_CLIENTS`; do · 4 unmodified lines
184	217	echo "a" \| \
185	218	timeout ${timeout_test} \
186	219	ip netns exec $ns \
187		− ./mptcp_connect -p $((I+10001)) -l -w 10 \
	220	+ ./mptcp_connect -p $((I+10001)) -l -w 20 \
188	221	-t ${timeout_poll} 0.0.0.0 >/dev/null &
189	222	done
190	223	wait_local_port_listen $ns $((NR_CLIENTS + 10001))
@@ -193,12 +226,11 @@ for I in `seq 1 $NR_CLIENTS`; do · 2 unmodified lines
193	226	echo "b" \| \
194	227	timeout ${timeout_test} \
195	228	ip netns exec $ns \
196		− ./mptcp_connect -p $((I+10001)) -w 10 \
	229	+ ./mptcp_connect -p $((I+10001)) -w 20 \
197	230	-t ${timeout_poll} 127.0.0.1 >/dev/null &
198	231	done
199		−sleep 1.5
200	232
201		−chk_msk_nr $((NR_CLIENTS*2)) "many msk socket present"
	233	+wait_msk_nr $((NR_CLIENTS*2)) "many msk socket present"
202	234	flush_pids
203	235
204	236	exit $ret

tools/testing/selftests/net/mptcp/mptcp_connect.c+1−1

@@ -265,7 +265,7 @@ static void sock_test_tcpulp(int sock, int proto, unsigned int line) · 264 unmodified lines
265	265	static int sock_listen_mptcp(const char * const listenaddr,
266	266	const char * const port)
267	267	{
268		− int sock;
	268	+ int sock = -1;
269	269	struct addrinfo hints = {
270	270	.ai_protocol = IPPROTO_TCP,
271	271	.ai_socktype = SOCK_STREAM,

tools/testing/selftests/net/mptcp/mptcp_inq.c+1−1

@@ -88,7 +88,7 @@ static void xgetaddrinfo(const char node, const char service, · 87 unmodified lines
88	88	static int sock_listen_mptcp(const char * const listenaddr,
89	89	const char * const port)
90	90	{
91		− int sock;
	91	+ int sock = -1;
92	92	struct addrinfo hints = {
93	93	.ai_protocol = IPPROTO_TCP,
94	94	.ai_socktype = SOCK_STREAM,

tools/testing/selftests/net/mptcp/mptcp_sockopt.c+1−1

@@ -136,7 +136,7 @@ static void xgetaddrinfo(const char node, const char service, · 135 unmodified lines
136	136	static int sock_listen_mptcp(const char * const listenaddr,
137	137	const char * const port)
138	138	{
139		− int sock;
	139	+ int sock = -1;
140	140	struct addrinfo hints = {
141	141	.ai_protocol = IPPROTO_TCP,
142	142	.ai_socktype = SOCK_STREAM,

tools/testing/selftests/net/tun.c+162−0

@@ -0,0 +1,162 @@
	1	+// SPDX-License-Identifier: GPL-2.0
	2	+
	3	+#define _GNU_SOURCE
	4	+
	5	+#include <errno.h>
	6	+#include <fcntl.h>
	7	+#include <stdio.h>
	8	+#include <stdlib.h>
	9	+#include <string.h>
	10	+#include <unistd.h>
	11	+#include <linux/if.h>
	12	+#include <linux/if_tun.h>
	13	+#include <linux/netlink.h>
	14	+#include <linux/rtnetlink.h>
	15	+#include <sys/ioctl.h>
	16	+#include <sys/socket.h>
	17	+
	18	+#include "../kselftest_harness.h"
	19	+
	20	+static int tun_attach(int fd, char *dev)
	21	+{
	22	+ struct ifreq ifr;
	23	+
	24	+ memset(&ifr, 0, sizeof(ifr));
	25	+ strcpy(ifr.ifr_name, dev);
	26	+ ifr.ifr_flags = IFF_ATTACH_QUEUE;
	27	+
	28	+ return ioctl(fd, TUNSETQUEUE, (void *) &ifr);
	29	+}
	30	+
	31	+static int tun_detach(int fd, char *dev)
	32	+{
	33	+ struct ifreq ifr;
	34	+
	35	+ memset(&ifr, 0, sizeof(ifr));
	36	+ strcpy(ifr.ifr_name, dev);
	37	+ ifr.ifr_flags = IFF_DETACH_QUEUE;
	38	+
	39	+ return ioctl(fd, TUNSETQUEUE, (void *) &ifr);
	40	+}
	41	+
	42	+static int tun_alloc(char *dev)
	43	+{
	44	+ struct ifreq ifr;
	45	+ int fd, err;
	46	+
	47	+ fd = open("/dev/net/tun", O_RDWR);
	48	+ if (fd < 0) {
	49	+ fprintf(stderr, "can't open tun: %s\n", strerror(errno));
	50	+ return fd;
	51	+ }
	52	+
	53	+ memset(&ifr, 0, sizeof(ifr));
	54	+ strcpy(ifr.ifr_name, dev);
	55	+ ifr.ifr_flags = IFF_TAP \| IFF_NAPI \| IFF_MULTI_QUEUE;
	56	+
	57	+ err = ioctl(fd, TUNSETIFF, (void *) &ifr);
	58	+ if (err < 0) {
	59	+ fprintf(stderr, "can't TUNSETIFF: %s\n", strerror(errno));
	60	+ close(fd);
	61	+ return err;
	62	+ }
	63	+ strcpy(dev, ifr.ifr_name);
	64	+ return fd;
	65	+}
	66	+
	67	+static int tun_delete(char *dev)
	68	+{
	69	+ struct {
	70	+ struct nlmsghdr nh;
	71	+ struct ifinfomsg ifm;
	72	+ unsigned char data[64];
	73	+ } req;
	74	+ struct rtattr *rta;
	75	+ int ret, rtnl;
	76	+
	77	+ rtnl = socket(AF_NETLINK, SOCK_DGRAM, NETLINK_ROUTE);
	78	+ if (rtnl < 0) {
	79	+ fprintf(stderr, "can't open rtnl: %s\n", strerror(errno));
	80	+ return 1;
	81	+ }
	82	+
	83	+ memset(&req, 0, sizeof(req));
	84	+ req.nh.nlmsg_len = NLMSG_ALIGN(NLMSG_LENGTH(sizeof(req.ifm)));
	85	+ req.nh.nlmsg_flags = NLM_F_REQUEST;
	86	+ req.nh.nlmsg_type = RTM_DELLINK;
	87	+
	88	+ req.ifm.ifi_family = AF_UNSPEC;
	89	+
	90	+ rta = (struct rtattr )(((char )&req) + NLMSG_ALIGN(req.nh.nlmsg_len));
	91	+ rta->rta_type = IFLA_IFNAME;
	92	+ rta->rta_len = RTA_LENGTH(IFNAMSIZ);
	93	+ req.nh.nlmsg_len += rta->rta_len;
	94	+ memcpy(RTA_DATA(rta), dev, IFNAMSIZ);
	95	+
	96	+ ret = send(rtnl, &req, req.nh.nlmsg_len, 0);
	97	+ if (ret < 0)
	98	+ fprintf(stderr, "can't send: %s\n", strerror(errno));
	99	+ ret = (unsigned int)ret != req.nh.nlmsg_len;
	100	+
	101	+ close(rtnl);
	102	+ return ret;
	103	+}
	104	+
	105	+FIXTURE(tun)
	106	+{
	107	+ char ifname[IFNAMSIZ];
	108	+ int fd, fd2;
	109	+};
	110	+
	111	+FIXTURE_SETUP(tun)
	112	+{
	113	+ memset(self->ifname, 0, sizeof(self->ifname));
	114	+
	115	+ self->fd = tun_alloc(self->ifname);
	116	+ ASSERT_GE(self->fd, 0);
	117	+
	118	+ self->fd2 = tun_alloc(self->ifname);
	119	+ ASSERT_GE(self->fd2, 0);
	120	+}
	121	+
	122	+FIXTURE_TEARDOWN(tun)
	123	+{
	124	+ if (self->fd >= 0)
	125	+ close(self->fd);
	126	+ if (self->fd2 >= 0)
	127	+ close(self->fd2);
	128	+}
	129	+
	130	+TEST_F(tun, delete_detach_close) {
	131	+ EXPECT_EQ(tun_delete(self->ifname), 0);
	132	+ EXPECT_EQ(tun_detach(self->fd, self->ifname), -1);
	133	+ EXPECT_EQ(errno, 22);
	134	+}
	135	+
	136	+TEST_F(tun, detach_delete_close) {
	137	+ EXPECT_EQ(tun_detach(self->fd, self->ifname), 0);
	138	+ EXPECT_EQ(tun_delete(self->ifname), 0);
	139	+}
	140	+
	141	+TEST_F(tun, detach_close_delete) {
	142	+ EXPECT_EQ(tun_detach(self->fd, self->ifname), 0);
	143	+ close(self->fd);
	144	+ self->fd = -1;
	145	+ EXPECT_EQ(tun_delete(self->ifname), 0);
	146	+}
	147	+
	148	+TEST_F(tun, reattach_delete_close) {
	149	+ EXPECT_EQ(tun_detach(self->fd, self->ifname), 0);
	150	+ EXPECT_EQ(tun_attach(self->fd, self->ifname), 0);
	151	+ EXPECT_EQ(tun_delete(self->ifname), 0);
	152	+}
	153	+
	154	+TEST_F(tun, reattach_close_delete) {
	155	+ EXPECT_EQ(tun_detach(self->fd, self->ifname), 0);
	156	+ EXPECT_EQ(tun_attach(self->fd, self->ifname), 0);
	157	+ close(self->fd);
	158	+ self->fd = -1;
	159	+ EXPECT_EQ(tun_delete(self->ifname), 0);
	160	+}
	161	+
	162	+TEST_HARNESS_MAIN

