On Fri, Feb 21, 2014 at 03:34:43AM +0000, Sharma, Shashank wrote: > Hi Ville/All, > > We gave a presentation on design on this framework, few months ago, in one of > our common forum with OTC folks. > We discussed, took review comments, and re-designed the framework, as per > the feedbacks.
Apparently I wasn't there. And in any case it would be better to discuss it on dri-devel where people outside Intel can give their opinion. > > We also discussed the benefits of providing the controls directly from /sysfs > over going for a UI manager based property settings. > So I don't understand where are we going wrong, can you please elaborate a > bit ? The most important issue from my POV is that it can't be part of the atomic modeset. Another issue is that it make the whole API inconsistent. Some stuff through ioctl, some stuff through sysfs, some stuff through whatever the next guy thinks of. It's not pretty. I've worked in the past with a driver where I had to poke at various standardish ioctls, custom ioctls, and sysfs to make it do anything useful, and I have no interest in repeating that experience. sysfs is especially painful since you have do the string<->binary conversions all over the place, and also you en up doing open+read/write+close cycles for every little thing. It also adds more entrypoints into the driver for us to worry about. That means extra worries about the power management stuff and locking at the very least. Also the rules of sysfs say "one item per file". The only allowed exception to this rule I know of is hardware provided blobs (like EDID, PCI ROM etc.). Your current implementation breaks this rule blatantly. > > This is just a basic design, and once go ahead with this, we can always work > on making hardware agnostic, as you recommended. > > IMHO, controls from /sysfs would be a very generic interface for all > linux/drm based platform, where any userspace can read/write and control > properties. > We don't even need a UI manager or a minimum executable to play around, just > a small script can do. But we can always write something on top of this, > to be included in any UI framework or property. If there's a real need to get at properties through sysfs, then we could think about exposing them all. But that may presents some issues where the current master suddenly gets confused about its state since someone else went behind its back and changed a bunch of stuff. > > Regards > Shashank > > -----Original Message----- > From: Ville Syrjälä [mailto:ville.syrj...@linux.intel.com] > Sent: Thursday, February 20, 2014 6:41 PM > To: Sharma, Shashank > Cc: intel-gfx@lists.freedesktop.org; Shankar, Uma; > dri-de...@lists.freedesktop.org > Subject: Re: [Intel-gfx] [PATCH 0/6] Intel Color Manager Framework > > On Thu, Feb 20, 2014 at 06:07:21PM +0530, Shashank Sharma wrote: > > Color manager is a new framework in i915 driver, which provides a > > unified interface for various color correction methods supported by > > intel hardwares. The high level overview of this change is: > > Would have been good to discuss this idea before implementing it. The plan is > to use kms properties for this kind of stuff which allows us to hook it up > with the upcoming atomic modeset API. Just yesterday there was some > discussion on #dri-devel about exposing user settable blob properties even > before the atomic modeset API lands (it was always the plan for the atomic > modeset API anyway). So based on a cursory glance, this looks like it's going > in the wrong direction. > > Also ideally the properties should be hardware agnostic, so a generic > userspace could use them regardless of the hardware/driver. Obviously that > might not be possible in all cases, but we should at least spend a bit of > effort on trying to make that happen for most properties. > > -- > Ville Syrjälä > Intel OTC -- Ville Syrjälä Intel OTC _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx