- 06 Mar, 2022 6 commits
-
-
Sebastian Andrzej Siewior authored
Since commit baebdf48 ("net: dev: Makes sure netif_rx() can be invoked in any context.") the function netif_rx() can be used in preemptible/thread context as well as in interrupt context. Use netif_rx(). Cc: Andrew Lunn <andrew@lunn.ch> Cc: Heiner Kallweit <hkallweit1@gmail.com> Cc: Radu Pirea <radu-nicolae.pirea@oss.nxp.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Russell King <linux@armlinux.org.uk> Signed-off-by:
Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Dongli Zhang says: ==================== tun/tap: use kfree_skb_reason() to trace dropped skb The commit c504e5c2 ("net: skb: introduce kfree_skb_reason()") has introduced the kfree_skb_reason() to help track the reason. The tun and tap are commonly used as virtio-net/vhost-net backend. This is to use kfree_skb_reason() to trace the dropped skb for those two drivers. Changed since v1: - I have renamed many of the reasons since v1. I make them as generic as possible so that they can be re-used by core networking and drivers. Changed since v2: - declare drop_reason as type "enum skb_drop_reason" - handle the drop in skb_list_walk_safe() case for tap driver, and kfree_skb_list_reason() is introduced Changed since v3 (only for PATCH 4/4): - rename to TAP_FILTER and TAP_TXFILTER - honor reverse xmas tree style declaration for 'drop_reason' in tun_net_xmit() Changed since v4: - make kfree_skb_list() static inline - add 'computation' to SKB_CSUM comment - change COPY_DATA to UCOPY_FAULT - add 'metadata' to DEV_HDR comment - expand comment on DEV_READY - change SKB_TRIM to NOMEM - chnage SKB_PULL to HDR_TRUNC Changed since v5: - rebase to net-next The following reasons are introduced. - SKB_DROP_REASON_SKB_CSUM - SKB_DROP_REASON_SKB_GSO_SEG - SKB_DROP_REASON_SKB_UCOPY_FAULT - SKB_DROP_REASON_DEV_HDR - SKB_DROP_REASON_FULL_RING - SKB_DROP_REASON_DEV_READY - SKB_DROP_REASON_NOMEM - SKB_DROP_REASON_HDR_TRUNC - SKB_DROP_REASON_TAP_FILTER - SKB_DROP_REASON_TAP_TXFILTER This is the output for TUN device. <idle>-0 [029] ..s1. 450.727651: kfree_skb: skbaddr=0000000023d235cc protocol=0 location=00000000a6748854 reason: NOT_SPECIFIED <idle>-0 [000] b.s3. 451.165671: kfree_skb: skbaddr=000000006b5de7cc protocol=4 location=000000007c2b9eae reason: FULL_RING <idle>-0 [000] b.s3. 453.149650: kfree_skb: skbaddr=000000006b5de7cc protocol=4 location=000000007c2b9eae reason: FULL_RING <idle>-0 [000] b.s3. 455.133576: kfree_skb: skbaddr=000000006b5de7cc protocol=4 location=000000007c2b9eae reason: FULL_RING <idle>-0 [000] b.s3. 457.117566: kfree_skb: skbaddr=000000006b5de7cc protocol=4 location=000000007c2b9eae reason: FULL_RING This is the output for TAP device. arping-7053 [006] ..s1. 1000.047753: kfree_skb: skbaddr=000000008618a587 protocol=2054 location=00000000743ad4a7 reason: FULL_RING <idle>-0 [022] ..s1. 1000.778514: kfree_skb: skbaddr=000000002c1e706c protocol=0 location=00000000a6748854 reason: NOT_SPECIFIED arping-7053 [006] ..s1. 1001.047830: kfree_skb: skbaddr=000000008618a587 protocol=2054 location=00000000743ad4a7 reason: FULL_RING arping-7053 [006] ..s1. 1002.047918: kfree_skb: skbaddr=000000008618a587 protocol=2054 location=00000000743ad4a7 reason: FULL_RING arping-7053 [006] ..s1. 1003.048017: kfree_skb: skbaddr=000000008618a587 protocol=2054 location=00000000743ad4a7 reason: FULL_RING ==================== Signed-off-by:
-
Dongli Zhang authored
The TUN can be used as vhost-net backend. E.g, the tun_net_xmit() is the interface to forward the skb from TUN to vhost-net/virtio-net. However, there are many "goto drop" in the TUN driver. Therefore, the kfree_skb_reason() is involved at each "goto drop" to help userspace ftrace/ebpf to track the reason for the loss of packets. The below reasons are introduced: - SKB_DROP_REASON_DEV_READY - SKB_DROP_REASON_NOMEM - SKB_DROP_REASON_HDR_TRUNC - SKB_DROP_REASON_TAP_FILTER - SKB_DROP_REASON_TAP_TXFILTER Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Joe Jin <joe.jin@oracle.com> Signed-off-by:
Dongli Zhang <dongli.zhang@oracle.com> Signed-off-by:
-
Dongli Zhang authored
No functional change. Just to split the if statement into different conditions to use kfree_skb_reason() to trace the reason later. Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Joe Jin <joe.jin@oracle.com> Signed-off-by:
David Ahern <dsahern@kernel.org> Signed-off-by:
-
Dongli Zhang authored
The TAP can be used as vhost-net backend. E.g., the tap_handle_frame() is the interface to forward the skb from TAP to vhost-net/virtio-net. However, there are many "goto drop" in the TAP driver. Therefore, the kfree_skb_reason() is involved at each "goto drop" to help userspace ftrace/ebpf to track the reason for the loss of packets. The below reasons are introduced: - SKB_DROP_REASON_SKB_CSUM - SKB_DROP_REASON_SKB_GSO_SEG - SKB_DROP_REASON_SKB_UCOPY_FAULT - SKB_DROP_REASON_DEV_HDR - SKB_DROP_REASON_FULL_RING Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Joe Jin <joe.jin@oracle.com> Signed-off-by:
-
Dust Li authored
This reverts commit a505cce6. Leon says: We already discussed that. SMC should be changed to use RDMA CQ pool API drivers/infiniband/core/cq.c. ib_poll_handler() has much better implementation (tracing, IRQ rescheduling, proper error handling) than this SMC variant. Since we will switch to ib_poll_handler() in the future, revert this patch. Link: https://lore.kernel.org/netdev/20220301105332.GA9417@linux.alibaba.com/Suggested-by:
Leon Romanovsky <leon@kernel.org> Suggested-by:
Karsten Graul <kgraul@linux.ibm.com> Signed-off-by:
Dust Li <dust.li@linux.alibaba.com> Reviewed-by:
Leon Romanovsky <leonro@nvidia.com> Signed-off-by:
-
- 05 Mar, 2022 34 commits
-
-
David S. Miller authored
Michael Chan says: ==================== bnxt_en: Updates. This patch series contains mainly NVRAM related features. More NVRAM error checking and logging are added when installing firmware packages. A new devlink hw health report is now added to report and diagnose NVRAM issues. Other miscellaneous patches include reporting correctly cards that don't support link pause, adding an internal unknown link state, and avoiding unnecessary link toggle during firmware reset. ==================== Signed-off-by: -
Vikas Gupta authored
Add an NVM test function for devlink hw reporter. In this function an NVM VPD area is read followed by a write. Test result is cached and if it is successful then the next test can be conducted only after HW_RETEST_MIN_TIME to avoid frequent writes to the NVM. Reviewed-by:
Edwin Peer <edwin.peer@broadcom.com> Signed-off-by:
Vikas Gupta <vikas.gupta@broadcom.com> Signed-off-by:
Michael Chan <michael.chan@broadcom.com> Signed-off-by:
-
Kalesh AP authored
This reporter will report NVM errors which are non-fatal. When we receive these NVM error events, we'll report it through this new hw health reporter. Reviewed-by:
Kalesh AP <kalesh-anakkur.purayil@broadcom.com> Signed-off-by:
-
Edwin Peer authored
Health reporter state should be maintained over resets. Previously reporters were destroyed if the device capabilities changed, but since none of the reporters depend on capabilities anymore, this logic should be removed. Signed-off-by:
-
Michael Chan authored
If the flow control settings have been changed, a subsequent FW reset may cause the ethernet link to toggle unnecessarily. This link toggle will increase the down time by a few seconds. The problem is caused by bnxt_update_phy_setting() detecting a false mismatch in the flow control settings between the stored software settings and the current FW settings after the FW reset. This mismatch is caused by the AUTONEG bit added to link_info->req_flow_ctrl in an inconsistent way in bnxt_set_pauseparam() in autoneg mode. The AUTONEG bit should not be added to link_info->req_flow_ctrl. Reviewed-by:
Colin Winegarden <colin.winegarden@broadcom.com> Reviewed-by:
Pavan Chebbi <pavan.chebbi@broadcom.com> Signed-off-by:
-
Michael Chan authored
Some cards are configured to never support link pause or PFC. Discover these cards and properly report no pause support to ethtool. Disable PFC settings from DCBNL if PFC is unsupported. Reviewed-by:
-
Edwin Peer authored
This will force link state to always be logged for initial NIC open. Signed-off-by:
-
Kalesh AP authored
Instead of always returning -ENOPKG, decode the firmware error code further when the HWRM_NVM_INSTALL_UPDATE firmware call fails. Return a more suitable error code to userspace and log an error in dmesg. Reviewed-by:
Somnath Kotur <somnath.kotur@broadcom.com> Reviewed-by:
-
Kalesh AP authored
FW returns error code "NVM_INSTALL_UPDATE_CMD_ERR_CODE_ANTI_ROLLBACK" in the response to indicate that HWRM_NVM_INSTALL_UPDATE command has failed due to Anti-rollback feature. Parse the error and return an appropriate error code to the user. Reviewed-by:
-
Kalesh AP authored
This is in anticipation of handling more "cmd_err" from FW in the next patch. Reviewed-by:
-
David S. Miller authored
Robert Hancock says: ==================== NAPI/GRO support for axienet driver Add support for NAPI and GRO receive in the Xilinx AXI Ethernet driver, and some other related cleanups. ==================== Signed-off-by: -
Robert Hancock authored
Add the ability to configure the RX/TX coalesce timer with ethtool. Change default setting to scale with the clock rate rather than being a fixed number of clock cycles. Signed-off-by:
Robert Hancock <robert.hancock@calian.com> Signed-off-by:
-
Robert Hancock authored
Now that NAPI has been implemented, the hardware interrupt mitigation mechanism is not needed to avoid excessive interrupt load in most cases. Reduce the default RX interrupt threshold to 1 to reduce introduced latency. This can be increased with ethtool if desired if some applications still want to reduce interrupts. Signed-off-by:
-
Robert Hancock authored
Implement NAPI and GRO receive. In addition to better performance, this also avoids handling RX packets in hard IRQ context, which reduces the IRQ latency impact to other devices. Signed-off-by:
-
Robert Hancock authored
When the RX or TX coalesce count is set to 1, there's no point in setting the delay timer value since an interrupt will already be raised on every packet, and the delay interrupt just causes extra pointless interrupts. Signed-off-by:
-
Robert Hancock authored
Simplify the DMA error handling process, and remove some duplicated code between the DMA error handling and the stop function. Signed-off-by:
-
Robert Hancock authored
Instead of using lp->ndev.parent to find the correct device to use for DMA API calls, just use the dev attribute in the device structure. Signed-off-by:
-
Robert Hancock authored
If a memory allocation error occurred during an attempt to refill a slot in the RX ring after the packet was received, the hardware tail pointer would still have been updated to point to or past the slot which remained marked as previously completed. This would likely result in the DMA engine raising an error when it eventually tried to use that slot again. If a slot cannot be refilled, then just stop processing and do not move the tail pointer past it. On the next attempt, we should skip receiving the packet from the empty slot and just try to refill it again. This failure mode has not actually been observed, but was found as part of other driver updates. Fixes: 8a3b7a25 ("drivers/net/ethernet/xilinx: added Xilinx AXI Ethernet driver") Signed-off-by:
-
David S. Miller authored
