On Saturday, August 14, 2021 6:54:40 PM CEST Phillip Potter wrote: > On Fri, 13 Aug 2021 at 13:42, Fabio M. De Francesco > > <fmdefrancesco@xxxxxxxxx> wrote: > > On Friday, August 13, 2021 12:05:36 PM CEST Martin Kaiser wrote: > > > Hi Dan and Phil, > > > [...] > > > > > Just my personal opinion, but I'd be inclined to strip out all DBG_88E > > > > > calls totally. If there are necessary functions being called such as > > > > > device_may_wakeup() we can always just keep this part and remove the > > > > > macro call (not checked this function out myself yet). Thanks. > > > > > > I'd agree with you, Phil. Most DBG_88E prints don't say anything useful. > > > > > > This comment from Greg made me drop the DBG_88E removal for now > > > > > > https://lore.kernel.org/linux-staging/20210803201511.29000-1-martin@xxxxxxxxx/T/#m05 > > > d82a 0ca8ed36180ebdc987114b4d892445c52d > > > > Hi Martin, > > > > I think you misunderstood what Greg was trying to convey with the above- > > mentioned message. > > > > Well, he doesn't like to feed developers with little spoons :-) > > > > I'm pretty sure that, by "Why not use the proper debugging calls instead of > > just deleting them?", he meant you should research, understand, and use the > > proper APIs for printing debug messages. > > > > Please check out pr_debug(), dev_dbg(), netdev_dbg(). Use them appropriately, > > according to the subsystem you're working in and to the different types of > > arguments they take. > > > > Thanks, > > > > Fabio > > > > > A compromise would be to remove only those DBG_88E prints which are > > > really not helpful. > > > > > > Best regards, > > > Martin > > The problem I see is that this driver is so littered with unnecessary > macro calls, how do we decide which ones to keep? In my mind, the > better option is to remove them all and then come up with some new > ones in the vein of netdev_dbg() and friends. I could be wrong of > course :-) I tried going down the route of keeping/converting some to > proper calls such as netdev_dbg() and the issue is a lot of the calls > don't have an obvious value anyway. > > Regards, > Phil I think that you'd better remove only the ones that "have no obvious value" and convert the others to using netdev_dbg(). Obviously, telling which have no value is at the discretion of whoever wants to carry on this work. Regards, Fabio
