Hi Matthieu, thanks for your attention. Our download/upload speed are not very high, they are 1,5M bps and 320Kbps respectively. The kernel is 2.4 so we cannot test the ueagle-atm. I've some questions about the various timers, because we've seen that our usb controller got the "STALL" status from modem when we have enabled the printk on "eu_uni_process_in_data: non-integral number of cells in incoming data" ... , otherwise we just lost a frame, but the modem continue work. So I think that the State Machine is very sensitive for code execution time. For example, when we have the printk on eu_err enabled, when we have non- integral number of cells, the watchdog often expired on TRANS_TIME_4_OPSTAT (80 ms) in STATE_OPERATIONAL_RX. So, how can we arrange the watchdog time to assure a stable connection? There are all these TRANS_TIME_xxx in eu_sm.h, how does it work? Also when we upload dsp code using "eu_upload_idma_swapp" that has a 1 msec wait, is it enough? Sometimes we have exhausted the "queued_urb_t" queue. We should wait more than 1 msec? ....
Regards, Jianliang Lu ---------- Original Message ---------------------------------- From: matthieu castet <[EMAIL PROTECTED]> Reply-To: [email protected] Date: Tue, 07 Jun 2005 19:36:46 +0200 >Hi, > >Selon Jianliang Lu <[EMAIL PROTECTED]>: > >> Hi all, >> we are testing the eagle-2.3.1 for the eagle-II modem on our PPC >> embedded system. During the test, that just send the ping to >> another system via adsl, we have had a lot of warning: >> "eu_uni_process_in_data: non-integral number of cells in incoming >> data" >> "eu_uni_process_in_data: incoming data length = 3f." >> "eu_uni_process_in_data: ATM_CELL_SIZE = 35." >> ..... >> so we lost a frame every time we have this warning, and some times >> our usb controller got also the "STALL" from the modem. >> >> We are using USB_BULK mode to receive the data, instead of >> isochronous mode to simplify the things. >What are the upload and download speed ? >Bulk mode don't work well with very high speed (> 4 Mbps). >see https://mail.gna.org/public/eagleusb-dev/2005-05/msg00000.html > >> >> I think that the firmware of the modem has some problems to >> assemble the eth packets, but our usb driver has also some >> problems that put the modem in stall on control mode, that take >> the handshake between usb controller and the modem. We are >> debugging it. >> >> I've also sent a patch about ppc to eagle-usb forum and I don't >> know if you have acepted it. >> >Yes it was applied. > >Matthieu > >PS : If you want, you could test the upcoming ueagle-atm : it is for 2.6 >kernel >only, use the kernel atm stack instead of facking an eth device and use a >generic library (usbatm) for managing the data stuff. > > > >_______________________________________________ >Eagleusb-dev mailing list >[email protected] >https://mail.gna.org/listinfo/eagleusb-dev >