Luiz Angelo Daros de Luca says: ==================== net: dsa: realtek: add rtl8_4t tag This patch series adds support for rtl8_4t tag. It is a variant of rtl8_4 tag, with identical values but placed at the end of the packet (before CRC). It forces checksum in software before adding the tag as those extra bytes at the end of the packet would be summed together with the rest of the payload. When the switch removes the tag before sending the packet to the network, that checksum will not match. It might be useful to diagnose or avoid checksum offload issues. With an ethertype tag like rtl8_4, the cpu port ethernet driver must work with cksum_start and chksum_offset to correctly calculate checksums. If not, the checksum field will be broken (it will contain the fake ip header sum). In those cases, using 'rtl8_4t' might be an alternative way to avoid checksum offload, either using runtime or device-tree property. Regards, Luiz v4-v5) - tags in alphabetical order in dsa_port.yaml - remove ret var from rtl8365mb_change_tag_protocol - Comment typos fixes v3-v4) - added rtl8_4 and rtl8_4t to dsa_port.yaml - removed generic considerations about checksum problems with DSA tags. They belong to Documentation/networking/dsa/dsa.rst v2-v3) - updated tag documentation (file header) - do not remove position and format from rtl8365mb_cpu - reinstate cpu to rtl8365mb - moved rtl8365mb_change_tag_protocol after rtl8365mb_cpu_config - do not modify rtl8365mb_cpu_config() logic - remove cpu arg from rtl8365mb_cpu_config(); get it from priv - dropped tag_protocol from rtl8365mb. It is now derived from cpu->position. - init cpu struct before dsa_register as default tag must be already defined before dsa_register() - fix formatting issues v1-v2) - remove mention to tail tagger, use trailing tagger. - use void* instead of char* for pointing to tag beginning - use memcpy to avoid problems with unaligned tags - calculate checksum if it still pending - keep in-use tag protocol in memory instead of reading from switch register ==================== Signed-off-by: -
Luiz Angelo Daros de Luca authored
The trailing tag is also supported by this family. The default is still rtl8_4 but now the switch supports changing the tag to rtl8_4t. Reintroduce the dropped cpu in struct rtl8365mb (removed by 6147631c). Signed-off-by:
Luiz Angelo Daros de Luca <luizluca@gmail.com> Reviewed-by:
Alvin Šipraga <alsi@bang-olufsen.dk> Reviewed-by:
Vladimir Oltean <olteanv@gmail.com> Signed-off-by:
-
Luiz Angelo Daros de Luca authored
Realtek switches supports the same tag both before ethertype or between payload and the CRC. Signed-off-by:
-
Luiz Angelo Daros de Luca authored
Realtek rtl8365mb DSA driver can use these two tag formats. Cc: devicetree@vger.kernel.org Signed-off-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Jakub Kicinski authored
Mat Martineau says: ==================== mptcp: Selftest refinements and a new test Patches 1 and 11 improve the printed output of the mptcp_join.sh selftest. Patches 2-8 add a test for the MP_FASTCLOSE option, including prerequisite changes like additional MPTCP MIBs. Patches 9-10 add some groundwork for upcoming tests. ==================== Link: https://lore.kernel.org/r/20220304193636.219315-1-mathew.j.martineau@linux.intel.comSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Geliang Tang authored
This patch updated the output info of chk_rm_nr. Renamed 'sf' to 'rmsf', which means 'remove subflow'. Added the display of whether the inverted namespaces has been used to check the mib counters. The new output looks like this: 002 remove multiple subflows syn[ ok ] - synack[ ok ] - ack[ ok ] rm [ ok ] - rmsf [ ok ] 003 remove single address syn[ ok ] - synack[ ok ] - ack[ ok ] add[ ok ] - echo [ ok ] rm [ ok ] - rmsf [ ok ] invert Signed-off-by:Geliang Tang <geliang.tang@suse.com> Signed-off-by:
Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by:
-
Geliang Tang authored
