- 26 Dec, 2023 11 commits
-
-
David S. Miller authored
Wen Gu says: ==================== net/smc: implement SMCv2.1 virtual ISM device support The fourth edition of SMCv2 adds the SMC version 2.1 feature updates for SMC-Dv2 with virtual ISM. Virtual ISM are created and supported mainly by OS or hypervisor software, comparable to IBM ISM which is based on platform firmware or hardware. With the introduction of virtual ISM, SMCv2.1 makes some updates: - Introduce feature bitmask to indicate supplemental features. - Reserve a range of CHIDs for virtual ISM. - Support extended GIDs (128 bits) in CLC handshake. So this patch set aims to implement these updates in Linux kernel. And it acts as the first part of SMC-D virtual ISM extension & loopback-ism [1]. [1] https://lore.kernel.org/netdev/1695568613-125057-1-git-send-email-guwen@linux.alibaba.com/ v8->v7: - Patch #7: v7 mistakenly changed the type of gid_ext in smc_clc_msg_accept_confirm to u64 instead of __be64 as previous versions when fixing the rebase conflicts. So fix this mistake. v7->v6: Link: https://lore.kernel.org/netdev/20231219084536.8158-1-guwen@linux.alibaba.com/ - Collect the Reviewed-by tag in v6; - Patch #3: redefine the struct smc_clc_msg_accept_confirm; - Patch #7: Because that the Patch #3 already adds '__packed' to smc_clc_msg_accept_confirm, so Patch #7 doesn't need to do the same thing. But this is a minor change, so I kept the 'Reviewed-by' tag. Other changes in previous versions but not yet acked: - Patch #1: Some minor changes in subject and fix the format issue (length exceeds 80 columns) compared to v3. - Patch #5: removes useless ini->feature_mask assignment in __smc_connect() and smc_listen_v2_check() compared to v4. - Patch #8: new added, compared to v3. v6->v5: Link: https://lore.kernel.org/netdev/1702371151-125258-1-git-send-email-guwen@linux.alibaba.com/ - Add 'Reviewed-by' label given in the previous versions: * Patch #4, #6, #9, #10 have nothing changed since v3; - Patch #2: * fix the format issue (Alignment should match open parenthesis) compared to v5; * remove useless clc->hdr.length assignment in smcr_clc_prep_confirm_accept() compared to v5; - Patch #3: new added compared to v5. - Patch #7: some minor changes like aclc_v2->aclc or clc_v2->clc compared to v5 due to the introduction of Patch #3. Since there were no major changes, I kept the 'Reviewed-by' label. Other changes in previous versions but not yet acked: - Patch #1: Some minor changes in subject and fix the format issue (length exceeds 80 columns) compared to v3. - Patch #5: removes useless ini->feature_mask assignment in __smc_connect() and smc_listen_v2_check() compared to v4. - Patch #8: new added, compared to v3. v5->v4: Link: https://lore.kernel.org/netdev/1702021259-41504-1-git-send-email-guwen@linux.alibaba.com/ - Patch #6: improve the comment of SMCD_CLC_MAX_V2_GID_ENTRIES; - Patch #4: remove useless ini->feature_mask assignment; v4->v3: https://lore.kernel.org/netdev/1701920994-73705-1-git-send-email-guwen@linux.alibaba.com/ - Patch #6: use SMCD_CLC_MAX_V2_GID_ENTRIES to indicate the max gid entries in CLC proposal and using SMC_MAX_V2_ISM_DEVS to indicate the max devices to propose; - Patch #6: use i and i+1 in smc_find_ism_v2_device_serv(); - Patch #2: replace the large if-else block in smc_clc_send_confirm_accept() with 2 subfunctions; - Fix missing byte order conversion of GID and token in CLC handshake, which is in a separate patch sending to net: https://lore.kernel.org/netdev/1701882157-87956-1-git-send-email-guwen@linux.alibaba.com/ - Patch #7: add extended GID in SMC-D lgr netlink attribute; v3->v2: https://lore.kernel.org/netdev/1701343695-122657-1-git-send-email-guwen@linux.alibaba.com/ - Rename smc_clc_fill_fce as smc_clc_fill_fce_v2x; - Remove ISM_IDENT_MASK from drivers/s390/net/ism.h; - Add explicitly assigning 'false' to ism_v2_capable in ism_dev_init(); - Remove smc_ism_set_v2_capable() helper for now, and introduce it in later loopback-ism implementation; v2->v1: - Fix sparse complaint; - Rebase to the latest net-next; ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Wen Gu authored
The System EID (SEID) is an internal EID that is used by the SMCv2 software stack that has a predefined and constant value representing the s390 physical machine that the OS is executing on. So it should be managed by SMC stack instead of ISM driver and be consistent for all ISMv2 device (including virtual ISM devices) on s390 architecture. Suggested-by:
Alexandra Winter <wintera@linux.ibm.com> Signed-off-by:
Wen Gu <guwen@linux.alibaba.com> Reviewed-and-tested-by:
Wenjia Zhang <wenjia@linux.ibm.com> Reviewed-by:
-
Wen Gu authored
The system EID (SEID) is an internal EID used by SMC-D to represent the s390 physical machine that OS is executing on. On s390 architecture, it predefined by fixed string and part of cpuid and is enabled regardless of whether underlay device is virtual ISM or platform firmware ISM. However on non-s390 architectures where SMC-D can be used with virtual ISM devices, there is no similar information to identify physical machines, especially in virtualization scenarios. So in such cases, SEID is forcibly disabled and the user-defined UEID will be used to represent the communicable space. Signed-off-by:
-
Wen Gu authored
Virtual ISM devices introduced in SMCv2.1 requires a 128 bit extended GID vs. the existing ISM 64bit GID. So the 2nd 64 bit of extended GID should be included in SMC-D linkgroup netlink attribute as well. Signed-off-by:
-
Wen Gu authored
According to virtual ISM support feature defined by SMCv2.1, GIDs of virtual ISM device are UUIDs defined by RFC4122, which are 128-bits long. So some adaptation work is required. And note that the GIDs of existing platform firmware ISM devices still remain 64-bits long. Signed-off-by:
-
Wen Gu authored
According to virtual ISM support feature defined by SMCv2.1, CHIDs in the range 0xFF00 to 0xFFFF are reserved for use by virtual ISM devices. And two helpers are introduced to distinguish virtual ISM devices from the existing platform firmware ISM devices. Signed-off-by:
-
Wen Gu authored
This introduces virtual ISM device support feature to SMCv2.1 as the first supplemental feature. Signed-off-by:
-
Wen Gu authored
This patch adds SMCv2.x supplemental features negotiation. Supported SMCv2.x supplemental features are represented by feature_mask in FCE field. The negotiation process is as follows. Server Client Proposal(features(c-mask bits)) <----------------------------------------- Accept(features(s-mask bits)) -----------------------------------------> Confirm(features(s&c-mask bits)) <----------------------------------------- Signed-off-by: -
Wen Gu authored
The structs of CLC accept and confirm messages for SMCv1 and SMCv2 are separately defined and often casted to each other in the code, which may increase the risk of errors caused by future divergence of them. So unify them into one struct for better maintainability. Suggested-by:
-
Wen Gu authored
There is a large if-else block in smc_clc_send_confirm_accept() and it is better to split it into two sub-functions. Suggested-by:
-
Wen Gu authored
Rename some functions or variables with 'fce' in their name but used in SMCv2.1 as 'fce_v2x' for clarity. Signed-off-by:
-
- 25 Dec, 2023 1 commit
-
-
Marek Behún authored
Commit 2f3ce7a5 ("net: sfp: rework the RollBall PHY waiting code") changed the long wait before accessing RollBall / FS modules into probing for PHY every 1 second, and trying 25 times. Wei Lei reports that this does not work correctly on FS modules: when initializing, they may report values different from 0xffff in PHY ID registers for some MMDs, causing get_phy_c45_ids() to find some bogus MMD. Fix this by adding the module_t_wait member back, and setting it to 4 seconds for FS modules. Fixes: 2f3ce7a5 ("net: sfp: rework the RollBall PHY waiting code") Reported-by:
Wei Lei <quic_leiwei@quicinc.com> Signed-off-by:
Marek Behún <kabel@kernel.org> Tested-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
- 23 Dec, 2023 28 commits
-
-
David S. Miller authored
Ioana Ciornei says: ==================== dpaa2-switch: small improvements This patch set consists of a series of small improvements on the dpaa2-switch driver ranging from adding some more verbosity when encountering errors to reorganizing code to be easily extensible. Changes in v3: - 4/8: removed the fixes tag and moved it to the commit message - 5/8: specified that there is no user-visible effect - 6/8: removed the initialization of the err variable Changes in v2: - No changes to the actual diff, only rephrased some commit messages and added more information. ==================== Signed-off-by: -
Ioana Ciornei authored
In case a DPAA2 switch interface joins a bridge, the FDB used on the port will be changed to the one associated with the bridge. What this means exactly is that any VLAN installed on the port will need to be removed and then installed back so that it points to the new FDB. Once this is done, the previous FDB will become unused (no VLAN to point to it). Even though no traffic will reach this FDB, it's best to just cleanup the state of the FDB by zeroing its egress flood domain. Reviewed-by:
Simon Horman <horms@kernel.org> Signed-off-by:
Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by:
-
Ioana Ciornei authored
Two different DPAA2 switch ports from two different DPSW instances cannot be under the same bridge. Instead of checking for this unsupported configuration in the CHANGEUPPER event, check it as early as possible in the PRECHANGEUPPER one. Reviewed-by:
-
Ioana Ciornei authored
Create separate functions, dpaa2_switch_port_prechangeupper and dpaa2_switch_port_changeupper, to be called directly when a DPSW port changes its upper device. This way we are not open-coding everything in the main event callback and we can easily extent, for example, with bond offload. Signed-off-by:
-
Ioana Ciornei authored
The DPSW object has multiple event sources multiplexed over the same IRQ. The driver has the capability to configure only some of these events to trigger the IRQ. The dpsw_get_irq_status() can clear events automatically based on the value stored in the 'status' variable passed to it. We don't want that to happen because we could get into a situation when we are clearing more events than we actually handled. Just resort to manually clearing the events that we handled. Also, since status is not used on the out path we remove its initialization to zero. This change does not have a user-visible effect because the dpaa2-switch driver enables and handles all the DPSW events which exist at the moment. Signed-off-by:
-
Ioana Ciornei authored
Commit 84cba729 ("dpaa2-switch: integrate the MAC endpoint support") added support for MAC endpoints in the dpaa2-switch driver but omitted to add the ENDPOINT_CHANGED irq to the list of interrupt sources. Fix this by extending the list of events which can raise an interrupt by extending the mask passed to the dpsw_set_irq_mask() firmware API. There is no user visible impact even without this patch since whenever a switch interface is connected/disconnected from an endpoint both events are set (LINK_CHANGED and ENDPOINT_CHANGED) and, luckily, the LINK_CHANGED event could actually raise the interrupt and thus get the MAC/PHY SW configuration started. Even with this, it's better to just not rely on undocumented firmware behavior which can change. Signed-off-by:
-
Ioana Ciornei authored
Print a netdev error when we hit a case in which a specific VLAN is already configured on the port. While at it, change the already existing netdev_warn into an _err for consistency purposes. Reviewed-by:
-
Ioana Ciornei authored
There is no restriction around the change of the MAC address on the switch ports, thus declare the interface netdevs IFF_LIVE_ADDR_CHANGE capable. Reviewed-by:
-
Ioana Ciornei authored
There is no point in updating the MAC address of a switch interface each time the link state changes, this only needs to happen in case the endpoint changes (the switch interface is [dis]connected from/to a MAC). Just move the call to dpaa2_switch_port_set_mac_addr() under DPSW_IRQ_EVENT_ENDPOINT_CHANGED. Reviewed-by:
-
David S. Miller authored
Roger Quadros says: ==================== net: ethernet: am65-cpsw: Add mqprio, frame preemption & coalescing This series adds mqprio qdisc offload in channel mode, Frame Preemption MAC merge support and RX/TX coalesing for AM65 CPSW driver. In v11 following changes were made - Fix patch "net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode" by including units.h Changelog information in each patch file. ==================== Signed-off-by: -
Grygorii Strashko authored
Add SW IRQ coalescing based on hrtimers for TX and RX data path which can be enabled by ethtool commands: - RX coalescing ethtool -C eth1 rx-usecs 50 - TX coalescing can be enabled per TX queue - by default enables coalesing for TX0 ethtool -C eth1 tx-usecs 50 - configure TX0 ethtool -Q eth0 queue_mask 1 --coalesce tx-usecs 100 - configure TX1 ethtool -Q eth0 queue_mask 2 --coalesce tx-usecs 100 - configure TX0 and TX1 ethtool -Q eth0 queue_mask 3 --coalesce tx-usecs 100 --coalesce tx-usecs 100 show configuration for TX0 and TX1: ethtool -Q eth0 queue_mask 3 --show-coalesce Comparing to gro_flush_timeout and napi_defer_hard_irqs, this patch allows to enable IRQ coalesing for RX path separately. Signed-off-by:
Grygorii Strashko <grygorii.strashko@ti.com> Signed-off-by:
Roger Quadros <rogerq@kernel.org> Signed-off-by:
-
Roger Quadros authored
Add driver support for viewing / changing the MAC Merge sublayer parameters and seeing the verification state machine's current state via ethtool. As hardware does not support interrupt notification for verification events we resort to polling on link up. On link up we try a couple of times for verification success and if unsuccessful then give up. The Frame Preemption feature is described in the Technical Reference Manual [1] in section: 12.3.1.4.6.7 Intersperced Express Traffic (IET – P802.3br/D2.0) Due to Silicon Errata i2208 [2] we set limit min IET fragment size to 124 (excluding 4 bytes mCRC). [1] AM62x TRM - https://www.ti.com/lit/ug/spruiv7a/spruiv7a.pdf [2] AM62x Silicon Errata - https://www.ti.com/lit/er/sprz487c/sprz487c.pdfSigned-off-by:
Vladimir Oltean <vladimir.oltean@nxp.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 (rate-limiting) on external port FIFOs. The MQPRIO Qdisc offload is expected to work with or without VLAN/priority tagged packets. The CPSW external Port FIFO has 8 Priority queues. The rate-limit can be set for each of these priority queues. Which Priority queue a packet is assigned to depends on PN_REG_TX_PRI_MAP register which maps header priority to switch priority. The header priority of a packet is assigned via the RX_PRI_MAP_REG which maps packet priority to header priority. The packet priority is either the VLAN priority (for VLAN tagged packets) or the thread/channel offset. For simplicity, we assign the same priority queue to all queues of a Traffic Class so it can be rate-limited correctly. 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:
-
Roger Quadros authored
Move register definitions to header file. No functional change. Signed-off-by:
-
Roger Quadros authored
Move this code around to avoid forward declaration. No functional change. Signed-off-by:
-
Roger Quadros authored
Handle offloading commands using switch-case in am65_cpsw_setup_taprio(). Move checks to am65_cpsw_taprio_replace(). Use NL_SET_ERR_MSG_MOD for error messages. Change error message from "Failed to set cycle time extension" to "cycle time extension not supported" Signed-off-by:
-
Roger Quadros authored
We will use this Kconfig option to not only enable TAS/EST offload but also other QoS features like Multiqueue priority descriptors and MAC-Merge/Frame Preemption. TI_AM65_CPSW_QOS seems a more appropriate Kconfig option name than TI_AM65_CPSW_TAS. Signed-off-by:
-
Roger Quadros authored
Build am65-cpsw-qos only if CONFIG_TI_AM65_CPSW_TAS is enabled. Signed-off-by:
-
Vladimir Oltean authored
Some devices do not support individual 'pmac' and 'emac' stats. For such devices, resort to 'aggregate' stats. Cc: Shuah Khan <shuah@kernel.org> Signed-off-by:
-
Vladimir Oltean authored
Some devices have errata due to which they cannot report ETH_ZLEN (60) in the rx-min-frag-size. This was foreseen of course, and lldpad has logic that when we request it to advertise addFragSize 0, it will round it up to the lowest value that is _actually_ supported by the hardware. The problem is that the selftest expects lldpad to report back to us the same value as we requested. Make the selftest smarter by figuring out on its own what is a reasonable value to expect. Cc: Shuah Khan <shuah@kernel.org> Signed-off-by:
-
David S. Miller authored
Hangbin Liu says: ==================== Convert net selftests to run in unique namespace (last part) Here is the last part of converting net selftests to run in unique namespace. This part converts all left tests. After the conversion, we can run the net sleftests in parallel. e.g. # ./run_kselftest.sh -n -t net:reuseport_bpf TAP version 13 1..1 # selftests: net: reuseport_bpf ok 1 selftests: net: reuseport_bpf mod 10... # Socket 0: 0 # Socket 1: 1 ... # Socket 4: 19 # Testing filter add without bind... # SUCCESS # ./run_kselftest.sh -p -n -t net:cmsg_so_mark.sh -t net:cmsg_time.sh -t net:cmsg_ipv6.sh TAP version 13 1..3 # selftests: net: cmsg_so_mark.sh ok 1 selftests: net: cmsg_so_mark.sh # selftests: net: cmsg_time.sh ok 2 selftests: net: cmsg_time.sh # selftests: net: cmsg_ipv6.sh ok 3 selftests: net: cmsg_ipv6.sh # ./run_kselftest.sh -p -n -c net TAP version 13 1..95 # selftests: net: reuseport_bpf_numa ok 3 selftests: net: reuseport_bpf_numa # selftests: net: reuseport_bpf_cpu ok 2 selftests: net: reuseport_bpf_cpu # selftests: net: sk_bind_sendto_listen ok 9 selftests: net: sk_bind_sendto_listen # selftests: net: reuseaddr_conflict ok 5 selftests: net: reuseaddr_conflict ... Here is the part 1 link: https://lore.kernel.org/netdev/20231202020110.362433-1-liuhangbin@gmail.com part 2 link: https://lore.kernel.org/netdev/20231206070801.1691247-1-liuhangbin@gmail.com part 3 link: https://lore.kernel.org/netdev/20231213060856.4030084-1-liuhangbin@gmail.com ==================== Signed-off-by:
-
Hangbin Liu authored
Add a variable RUN_IN_NETNS if the user wants to run all the selected tests in namespace in parallel. With this, we can save a lot of testing time. Note that some tests may not fit to run in namespace, e.g. net/drop_monitor_tests.sh, as the dwdump needs to be run in init ns. I also added another parameter -p to make all the logs reported separately instead of mixing them in the stdout or output.log. Nit: the NUM in run_one is not used, rename it to test_num. Acked-by:
David Ahern <dsahern@kernel.org> Signed-off-by:
Hangbin Liu <liuhangbin@gmail.com> Signed-off-by:
-
Hangbin Liu authored
pmtu test use /bin/sh, so we need to source ./lib.sh instead of lib.sh Here is the test result after conversion. # ./pmtu.sh TEST: ipv4: PMTU exceptions [ OK ] TEST: ipv4: PMTU exceptions - nexthop objects [ OK ] TEST: ipv6: PMTU exceptions [ OK ] TEST: ipv6: PMTU exceptions - nexthop objects [ OK ] ... TEST: ipv4: list and flush cached exceptions - nexthop objects [ OK ] TEST: ipv6: list and flush cached exceptions [ OK ] TEST: ipv6: list and flush cached exceptions - nexthop objects [ OK ] TEST: ipv4: PMTU exception w/route replace [ OK ] TEST: ipv4: PMTU exception w/route replace - nexthop objects [ OK ] TEST: ipv6: PMTU exception w/route replace [ OK ] TEST: ipv6: PMTU exception w/route replace - nexthop objects [ OK ] Signed-off-by:
-
Hangbin Liu authored
The setup_loopback and setup_veth use their own way to create namespace. So let's just re-define server_ns/client_ns to unique name. At the same time update the namespace name in gro.sh and toeplitz.sh. As I don't have env to run toeplitz.sh. Here is only the gro test result. # ./gro.sh running test ipv4 data Expected {200 }, Total 1 packets Received {200 }, Total 1 packets. ... Gro::large test passed. All Tests Succeeded! Acked-by: -
Hangbin Liu authored
Here is the test result after conversion. # ./xfrm_policy.sh PASS: policy before exception matches PASS: ping to .254 bypassed ipsec tunnel (exceptions) PASS: direct policy matches (exceptions) PASS: policy matches (exceptions) PASS: ping to .254 bypassed ipsec tunnel (exceptions and block policies) PASS: direct policy matches (exceptions and block policies) PASS: policy matches (exceptions and block policies) PASS: ping to .254 bypassed ipsec tunnel (exceptions and block policies after hresh changes) PASS: direct policy matches (exceptions and block policies after hresh changes) PASS: policy matches (exceptions and block policies after hresh changes) PASS: ping to .254 bypassed ipsec tunnel (exceptions and block policies after hthresh change in ns3) PASS: direct policy matches (exceptions and block policies after hthresh change in ns3) PASS: policy matches (exceptions and block policies after hthresh change in ns3) PASS: ping to .254 bypassed ipsec tunnel (exceptions and block policies after htresh change to normal) PASS: direct policy matches (exceptions and block policies after htresh change to normal) PASS: policy matches (exceptions and block policies after htresh change to normal) PASS: policies with repeated htresh change Acked-by:
-
Hangbin Liu authored
Here is the test result after conversion. # ./stress_reuseport_listen.sh listen 24000 socks took 0.47714 Acked-by:
-
Hangbin Liu authored
When running the test in namespace, the debugfs may not load automatically. So add a checking to make sure debugfs loaded. Here is the test result after conversion. # ./rtnetlink.sh PASS: policy routing PASS: route get ... PASS: address proto IPv4 PASS: address proto IPv6 Acked-by:
-
Hangbin Liu authored
This test will move the device to netns 1. Add a new test_ns to do this. Here is the test result after conversion. # ./netns-name.sh netns-name.sh [ OK ] Acked-by:
-
