- 16 Aug, 2023 40 commits
-
-
Abel Wu authored
Now there are two indicators of socket memory pressure sit inside struct mem_cgroup, socket_pressure and tcpmem_pressure, indicating memory reclaim pressure in memcg->memory and ->tcpmem respectively. When in legacy mode (cgroupv1), the socket memory is charged into ->tcpmem which is independent of ->memory, so socket_pressure has nothing to do with socket's pressure at all. Things could be worse by taking socket_pressure into consideration in legacy mode, as a pressure in ->memory can lead to premature reclamation/throttling in socket. While for the default mode (cgroupv2), the socket memory is charged into ->memory, and ->tcpmem/->tcpmem_pressure are simply not used. So {socket,tcpmem}_pressure are only used in default/legacy mode respectively for indicating socket memory pressure. This patch fixes the pieces of code that make mixed use of both. Fixes: 8e8ae645 ("mm: memcontrol: hook up vmpressure to socket pressure") Signed-off-by:Abel Wu <wuyun.abel@bytedance.com> Acked-by:
Shakeel Butt <shakeelb@google.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Louis Peens authored
Take over maintainership of the nfp driver from Simon as he is moving away from Corigine. Signed-off-by:
Louis Peens <louis.peens@corigine.com> Acked-by:
Simon Horman <simon.horman@corigine.com> Signed-off-by:
-
Grygorii Strashko authored
This patch adds MQPRIO Qdisc offload in full 'channel' mode which allows not only setting up pri:tc mapping, but also configuring TX shapers on external port FIFOs. The K3 CPSW MQPRIO Qdisc offload is expected to work with VLAN/priority tagged packets. Non-tagged packets have to be mapped only to TC0. - TX traffic classes must be rated starting from TC that has highest priority and with no gaps - Traffic classes are used starting from 0, that has highest priority - min_rate defines Committed Information Rate (guaranteed) - max_rate defines Excess Information Rate (non guaranteed) and offloaded as (max_rate[i] - tcX_min_rate[i]) - VLAN/priority tagged packets mapped to TC0 will exit switch with VLAN tag priority 0 The configuration example: ethtool -L eth1 tx 5 ethtool --set-priv-flags eth1 p0-rx-ptype-rrobin off tc qdisc add dev eth1 parent root handle 100: mqprio num_tc 3 \ map 0 0 1 2 0 0 0 0 0 0 0 0 0 0 0 0 \ queues 1@0 1@1 1@2 hw 1 mode channel \ shaper bw_rlimit min_rate 0 100mbit 200mbit max_rate 0 101mbit 202mbit tc qdisc replace dev eth2 handle 100: parent root mqprio num_tc 1 \ map 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 queues 1@0 hw 1 ip link add link eth1 name eth1.100 type vlan id 100 ip link set eth1.100 type vlan egress 0:0 1:1 2:2 3:3 4:4 5:5 6:6 7:7 In the above example two ports share the same TX CPPI queue 0 for low priority traffic. 3 traffic classes are defined for eth1 and mapped to: TC0 - low priority, TX CPPI queue 0 -> ext Port 1 fifo0, no rate limit TC1 - prio 2, TX CPPI queue 1 -> ext Port 1 fifo1, CIR=100Mbit/s, EIR=1Mbit/s TC2 - prio 3, TX CPPI queue 2 -> ext Port 1 fifo2, CIR=200Mbit/s, EIR=2Mbit/s Signed-off-by:
Grygorii Strashko <grygorii.strashko@ti.com> Signed-off-by:
Roger Quadros <rogerq@kernel.org> Reviewed-by:
Simon Horman <horms@kernel.org> Signed-off-by:
-
David S. Miller authored
Eric Dumazet says: ==================== inet: socket lock and data-races avoidance In this series, I converted 20 bits in "struct inet_sock" and made them truly atomic. This allows to implement many IP_ socket options in a lockless fashion (no need to acquire socket lock), and fixes data-races that were showing up in various KCSAN reports. I also took care of IP_TTL/IP_MINTTL, but left few other options for another series. v4: Rebased after recent mptcp changes. Added Reviewed-by: tags from Simon (thanks !) v3: fixed patch 7, feedback from build bot about ipvs set_mcast_loop() v2: addressed a feedback from a build bot in patch 9 by removing unused issk variable in mptcp_setsockopt_sol_ip_set_transparent() Added Acked-by: tags from Soheil (thanks !) ==================== Signed-off-by: -
Eric Dumazet authored
inet->min_ttl is already read with READ_ONCE(). Implementing IP_MINTTL socket option set/read without holding the socket lock is easy. Signed-off-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Soheil Hassas Yeganeh <soheil@google.com> Reviewed-by:
-
Eric Dumazet authored
ip_select_ttl() is racy, because it reads inet->uc_ttl without proper locking. Add READ_ONCE()/WRITE_ONCE() annotations while allowing IP_TTL socket option to be set/read without holding the socket lock. Signed-off-by:
-
Eric Dumazet authored
Make room in struct inet_sock by removing this bit field, using one available bit in inet_flags instead. Also move local_port_range to fill the resulting hole, saving 8 bytes on 64bit arches. Signed-off-by:
Matthieu Baerts <matthieu.baerts@tessares.net> Signed-off-by:
-
Eric Dumazet authored
IP_BIND_ADDRESS_NO_PORT socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
IP_NODEFRAG socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
We move single bit fields to inet->inet_flags to avoid races. Signed-off-by:
-
Eric Dumazet authored
IP_TRANSPARENT socket option can now be set/read without locking the socket. v2: removed unused issk variable in mptcp_setsockopt_sol_ip_set_transparent() v4: rebased after commit 3f326a82 ("mptcp: change the mpc check helper to return a sk") Signed-off-by:
-
Eric Dumazet authored
IP_MULTICAST_ALL socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
IP_MULTICAST_LOOP socket option can now be set/read without locking the socket. v3: fix build bot error reported in ipvs set_mcast_loop() Signed-off-by:
-
Eric Dumazet authored
IP_HDRINCL socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
IP_FREEBIND socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
IP_RECVERR_RFC4884 socket option can now be set/read without locking the socket. Signed-off-by:
-
Eric Dumazet authored
IP_RECVERR socket option can now be set/get without locking the socket. This patch potentially avoid data-races around inet->recverr. Signed-off-by:
-
Eric Dumazet authored
Now we have inet->inet_flags, we can set following options without having to hold the socket lock: IP_PKTINFO, IP_RECVTTL, IP_RECVTOS, IP_RECVOPTS, IP_RETOPTS, IP_PASSSEC, IP_RECVORIGDSTADDR, IP_RECVFRAGSIZE. ip_sock_set_pktinfo() no longer hold the socket lock. Similarly we can get the following options whithout holding the socket lock: IP_PKTINFO, IP_RECVTTL, IP_RECVTOS, IP_RECVOPTS, IP_RETOPTS, IP_PASSSEC, IP_RECVORIGDSTADDR, IP_CHECKSUM, IP_RECVFRAGSIZE. Signed-off-by:
-
Eric Dumazet authored
Various inet fields are currently racy. do_ip_setsockopt() and do_ip_getsockopt() are mostly holding the socket lock, but some (fast) paths do not. Use a new inet->inet_flags to hold atomic bits in the series. Remove inet->cmsg_flags, and use instead 9 bits from inet_flags. Signed-off-by:
-
David S. Miller authored
Ruan Jinjie says: ==================== net: Remove redundant of_match_ptr() macro Since these net drivers depend on CONFIG_OF, there is no need to wrap the macro of_match_ptr() here. Changes in v3: - Collect responses from v1 and v2. ==================== Signed-off-by: -
Ruan Jinjie authored
The driver depends on CONFIG_OF, it is not necessary to use of_match_ptr() here. Signed-off-by:
Ruan Jinjie <ruanjinjie@huawei.com> Reviewed-by:
-
Ruan Jinjie authored
The driver depends on CONFIG_OF, it is not necessary to use of_match_ptr() here. Signed-off-by:
-
Ruan Jinjie authored
The driver depends on CONFIG_OF, it is not necessary to use of_match_ptr() here. Signed-off-by:
Linus Walleij <linus.walleij@linaro.org> Signed-off-by:
-
Ruan Jinjie authored
The driver depends on CONFIG_OF, it is not necessary to use of_match_ptr() here. Signed-off-by:
Vladimir Oltean <olteanv@gmail.com> Signed-off-by:
-
Ruan Jinjie authored
The driver depends on CONFIG_OF, it is not necessary to use of_match_ptr() here. Signed-off-by:
-
Li Zetao authored
Use the module_misc_device macro to simplify the code, which is the same as declaring with module_init() and module_exit(). Signed-off-by:
Li Zetao <lizetao1@huawei.com> Reviewed-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by:
-
David S. Miller authored
Jijie Shao says: ==================== hns3: refactor registers information for ethtool -d refactor registers information for ethtool -d ==================== Signed-off-by: -
Jijie Shao authored
In the original RPU query command, the status register values of multiple RPU tunnels are accumulated by default, which is unreasonable. This patch Fix it by querying the specified tunnel ID. The tunnel number of the device can be obtained from firmware during initialization. Fixes: ddb54554 ("net: hns3: add DFX registers information for ethtool -d") Signed-off-by:
Jijie Shao <shaojijie@huawei.com> Reviewed-by:
Leon Romanovsky <leonro@nvidia.com> Signed-off-by:
-
Jijie Shao authored
The dump register function is being refactored. The third step in refactoring is to support tlv info in regs data for HNS3 PF driver. Currently, if we use "ethtool -d" to dump regs value, the output is as follows: offset1: 00 01 02 03 04 05 ... offset2：10 11 12 13 14 15 ... ...... We can't get the value of a register directly. This patch deletes the original separator information and add tag_len_value information in regs data. ethtool can parse register data in key-value format by -d command. a patch will be added to the ethtool to parse regs data in the following format: reg1 : value2 reg2 : value2 ...... Signed-off-by:
-
Jijie Shao authored
The dump register function is being refactored. The second step in refactoring is to support tlv info in regs data for HNS3 PF driver. Currently, if we use "ethtool -d" to dump regs value, the output is as follows: offset1: 00 01 02 03 04 05 ... offset2：10 11 12 13 14 15 ... ...... We can't get the value of a register directly. This patch deletes the original separator information and add tag_len_value information in regs data. ethtool can parse register data in key-value format by -d command. a patch will be added to the ethtool to parse regs data in the following format: reg1 : value2 reg2 : value2 ...... Signed-off-by:
-
Jijie Shao authored
The dump register function is being refactored. The first step in refactoring is put the dump regs function into a separate file. Signed-off-by:
-
David S. Miller authored
Wei Fang says: ==================== net: fec: add XDP_TX feature support This patch set is to support the XDP_TX feature of FEC driver, the first patch is add initial XDP_TX support, and the second patch improves the performance of XDP_TX by not using xdp_convert_buff_to_frame(). Please refer to the commit message of each patch for more details. ==================== Acked-by:
Jesper Dangaard Brouer <hawk@kernel.org> Signed-off-by:
-
Wei Fang authored
