TG3 network data corruption regression 2.6.24/2.6.23.4
Subsystems (3)
I am experiencing network data corruption with a 3Com 3C996B-T NIC (Broadcom NetXtreme BCM5701; driver tg3.ko). I have identified the following patch as the trigger:
commit fb93134dfc2a6e6fbedc7c270a31da03fce88db9
Author: Herbert Xu <[email protected]> Date: Wed Nov 14 15:45:21 2007 -0800
[TCP]: Fix size calculation in sk_stream_alloc_pskb
We round up the header size in sk_stream_alloc_pskb so that TSO packets get zero tail room. Unfortunately this rounding up is not coordinated with the select_size() function used by TCP to calculate the second parameter of sk_stream_alloc_pskb.
As a result, we may allocate more than a page of data in the non-TSO case when exactly one page is desired.
In fact, rounding up the head room is detrimental in the non-TSO case because it makes memory that would otherwise be available to the payload head room. TSO doesn't need this either, all it wants is the guarantee that there is no tail room.
So this patch fixes this by adjusting the skb_reserve call so that exactly the requested amount (which all callers have calculated in a precise way) is made available as tail room.
This patch was included in 2.6.24 and 2.6.23.4 -stable. I am experiencing data corruption with kernels 2.6.23.4 - 2.6.23.16, 2.6.24 - 2.6.24.2, and 2.6.25-rc2-git1. I have verified that reverting the above patch (by hand) makes the data corruption go away on all affected kernels (note that in 2.6.25 the function is sk_stream_alloc_skb() in net/ipv4/tcp.c rather than sk_stream_alloc_pskb() in include/net/sock.h).
(Also note that when testing 2.6.23 - 2.6.23.4, I had to apply the individual patch "TG3: Fix performance regression on 5705." from 2.6.23.5.)
I do not get data corruption when substituting a SysKonnect 9D21 NIC (which also uses the tg3.ko driver) or a Intel PRO/1000 82546GB NIC (which uses the e1000.ko driver).
In addition to the 3Com NIC, my computer has a SCSI HBA with an attached tape drive. The network data corruption happens only when reading from or writing to the tape drive. I have tried both a LSI MPT Fusion Ultra320 SCSI HBA (mptspi.ko) and a LSI 53c1010 Ultra160 HBA (sym53c8xx.ko) with the same results. The NIC and SCSI HBA are on separate PCI-X buses and do not share IRQs. I am using two completely separate test programs to access the SCSI tape drive and test network data integrity, so one would expect no interaction between the two tests other than CPU scheduling and DMA bandwidth. There is no disk I/O generated by either test program.
The test program that I am using to debug this problem does the following:
Computer A (kernel 2.6.24.2; 3Com 3C996B-T NIC):
malloc a 64 KB buf aligned to a 4 KB boundary
loop {
fill 64 KB buf with count data pattern
send(64 KB, MSG_MORE) <--- eventually sends corrupted data
}
(SCSI tape drive test program runs separately in the background)
Computer B (kernel 2.6.12):
malloc a 64 KB buf aligned to a 4 KB boundary
loop {
recv(64 KB, MSG_WAITALL)
verify count data pattern in 64 KB buf
}
After running for a few seconds, the verify on computer B detects data corruption in the last 4 bytes of the 64 KB buffer. The last 48 bytes of the corrupted 64 KB buffer look like this:
D0 D1 D2 D3 | D4 D5 D6 D7 | D8 D9 DA DB | DC DD DE DF E0 E1 E2 E3 | E4 E5 E6 E7 | E8 E9 EA EB | EC ED EE EF F0 F1 F2 F3 | F4 F5 F6 F7 | F8 F9 FA FB | F4 F5 F6 F7
The last 4 bytes should be "FC FD FE FF" but instead are corrupted to "F4 F5 F6 F7", a sequence which came earlier in the data stream. The data corruption always occurs at this same buffer offset and with the same 4 earlier bytes duplicated. However, it occurs on a different iteration of the send()/recv() loop each time the test is run.
When I reverse the test so that Computer A does recv() and Computer B does send(), the test passes with no data corruption. Therefore, it appears that the data corruption happens on send() but not recv().
The motherboard that I am using is a Commell LV-672. This motherboard has a PCI-express x16 slot but no PCI-X slots. To plug in the PCI-X NIC and SCSI HBA, I am using a SuperMicro CSE-RR2UE-AX riser card which plugs into the PCI-express slot on the motherboard and provides 3 PCI-X slots (two slots together on one PCI-X bus and one slot on its own PCI-X bus). The data corruption happens with every combination of the 2 cards in the 3 slots.
I assume that the above patch is just exposing some way in which the tg3 driver or the BCM5701 chip are broken. For now, I am just reverting the above patch for kernels that I use until a better solution is forthcoming. I expect that this problem will be difficult for other developers to reproduce, but I can test any patches that anyone wants to send me. [ In the meantime, should we revert the patch for 2.6.23.x and 2.6.24.x -stable, or wait for a fix to tg3? ]
I am not sure if it is relevant, but I am also getting the following messages sometimes during testing:
Clocksource tsc unstable (delta = 64002086 ns)
Time: pit clocksource has been installed.
This seems bogus because the CPU is a Intel Pentium 4 with HyperThreading, so the tsc should be reliable.
---
network MTU = 1500
lspci
00:00.0 Host bridge: Intel Corporation 82915G/P/GV/GL/PL/910GL Memory Controller Hub (rev 0e)
00:01.0 PCI bridge: Intel Corporation 82915G/P/GV/GL/PL/910GL PCI Express Root Port (rev 0e)
00:02.0 VGA compatible controller: Intel Corporation 82915G/GV/910GL Integrated Graphics Controller (rev 0e)
00:1d.0 USB Controller: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) USB UHCI #1 (rev 04)
00:1d.1 USB Controller: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) USB UHCI #2 (rev 04)
00:1d.2 USB Controller: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) USB UHCI #3 (rev 04)
00:1d.3 USB Controller: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) USB UHCI #4 (rev 04)
00:1d.7 USB Controller: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) USB2 EHCI Controller (rev 04)
00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev d4) 00:1f.0 ISA bridge: Intel Corporation 82801FB/FR (ICH6/ICH6R) LPC Interface Bridge (rev 04)
00:1f.1 IDE interface: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) IDE Controller (rev 04)
00:1f.3 SMBus: Intel Corporation 82801FB/FBM/FR/FW/FRW (ICH6 Family) SMBus Controller (rev 04)
01:00.0 PCI bridge: Intel Corporation 6700PXH PCI Express-to-PCI Bridge A (rev 09)
01:00.2 PCI bridge: Intel Corporation 6700PXH PCI Express-to-PCI Bridge B (rev 09)
02:02.0 SCSI storage controller: LSI Logic / Symbios Logic 53c1030 PCI-X Fusion-MPT Dual Ultra320 SCSI (rev 08)
03:01.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5701 Gigabit Ethernet (rev 15)
04:0d.0 FireWire (IEEE 1394): Agere Systems FW323 (rev 61)
cat /proc/interrupts
CPU0 CPU1
0: 89 0 IO-APIC-edge timer 1: 78 0 IO-APIC-edge i8042 3: 17 0 IO-APIC-edge serial 8: 0 0 IO-APIC-edge rtc 9: 0 0 IO-APIC-fasteoi acpi 12: 5 0 IO-APIC-edge i8042 14: 465 0 IO-APIC-edge ide0 16: 0 0 IO-APIC-fasteoi uhci_hcd:usb5 17: 149220 0 IO-APIC-fasteoi eth0 18: 10007 0 IO-APIC-fasteoi uhci_hcd:usb4, ioc0 19: 29 0 IO-APIC-fasteoi uhci_hcd:usb3 23: 2 0 IO-APIC-fasteoi ehci_hcd:usb1, uhci_hcd:usb2 NMI: 0 0 Non-maskable interrupts LOC: 7457 10023 Local timer interrupts RES: 1962 14316 Rescheduling interrupts CAL: 40 49 function call interrupts TLB: 39 76 TLB shootdowns TRM: 0 0 Thermal event interrupts SPU: 0 0 Spurious interrupts ERR: 0 MIS: 0
(eth0 == tg3; ioc0 == LSI SCSI HBA)
ifconfig
eth0 Link encap:Ethernet HWaddr 00:02:A5:E7:3C:2D
inet addr:192.168.1.1 Bcast:192.168.1.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:77198 errors:0 dropped:0 overruns:0 frame:0
TX packets:3488350 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:5403873 (5.1 MiB) TX bytes:1000276920 (953.9 MiB)
Interrupt:17lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:16436 Metric:1 RX packets:0 errors:0 dropped:0 overruns:0 frame:0 TX packets:0 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0
RX bytes:0 (0.0 B) TX bytes:0 (0.0 B)
information from ethtool
driver: tg3
version: 3.87
firmware-version:
bus-info: 0000:03:01.0
Settings for eth0:
Supported ports: [ TP ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full 1000baseT/Half 1000baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full 1000baseT/Half 1000baseT/Full
Advertised auto-negotiation: Yes
Speed: 1000Mb/s
Duplex: Full
Port: Twisted Pair
PHYAD: 1
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: g
Wake-on: g
Current message level: 0x000000ff (255)
Link detected: yes
Offload parameters for eth0:
rx-checksumming: on
tx-checksumming: on
scatter-gather: on
tcp segmentation offload: off
udp fragmentation offload: off
generic segmentation offload: off
NIC statistics:
rx_octets: 5403873
rx_fragments: 0
rx_ucast_packets: 77197
rx_mcast_packets: 0
rx_bcast_packets: 1
rx_fcs_errors: 0
rx_align_errors: 0
rx_xon_pause_rcvd: 0
rx_xoff_pause_rcvd: 0
rx_mac_ctrl_rcvd: 0
rx_xoff_entered: 0
rx_frame_too_long_errors: 0
rx_jabbers: 0
rx_undersize_packets: 0
rx_in_length_errors: 0
rx_out_length_errors: 0
rx_64_or_less_octet_packets: 2
rx_65_to_127_octet_packets: 77196
rx_128_to_255_octet_packets: 0
rx_256_to_511_octet_packets: 0
rx_512_to_1023_octet_packets: 0
rx_1024_to_1522_octet_packets: 0
rx_1523_to_2047_octet_packets: 0
rx_2048_to_4095_octet_packets: 0
rx_4096_to_8191_octet_packets: 0
rx_8192_to_9022_octet_packets: 0
tx_octets: 1000276920
tx_collisions: 0
tx_xon_sent: 0
tx_xoff_sent: 0
tx_flow_control: 0
tx_mac_errors: 0
tx_single_collisions: 0
tx_mult_collisions: 0
tx_deferred: 0
tx_excessive_collisions: 0
tx_late_collisions: 0
tx_collide_2times: 0
tx_collide_3times: 0
tx_collide_4times: 0
tx_collide_5times: 0
tx_collide_6times: 0
tx_collide_7times: 0
tx_collide_8times: 0
tx_collide_9times: 0
tx_collide_10times: 0
tx_collide_11times: 0
tx_collide_12times: 0
tx_collide_13times: 0
tx_collide_14times: 0
tx_collide_15times: 0
tx_ucast_packets: 3488350
tx_mcast_packets: 0
tx_bcast_packets: 0
tx_carrier_sense_errors: 0
tx_discards: 0
tx_errors: 0
dma_writeq_full: 0
dma_write_prioq_full: 0
rxbds_empty: 0
rx_discards: 0
rx_errors: 0
rx_threshold_hit: 11
dma_readq_full: 2188114
dma_read_prioq_full: 162588
tx_comp_queue_full: 0
ring_set_send_prod_index: 2901128
ring_status_update: 218885
nic_irqs: 146494
nic_avoided_irqs: 72391
nic_tx_threshold_hit: 103584Tony Battersby
Cybernetics
Quoted 3 lines · Tony Battersby
I am experiencing network data corruption with a 3Com 3C996B-T NIC (Broadcom NetXtreme BCM5701; driver tg3.ko). I have identified the following patch as the trigger:
Assuming this problem is unique to the 5701, I'm not sure how it is exposed by Herbert's patch. One thing unique on the 5701 is that it double-copies all RX packets so that the data starts at offset 2, but that's quite unrelated to the patch below.
Quoted 11 lines
commit
fb93134dfc2a6e6fbedc7c270a31da03fce88db9Author: Herbert Xu <[email protected]> Date: Wed Nov 14 15:45:21 2007 -0800[TCP]: Fix size calculation in
sk_stream_alloc_pskbI do not get data corruption when substituting a SysKonnect 9D21 NIC (which also uses the tg3.ko driver)
What Broadcom chip is on the Syskonnect card?
David Miller
maintainerFrom: "Michael Chan" <[email protected]>
Date: Mon, 18 Feb 2008 16:32:00 -0800
Quoted 9 lines · Tony Battersby
Quoted 3 lines
I am experiencing network data corruption with a 3Com 3C996B-T NIC (Broadcom NetXtreme BCM5701; driver tg3.ko). I have identified the following patch as the trigger:
Assuming this problem is unique to the 5701, I'm not sure how it is exposed by Herbert's patch. One thing unique on the 5701 is that it double-copies all RX packets so that the data starts at offset 2, but that's quite unrelated to the patch below.
One consequence of Herbert's change is that the chip will see a different datastream. The initial skb->data linear area will be smaller, and the transition to the fragmented area of pages will be quicker.
Quoted 5 lines · David Miller
One consequence of Herbert's change is that the chip will see a different datastream. The initial
skb->datalinear area will be smaller, and the transition to the fragmented area of pages will be quicker.
I see. Perhaps when we get to the end of the data-stream, there is a tiny frag that the chip cannot handle. That's the only thing I can think of.
Please try this patch to see if the problem goes away. This will disable SG on 5701 so we always get linear SKBs.
drivers/net/tg3.c +3 -0
@@ -12717,6 +12717,9 @@ static int __devinit tg3_init_one(struct pci_dev *pdev, }else tp->tg3_flags &= ~TG3_FLAG_RX_CHECKSUMS ; +if (GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 ) + dev->features &= ~(NETIF_F_IP_CSUM |NETIF_F_SG ); +/* flow control autonegotiation is default behavior */ tp->tg3_flags |=TG3_FLAG_PAUSE_AUTONEG ; tp->link_config.flowctrl =TG3_FLOW_CTRL_TX |TG3_FLOW_CTRL_RX ;
Quoted 26 lines · Michael Chan
Quoted 6 lines · David Miller
One consequence of Herbert's change is that the chip will see a different datastream. The initial
skb->datalinear area will be smaller, and the transition to the fragmented area of pages will be quicker.I see. Perhaps when we get to the end of the data-stream, there is a tiny frag that the chip cannot handle. That's the only thing I can think of.
Please try this patch to see if the problem goes away. This will disable SG on 5701 so we always get linear SKBs.
drivers/net/tg3.c +0
-0
drivers/net/tg3.c static int __devinit tg3_init_one(struct pci_dev *pdev, · line 12725
Quoted 9 lines
drivers/net/tg3.c +3
-0@@ -12717,6 +12717,9 @@ static int __devinit tg3_init_one(struct pci_dev *pdev, }else tp->tg3_flags &= ~TG3_FLAG_RX_CHECKSUMS ; +if (GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 ) + dev->features &= ~(NETIF_F_IP_CSUM |NETIF_F_SG ); +/* flow control autonegotiation is default behavior */ tp->tg3_flags |=TG3_FLAG_PAUSE_AUTONEG ; tp->link_config.flowctrl =TG3_FLOW_CTRL_TX |TG3_FLOW_CTRL_RX ;
This patch does appear to fix the data corruption (tested with 2.6.24.2). However, it results in performance problems with the iSCSI application that I am trying to run on this machine.
The test program that I described in the previous message still gets good performance in both directions. "iperf -r" gets good performance in both directions (940 Mbits/s or 117 MB/s). However, my target-mode iSCSI application (which obviously generates rx/tx traffic patterns more complicated than the synthetic tests) gets very poor performance in one direction but good performance in the other direction. iSCSI performance drops to 6 - 15 MB/s when the 3Com NIC is doing heavy rx with light tx, but remains at a decent 115 MB/s when the 3Com NIC is doing heavy tx with light rx. When I revert Herbert's patch instead of applying the patch above, I get 115 MB/s in both cases. (With a stock unpatched kernel, the test fails almost immediately because the iSCSI control PDUs are corrupted, causing the TCP connection to be dropped.)
The SysKonnect NIC that does not exhibit this problem has a chip that says "BCM5411KQM" "TT0128 P2Q" and "56975E".
Tony
Quoted 3 lines · Tony Battersby
iSCSI
performance drops to 6 - 15 MB/s when the 3Com NIC is doing heavy rx with light tx,
That's strange. The patch should only affect TX performance slightly since we are just turning off SG for TX. Please take an ethereal trace to see what's happening and compare with a good trace.
Quoted 10 lines
but remains at a decent 115 MB/s when the 3Com NIC is
doing heavy tx with light rx. When I revert Herbert's patch instead of
applying the patch above, I get 115 MB/s in both cases. (With a stock unpatched kernel, the test fails almost immediately because the iSCSI control PDUs are corrupted, causing the TCP connection to be dropped.)The SysKonnect NIC that does not exhibit this problem has a chip that says "BCM5411KQM" "TT0128 P2Q" and "56975E".
I think this is the 5700, but please send me the tg3 output that identifies the chip and the revision. Something like this:
eth2: Tigon3 [partno(BCM95705) rev 3003 PHY(5705)] (PCI:66MHz:32-bit) 10/100/1000Base-T Ethernet 00:10:18:04:57:0d
eth2: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] WireSpeed[0] TSOcap[1]
Quoted 8 lines · Michael Chan
Quoted 2 lines
The SysKonnect NIC that does not exhibit this problem has a chip that says "BCM5411KQM" "TT0128 P2Q" and "56975E".
I think this is the 5700, but please send me the tg3 output that identifies the chip and the revision. Something like this:
eth2: Tigon3 [partno(BCM95705) rev 3003 PHY(5705)] (PCI:66MHz:32-bit) 10/100/1000Base-T Ethernet 00:10:18:04:57:0d
eth2: RXcsums[1] LinkChgREG[0] MIirq[0] ASF[0] WireSpeed[0] TSOcap[1]
Here is the dmesg output for the SysKonnect NIC:
eth0: Tigon3 [partno(SK-9D21) rev 7104 PHY(5411)] (PCI:66MHz:64-bit) 10/100/1000Base-T Ethernet 00:00:5a:9d:0c:4a
eth0: RXcsums[1] LinkChgREG[1] MIirq[1] ASF[0] WireSpeed[0] TSOcap[0]
eth0: dma_rwctrl[76ff000f] dma_mask[64-bit]
Tony
Quoted 12 lines · Michael Chan
Quoted 4 lines · Tony Battersby
iSCSI
performance drops to 6 - 15 MB/s when the 3Com NIC is doing heavy rx with light tx,That's strange. The patch should only affect TX performance slightly since we are just turning off SG for TX. Please take an ethereal trace to see what's happening and compare with a good trace.
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
I do not regularly use wireshark or look at network packet dumps, so I am not really sure what to look for. Given the above information, do you still believe that there is value in examining the packet dump?
Tony
Quoted 11 lines · Tony Battersby
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
I do not regularly use wireshark or look at network packet dumps, so I am not really sure what to look for. Given the above information, do you still believe that there is value in examining the packet dump?
Can you confirm whether you're getting TCP checksum errors on the other side that is receiving packets from the 5701? You can just check statistics using netstat -s. I suspect that after we turn off SG, checksum is no longer offloaded and we are getting lots of TCP checksum errors instead that are slowing the performance.
Quoted 20 lines · Tony Battersby
Quoted 12 lines · Michael Chan
Quoted 4 lines · Tony Battersby
iSCSI
performance drops to 6 - 15 MB/s when the 3Com NIC is doing heavy rx with light tx,That's strange. The patch should only affect TX performance slightly since we are just turning off SG for TX. Please take an ethereal trace to see what's happening and compare with a good trace.
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
Quoted 5 lines
I do not regularly use wireshark or look at network packet dumps, so I am not really sure what to look for. Given the above information, do you still believe that there is value in examining the packet dump?
Tony
Hi Tony. Can you give us the output of :
sudo lspci -vvv -xxxx -s 03:01.0'
(assuming that is still the correct address of the 3Com NIC.)
Also, after some digging, I found that the 5701 can run into trouble if a 64-bit DMA read terminates early and then completes as a 32-bit transfer. The problem is reportedly very rare, but the failure mode looks like a match. Can you apply the following patch and see if it helps your performance / corruption problems?
drivers/net/tg3.c +2 -0
@@ -11409,6 +11409,8 @@ static int __devinit tg3_get_invariants(struct tg3 *tp) tp->tg3_flags |=TG3_FLAG_PCI_HIGH_SPEED ;if ((pci_state_reg &PCISTATE_BUS_32BIT ) !=0 ) tp->tg3_flags |=TG3_FLAG_PCI_32BIT ; +else if (GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 ) + tp->grc_mode |=GRC_MODE_FORCE_PCI32BIT ;/* Chip-specific fixup from Broadcom driver */ if ((tp->pci_chip_rev_id == CHIPREV_ID_5704_A0) &&
Quoted 6 lines · Tony Battersby
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
Interesting. So the workload that regressed is mostly RX with a little TX traffic? Can you try to reproduce this with something like netperf to eliminate other variables?
This is all very puzzling since the patch in question shouldn't change an RX load at all.
Thanks,
Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txtQuoted 15 lines · Michael Chan
Quoted 11 lines · Tony Battersby
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
I do not regularly use wireshark or look at network packet dumps, so I am not really sure what to look for. Given the above information, do you still believe that there is value in examining the packet dump?
Quoted 8 lines
Can you confirm whether you're getting TCP checksum errors on the other side that is receiving packets from the 5701? You can just check statistics using netstat -s. I suspect that after we turn off SG, checksum is no longer offloaded and we are getting lots of TCP checksum errors instead that are slowing the performance.
Confirmed. With a 100 MB read/write test, netstat -s shows 75 bad segments received, and performance in the one direction is about 5 MB/s. When I switch to the SysKonnect NIC, netstat -s shows 0 bad segments received, and performance is 115 MB/s. So that solves that mystery - there is still data corruption, but the software-computed TCP checksum causes the bad packets to be retransmitted rather than being passed on to the application.
Tony
Quoted 9 lines · Herbert Xu
Quoted 6 lines · Tony Battersby
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
Quoted 9 lines
Interesting. So the workload that regressed is mostly RX with a little TX traffic? Can you try to reproduce this with something like netperf to eliminate other variables?
This is all very puzzling since the patch in question shouldn't change an RX load at all.
Thanks,
We have established that the slowdown was caused by TCP checksum errors and retransmits. I assume that the slowdown in my test was due to the light TX rather than the heavy RX. I am no TCP protocol expert, but perhaps heavy TX (such as iperf) might not be affected as much because the wire stays busy while waiting for the retransmit, whereas with my light TX iSCSI load, the wire goes idle while waiting for the retransmit because the iSCSI state machine is stalled.
Tony
Quoted 4 lines · Matt Carlson
Hi Tony. Can you give us the output of :
sudo lspci -vvv -xxxx -s 03:01.0'
03:01.0 Ethernet controller: Broadcom Corporation NetXtreme BCM5701 Gigabit Ethernet (rev 15)
Subsystem: Compaq Computer Corporation NC7770 Gigabit Server Adapter (PCI-X, 10/100/1000-T)
Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 64 (16000ns min), Cache Line Size: 64 bytes
Interrupt: pin A routed to IRQ 17
Region 0: Memory at df7f0000 (64-bit, non-prefetchable) [size=64K]
[virtual] Expansion ROM at dfc00000 [disabled] [size=64K]
Capabilities: [40] PCI-X non-bridge device
Command: DPERE- ERO- RBC=512 OST=1
Status: Dev=03:01.1 64bit+ 133MHz+ SCD- USC- DC=simple DMMRBC=512 DMOST=1 DMCRS=8 RSCEM- 266MHz- 533MHz-
Capabilities: [48] Power Management version 2
Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot+,D3cold+)
Status: D0 PME-Enable- DSel=0 DScale=1 PME-
Capabilities: [50] Vital Product Data <?>
Capabilities: [58] Message Signalled Interrupts: Mask- 64bit+ Queue=0/3 Enable-
Address: 063000119b608000 Data: 0423
00: e4 14 45 16 06 00 b0 02 15 00 00 02 10 40 00 00
10: 04 00 7f df 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 11 0e 7c 00
30: 00 00 00 00 40 00 00 00 00 00 00 00 0b 01 40 00
40: 07 48 00 00 09 03 03 00 01 50 02 c0 00 20 00 64
50: 03 58 00 00 08 10 21 08 05 00 86 00 00 80 60 9b
60: 11 00 30 06 23 04 00 00 98 02 05 01 0f 00 db 76
70: 8a 10 00 00 c7 00 00 80 50 00 00 00 00 00 00 00
80: 03 58 00 00 00 00 00 00 34 80 13 04 82 10 00 00
90: 09 06 00 01 00 00 00 00 00 00 00 00 c6 01 00 00
a0: 00 00 00 00 fe 02 00 00 00 00 00 00 af 01 00 00
b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
f0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
Quoted 14 lines
Also, after some digging, I found that the 5701 can run into trouble if a 64-bit DMA read terminates early and then completes as a 32-bit transfer. The problem is reportedly very rare, but the failure mode looks like a match. Can you apply the following patch and see if it helps your performance / corruption problems?
drivers/net/tg3.c +0
-0
drivers/net/tg3.c static int __devinit tg3_get_invariants(struct tg3 *tp) · line 11416
Quoted 8 lines
drivers/net/tg3.c +2
-0@@ -11409,6 +11409,8 @@ static int __devinit tg3_get_invariants(struct tg3 *tp) tp->tg3_flags |=TG3_FLAG_PCI_HIGH_SPEED ;if ((pci_state_reg &PCISTATE_BUS_32BIT ) !=0 ) tp->tg3_flags |=TG3_FLAG_PCI_32BIT ; +else if (GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 ) + tp->grc_mode |=GRC_MODE_FORCE_PCI32BIT ;/* Chip-specific fixup from Broadcom driver */ if ((tp->pci_chip_rev_id == CHIPREV_ID_5704_A0) &&
Sorry, this didn't help. I still get data corruption with hardware checksumming or poor performance with software checksumming.
Tony
Update:
Herbert's patch alters the arguments to alloc_skb_fclone() and skb_reserve() from within sk_stream_alloc_pskb(). This changes the skb_headroom() and skb_tailroom() of the returned skb. I decided to see if I could detect the precise point at which data corruption started to happen. The result is this table:
(sk_stream_alloc_pskb() called with size == 1448;sk->sk_prot->max_header == 160)
skb_headroom skb_tailroom test result note 216 1448 fail [1] 344 1448 fail 340 1452 pass 336 1456 pass 332 1460 pass 328 1464 fail 324 1468 pass 320 1472 pass 316 1476 pass 312 1480 fail 308 1484 pass 304 1488 pass 300 1492 pass 296 1496 fail 292 1500 pass 288 1504 pass 284 1508 pass 280 1512 fail 276 1516 pass 272 1520 pass 268 1524 pass 264 1528 fail 260 1532 pass 256 1536 pass [2]
Notes:
[1] Kernels 2.6.23.4 - 2.6.23.16 and 2.6.24 - current with Herbert's patch
[2] Kernels 2.6.23.3 and before without Herbert's patch
Note that the first row has skb_headroom + skb_tailroom == 1664; the remaining rows have skb_headroom + skb_tailroom == 1792.
Quoted 1 line
From these results, it looks like a data alignment issue. Herbert's
patch unfortunately just happened to change the alignment in a way that made it break.
Tony
The following patch fixes the problem for me. Do we want to accept this patch and call it a day or continue investigating the source of the problem?
Patch applies to 2.6.24.2, but doesn't apply to 2.6.25-rc. If everyone agrees that this is the right solution, I will resubmit with a proper subject line and description.
Tony
include/net/sock.h renamed from include/net/sock.h.orig +4 -2
@@ -1236,8 +1236,10 @@ static inline struct sk_buff *sk_stream_ {struct sk_buff *skb; -/* The TCP header must be at least 32-bit aligned. */ - size =ALIGN (size,4 ); +/* The TCP header must be at least 32-bit aligned, but some chipsets +* such as Broadcom BCM5701 require at least 16-byte alignment. +*/ + size =ALIGN (size,16 ); skb =alloc_skb_fclone (size + sk->sk_prot->max_header, gfp);if (skb) {
David Miller
maintainerFrom: Tony Battersby <[email protected]>
Date: Wed, 20 Feb 2008 18:04:09 -0500
Quoted 6 lines
The following patch fixes the problem for me. Do we want to accept this patch and call it a day or continue investigating the source of the problem?
Patch applies to 2.6.24.2, but doesn't apply to 2.6.25-rc. If everyone agrees that this is the right solution, I will resubmit with a proper subject line and description.
A chipset bug, if it even exists, should be worked around in the driver for that hardware. We shouldn't make every other piece of hardware in the world suffer too.
Quoted 6 lines · David Miller
From: Tony Battersby <[email protected]>
Date: Wed, 20 Feb 2008 18:04:09 -0500Quoted 3 lines
The following patch fixes the problem for me. Do we want to accept this patch and call it a day or continue investigating the source of the problem?
Quoted 9 lines
Quoted 3 lines
Patch applies to 2.6.24.2, but doesn't apply to 2.6.25-rc. If everyone agrees that this is the right solution, I will resubmit with a proper subject line and description.
A chipset bug, if it even exists, should be worked around in the driver for that hardware. We shouldn't make every other piece of hardware in the world suffer too.
Yes, we should workaround this in the TG3 driver once we understand what the problem is and how to workaround it. We are still looking through the errata list to sort this out. It looks like it is the starting DMA address of the TX buffer that is causing the problem.
Hi Tony. Sorry for the radio silence.
Michael and I have discussed this problem a bit. Another possibility is that the chip may be having difficulty with non-dword aligned TX buffers. Since we already know the RX side has the same problem, it isn't so far-fetched to think that perhaps it can affect the TX side too. Can you give the following patch a try and see if the corruption still happens?
drivers/net/tg3.c +14 -1
@@ -4135,11 +4135,20 @@ static int tigon3_dma_hwbug_workaround(struct tg3 *tp, struct sk_buff *skb, u32 last_plus_one, u32 *start, u32 base_flags, u32 mss) { - struct sk_buff *new_skb = skb_copy(skb, GFP_ATOMIC); + struct sk_buff *new_skb;dma_addr_t new_addr =0 ;u32 entry = *start;int i, ret =0 ; +if (GET_ASIC_REV (tp->pci_chip_rev_id) !=ASIC_REV_5701 ) + new_skb =skb_copy (skb,GFP_ATOMIC ); +else { +int more_headroom =4 - (skb->mac_header &3 ); + + new_skb =skb_copy_expand (skb,skb_headroom (skb) + more_headroom, +skb_tailroom (skb),GFP_ATOMIC ); + } +if (!new_skb) { ret =-1 ; }else {
@@ -4465,6 +4474,10 @@ static int tg3_start_xmit_dma_bug(struct sk_buff *skb, struct net_device *dev)if (tg3_4g_overflow_test (mapping, len)) would_hit_hwbug =1 ; +/* Force the 5701 into the double copy path. */ +if (GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 ) + would_hit_hwbug =1 ; +tg3_set_txd (tp, entry, mapping, len, base_flags, (skb_shinfo (skb)->nr_frags ==0 ) | (mss <<1 ));
Quoted 33 lines · Tony Battersby
Quoted 9 lines · Herbert Xu
Quoted 6 lines · Tony Battersby
Update: when I revert Herbert's patch in addition to applying your patch, the iSCSI performance goes back up to 115 MB/s again in both directions. So it looks like turning off SG for TX didn't itself cause the performance drop, but rather that the performance drop is just another manifestation of whatever bug is causing the data corruption.
Quoted 9 lines
Interesting. So the workload that regressed is mostly RX with a little TX traffic? Can you try to reproduce this with something like netperf to eliminate other variables?
This is all very puzzling since the patch in question shouldn't change an RX load at all.
Thanks,
We have established that the slowdown was caused by TCP checksum errors and retransmits. I assume that the slowdown in my test was due to the light TX rather than the heavy RX. I am no TCP protocol expert, but perhaps heavy TX (such as iperf) might not be affected as much because the wire stays busy while waiting for the retransmit, whereas with my light TX iSCSI load, the wire goes idle while waiting for the retransmit because the iSCSI state machine is stalled.
Tony
--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Tony Battersby
Tested-byQuoted 10 lines · Matt Carlson
Hi Tony. Sorry for the radio silence.
Michael and I have discussed this problem a bit. Another possibility is that the chip may be having difficulty with non-dword aligned TX buffers. Since we already know the RX side has the same problem, it isn't so far-fetched to think that perhaps it can affect the TX side too. Can you give the following patch a try and see if the corruption still happens?
Thanks, your patch fixes the problem (tested on 2.6.24.4). However, I had to change "(skb->mac_header & 3)" in your patch to "((long) skb->mac_header & 3)" since mac_header is a pointer rather than an int on 32-bit systems.
David Miller
maintainerFrom: Tony Battersby <[email protected]>
Date: Tue, 15 Apr 2008 11:39:27 -0400
Quoted 6 lines
Thanks, your patch fixes the problem (tested on 2.6.24.4). However, I had to change "(
skb->mac_header& 3)" in your patch to "((long)skb->mac_header& 3)" sincemac_headeris a pointer rather than an int on 32-bit systems.Tested-by Tony Battersby <[email protected]>
Thanks for testing.
Matt, skb->mac_header is either a pointer or an integer offset depending upon whether we are building 32-bit or 64-bit.
Testing skb->mac_header is therefore wrong, because it's an offset from a pointer in the 64-bit case and therefore it's alignment does not indicate correctly the actual final alignment of skb->head + skb->max_header.
Therefore you should test skb_mac_header(skb) and cast it with (unsigned long).
Please respin this fix with that correction so I can apply it and get this bug fixed, thanks!
Quoted 10 lines · David Miller
Matt,
skb->mac_headeris either a pointer or an integer offset depending upon whether we are building 32-bit or 64-bit.Testing
skb->mac_headeris therefore wrong, because it's an offset from a pointer in the 64-bit case and therefore it's alignment does not indicate correctly the actual final alignment ofskb->head+skb->max_header.Therefore you should test
skb_mac_header(skb)and cast it with (unsigned long).
Isn't it better to test for skb->data? That's where we tell the hardware to start transmitting.
Quoted 5 lines
Please respin this fix with that correction so I can apply it and get this bug fixed, thanks!
We think that this problem is unique in Tony's environment because of the PCIE-to-PCI bridge that he is using. We therefore want to test for that bridge and apply the workaround only when it's present. We've never seen this problem in the last 6 or 7 years during the lifetime of the 5701.
We'll try to get this done ASAP.
Thanks.
Quoted 28 lines · Michael Chan
Quoted 10 lines · David Miller
Matt,
skb->mac_headeris either a pointer or an integer offset depending upon whether we are building 32-bit or 64-bit.Testing
skb->mac_headeris therefore wrong, because it's an offset from a pointer in the 64-bit case and therefore it's alignment does not indicate correctly the actual final alignment ofskb->head+skb->max_header.Therefore you should test
skb_mac_header(skb)and cast it with (unsigned long).Isn't it better to test for
skb->data? That's where we tell the hardware to start transmitting.Quoted 5 lines
Please respin this fix with that correction so I can apply it and get this bug fixed, thanks!
We think that this problem is unique in Tony's environment because of the PCIE-to-PCI bridge that he is using. We therefore want to test for that bridge and apply the workaround only when it's present. We've never seen this problem in the last 6 or 7 years during the lifetime of the 5701.
Quoted 3 lines
We'll try to get this done ASAP.
Thanks.
Tony,
Below is a patch that attempts to limit the workaround to the bridge you have on your system. Can you test it and verify that the workaround is still enabled?
drivers/net/tg3.c +49 -2
@@ -4135,11 +4135,21 @@ static int tigon3_dma_hwbug_workaround(struct tg3 *tp, struct sk_buff *skb, u32 last_plus_one, u32 *start, u32 base_flags, u32 mss) { - struct sk_buff *new_skb = skb_copy(skb, GFP_ATOMIC); + struct sk_buff *new_skb;dma_addr_t new_addr =0 ;u32 entry = *start;int i, ret =0 ; +if (GET_ASIC_REV (tp->pci_chip_rev_id) !=ASIC_REV_5701 ) + new_skb =skb_copy (skb,GFP_ATOMIC ); +else { +int more_headroom =4 - ((unsigned long )skb->data &3 ); + + new_skb =skb_copy_expand (skb, +skb_headroom (skb) + more_headroom, +skb_tailroom (skb),GFP_ATOMIC ); + } +if (!new_skb) { ret =-1 ; }else {
@@ -4462,7 +4472,9 @@ static int tg3_start_xmit_dma_bug(struct sk_buff *skb, struct net_device *dev) would_hit_hwbug =0 ; -if (tg3_4g_overflow_test (mapping, len)) +if (tp->tg3_flags3 &TG3_FLG3_5701_DMA_BUG ) + would_hit_hwbug =1 ; +else if (tg3_4g_overflow_test (mapping, len)) would_hit_hwbug =1 ; tg3_set_txd(tp, entry, mapping, len, base_flags,
@@ -11339,6 +11351,41 @@ static int __devinit tg3_get_invariants(struct tg3 *tp) } } +if ((GET_ASIC_REV (tp->pci_chip_rev_id) ==ASIC_REV_5701 )) { +static struct tg3_dev_id { +u32 vendor; +u32 device; + } bridge_chipsets[] = { + {PCI_VENDOR_ID_INTEL ,PCI_DEVICE_ID_INTEL_PXH_0 }, + {PCI_VENDOR_ID_INTEL ,PCI_DEVICE_ID_INTEL_PXH_1 }, + { }, + }; +struct tg3_dev_id *pci_id = &bridge_chipsets[0 ]; +struct pci_dev *bridge =NULL ; + +while (pci_id->vendor !=0 && + !(tp->tg3_flags3 &TG3_FLG3_5701_DMA_BUG )) { +while (1 ) { + bridge =pci_get_device (pci_id->vendor, + pci_id->device, + bridge); +if (!bridge) { + pci_id++; +break ; + } +if (bridge->subordinate && + (bridge->subordinate->number <= + tp->pdev->bus->number) && + (bridge->subordinate->subordinate >= + tp->pdev->bus->number)) { + tp->tg3_flags3 |=TG3_FLG3_5701_DMA_BUG ; +pci_dev_put (bridge); +break ; + } + } + } + } +/* The EPB bridge inside 5714, 5715, and 5780 cannot support * DMA addresses > 40-bit. This bridge may have other additional * 57xx devices behind it in some 4-port NIC designs for example.
drivers/net/tg3.h +1 -0
@@ -2476,6 +2476,7 @@ struct tg3 {#define TG3_FLG3_NO_NVRAM_ADDR_TRANS 0x00000001#define TG3_FLG3_ENABLE_APE 0x00000002#define TG3_FLG3_5761_5784_AX_FIXES 0x00000004 +#define TG3_FLG3_5701_DMA_BUG 0x00000008struct timer_list timer;u16 timer_counter;
Tony Battersby
Tested-byQuoted 33 lines · Matt Carlson
Quoted 31 lines · Michael Chan
Quoted 11 lines · David Miller
Matt,
skb->mac_headeris either a pointer or an integer offset depending upon whether we are building 32-bit or 64-bit.Testing
skb->mac_headeris therefore wrong, because it's an offset from a pointer in the 64-bit case and therefore it's alignment does not indicate correctly the actual final alignment ofskb->head+skb->max_header.Therefore you should test
skb_mac_header(skb)and cast it with (unsigned long).Isn't it better to test for
skb->data? That's where we tell the hardware to start transmitting.Quoted 5 lines
Please respin this fix with that correction so I can apply it and get this bug fixed, thanks!
We think that this problem is unique in Tony's environment because of the PCIE-to-PCI bridge that he is using. We therefore want to test for that bridge and apply the workaround only when it's present. We've never seen this problem in the last 6 or 7 years during the lifetime of the 5701.
We'll try to get this done ASAP.
Quoted 11 lines
Quoted 2 lines
Thanks.
Tony,
Below is a patch that attempts to limit the workaround to the bridge you have on your system. Can you test it and verify that the workaround is still enabled?
This new patch also passes the test. Thumbs up!
Tony
David Miller
maintainerFrom: "Michael Chan" <[email protected]>
Date: Wed, 16 Apr 2008 08:40:25 -0700
Quoted 2 lines
Isn't it better to test for
skb->data? That's where we tell the hardware to start transmitting.
That's correct.