usrp-users@lists.ettus.com
Discussion and technical support related to USRP, UHD, RFNoC
View all threadsUSRP x300 10GbE Sequence Error Issue
Hi everyone, I am having an issue with TX sequence errors on my USRP x300. I have already read the instructions (shocking, isn't it??) from the documentation here: http://files.ettus.com/manual/page_usrp_x3x0_config.html#x3x0cfg_hostpc_netcfg_ip
This is my hardware specs:
Intel quad-core Haswell CPU.
USRP x300.
The same Intel 10GbE NIC that Ettus recommends on the website (using Port 1, if that makes any difference, so default IP of 192.168.40.1).
The Intel NIC is connected to the motherboard's PCI-E 3.0 x16 lane.
Software settings (Per the instructions from the documentation):
Linux kernel 3.12.26 (higher than 3.11, per recommendation).
The rmem_max and wmem_max are both set to 33554432.
Set the MTU on the NIC to 9000 since I am using the 10GbE card
Firewall settings are permitted for USRP-related traffic
Now, I am transmitting a waveform at sampling rate of 100 MHz. Looking at top, my CPU resources never goes over 30%. However, I get an "S" (sequence error on TX) very frequently. HOWEVER, if I set the MTU of the NIC to 1500 (the default MTU), then the sequence errors goes away. I don't understand why this is happening. Is there any particular setting or "gotchas" that I need to do to remove the sequence errors when I set my NIC's MTU to 9000?
Regards,
Joe
Hi Joe,
If changing the MTU size changes behavior and prevents errors, then I think
you have your answer. Using jumbo frames must be causing packets to either
be dropped or reordered. When the sequence errors are detected, do you see
a single S at a time or are there bursts of 3? I ask because a single
packet being dropped will yield a single S, but reordering of packets would
yield a burst of 3 (one for the skip forward, one for the step back, and
one for the skip forward to resume the sequence).
Regards,
Michael
On Tue, Feb 10, 2015 at 10:48 PM, Joe Leech via USRP-users <
usrp-users@lists.ettus.com> wrote:
Hi everyone, I am having an issue with TX sequence errors on my USRP x300.
I have already read the instructions (shocking, isn't it??) from the
documentation here:
http://files.ettus.com/manual/page_usrp_x3x0_config.html#x3x0cfg_hostpc_netcfg_ip
This is my hardware specs:
Intel quad-core Haswell CPU.
USRP x300.
The same Intel 10GbE NIC that Ettus recommends on the website (using Port
1, if that makes any difference, so default IP of 192.168.40.1).
The Intel NIC is connected to the motherboard's PCI-E 3.0 x16 lane.
Software settings (Per the instructions from the documentation):
Linux kernel 3.12.26 (higher than 3.11, per recommendation).
The rmem_max and wmem_max are both set to 33554432.
Set the MTU on the NIC to 9000 since I am using the 10GbE card
Firewall settings are permitted for USRP-related traffic
Now, I am transmitting a waveform at sampling rate of 100 MHz. Looking at
top, my CPU resources never goes over 30%. However, I get an "S" (sequence
error on TX) very frequently. HOWEVER, if I set the MTU of the NIC to 1500
(the default MTU), then the sequence errors goes away. I don't understand
why this is happening. Is there any particular setting or "gotchas" that I
need to do to remove the sequence errors when I set my NIC's MTU to 9000?
Regards,
Joe
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
Hi Michael,
Thanks for the response! I believe I see single "S", although sometimes I
get them quite frequently since my waveform is 1ms so it's difficult to
tell if I can visually distinguish a single S vs. 3S's. FYI, I queried
the RX metadata and nothing was out of sequence;
does reordering of packets cause out of sequence flag to be toggled?
Because that flag never toggles for me so I've been assuming the packets
came in the correct order (dangerous assumption, I know!).
Regards,
Joe
Date: Wed, 11 Feb 2015 12:32:18 -0800
Subject: Re: [USRP-users] USRP x300 10GbE Sequence Error Issue
From: michael.west@ettus.com
To: think4das11@outlook.com
CC: usrp-users@lists.ettus.com
Hi Joe,
If changing the MTU size changes behavior and prevents errors, then I think you have your answer. Using jumbo frames must be causing packets to either be dropped or reordered. When the sequence errors are detected, do you see a single S at a time or are there bursts of 3? I ask because a single packet being dropped will yield a single S, but reordering of packets would yield a burst of 3 (one for the skip forward, one for the step back, and one for the skip forward to resume the sequence).
Regards,Michael
On Tue, Feb 10, 2015 at 10:48 PM, Joe Leech via USRP-users usrp-users@lists.ettus.com wrote:
Hi everyone, I am having an issue with TX sequence errors on my USRP x300. I have already read the instructions (shocking, isn't it??) from the documentation here: http://files.ettus.com/manual/page_usrp_x3x0_config.html#x3x0cfg_hostpc_netcfg_ip
This is my hardware specs:
Intel quad-core Haswell CPU.
USRP x300.
The same Intel 10GbE NIC that Ettus recommends on the website (using Port 1, if that makes any difference, so default IP of 192.168.40.1).
The Intel NIC is connected to the motherboard's PCI-E 3.0 x16 lane.
Software settings (Per the instructions from the documentation):
Linux kernel 3.12.26 (higher than 3.11, per recommendation).
The rmem_max and wmem_max are both set to 33554432.
Set the MTU on the NIC to 9000 since I am using the 10GbE card
Firewall settings are permitted for USRP-related traffic
Now, I am transmitting a waveform at sampling rate of 100 MHz. Looking at top, my CPU resources never goes over 30%. However, I get an "S" (sequence error on TX) very frequently. HOWEVER, if I set the MTU of the NIC to 1500 (the default MTU), then the sequence errors goes away. I don't understand why this is happening. Is there any particular setting or "gotchas" that I need to do to remove the sequence errors when I set my NIC's MTU to 9000?
Regards,
Joe
USRP-users mailing list
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
Hi Joe,
Dropped or reordered packets on the TX side result in an 'S' (sequence
error), where dropped or reordered packets on the RX side result in a 'D'
(dropped packet). If you are only seeing S's, then your problem is limited
to the TX side. It sounds like you are experiencing dropped TX packets
when using jumbo frames. This could be happening anywhere in the network
stack between the host and USRP and would take significant time and effort
to track down. But if you can successfully send and receive using an MTU
size of 1500, I would say to just go with that.
Regards,
Michael
On Wed, Feb 11, 2015 at 3:55 PM, Joe Leech think4das11@outlook.com wrote:
Hi Michael,
Thanks for the response! I believe I see single "S", although sometimes I
get them quite frequently since my waveform is 1ms so it's difficult to
tell if I can visually distinguish a single S vs. 3S's. FYI, I queried the
RX metadata and nothing was out of sequence; does reordering of packets
cause out of sequence flag to be toggled? Because that flag never toggles
for me so I've been assuming the packets came in the correct order
(dangerous assumption, I know!).
Regards,
Joe
Date: Wed, 11 Feb 2015 12:32:18 -0800
Subject: Re: [USRP-users] USRP x300 10GbE Sequence Error Issue
From: michael.west@ettus.com
To: think4das11@outlook.com
CC: usrp-users@lists.ettus.com
Hi Joe,
If changing the MTU size changes behavior and prevents errors, then I
think you have your answer. Using jumbo frames must be causing packets to
either be dropped or reordered. When the sequence errors are detected, do
you see a single S at a time or are there bursts of 3? I ask because a
single packet being dropped will yield a single S, but reordering of
packets would yield a burst of 3 (one for the skip forward, one for the
step back, and one for the skip forward to resume the sequence).
Regards,
Michael
On Tue, Feb 10, 2015 at 10:48 PM, Joe Leech via USRP-users <
usrp-users@lists.ettus.com> wrote:
Hi everyone, I am having an issue with TX sequence errors on my USRP x300.
I have already read the instructions (shocking, isn't it??) from the
documentation here:
http://files.ettus.com/manual/page_usrp_x3x0_config.html#x3x0cfg_hostpc_netcfg_ip
This is my hardware specs:
Intel quad-core Haswell CPU.
USRP x300.
The same Intel 10GbE NIC that Ettus recommends on the website (using Port
1, if that makes any difference, so default IP of 192.168.40.1).
The Intel NIC is connected to the motherboard's PCI-E 3.0 x16 lane.
Software settings (Per the instructions from the documentation):
Linux kernel 3.12.26 (higher than 3.11, per recommendation).
The rmem_max and wmem_max are both set to 33554432.
Set the MTU on the NIC to 9000 since I am using the 10GbE card
Firewall settings are permitted for USRP-related traffic
Now, I am transmitting a waveform at sampling rate of 100 MHz. Looking at
top, my CPU resources never goes over 30%. However, I get an "S" (sequence
error on TX) very frequently. HOWEVER, if I set the MTU of the NIC to 1500
(the default MTU), then the sequence errors goes away. I don't understand
why this is happening. Is there any particular setting or "gotchas" that I
need to do to remove the sequence errors when I set my NIC's MTU to 9000?
Regards,
Joe
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com