As suggested by Jesper and Alexander, we can avoid converting xdp_buff to xdp_frame in case of XDP_TX to save a bunch of CPU cycles, so that we can further improve the XDP_TX performance. Before this patch on i.MX8MP-EVK board, the performance shows as follows. root@imx8mpevk:~# ./xdp2 eth0 proto 17: 353918 pkt/s proto 17: 352923 pkt/s proto 17: 353900 pkt/s proto 17: 352672 pkt/s proto 17: 353912 pkt/s proto 17: 354219 pkt/s After applying this patch, the performance is improved. root@imx8mpevk:~# ./xdp2 eth0 proto 17: 369261 pkt/s proto 17: 369267 pkt/s proto 17: 369206 pkt/s proto 17: 369214 pkt/s proto 17: 369126 pkt/s proto 17: 369272 pkt/s Signed-off-by:
Wei Fang <wei.fang@nxp.com> Suggested-by:
Alexander Lobakin <aleksander.lobakin@intel.com> Suggested-by:
-
Wei Fang authored
The XDP_TX feature is not supported before, and all the frames which are deemed to do XDP_TX action actually do the XDP_DROP action. So this patch adds the XDP_TX support to FEC driver. I tested the performance of XDP_TX in XDP_DRV mode and XDP_SKB mode respectively on i.MX8MP-EVK platform, and as suggested by Jesper, I also tested the performance of XDP_REDIRECT on the same platform. And the test steps and results are as follows. XDP_TX test: Step 1: One board is used as generator and connects to switch,and the FEC port of DUT also connects to the switch. Both boards with flow control off. Then the generator runs the pktgen_sample03_burst_single_flow.sh script to generate and send burst traffic to DUT. Note that the size of packet was set to 64 bytes and the procotol of packet was UDP in my test scenario. In addition, the SMAC of the packet need to be different from the MAC of the generator, because the xdp2 program will swap the DMAC and SMAC of the packet and send it back to the generator. If the SMAC of the generated packet is the MAC of the generator, the generator will receive the returned traffic which increase the CPU loading and significantly degrade the transmit speed of the generator, and finally it affects the test of XDP_TX performance. Step 2: The DUT runs the xdp2 program to transmit received UDP packets back out on the same port where they were received. root@imx8mpevk:~# ./xdp2 eth0 proto 17: 353918 pkt/s proto 17: 352923 pkt/s proto 17: 353900 pkt/s proto 17: 352672 pkt/s proto 17: 353912 pkt/s proto 17: 354219 pkt/s root@imx8mpevk:~# ./xdp2 -S eth0 proto 17: 160604 pkt/s proto 17: 160708 pkt/s proto 17: 160564 pkt/s proto 17: 160684 pkt/s proto 17: 160640 pkt/s proto 17: 160720 pkt/s The above results show that the XDP_TX performance of XDP_DRV mode is much better than XDP_SKB mode, more than twice that of XDP_SKB mode, which is in line with our expectation. XDP_REDIRECT test: Step1: Both the generator and the FEC port of the DUT connet to the switch port. All the ports with flow control off, then the generator runs the pktgen script to generate and send burst traffic to DUT. Note that the size of packet was set to 64 bytes and the procotol of packet was UDP in my test scenario. Step2: The DUT runs the xdp_redirect program to redirect the traffic from the FEC port to the FEC port itself. root@imx8mpevk:~# ./xdp_redirect eth0 eth0 Redirecting from eth0 (ifindex 2; driver fec) to eth0 (ifindex 2; driver fec) Summary 232,302 rx/s 0 err,drop/s 232,344 xmit/s Summary 234,579 rx/s 0 err,drop/s 234,577 xmit/s Summary 235,548 rx/s 0 err,drop/s 235,549 xmit/s Summary 234,704 rx/s 0 err,drop/s 234,703 xmit/s Summary 235,504 rx/s 0 err,drop/s 235,504 xmit/s Summary 235,223 rx/s 0 err,drop/s 235,224 xmit/s Summary 234,509 rx/s 0 err,drop/s 234,507 xmit/s Summary 235,481 rx/s 0 err,drop/s 235,482 xmit/s Summary 234,684 rx/s 0 err,drop/s 234,683 xmit/s Summary 235,520 rx/s 0 err,drop/s 235,520 xmit/s Summary 235,461 rx/s 0 err,drop/s 235,461 xmit/s Summary 234,627 rx/s 0 err,drop/s 234,627 xmit/s Summary 235,611 rx/s 0 err,drop/s 235,611 xmit/s Packets received : 3,053,753 Average packets/s : 234,904 Packets transmitted : 3,053,792 Average transmit/s : 234,907 Compared the performance of XDP_TX with XDP_REDIRECT, XDP_TX is also much better than XDP_REDIRECT. It's also in line with our expectation. Signed-off-by:
Jakub Kicinski <kuba@kernel.org> Reviewed-by:
Larysa Zaremba <larysa.zaremba@intel.com> Acked-by:
-
Zhengchao Shao authored
When run command "ip netns delete client", device link1_1 has been deleted. So, it is no need to delete link1_1 again. Remove it. Signed-off-by:
Zhengchao Shao <shaozhengchao@huawei.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxJakub Kicinski authored
Saeed Mahameed says: ==================== mlx5-updates-2023-08-14 1) Handle PTP out of order CQEs issue 2) Check FW status before determining reset successful 3) Expose maximum supported SFs via devlink resource 4) MISC cleanups * tag 'mlx5-updates-2023-08-14' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux: net/mlx5: Don't query MAX caps twice net/mlx5: Remove unused MAX HCA capabilities net/mlx5: Remove unused CAPs net/mlx5: Fix error message in mlx5_sf_dev_state_change_handler() net/mlx5: Remove redundant check of mlx5_vhca_event_supported() net/mlx5: Use mlx5_sf_start_function_id() helper instead of directly calling MLX5_CAP_GEN() net/mlx5: Remove redundant SF supported check from mlx5_sf_hw_table_init() net/mlx5: Use auxiliary_device_uninit() instead of device_put() net/mlx5: E-switch, Add checking for flow rule destinations net/mlx5: Check with FW that sync reset completed successfully net/mlx5: Expose max possible SFs via devlink resource net/mlx5e: Add recovery flow for tx devlink health reporter for unhealthy PTP SQ net/mlx5e: Make tx_port_ts logic resilient to out-of-order CQEs net/mlx5: Consolidate devlink documentation in devlink/mlx5.rst ==================== Link: https://lore.kernel.org/r/20230814214144.159464-1-saeed@kernel.orgSigned-off-by:
-
Jakub Kicinski authored
Jakub Kicinski says: ==================== net: warn about attempts to register negative ifindex Follow up to the recently posted fix for OvS lacking input validation: https://lore.kernel.org/all/20230814203840.2908710-1-kuba@kernel.org/ Warn about negative ifindex more explicitly and misc YNL updates. ==================== Link: https://lore.kernel.org/r/20230814205627.2914583-1-kuba@kernel.orgSigned-off-by:
-
Jakub Kicinski authored
When developing specs its useful to know which attr space YNL was trying to find an attribute in on key error. Instead of printing: KeyError: 0 add info about the space: Exception: Space 'vport' has no attribute with value '0' Reviewed-by:
Donald Hunter <donald.hunter@gmail.com> Link: https://lore.kernel.org/r/20230814205627.2914583-4-kuba@kernel.orgSigned-off-by:
-
Jakub Kicinski authored
Add NEW to the spec, it was useful testing the fix for OvS input validation. Reviewed-by:
-
Jakub Kicinski authored
Since the xarray changes we mix returning valid ifindex and negative errno in a single int returned from dev_index_reserve(). This depends on the fact that ifindexes can't be negative. Otherwise we may insert into the xarray and return a very large negative value. This in turn may break ERR_PTR(). OvS is susceptible to this problem and lacking validation (fix posted separately for net). Reject negative ifindex explicitly. Add a warning because the input validation is better handled by the caller. Reviewed-by:
-
