Karl Reichert wrote: > Jan Kiszka wrote: >> Karl Reichert wrote: >>> Jan Kiszka wrote: >>>> Karl Reichert wrote: >>>>> Very good news: I found the reason for this behavoir/bug now! >>>>> >>>>> I'm using the attaced rtnet_start script to start the slave (can't use >>>> the provided one because I don't want RTcfg). >>>>> The last command is 'tdmacfg rteth0 slot 0 2300 -s 100 -l rtnet.log' >>>>> >>>>> If I use this script as attached, I get this weird behavoir that the >>>> slave request and calibration reply in a cycle from the past. >>>>> If I comment this last comment and enter it manually on console (after >>>> sucessfull run of rtnet_start), everything works fine. Also a 'sleep 3' >>>> between the last command and the precending works fine. >>>> >>>> Private follow-up discussion revealed that this is most problably the >>>> same issue as already described here: >>>> >>>> http://article.gmane.org/gmane.linux.real-time.rtnet.user/1891/ >>>> >>>> i.e. it is rt_e1000-specific. >>> Seems like. I would suggest that I remove my wiki entry which said >>> this is a tdmacfg issue and create a new which deals about rt_e1000. >>> There I will describe the workaround so that other rt_e1000 users won't >>> face the same issues. >> Ack. > > Please see http://www.xenomai.org/index.php/RTnet:rt_e1000 and edit if needed. > >>> Do you agree? Or do you think this is sth that needs a change on >>> rt_e1000 driver sourcecode? >> Well, the ultimate and most comfortable solution would be inside the >> driver - if nothing helps a delays before returning from the open handler. >> >> Sigh, I just re-read my last mail on this issue >> (http://thread.gmane.org/gmane.linux.real-time.rtnet.user/1891). >> Obviously, nothing happened to this know issue for more than a year now. >> >> OK, if this slow line setup of the [rt_]e1000 is a common issue, we >> really need that msleep in the setup path, something like this: >> >> Index: drivers/e1000/e1000_main.c >> =================================================================== >> --- drivers/e1000/e1000_main.c (Revision 1136) >> +++ drivers/e1000/e1000_main.c (Arbeitskopie) >> @@ -1202,6 +1202,9 @@ e1000_open(struct rtnet_device *netdev) >> e1000_check_mng_mode(&adapter->hw)) >> e1000_get_hw_control(adapter); >> >> + /* Wait for the hardware to come up */ >> + msleep(2000); >> + >> return E1000_SUCCESS; >> >> err_up: >> >> >> Does it work for you (alternatively, using msleep(3000))? >> > > msleep(2000) already works (no more sleep in startup script needed), but I > would suggest to use msleep(3000). I own two cards, one works with > msleep(2000), the other one needs msleep(3000). So it seems sleeping 2000 ms > is very short, but stick to 3000 ms.
I finally committed msleep(3000), will be part of next release (0.9.10). > > If you add this patch to current RTnet version (SVN version), please edit the > created wiki page (mark as obsolete or sth similiar). Thanks! Is this page already reference by some other page? I didn't find anything. However, maybe we could create a "known issues (with/without fixes)" out of it. Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
