- 08 Apr, 2024 38 commits
-
-
Eric Dumazet authored
Print these additional fields in skb_dump() to ease debugging. - mac_len - csum_start (in v2, at Willem suggestion) - csum_offset (in v2, at Willem suggestion) - priority - mark - alloc_cpu - vlan_all - encapsulation - inner_protocol - inner_mac_header - inner_network_header - inner_transport_header Signed-off-by:
Eric Dumazet <edumazet@google.com> Reviewed-by:
Willem de Bruijn <willemb@google.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Andrew Lunn says: ==================== net: Clean up some EEE code Previous patches have reworked the API between phylib and MAC drivers with respect to EEE, pushing most of the work into phylib. These two patches rework two drivers to make use of the new API, and fix their EEE implementation, so that EEE is configured in the MAC based on what is actually negotiated during autoneg. Compile tested only. ==================== Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Andrew Lunn authored
The enabling/disabling of EEE in the MAC should happen as a result of auto negotiation. So move the enable/disable into lan743x_phy_link_status_change() which gets called by phylib when there is a change in link status. lan743x_ethtool_set_eee() now just programs the hardware with the LTI timer value, and passed everything else to phylib, so it can correctly setup the PHY. lan743x_ethtool_get_eee() relies on phylib doing most of the work, the MAC driver just adds the LTI timer value. Signed-off-by:
Oleksij Rempel <o.rempel@pengutronix.de> Signed-off-by:
-
Andrew Lunn authored
The enabling/disabling of EEE in the MAC should happen as a result of auto negotiation. So move the enable/disable into lan783xx_phy_link_status_change() which gets called by phylib when there is a change in link status. lan78xx_set_eee() now just programs the hardware with the LPI timer value, and passed everything else to phylib, so it can correctly setup the PHY. lan743x_get_eee() relies on phylib doing most of the work, the MAC driver just adds the LPI timer value. Call phy_support_eee() to indicate the MAC does actually support EEE. Signed-off-by:
-
Jason Xing authored
The reason codes are handled in two ways nowadays (quoting Mat Martineau): 1. Sending in the MPTCP option on RST packets when there is no subflow context available (these use subflow_add_reset_reason() and directly call a TCP-level send_reset function) 2. The "normal" way via subflow->reset_reason. This will propagate to both the outgoing reset packet and to a local path manager process via netlink in mptcp_event_sub_closed() RFC 8684 defines the skb reset reason behaviour which is not required even though in some places: A host sends a TCP RST in order to close a subflow or reject an attempt to open a subflow (MP_JOIN). In order to let the receiving host know why a subflow is being closed or rejected, the TCP RST packet MAY include the MP_TCPRST option (Figure 15). The host MAY use this information to decide, for example, whether it tries to re-establish the subflow immediately, later, or never. Since the commit dc87efdb ("mptcp: add mptcp reset option support") introduced this feature about three years ago, we can fully use it. There remains some places where we could insert reason into skb as we can see in this patch. Many thanks to Mat and Paolo for help:) Signed-off-by:Jason Xing <kernelxing@tencent.com> Acked-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
-
Guillaume Nault authored
Add a "scope" parameter to ip_route_output() so that callers don't have to override the tos parameter with the RTO_ONLINK flag if they want a local scope. This will allow converting flowi4_tos to dscp_t in the future, thus allowing static analysers to flag invalid interactions between "tos" (the DSCP bits) and ECN. Only three users ask for local scope (bonding, arp and atm). The others continue to use RT_SCOPE_UNIVERSE. While there, add a comment to warn users about the limitations of ip_route_output(). Signed-off-by:
Guillaume Nault <gnault@redhat.com> Acked-by: Leon Romanovsky <leonro@nvidia.com> # infiniband Signed-off-by:
-
Venkat Venkatsubra authored
In case of stacked devices, to help propagate the down link state from the parent/root device (to this leaf device), handle NETDEV_DOWN event like it is done now for NETDEV_UP. In the below example, ens5 is the host interface which is the parent of the ipvlan interface eth0 in the container. Host: [root@gkn-podman-x64 ~]# ip link set ens5 down [root@gkn-podman-x64 ~]# ip -d link show dev ens5 3: ens5: <BROADCAST,MULTICAST> mtu 9000 qdisc mq state DOWN ... [root@gkn-podman-x64 ~]# Container: [root@testnode-ol8 /]# ip -d link show dev eth0 2: eth0@if3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 state UNKNOWN ... ipvlan mode l2 bridge ... [root@testnode-ol8 /]# eth0's state continues to show up as UP even though ens5 is now DOWN. For macvlan the handling of NETDEV_DOWN event was added in commit 80fd2d6c ("macvlan: Change status when lower device goes down"). Reported-by:Gia-Khanh Nguyen <gia-khanh.nguyen@oracle.com> Signed-off-by:
Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com> Reviewed-by:
-
Eric Dumazet authored
Although the code is correct, the following line copy_from_sockptr(&req_u.req, optval, len)); triggers this warning : memcpy: detected field-spanning write (size 28) of single field "dst" at include/linux/sockptr.h:49 (size 16) Refactor the code to be more explicit. Reported-by:
syzbot <syzkaller@googlegroups.com> Signed-off-by:
Kees Cook <keescook@chromium.org> Reviewed-by:
-
Niklas Schnelle authored
In a future patch HAS_IOPORT=n will disable inb()/outb() and friends at compile time. We thus need to add HAS_IOPORT as dependency for those drivers requiring them. For the DEFXX driver the use of I/O ports is optional and we only need to fence specific code paths. It also turns out that with HAS_IOPORT handled explicitly HAMRADIO does not need the !S390 dependency and successfully builds the bpqether driver. Acked-by:
Marc Kleine-Budde <mkl@pengutronix.de> Acked-by:
Jakub Kicinski <kuba@kernel.org> Acked-by:
Maciej W. Rozycki <macro@orcam.me.uk> Co-developed-by:
Arnd Bergmann <arnd@kernel.org> Signed-off-by:
Niklas Schnelle <schnelle@linux.ibm.com> Signed-off-by:
-
David S. Miller authored
Matthieu Baerts says: ==================== selftests: mptcp: cleanups and 'ip mptcp' support Here are some patches from Geliang, doing different cleanups, and supporting 'ip mptcp' in more MPTCP selftests. Patch 1 checks that TC is available in selftests requiring it. Patch 2 adds 'ms' units in TC commands, to avoid confusions. Patches 3-9 are some prerequisites for patch 10: some export code from mptcp_join.sh to mptcp_lib.sh, to be re-used in pm_netlink.sh, mptcp_sockopt.sh and simult_flows.sh ; and others add helpers to pm_netlink.sh to easily support both 'ip mptcp' and 'pm_nl_ctl' tools to interact with the in-kernel MPTCP path-manager. Patch 10 adds a '-i' parameter in mptcp_sockopt.sh, pm_netlink.sh, and simult_flows.sh to use 'ip mptcp' tool instead of 'pm_nl_ctl'. Patch 11 fixes some ShellCheck warnings in pm_netlink.sh, in order to drop a ShellCheck's 'disable' instruction. ==================== Signed-off-by:
Matthieu Baerts (NGI0) <matttbe@kernel.org> Signed-off-by:
-
Geliang Tang authored
Now there are only a few of variables are not using double quotes. Modifying them, then "shellcheck disable=SC2086" can be dropped. Signed-off-by:
Geliang Tang <tanggeliang@kylinos.cn> Reviewed-by:
-
Geliang Tang authored
This patch adds '-i' option for mptcp_sockopt.sh, pm_netlink.sh, and simult_flows.sh, to use 'ip mptcp' command in the tests instead of 'pm_nl_ctl'. Update usage() correspondingly. Signed-off-by:
-
Geliang Tang authored
Use those newly added pm_nl endpoint ops helpers to replace all 'pm_nl_ctl' commands with 'limits', 'add', 'del', 'flush', 'show' and 'set' arguments in scripts mptcp_sockopt.sh and simult_flows.sh. In pm_netlink.sh, add wrappers of there helpers to make the function names shorter. Then use the wrappers to replace all 'pm_nl_ctl' commands. Signed-off-by:
-
Geliang Tang authored
This patch exports six endpoint operation helpers with pm_nl_ prefix, pm_nl_set_limits(), pm_nl_add_endpoint(), pm_nl_del_endpoint(), pm_nl_flush_endpoint(), pm_nl_show_endpoints() and pm_nl_change_endpoint() into mptcp_lib.sh as public functions, and renamed each of them with a mptcp_lib_ prefix. Then these old pm_nl_ prefix helpers in mptcp_join.sh can be wrappers of mptcp_lib_ prefix ones. Signed-off-by:
-
Geliang Tang authored
This patch uses 'case' statements to simplify pm_nl_add_endpoint() and pm_nl_check_endpoint(). And simplify pm_nl_check_endpoint() with check_output() helper. Also update pm_nl_del_endpoint() to avoid the 'double quote' shellcheck warning. Signed-off-by:
-
Geliang Tang authored
The output formats of 'ip mptcp' commands are much different from that of 'pm_nl_ctl' commands. A new 'change_address' helper is added here, to change the flag of an address. This is a bit similar to mptcp_join.sh's pm_nl_change_endpoint(). Usage: Address ID - pm_nl_change_endpoint $ns id $id $flags IP address - change_address $ns $addr $flags Use this new helper in pm_netlink.sh to replace all 'pm_nl_ctl set' commands. Signed-off-by:
-
Geliang Tang authored
The output formats of 'ip mptcp' commands are much different from that of 'pm_nl_ctl' commands. This patch adds a new helper format_endpoints() to format the outputs of 'ip mptcp' and 'pm_nl_ctl' with 'endpoints' arguments to hide these differences. A new helper named get_endpoint() has also been added to show a specific endpoint identified by the given address ID, similar to mptcp_join.sh's pm_nl_show_endpoints() helper, but showing all entries. Use these two helpers in mptcp_join.sh and pm_netlink.sh to replace all 'pm_nl_ctl get' commands and outputs of 'pm_nl_ctl dump/get'. Suggested-by:
-
Geliang Tang authored
The output format of 'ip mptcp limits' command is much different from that of 'pm_nl_ctl limits' command. This patch adds format_limits() helper to format the outputs of these two commands to hide the difference. get_limits() has been added to show the limits. Use these two helpers in pm_netlink.sh to replace all 'pm_nl_ctl limits' commands and outputs. Signed-off-by:
-
Geliang Tang authored
This patch exports ip_mptcp into mptcp_lib.sh as a public variable, named MPTCP_LIB_IP_MPTCP. Add a helper mptcp_lib_set_ip_mptcp() to set it, and a helper mptcp_lib_is_ip_mptcp() to test whether it is set. Use these two helpers in mptcp_join.sh. This patch is prepared for coming commits. Signed-off-by:
-
Geliang Tang authored
'delay 1' in tc-netem is confusing, not sure if it's a delay of 1 second or 1 millisecond. This patch explicitly adds millisecond units to make these commands clearer. Signed-off-by:
-
Geliang Tang authored
tc are used in some test scripts: mptcp_connect.sh, mptcp_join.sh and simult_flows.sh. It makes sense to check if tc is installed before running these scripts, just like other tools. So this patch add 'tc' check for mptcp_lib_check_tools(), and check it in these test scripts. Signed-off-by:
-
Eric Dumazet authored
tp->recvmsg_inq is used from tcp recvmsg() thus should be in tcp_sock_read_rx group. tp->tcp_clock_cache and tp->tcp_mstamp are written both in rx and tx paths, thus are better placed in tcp_sock_write_txrx group. Signed-off-by:
-
Jose Ignacio Tornos Martinez authored
If the mac address can not be read from the device registers or the devicetree, a random address is generated, but this was already done from usbnet_probe, so it is not necessary to call eth_hw_addr_random from here again to generate another random address. Indeed, when reset was also executed from bind, generate another random mac address invalidated the check from usbnet_probe to configure if the assigned mac address for the interface was random or not, because it is comparing with the initial generated random address. Now, with only a reset from open operation, it is just a harmless simplification. Signed-off-by:
Jose Ignacio Tornos Martinez <jtornosm@redhat.com> Signed-off-by:
-
Michal Swiatkowski authored
From Arnd comments: "The memcpy() in the ip_tunnel_info_opts_set() causes a string.h fortification warning, with at least gcc-13: In function 'fortify_memcpy_chk', inlined from 'ip_tunnel_info_opts_set' at include/net/ip_tunnels.h:619:3, inlined from 'pfcp_encap_recv' at drivers/net/pfcp.c:84:2: include/linux/fortify-string.h:553:25: error: call to '__write_overflow_field' declared with attribute warning: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror=attribute-warning] 553 | __write_overflow_field(p_size_field, size);" It is a false-positivie caused by ambiguity of the union. However, as Arnd noticed, copying here is unescessary. The code can be simplified to avoid calling ip_tunnel_info_opts_set(), which is doing copying, setting flags and options_len. Set correct flags and options_len directly on tun_info. Fixes: 6dd514f4 ("pfcp: always set pfcp metadata") Reported-by:Arnd Bergmann <arnd@arndb.de> Closes: https://lore.kernel.org/netdev/701f8f93-f5fb-408b-822a-37a1d5c424ba@app.fastmail.com/Acked-by:
Michal Swiatkowski <michal.swiatkowski@linux.intel.com> Reviewed-by:
Simon Horman <horms@kernel.org> Signed-off-by:
-
David S. Miller authored
Jakub Kicinski says: ==================== selftests: net: groundwork for YNL-based tests Currently the options for writing networking tests are C, bash or some mix of the two. YAML/Netlink gives us the ability to easily interface with Netlink in higher level laguages. In particular, there is a Python library already available in tree, under tools/net. Add the scaffolding which allows writing tests using this library. The "scaffolding" is needed because the library lives under tools/net and uses YAML files from under Documentation/. So we need a small amount of glue code to find those things and add them to TEST_FILES. This series adds both a basic SW sanity test and driver test which can be run against netdevsim or a real device. When I develop core code I usually test with netdevsim, then a real device, and then a backport to Meta's kernel. Because of the lack of integration, until now I had to throw away the (YNL-based) test script and netdevsim code. Running tests in tree directly: $ ./tools/testing/selftests/net/nl_netdev.py KTAP version 1 1..2 ok 1 nl_netdev.empty_check ok 2 nl_netdev.lo_check # Totals: pass:2 fail:0 xfail:0 xpass:0 skip:0 error:0 in tree via make: $ make -C tools/testing/selftests/ TARGETS=net \ TEST_PROGS=nl_netdev.py TEST_GEN_PROGS="" run_tests [ ... ] and installed externally, all seem to work: $ make -C tools/testing/selftests/ TARGETS=net \ install INSTALL_PATH=/tmp/ksft-net $ /tmp/ksft-net/run_kselftest.sh -t net:nl_netdev.py [ ... ] For driver tests I followed the lead of net/forwarding and get the device name from env and/or a config file. v3: - fix up netdevsim C - various small nits in other patches (see changelog in patches) v2: https://lore.kernel.org/all/20240403023426.1762996-1-kuba@kernel.org/ - don't add to TARGETS, create a deperate variable with deps - support and use with - support and use passing arguments to tests v1: https://lore.kernel.org/all/20240402010520.1209517-1-kuba@kernel.org/ ==================== Signed-off-by:
-
Jakub Kicinski authored
Add a very simple test to make sure drivers report expected stats. Drivers which implement FEC or pause configuration should report relevant stats. Qstats must be reported, at least packet and byte counts, and they must match total device stats. Tested with netdevsim, bnxt, in-tree and installed. Signed-off-by:
-
Jakub Kicinski authored
Add drivers/net as a target for mixed-use tests. The setup is expected to work similarly to the forwarding tests. Since we only need one interface (unlike forwarding tests) read the target device name from NETIF. If not present we'll try to run the test against netdevsim. Reviewed-by:
Petr Machata <petrm@nvidia.com> Signed-off-by:
-
Jakub Kicinski authored
Real devices should implement qstats. Devices which support pause or FEC configuration should also report the relevant stats. nsim was missing FEC stats completely, some of the qstats and pause stats required toggling a debugfs knob. Note that the tests which used pause always initialize the setting so they shouldn't be affected by the different starting value. Signed-off-by:
-
Jakub Kicinski authored
Add a trivial test using YNL. $ ./tools/testing/selftests/net/nl_netdev.py KTAP version 1 1..2 ok 1 nl_netdev.empty_check ok 2 nl_netdev.lo_check Instantiate the family once, it takes longer than the test itself. Reviewed-by:
-
Jakub Kicinski authored
Add glue code for accessing the YNL library which lives under tools/net and YAML spec files from under Documentation/. Automatically figure out if tests are run in tree or not. Since we'll want to use this library both from net and drivers/net test targets make the library a target as well, and automatically include it when net or drivers/net are included. Making net/lib a target ensures that we end up with only one copy of it, and saves us some path guessing. Add a tiny bit of formatting support to be able to output KTAP from the start. Reviewed-by:
-
git://git.open-mesh.org/linux-mergeDavid S. Miller authored
Simon Wunderlich says: ==================== This cleanup patchset includes the following patches: - bump version strings, by Simon Wunderlich - prefer kfree_rcu() over call_rcu() with free-only callbacks, by Dmitry Antipov - bypass empty buckets in batadv_purge_orig_ref(), by Eric Dumazet ==================== Signed-off-by: -
