Date: Mon, 9 Apr 2001 11:09:50 -0500 From: Joe Thomas <joe_thomas@cnt.com> To: 'Alex Huppenthal' <alex@aspenworks.com>, "Archinaco, Chris" <Chris.Archinaco@marconi.com> Cc: freebsd-atm@FreeBSD.ORG Subject: RE: SR100069960 - Linux resources Message-ID: <67DB78CE91D7D41190A100508BFDF5E051EF1A@esply03.cnt.com>
next in thread | raw e-mail | index | archive | help
Well, having a MTU of size 48 is going to be the cause of the fragmentation problem. I don't know why the interface would come up with that value since it's supposed to be set at 9180. [This is done in the sys/netatm/atm_if.c module when the IP interfaces are defined and associated with the physical ATM interface (such as hfa0 for a FORE adapter).] Yeah, it can't be set with the ifconfig command because the SIOCSIFMTU case isn't supported. As I recall, we chose to not allow it because of the difficulty in propagating the MTU everywhere. With UNI, the MTU size is also used to define forward and backward packet sizes in UNI SETUP messages so that the ATM network can ensure there's enough buffering available. If one changed the MTU size on the fly, you would be expected to close down any open VCs across that interface and reopen them with the new value. This really confuses me as to how/why the MTU is being set to 48. Can you check for ATM_NIF_MTU in sys/netatm/atm_if.*? It should be defined to 9180 in the .h file and you should see a line like atm_if.c: ifp->if_mtu = ATM_NIF_MTU; in the .c file. The other thing that might be interesting to try is to bump up the number of IP/ATM interfaces in your rc.conf file. Somewhere there'll be a line similar to 'atm_netif_hfa0="atm 1"' where hfa0 is the physical ATM interface (hea0 for Efficient), 'atm' is the name to be used for the IP interface (what'll see with ifconfig), and 1 is the number of logical IP interfaces to create. In this example, one IP interface named 'atm0' will be created. Increment this by one or two and then reboot. Right after the system comes up, check the MTU sizes with 'ifconfig -a' again. If the new interfaces have the correct MTU size (9180), then something's happening to change the MTU after the system is up. If the new interfaces have the MTU size of 48, then the value being set when the interface is created is wrong. This might at least give folks an idea of where to start looking... Joe -----Original Message----- From: Alex Huppenthal [mailto:alex@aspenworks.com] Sent: Friday, April 06, 2001 7:37 PM To: Joe Thomas; Archinaco, Chris Cc: freebsd-atm@FreeBSD.ORG Subject: Re: SR100069960 - Linux resources MTU shows up as 48. Can't change the MTU with ifconfig -Alex Thanks to everyone for the insight and information today. ----- Original Message ----- From: "Joe Thomas" <joe_thomas@cnt.com> To: "'Alex Huppenthal'" <alex@aspenworks.com>; "Archinaco, Chris" <Chris.Archinaco@marconi.com> Cc: <freebsd-atm@FreeBSD.ORG> Sent: Friday, April 06, 2001 10:18 AM Subject: RE: SR100069960 - Linux resources > > The microcode was embedded as a .h structure sometime in mid/late > FBSD 3.x so no more having to download the "correct" version. To > the best of my knowledge, the microcode used has not changed since > we released the driver. Any improvement you see in your problem by > backing down to FBSD 3.5 will be because of IP changes between 3.5 > and 4.X, not from microcode or device driver changes. [The problems > in the device driver between 3.5/4.0 and 4.1+ were because of mbuf > structure changes which caused some problems with how the driver > reserves space for passing tokens/etc. with the buffers, and with > a corresponding change that caused a problem with how we used some > space in the pkthdr structure where our assumptions about who and > when those fields were used were no longer valid. The driver doesn't > and never did try to do IP fragmentation. Someone mentioned examining > the MTU size. This would be on the IPATM interface -- check with > 'ifconfig -a'.] > > Joe > > -----Original Message----- > From: Alex Huppenthal [mailto:alex@aspenworks.com] > Sent: Friday, April 06, 2001 11:07 AM > To: Joe Thomas; Archinaco, Chris > Cc: freebsd-atm@FreeBSD.ORG > Subject: Re: SR100069960 - Linux resources > > > Yes, thank you for your suggestion on how to debug the driver via mbuf > dumps. I can only hope I'm able to do that. :-) Its been many moons since I > did any kernal debugging. Perhaps Chis at Marconi can shed some light on > microcode revs too. Perhaps I'm using downloadable microcode that's not the > right revision. > > I know people have said 3.5 FreeBSD runs the ATM PCA200e well. Perhaps I > should just move back to that OS.. Forward sounds better though. > > > ----- Original Message ----- > From: "Joe Thomas" <joe_thomas@cnt.com> > To: "'Alex Huppenthal'" <alex@aspenworks.com>; "Archinaco, Chris" > <Chris.Archinaco@marconi.com> > Cc: <freebsd-atm@FreeBSD.ORG> > Sent: Friday, April 06, 2001 9:49 AM > Subject: RE: SR100069960 - Linux resources > > > > > > The API for the 200E line of adapters