Hi Gaetan Sorry for very late reply, we were busy working on 21.11 integration. Although we have adopted this code internally for us but I am sharing the patch to opensource for community benefit. This is specific case of AZURE setup with our very customized complex environment. Let me share the logs with traceback first SECONDARY PROCESS timestamp=1633598184 TCZ0.0.0 Cycle 152 (Build 1832) signal 11 (Segmentation fault), address is 0x31117bbce6c8 from 0x47d08b1 [bt]: ( 1) _Z18snprintf_backtraceRPciiP9siginfo_tPv (+ 0xf4) - sp = 0x7fffef3fd110, ip = 0x3acdc54 [bt]: ( 2) _Z13crit_err_hdlriP9siginfo_tPv (+ 0x159) - sp = 0x7fffef3fdc20, ip = 0x3acdf29 [bt]: ( 3) _ZN13SignalAdapter12handleSignalEiP9siginfo_tPv (+ 0x104) - sp = 0x7fffef3fdf00, ip = 0x274d4c4 [bt]: ( 4) _L_unlock_18 (+ 0x2c) - sp = 0x7fffef3fdf80, ip = 0x7ffff7bce630 [bt]: ( 5) rte_eth_dev_attach_secondary (+ 0x21) - sp = 0x7fffef3fec50, ip = 0x47d08b1 [bt]: ( 6) rte_eth_from_ring (+ 0x3438) - sp = 0x7fffef3fec80, ip = 0x4e49da8 [bt]: ( 7) _init (+ 0xa1b8) - sp = 0x7fffef3feec0, ip = 0x12e0368 [bt]: ( 8) local_dev_probe (+ 0xac) - sp = 0x7fffef3feef0, ip = 0x478fd2c [bt]: ( 9) rte_uuid_unparse (+ 0x274) - sp = 0x7fffef3fef30, ip = 0x47a3e94 [bt]: (10) rte_eal_vfio_get_vf_token (+ 0xd7) - sp = 0x7fffef3ff110, ip = 0x47b04b7 [bt]: (11) eal_hugepage_info_read (+ 0x602) - sp = 0x7fffef3ff170, ip = 0x47b2cd2 [bt]: (12) start_thread (+ 0xc5) - sp = 0x7fffef3ff220, ip = 0x7ffff7bc6ea5 [bt]: (13) clone (+ 0x6d) - sp = 0x7fffef3ff2c0, ip = 0x7ffff004096d EAL: Fail to recv reply for request /var/run/dpdk/oracusbc/mp_socket:eal_dev_mp_request EAL: Cannot send request to primary EAL: Failed to send hotplug request to primary net_failsafe: Failed to probe devargs net_tap_vsc0 EAL: Fail to recv reply for request /var/run/dpdk/oracusbc/mp_socket:eal_dev_mp_request EAL: Cannot send request to primary EAL: Failed to send hotplug request to primary net_failsafe: Failed to probe devargs net_tap_vsc1 EAL: No legacy callbacks, legacy socket not created EAL: Drop mp reply: eal_dev_mp_request PRIMARY PROCESS timestamp=1633598196 TCZ0.0.0 Cycle 152 (Build 1832) signal 11 (Segmentation fault), address is 0x38 from 0x9d8fbe [bt]: ( 1) _Z18snprintf_backtraceRPciiP9siginfo_tPv (+ 0xf4) - sp = 0x7fffecf41150, ip = 0x100dd44 [bt]: ( 2) _Z13crit_err_hdlriP9siginfo_tPv (+ 0x159) - sp = 0x7fffecf41c60, ip = 0x100e019 [bt]: ( 3) _ZN13SignalAdapter12handleSignalEiP9siginfo_tPv (+ 0x104) - sp = 0x7fffecf41f40, ip = 0xff4894 [bt]: ( 4) _L_unlock_18 (+ 0x2c) - sp = 0x7fffecf41fc0, ip = 0x7ffff61d9630 [bt]: ( 5) failsafe_eth_dev_close (+ 0x65e) - sp = 0x7fffecf42c90, ip = 0x9d8fbe [bt]: ( 6) rte_eth_link_get_nowait (+ 0x6a) - sp = 0x7fffecf42cf0, ip = 0x62fa0a [bt]: ( 7) _ZN11StatsThread9statsLoopEP10CustomObject (+ 0x33e) - sp = 0x7fffecf42d20, ip = 0xedea2e [bt]: ( 8) _ZN11StatsThread9statsLoopEP10CustomObject (+ 0x8dc) - sp = 0x7fffecf42d90, ip = 0xedefcc [bt]: ( 9) ThreadFunction (+ 0xe6) - sp = 0x7fffecf42db0, ip = 0x7ffff6b477e6 [bt]: (10) start_thread (+ 0xc5) - sp = 0x7fffecf42de0, ip = 0x7ffff61d1ea5 [bt]: (11) clone (+ 0x6d) - sp = 0x7fffecf42e80, ip = 0x7ffff0a6b96d DPDK 20.11.2 core mask is 00000000000000000000000000004000 DPDK Custom Process initialized with 2 ports the min max TxQ is maxTxQueues 16 Using 1 RxQs for port 0 (# F-core=1) Using 1 RxQs for port 3 (# F-core=1) Core 14 (port=0, rxQ=0) kni_ring=(nil) Core 14 (port=3, rxQ=0) kni_ring=(nil) Core 14 txN = 0 Thread for core 14 using ring from usbc of 0x31117b29bb00 Ring size must be powers of 2, adjusting from 8196 to 16384 Thread for core 14 using ring from MEDIA of 0x31117b27b840 Encaps Memory Zone= 48044 sizeof encaps = 60 Trace Memory Zone= 272 Policy Memory Zone= 8196 sizeof policy = 240 link status for port 0 is 1 link status for port 3 is 1 PORT 0 supports 16 rx queues and 16 tx queues (driver_name = net_failsafe, driver_type = 16) PORT 0 is polling for link-change, interrupts disabled [DPDK] tap_flow_create(): Kernel refused TC filter rule creation (17): File exists [DPDK] net_failsafe: Failed to create flow on sub_device 1 add_flow(): create() fails for port 0; Reason: overlapping rules or Kernel too old for flower support Error adding broadcast flow PORT 3 supports 16 rx queues and 16 tx queues (driver_name = net_failsafe, driver_type = 16) PORT 3 is polling for link-change, interrupts disabled [DPDK] EAL: Failed to hotplug add device on primary [DPDK] tap_flow_create(): Kernel refused TC filter rule creation (17): File exists [DPDK] net_failsafe: Failed to create flow on sub_device 1 add_flow(): create() fails for port 3; Reason: overlapping rules or Kernel too old for flower support Error adding broadcast flow Cmd Thread is available Capture object initialized init :Stats Thread is available ifLinkUpdate: Sending OperStatus for port=0 stat=1 ifLinkUpdate: Port 0 Link Change - speed 40000 Mbps - full-duplex [DPDK] EAL: Fail to recv reply for request /var/run/dpdk/oracusbc/mp_socket_2934_298e9db8d1:eal_dev_mp_request [DPDK] EAL: rte_mp_request_sync failed [DPDK] EAL: Failed to send hotplug request to secondary [DPDK] EAL: Fail to recv reply for request /var/run/dpdk/oracusbc/mp_socket_2934_298e9db8d1:eal_dev_mp_request [DPDK] EAL: rte_mp_request_sync failed [DPDK] EAL: Failed to hotplug add device on primary [DPDK] Invalid port_id=2 [DPDK] net_failsafe: Operation rte_eth_stats_get failed for sub_device 1 with error -19 There is some race at secondary process and primary got crashed because its data-structures and partially filled. Let me know if you need GDB analysis, I can share with next reply if you are still unsatisfied. GDB analysis will be bigger. Thanks! Regards Vipul -----Original Message----- From: Gaëtan Rivet Sent: Monday, November 22, 2021 3:53 PM To: Vipul Ashri ; dev@dpdk.org Cc: stable@dpdk.org Subject: [External] : Re: [PATCH v2] net/failsafe: link_update request crashing at boot On Thu, Oct 21, 2021, at 23:42, vipul.ashri@oracle.com wrote: > From: Vipul Ashri > > > failsafe crashed while sending early link_update request during boot > time initialization. > Based on debugging we found failsafe device was good but sub- devices > were progressing towards initialization and SUBOPS macro where > expanding macro gives [partial_dev]->dev_ops->link_update() > execution of which triggered crash because dev_ops==0. similar crash > seen at failsafe_eth_dev_close() > > Failsafe driver need a separate check for subdevices similar to > "RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);" which is called > to almost every eth_dev function. > > Fixes: a46f8d5 ("net/failsafe: add fail-safe PMD") > Cc: stable@dpdk.org > Signed-off-by: Vipul Ashri > Hello Vipul, I'm sorry for the delay, I missed your fix on the mailing list. IIUC, the issue is that failsafe finished init and received an ethdev operation call, but one of its sub-device, although marked DEV_ACTIVE, has its eth_dev->dev_ops field NULL. It is really surprising to me, because there aren't many ways for a sub-device to become DEV_ACTIVE. The only two ways are * by executing 'fs_dev_configure()', which will first execute rte_eth_dev_configure() on the sub-device, and on error would stop *without* setting DEV_ACTIVE. rte_eth_dev_configure() will itself execute RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV), so it would return negative errno and fs_dev_configure() would abort. * by executing 'fs_dev_remove()' and the sub-device was 'DEV_STARTED' to begin with, then it is retrograded to DEV_ACTIVE once stopped. So I don't understand yet how it is possible for a sub-device to become DEV_ACTIVE while its eth_dev->dev_ops are NULL. It seems more like a bug, memory corruption or just an unexpected execution pattern. Could describe in more detail the execution? In particular, setting the EAL log-level to debug with the option: ' --log-level pmd.net.failsafe:debug ' for example while using testpmd or your DPDK app. It should show ethdev level accesses to the sub-devices, and error values. Best regards, -- Gaetan Rivet