On Mon, Dec 05, 2022 at 03:13:46PM +0100, Andrzej Pietrasiewicz wrote: > Hi Greg, > > W dniu 5.12.2022 o 13:20, Greg KH pisze: > > On Mon, Dec 05, 2022 at 01:14:21PM +0100, Andrzej Pietrasiewicz wrote: > > > Hi All, > > > > > > include/linux/usb/composite.h contains: > > > > > > /* messaging utils */ > > > #define DBG(d, fmt, args...) \ > > > dev_dbg(&(d)->gadget->dev , fmt , ## args) > > > #define VDBG(d, fmt, args...) \ > > > dev_vdbg(&(d)->gadget->dev , fmt , ## args) > > > #define ERROR(d, fmt, args...) \ > > > dev_err(&(d)->gadget->dev , fmt , ## args) > > > #define WARNING(d, fmt, args...) \ > > > dev_warn(&(d)->gadget->dev , fmt , ## args) > > > #define INFO(d, fmt, args...) \ > > > dev_info(&(d)->gadget->dev , fmt , ## args) > > > > > > Gadget functions do use these, but not consistently: > > <snip> > > > > Questions: > > > > > > 1) Should we make them use the messaging utils consitently? > > > > Yes, converting to use the dev_*() calls is good to do. > > Heh, I sent this RFC to prevent learning the hard way (by actually creating > incorrect patches), that we want to be consistent, but using dev_*() calls > rather than composite.h utils. That's ok. > > Which brings an interesting question: should the ultimate goal be to remove the > messaging utils altogether from composite.h? It _seems_ their purpose is mainly > to wrap dereferencing a pointer two pointers away: &(d)->gadget->dev to make > invocations shorter. With the default of 100 columns in checkpatch nowadays it > is maybe a less important goal? Or maybe we should prevent such long > dereferencing by introducing helper variables just like below in > u_audio_start_capture()? Yes, that should be the ultimate goal. We did that in the USB drivers a decade or so ago, but no one spent the time to do the same for the USB gadget code. thanks, greg k-h