Andy Shevchenko authored
Replace open coded variant of device_is_compatible(). Signed-off-by:
Andy Shevchenko <andriy.shevchenko@linux.intel.com> Reviewed-by:
-
Eric Dumazet authored
Commit 6025b913 ("net: dqs: add NIC stall detector based on BQL") added three sysfs files. Use the recommended sysfs_emit() helper. Signed-off-by:
-
Alexander Lobakin authored
Structures which are about to be copied to userspace shouldn't have uninitialized fields or paddings. memset() the whole &ip_tunnel_parm in ip_tunnel_parm_to_user() before filling it with the kernel data. The compilers will hopefully combine writes to it. Fixes: 117aef12 ("ip_tunnel: use a separate struct to store tunnel params in the kernel") Reported-by:
Dan Carpenter <dan.carpenter@linaro.org> Closes: https://lore.kernel.org/netdev/5f63dd25-de94-4ca3-84e6-14095953db13@moroto.mountainSigned-off-by:
Alexander Lobakin <aleksander.lobakin@intel.com> Reviewed-by:
-
Eric Dumazet authored
No longer hold RTNL while calling ip6addrlbl_dump() ("ip addrlabel show") ip6addrlbl_dump() was already mostly relying on RCU anyway. Add READ_ONCE()/WRITE_ONCE() annotations around net->ipv6.ip6addrlbl_table.seq Note that ifal_seq value is currently ignored in iproute2, and a bit weak. We might user later cb->seq / nl_dump_check_consistent() protocol if needed. Also change return value for a completed dump, so that NLMSG_DONE can be appended to current skb, saving one recvmsg() system call. v2: read net->ipv6.ip6addrlbl_table.seq once, (David Ahern) Signed-off-by:David Ahern <dsahern@kernel.org> Signed-off-by:
-
Eric Dumazet authored
fqdir_free_fn() is using very expensive rcu_barrier() When one netns is dismantled, we often call fqdir_exit() multiple times, typically lauching fqdir_free_fn() twice. Delaying by one second fqdir_free_fn() helps to reduce the number of rcu_barrier() calls, and lock contention on rcu_state.barrier_mutex. Signed-off-by:
-
Breno Leitao authored
Commit 3e2f544d ("net: get stats64 if device if driver is configured") moved the callback to dev_get_tstats64() to net core, so, unless the driver is doing some custom stats collection, it does not need to set .ndo_get_stats64. Since this driver is now relying in NETDEV_PCPU_STAT_TSTATS, then, it doesn't need to set the dev_get_tstats64() generic .ndo_get_stats64 function pointer. Signed-off-by:
Breno Leitao <leitao@debian.org> Reviewed-by:
-
Breno Leitao authored
With commit 34d21de9 ("net: Move {l,t,d}stats allocation to core and convert veth & vrf"), stats allocation could be done on net core instead of in this driver. With this new approach, the driver doesn't have to bother with error handling (allocation failure checking, making sure free happens in the right spot, etc). This is core responsibility now. Remove the allocation in the ip6_vti and leverage the network core allocation instead. Signed-off-by:
-
- 06 Apr, 2024 2 commits
-
-
David S. Miller authored
Maxime Chevallier says: ==================== Introduce PHY listing and link_topology tracking This is V11 for the link topology addition, allowing to track all PHYs that are linked to netdevices. This V11 addresses the various netlink-related issues that were raised by Jakub, and fixes some typos in the documentation. As a remainder, here's what the PHY listings would look like : - eth0 has a 88x3310 acting as media converter, and an SFP module with an embedded 88e1111 PHY - eth2 has a 88e1510 PHY PHY for eth0: PHY index: 1 Driver name: mv88x3310 PHY device name: f212a600.mdio-mii:00 Downstream SFP bus name: sfp-eth0 PHY id: 0 Upstream type: MAC PHY for eth0: PHY index: 2 Driver name: Marvell 88E1111 PHY device name: i2c:sfp-eth0:16 PHY id: 21040322 Upstream type: PHY Upstream PHY index: 1 Upstream SFP name: sfp-eth0 PHY for eth2: PHY index: 1 Driver name: Marvell 88E1510 PHY device name: f212a200.mdio-mii:00 PHY id: 21040593 Upstream type: MAC Ethtool patches : https://github.com/minimaxwell/ethtool/tree/link-topo-v6 ==================== Signed-off-by:
-
Maxime Chevallier authored
Some netlink commands are target towards ethernet PHYs, to control some of their features. As there's several such commands, add the ability to pass a PHY index in the ethnl request, which will populate the generic ethnl_req_info with the relevant phydev when the command targets a PHY. Signed-off-by:
Maxime Chevallier <maxime.chevallier@bootlin.com> Reviewed-by:
-
