Date:	Sun, 15 Jun 2014 12:56:01 -0700 (PDT)
From:	Joseph Gasparakis <joseph.gasparakis@...el.com>
To:	Tom Herbert <therbert@...gle.com>
cc:	Or Gerlitz <or.gerlitz@...il.com>,
	David Miller <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Joseph Gasparakis <joseph.gasparakis@...el.com>,
	Geert Uytterhoeven <geert@...ux-m68k.org>,
	Pravin B Shelar <pshelar@...ira.com>
Subject: Re: [PATCH v2 0/5] Checksum fixes



On Sun, 15 Jun 2014, Tom Herbert wrote:

> On Sun, Jun 15, 2014 at 5:29 AM, Or Gerlitz <or.gerlitz@...il.com> wrote:
> > On Sun, Jun 15, 2014 at 11:01 AM, David Miller <davem@...emloft.net> wrote:
> >>
> >> From: Tom Herbert <therbert@...gle.com>
> >> Date: Sat, 14 Jun 2014 23:23:44 -0700 (PDT)
> >>
> >> > Fixes related to some recent checksum modifications.
> >> >
> >> > - Fix GSO constants to match NETIF flags
> >> > - Fix logic in saving checksum complete in __skb_checksum_complete
> >> > - Call __skb_checksum_complete from UDP if we are checksumming over
> >> >   whole packet in order to save checksum.
> >> > - Fixes to VXLAN to work correctly with checksum complete
> >>
> >> Series applied, thanks Tom.
> >
> >
> >
> > OK, so I gave the net tree which contains these fixes (e.g up to
> > commit b58537a "net: sctp: fix permissions for rto_alpha and rto_beta
> > knobs")
> > a try to see how vxlan offloading goes. Things (HW TX/RX checksum, HW
> > LSO and GRO come into play) looks basically OK.  I tested with
> > Mellanox ConnectX3-pro card and the mlx4 driver with the offloads
> > enabled.
> >
> Thanks for testing the combinations.
> 
> > I quickly passed through the four different setups below
> >
> > 1. veth --> OVS --> OVS VXLAN port --> IP stack ---> mlx4 --> HW
> >
> > 2. veth --> Bridge --> VXLAN NIC  --> IP stack ---> mlx4 --> HW
> >
> > 3. bridge --> VXLAN NIC  --> IP stack ---> mlx4 --> HW
> >
> > 4. VXLAN NIC  --> IP stack ---> mlx4 --> HW
> >
> > I didn't test yet the path from VM, e.g through vhost and tap to OVS
> > and then to VXLAN, will do in the coming days
> >
> > @ some point, I did stepped on something... the below trace
> >
> Possibly related to truesize?

Although it seems like Or did a more thorough testing, just to let you 
know I also tested my basic VXLAN to VXLAN with i40e and I can confirm I 
see again GSOs with VXLAN. Thank you for taking care of the issue.

> 
> > Or.
> >
> > ------------[ cut here ]------------
> > WARNING: CPU: 1 PID: 11257 at net/core/skbuff.c:3884
> > skb_try_coalesce+0x246/0x371()
> > Modules linked in: veth bridge stp llc mlx4_en mlx4_ib mlx4_core vxlan
> > netconsole nfsv3 nfs_acl auth_rpcgss oid_registry nfsv4 nfs lockd
> > autofs4 sunrpc crc32c_generic ib_ucm rdma_ucm rdma_cm iw_cm ib_ipoib
> > ib_cm ib_uverbs ib_umad ib_sa ib_mad ib_core ib_addr ipv6 dm_mirror
> > dm_region_hash dm_log dm_mod igb ptp pps_core joydev sg microcode
> > pcspkr rng_core ehci_pci ehci_hcd ioatdma dca shpchp button sr_mod
> > ext3 jbd floppy usb_storage sd_mod crc_t10dif crct10dif_common
> > ata_piix libata scsi_mod uhci_hcd radeon ttm drm_kms_helper hwmon
> > [last unloaded: veth]
> > CPU: 1 PID: 11257 Comm: netserver Tainted: G        W I   3.15.0+ #152
> > Hardware name: Supermicro X7DWU/X7DWU, BIOS  1.1 04/30/2008
> >  0000000000000f2c ffff88022fc83ac8 ffffffff813d6986 0000000000000f2c
> >  0000000000000000 ffff88022fc83b08 ffffffff81039c7c ffff88021ea37e00
> >  ffffffff81333adb ffff88021ea37f00 ffff88021ea37b00 ffff88022fc83b84
> > Call Trace:
> >  <IRQ>  [<ffffffff813d6986>] dump_stack+0x51/0x6b
> >  [<ffffffff81039c7c>] warn_slowpath_common+0x7c/0x96
> >  [<ffffffff81333adb>] ? skb_try_coalesce+0x246/0x371
> >  [<ffffffff81039cab>] warn_slowpath_null+0x15/0x17
> >  [<ffffffff81333adb>] skb_try_coalesce+0x246/0x371
> >  [<ffffffff8133d3d2>] ? netif_receive_skb_internal+0xed/0xed
> >  [<ffffffff81380d49>] tcp_try_coalesce+0x4d/0xa1
> >  [<ffffffff81380df5>] tcp_queue_rcv+0x58/0x105
> >  [<ffffffff81384282>] tcp_rcv_established+0x3a3/0x5f9
> >  [<ffffffff8138b636>] ? tcp_v4_rcv+0x476/0x95b
> >  [<ffffffff8138ae97>] tcp_v4_do_rcv+0x102/0x42b
> >  [<ffffffff813dae75>] ? _raw_spin_lock_nested+0x3a/0x41
> >  [<ffffffff8138b662>] tcp_v4_rcv+0x4a2/0x95b
> >  [<ffffffff8136d907>] ? ip_local_deliver_finish+0x2f/0x289
> >  [<ffffffff8136da42>] ip_local_deliver_finish+0x16a/0x289
> >  [<ffffffff8136d907>] ? ip_local_deliver_finish+0x2f/0x289
> >  [<ffffffff8136dbd4>] ip_local_deliver+0x73/0x7a
> >  [<ffffffff8136d5d7>] ip_rcv_finish+0x417/0x42f
> >  [<ffffffff8136d8a0>] ip_rcv+0x2b1/0x2e9
> >  [<ffffffff8133d09c>] __netif_receive_skb_core+0x47e/0x4d4
> >  [<ffffffff8133ccde>] ? __netif_receive_skb_core+0xc0/0x4d4
> >  [<ffffffff8133d147>] __netif_receive_skb+0x55/0x5a
> >  [<ffffffff8133d1fb>] process_backlog+0xaf/0x199
> >  [<ffffffff8133d697>] net_rx_action+0xac/0x1f2
> >  [<ffffffff8103e3cd>] __do_softirq+0x12b/0x28e
> >  [<ffffffff8132e88d>] ? release_sock+0x1c0/0x1c9
> >  [<ffffffff813e41cc>] do_softirq_own_stack+0x1c/0x30
> >  <EOI>  [<ffffffff8103e074>] do_softirq+0x31/0x4a
> >  [<ffffffff8103e1ba>] __local_bh_enable_ip+0x92/0xa5
> >  [<ffffffff813db512>] _raw_spin_unlock_bh+0x34/0x38
> >  [<ffffffff8132e88d>] release_sock+0x1c0/0x1c9
> >  [<ffffffff8137b6b8>] tcp_recvmsg+0x907/0xa4c
> >  [<ffffffff8135e794>] ? sch_direct_xmit+0x9e/0x213
> >  [<ffffffff8139d5d3>] inet_recvmsg+0xd1/0xeb
> >  [<ffffffff8132aa8a>] sock_recvmsg+0x94/0xb2
> >  [<ffffffff81073034>] ? trace_hardirqs_on+0xd/0xf
> >  [<ffffffff813db541>] ? _raw_spin_unlock_irq+0x2b/0x38
> >  [<ffffffff81126726>] ? __fdget+0xe/0x10
> >  [<ffffffff8132ab67>] SyS_recvfrom+0xbf/0x10f
> >  [<ffffffff81085e08>] ? rcu_irq_exit+0x7d/0x8f
> >  [<ffffffff813dbaa0>] ? retint_restore_args+0xe/0xe
> >  [<ffffffff8132e6fd>] ? release_sock+0x30/0x1c9
> >  [<ffffffff8132e6fd>] ? release_sock+0x30/0x1c9
> >  [<ffffffff813e2962>] system_call_fastpath+0x16/0x1b
> > ---[ end trace 665483c39b84f51c ]---
> 
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists