Hi --
I'm seeing a panic while running NICDRV. It's happening when running
test09 (Nemo_Trunk) of the testsuite. I didn't see this before though.
Just want to check if it's my problem or not. It does seem like a race
between TX interrupt calling mac_tx_update() and thread removing TX src
(mac src client).
> $C
ffffff0003ec59f0 vpanic()
ffffff0003ec5a20 mutex_panic+0x73(fffffffffb92d478, ffffffcc53dc82c0)
ffffff0003ec5a80 mutex_vector_enter+0x446(ffffffcc53dc82c0)
ffffff0003ec5ad0 mac_tx_srs_wakeup+0x2b(ffffffcc53dc82c0, 0)
ffffff0003ec5b10 i_mac_tx_srs_notify+0x79(ffffff0153f3a738, 0)
ffffff0003ec5b30 mac_tx_update+0x18(ffffff0153f3a738)
ffffff0003ec5b60 bfe_tx_done+0x34(ffffff0153b0dd00, 1000000)
ffffff0003ec5bb0 bfe_interrupt+0xbb(ffffff0153b0dd00, 0)
ffffff0003ec5c00 av_dispatch_autovect+0x7c(15)
ffffff0003ec5c40 dispatch_hardint+0x33(15, 0)
ffffff00045888d0 switch_sp_and_call+0x13()
ffffff0004588920 do_interrupt+0x9e(ffffff0004588930, ffffffcc53dc88b8)
ffffff0004588930 _interrupt+0xba()
ffffff0004588a30 mac_srs_free+0x5a(ffffffcc53dc82c0)
ffffff0004588a90 mac_srs_group_teardown+0xf5(ffffff6972044130,
ffffff9a92f95428
, 1)
ffffff0004588af0 mac_datapath_teardown+0x8e(ffffff6972044130,
ffffff9a92f95428,
1)
ffffff0004588b40 mac_unicast_remove+0xfa(ffffff6972044130, ffffff84bda00c40)
ffffff0004588b70 aggr_port_delete+0xc2(ffffff84bd9ea790)
ffffff0004588be0 aggr_grp_delete+0x1aa(11c)
ffffff0004588c20 aggr_ioc_delete+0x27(ffffff84c246dd60, 8046f74, 100003,
ffffff7302534830, ffffff0004588de4)
ffffff0004588cc0 drv_ioctl+0x137(b200000000, a660002, 8046f74, 100003,
ffffff7302534830, ffffff0004588de4)
ffffff0004588d00 cdev_ioctl+0x45(b200000000, a660002, 8046f74, 100003,
ffffff7302534830, ffffff0004588de4)
ffffff0004588d40 spec_ioctl+0x83(ffffff01513c7200, a660002, 8046f74,
100003,
ffffff7302534830, ffffff0004588de4, 0)
ffffff0004588dc0 fop_ioctl+0x7b(ffffff01513c7200, a660002, 8046f74, 100003,
ffffff7302534830, ffffff0004588de4, 0)
ffffff0004588ec0 ioctl+0x18e(3, a660002, 8046f74)
ffffff0004588f10 _sys_sysenter_post_swapgs+0x14b()
Here's the cclient list
> ffffff0153f3a738::print mac_impl_t mi_clients_list | ::list
mac_client_impl_t mci_client_next | ::print mac_client_impl_t mci_flent
| ::print flow_entry_t fe_tx_srs
fe_tx_srs = 0xffffffcc53dc82c0
fe_tx_srs = 0
And the mac src (TX) is freed already by the pinned thread perhaps.
> 0xffffffcc53dc82c0::print mac_soft_ring_set_t srs_lock
{
srs_lock._opaque = [ 0xdeadbeefdeadbeef ]
}
>
I don't see any detach in progress so my driver didn't call
mac_unregister while interrupt was still active.
From test09 logs :
stdout| Configuring nfs . DONE
stdout| Distributing the test binary . . DONE
stdout| Configuring mount points . . DONE
stdout| Starting Corrupt . . DONE
stdout| Corrupt Started. Running for 600 seconds
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
stdout| trunk_switch: aggr1 deleted
stdout| trunk_switch: aggr1 plumbed
I'm running snv_113.
Thanks,
--
http://saurabhslr.blogspot.com/