tools/testing/selftests/net/udpgso_bench.sh+1−1

@@ -120,7 +120,7 @@ run_all() { · 119 unmodified lines
120	120	run_udp "${ipv4_args}"
121	121
122	122	echo "ipv6"
123		− run_tcp "${ipv4_args}"
	123	+ run_tcp "${ipv6_args}"
124	124	run_udp "${ipv6_args}"
125	125	}
126	126

tools/testing/selftests/tc-testing/tc-tests/actions/gact.json+77−0

@@ -609,5 +609,82 @@ · 608 unmodified lines
609	609	"teardown": [
610	610	"$TC actions flush action gact"
611	611	]
	612	+ },
	613	+ {
	614	+ "id": "7f52",
	615	+ "name": "Try to flush action which is referenced by filter",
	616	+ "category": [
	617	+ "actions",
	618	+ "gact"
	619	+ ],
	620	+ "plugins": {
	621	+ "requires": "nsPlugin"
	622	+ },
	623	+ "setup": [
	624	+ [
	625	+ "$TC actions flush action gact",
	626	+ 0,
	627	+ 1,
	628	+ 255
	629	+ ],
	630	+ "$TC qdisc add dev $DEV1 ingress",
	631	+ "$TC actions add action pass index 1",
	632	+ "$TC filter add dev $DEV1 protocol all ingress prio 1 handle 0x1234 matchall action gact index 1"
	633	+ ],
	634	+ "cmdUnderTest": "$TC actions flush action gact",
	635	+ "expExitCode": "1",
	636	+ "verifyCmd": "$TC actions ls action gact",
	637	+ "matchPattern": "total acts 1.action order [0-9]: gact action pass.*index 1 ref 2 bind 1",
	638	+ "matchCount": "1",
	639	+ "teardown": [
	640	+ "$TC qdisc del dev $DEV1 ingress",
	641	+ [
	642	+ "sleep 1; $TC actions flush action gact",
	643	+ 0,
	644	+ 1
	645	+ ]
	646	+ ]
	647	+ },
	648	+ {
	649	+ "id": "ae1e",
	650	+ "name": "Try to flush actions when last one is referenced by filter",
	651	+ "category": [
	652	+ "actions",
	653	+ "gact"
	654	+ ],
	655	+ "plugins": {
	656	+ "requires": "nsPlugin"
	657	+ },
	658	+ "setup": [
	659	+ [
	660	+ "$TC actions flush action gact",
	661	+ 0,
	662	+ 1,
	663	+ 255
	664	+ ],
	665	+ "$TC qdisc add dev $DEV1 ingress",
	666	+ [
	667	+ "$TC actions add action pass index 1",
	668	+ 0,
	669	+ 1,
	670	+ 255
	671	+ ],
	672	+ "$TC actions add action reclassify index 2",
	673	+ "$TC actions add action drop index 3",
	674	+ "$TC filter add dev $DEV1 protocol all ingress prio 1 handle 0x1234 matchall action gact index 3"
	675	+ ],
	676	+ "cmdUnderTest": "$TC actions flush action gact",
	677	+ "expExitCode": "0",
	678	+ "verifyCmd": "$TC actions ls action gact",
	679	+ "matchPattern": "total acts 1.action order [0-9]: gact action drop.*index 3 ref 2 bind 1",
	680	+ "matchCount": "1",
	681	+ "teardown": [
	682	+ "$TC qdisc del dev $DEV1 ingress",
	683	+ [
	684	+ "sleep 1; $TC actions flush action gact",
	685	+ 0,
	686	+ 1
	687	+ ]
	688	+ ]
612	689	}
613	690	]