is available as > > ftp://ftp.fore.com/pub/docs/port/aali.pdf. I just checked and > > it is dated May 1997. This document never did list what the > > changes from microcode rev to rev were, they simply put out > > a new document. Quite some time ago, Fore posted info on obtaining > > driver source and documentation for their new HE line of adapters. > > The HE is not really an API since you'd be talking straight to the > > adapters registers, etc. as opposed to talking to a microcode app > > on the 200E line. I assume someone actively doing HE development > > on Linux could point you to the right specs, else someone at Macroni, > > and if all else fails, I could go through my archives but that would > > take some time to do... > > > > [Having done the initial FreeBSD FORE driver, I can attempt to > > answer questions as my job/time permits. I do not have any access > > to ATM equipment which would allow me to test anything, let alone > > do any kind of debug/development work. Fortunately, there have been > > several other folks who are active on the list that have stepped up > > to the task so hats off to them! Did you see the comments I replied > > with yesterday about examining IP mbufs as they come into ipatm_output()? > > This would be the place to start determining who's doing the fragmenting.] > > > > Joseph Thomas > > > > -----Original Message----- > > From: Alex Huppenthal [mailto:alex@aspenworks.com] > > Sent: Friday, April 06, 2001 10:32 AM > > To: Archinaco, Chris > > Cc: freebsd-atm@FreeBSD.ORG > > Subject: Re: SR100069960 - Linux resources > > > > > > Chris, > > > > Thanks a million for this info. I'm still stuggling a bit with the BSD > > driver and wonder if you can shed some light on how the ForeRunner card > > actually operates now days. Is there a place we can look to find > > documentation on the API to the card? Perhaps including a list of > functional > > changes made to PCA200E recently? > > > > I'm excited about using this card in a number of installations. > Currently, > > the FreeBSD version of the driver seems to fragment PDUs in a way which > > Xedia routers (lucent) interpret as a 'tiny fragment' attack. At that > point > > they discard the entire PDU. > > > > New development is going on and I just wonder if you have information > which > > could verify how things work. > > > > I'm copying the group which is currently very interested in making > FreeBSD > > a high performance platform using your ATM PCI solution. > > > > Thank you, > > > > Alex > > > > PS: to listen in on the discussion about FreeBSD and ATM.. Send an email > to: > > mailto:subscribe-freebsd-atm@freebsd.org > > > > > > ----- Original Message ----- > > From: "Archinaco, Chris" <Chris.Archinaco@marconi.com> > > To: <alex@aspenworks.com> > > Sent: Wednesday, March 14, 2001 2:41 PM > > Subject: FW: SR100069960 - Linux resources > > > > > > > > Hi Alex, > > > > > > > > Here are the Linux resources I was telling you about. I asked the > > > > TSE who was running the PCA on Linux about FreeBSD but he told me that > > he > > > > had never actually tried it with our adapter. Let me know if you have > > any > > > > problems getting the source via Tactics Online. Let me know if you > have > > > > any questions. > > > > > > > > > > > > LINUX Adapter code for Marconi ATM adapters. > > > > The following are listed as a convenience for persons interested in > > LINUX > > > > drivers available for Marconi ATM Cards. Marconi assumes no liability > > for > > > > the utilization of any of the software referenced. > > > > The referencing of this software should not be construed as an > > endorsement > > > > by Marconi for the use of the software or appropriateness of purpose > for > > > > any particular task. > > > > These are not the only drivers available. > > > > > > > > HE-series card driver: > > > > <ftp://ftp.cmf.nrl.navy.mil/pub/chas/linux-atm>; > > > > > > > > PCA200E card driver: > > > > <http://os.inf.tu-dresden.de/pca200e/>; > > > > > > > > LE25/155 card driver: > > > > <http://frontpage.hypermall.com/linuxatm/forerunner_le_25.htm>; > > > > > > > > Linux ATM Resources (Help on installation): > > > > <http://frontpage.hypermall.com/linuxatm/>; > > > > > > > > Hope this helps, > > > > > > > > Chris > > > > > > > > Chris Archinaco > > > > Level 2 Technical Support Engineer - Edge Products > > > > Marconi Services > > > > phone: (800) 671-3673 > > > > fax: (724) 742-6275 > > > > email: chris.archinaco@marconi.com > > > > URL: http://www.marconi.com > > > > Tactics Online: http://tactics.marconi.com > > > > > > > > > > > > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > > with "unsubscribe freebsd-atm" in the body of the message > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-atm" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-atm" in the body of the message To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-atm" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?67DB78CE91D7D41190A100508BFDF5E051EF1A>