|
From: Yan, L. (N. - CN/Hangzhou) <lim...@ns...> - 2013-03-11 02:32:15
|
Hi, John
Thanks. The buffer in my own sending side shall be linear. I allocated a skb buffer and reserved enough space(but not more than 1472) and copy my data into skb->data. I'm not sure if the non-linear is caused by kernel sending side or kernel receiving side, or caused by driver sending side or driver receiving side? I'll test more.
---------------------------------------------------------------------
DMXMSG Trouble Shooting Guide:
https://sharenet-ims.inside.nokiasiemensnetworks.com/Overview/D426954268
---- ☺ ----
Best Regards.
Liming
-----Original Message-----
From: ext Ronciak, John [mailto:joh...@in...]
Sent: Saturday, March 09, 2013 7:23 AM
To: Yan, Liming (NSN - CN/Hangzhou); e10...@li...
Subject: RE: [E1000-devel] my skb not linear when whole packet len only 513 bytes
Another point is that you could try to issue a call to skb_linearize prior to sending the skb down to the driver.
The 512 byte issue sounds like buffer size being used but as I said I don't know what's sending the packets down the stack.
Cheers,
John
> -----Original Message-----
> From: Ronciak, John [mailto:joh...@in...]
> Sent: Friday, March 08, 2013 1:50 PM
> To: Yan, Liming (NSN - CN/Hangzhou); e10...@li...
> Subject: Re: [E1000-devel] my skb not linear when whole packet len only
> 513 bytes
>
> The buffer containing the data is not in a single buffer coming down to
> the driver. I'm not sure what kind of stack you supplying the data but
> maybe you should try to have all the data in a single larger buffer
> that is then used in the skb.
>
> Cheers,
> John
>
>
> > -----Original Message-----
> > From: Yan, Liming (NSN - CN/Hangzhou) [mailto:lim...@ns...]
> > Sent: Friday, March 08, 2013 1:20 AM
> > To: e10...@li...
> > Subject: [E1000-devel] my skb not linear when whole packet len only
> > 513 bytes
> >
> > Hi,
> > I'm sending non-IP packets with dev_queue_xmit. It's not linear as
> > expected in receiving side when whole packet len only 513
> > bytes(include mac header). 513 bytes is a boundary because if I use
> 512, it's ok.
> > Anybody know what may cause this problem? Thanks. My MTU is 1500.
> >
> > In sending side, I print the skb, seems it's normal:
> > >> Mar 9 22:41:03.972683 info CLA-0 kernel: [251326.875463]: before
> > dev_queue_xmit: skb->data_len=0, skb->len=513.
> >
> > In receiving side, I register my softirq handler with own ether-type.
> > The skb I got from kernel is:
> > >> Mar 9 22:41:03.973199 info CLA-1 kernel: [11139.235550]: skb-
> > >data_len=1,skb->len=499.
> > So skb_is_nonlinear return true when skb->data_len>0.
> >
> > Kernel version: 2.6.34.13
> >
> > 82599 firmware and driver version:
> > # ethtool -i bi0
> > driver: igb
> > version: 2.4.12
> > firmware-version: 2.1-3
> > bus-info: 0000:02:00.0
> >
> > # ifconfig bi0
> > bi0 Link encap:Ethernet HWaddr 00:A0:A5:64:DB:FA
> > UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
> > RX packets:46846508 errors:0 dropped:0 overruns:0 frame:0
> > TX packets:47748723 errors:0 dropped:0 overruns:0 carrier:0
> > collisions:0 txqueuelen:1000
> > RX bytes:5803096993 (5.4 GiB) TX bytes:55743743072 (51.9
> > GiB)
> >
> > ---------------------------------------------------------------------
> > Best Regards.
> > Liming
>
>
> -----------------------------------------------------------------------
> -------
> Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester
> Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the
> endpoint security space. For insight on selecting the right partner to
> tackle endpoint security challenges, access the full report.
> http://p.sf.net/sfu/symantec-dev2dev
> _______________________________________________
> E1000-devel mailing list
> E10...@li...
> https://lists.sourceforge.net/lists/listinfo/e1000-devel
> To learn more about Intel® Ethernet, visit
> http://communities.intel.com/community/wired
|