This patch added five more arguments for chk_join_nr(). The default values of them are all zero. The first two, csum_ns1 and csum_ns1, are passed to chk_csum_nr(), to check the mib counters of the checksum errors in ns1 and ns2. A '+' can be added into this two arguments to represent that multiple checksum errors are allowed when doing this check. For example, chk_csum_nr "" +2 +2 indicates that two or more checksum errors are allowed in both ns1 and ns2. The remaining two, fail_nr and rst_nr, are passed to chk_fail_nr() and chk_rst_nr() respectively, to check the sending and receiving mib counters of MP_FAIL and MP_RST. Also did some cleanups in chk_fail_nr(), renamed two local variables and updated the output message. Signed-off-by: -
Geliang Tang authored
This patch added the invert bytes check for the output data in check_transfer(). Instead of the file mismatch error: [ FAIL ] file received by server does not match (in, out): -rw------- 1 root root 45643832 Jan 16 15:04 /tmp/tmp.9xpM6Paivv Trailing bytes are: MPTCP_TEST_FILE_END_MARKER -rw------- 1 root root 45643832 Jan 16 15:04 /tmp/tmp.wnz1Yp4u7Z Trailing bytes are: MPTCP_TEST_FILE_END_MARKER Print out the inverted bytes like this: file received by server has inverted byte at 7454789 file received by server has inverted byte at 7454790 file received by server has inverted byte at 7454791 file received by server has inverted byte at 7454792 Signed-off-by:
-
Geliang Tang authored
This patch added the self test for MP_FASTCLOSE. Reused the argument addr_nr_ns2 of do_transfer() to pass the extra arguments '-I 2' to mptcp_connect commands. Then mptcp_connect disconnected the connections to trigger the MP_FASTCLOSE sending and receiving. Used chk_fclose_nr to check the MP_FASTCLOSE mibs and used chk_rst_nr to check the MP_RST mibs. This test used the test_linkfail value to make 1024KB test files. The output looks like this: Created /tmp/tmp.XB8sfv1hJ0 (size 1024 KB) containing data sent by client Created /tmp/tmp.RtTDbzqrXI (size 1024 KB) containing data sent by server 001 fastclose test syn[ ok ] - synack[ ok ] - ack[ ok ] ctx[ ok ] - fclzrx[ ok ] rtx[ ok ] - rstrx [ ok ] invert Signed-off-by: -
Geliang Tang authored
This patch reused the test_linkfail values above 2 to make test files with the given sizes (KB) for both the client side and the server side. It's useful for the test cases using different file sizes. Signed-off-by:
-
Geliang Tang authored
Instead of using a global variable mptcp_connect, this patch added a new local variable extra_args in do_transfer() to store the extra arguments passing to the mptcp_connect commands. This patch also renamed the speed level 'least' to 'speed_*'. This more flexible way can avoid the need to add new speed levels in the future. Signed-off-by:
-
Geliang Tang authored
This patch added a new function chk_rst_nr() to check the numbers of the MP_RST sending and receiving mibs. Showed in the output whether the inverted namespaces check order is used. Since if we pass -Cz to mptcp_join.sh, the MP_RST information is showed twice. Signed-off-by:
-
Geliang Tang authored
This patch added two more mibs for MP_RST, MPTCP_MIB_MPRSTTX for the MP_RST sending and MPTCP_MIB_MPRSTRX for the MP_RST receiving. Signed-off-by:
-
Geliang Tang authored
This patch added a new function chk_fclose_nr() to check the numbers of the MP_FASTCLOSE sending and receiving mibs. Signed-off-by:
-
Geliang Tang authored
This patch added two more mibs for MP_FASTCLOSE, MPTCP_MIB_MPFASTCLOSETX for the MP_FASTCLOSE sending and MPTCP_MIB_MPFASTCLOSERX for receiving. Also added a debug log for MP_FASTCLOSE receiving, printed out the recv_key of MP_FASTCLOSE in mptcp_parse_option to show that MP_RST is received. Signed-off-by:
-
Geliang Tang authored
The number of self tests in mptcp_join.sh will soon be more than 100, the output alignment is no longer OK. This patch adjusted it. Signed-off-by:
-
