> > Hello. > > > > On 05-03-2013 12:17, Sekhar Nori wrote: > > > >> On 3/4/2013 11:32 PM, Danny Marsh wrote: > >>> I have read many posts and I am still unclear if there is a solution > to > >>> our issue. > >>> > >>> We are using a custom DM6446 board running Linux git kernel 3.0.0 with > a > >>> TUSB2046B <http://www.ti.com/product/TUSB2046B> hub. > > > >>> We are trying to hook up 2 usbnet devices (using the cdc_ethernet > >>> driver). > > > >>> Each device needs the following EPs. > > > >>> TUSB2046B <http://www.ti.com/product/TUSB2046B> - 1 Interrupt EP > > > >>> USB Device 1 - 1 Interrupt EP + 2 Bulk EPs > > > >>> USB Device 2 - 1 Interrupt EP + 2 Bulk EPs > > > >>> 1. Is there a hardware limitation in the DM6446 that prevents us from > >>> doing this? > > > >> DM644x has 4 endpoints. > > > > 5, counting the control one. > > > >> You are definitely shooting above that mark from > >> this description. > > > > All bulk endpoint transfers are sheduled using the single MUSB EP 1, > > AFAIR. > > I just wanted help verifying the number of EPs we are using. Is this > correct 5 EPs? > > DM6446 - 1 EP > TUSB Hub - 1 Interrupt EP > USB Device 1 - 1 Interrupt EP > USB Device 2 - 1 Interrupt EP > All Bulk Transfers - 1 EP > > > > >>> 2. Is there a software work around ("Interrupt endpoint scheduling") > >>> that works with devices that need Interrupt EPs? If so what kernel > >>> version do I need? > > > >> "Interrupt endpoint scheduling" was implemented for very early v2.6.10 > >> based MV kernels that TI used to ship. Recently Ravi up-ported that to > >> v3.3 but was only tested it on DA850 as per my knowledge. Here is the > >> link: > >> http://arago-project.org/git/projects/?p=linux- > davinci.git;a=commit;h=0795c14aa91650d778a27fe7b2ef23e2d9ff8c89 > > > > This code seems to have the same mistake I fixed for 2.6.18 MV > > release -- > > it tries to schedule several URBs concurrently on the same endpoint. It > > seems TI engineers have learned nothing from my work. :-( > > I will wait for Ravi's reply if the "Interrupt endpoint scheduling was > tested for the DM6446 or if there is some type of implementation issue. >
This patch was verified only on da830, not verified on DM6446. Did you checked with this patch and face any issue. Can you list out totally how many tx/rx bulk/interrupt pipes used? The limitation is only 4 Tx/Rx eps is available. This patch will help, if you can connect more interrupt devices (like HID devices) but connecting devices which uses more bulk pipes could be limitation. -- Ravi B _______________________________________________ Davinci-linux-open-source mailing list [email protected] http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
