Hi, Jonathan >-----Original Message----- >From: Jonathan Corbet [mailto:corbet@xxxxxxx] >Sent: Monday, 17 December, 2012 23:29 >To: Albert Wang >Cc: g.liakhovetski@xxxxxx; linux-media@xxxxxxxxxxxxxxx; Libin Yang >Subject: Re: [PATCH V3 10/15] [media] marvell-ccic: split mcam-core into 2 parts for >soc_camera support > >On Sun, 16 Dec 2012 14:12:11 -0800 >Albert Wang <twang13@xxxxxxxxxxx> wrote: > >> > - Is the soc_camera mode necessary? Is there something you're trying >> > to do that can't be done without it? Or, at least, does it add >> > sufficient benefit to be worth this work? It would be nice if the >> > reasoning behind this change were put into the changelog. >> > >> [Albert Wang] We just want to add one more option for user. :) >> And we split it to 2 parts because we want to keep the original mode. >> >> > - If the soc_camera change is deemed to be worthwhile, is there >> > anything preventing you from doing it 100% so it's the only mode >> > used? >> > >> [Albert Wang] No, but current all Marvell platform have used the soc_camera in camera >driver. :) >> So we just hope the marvell-ccic can have this option. :) > >OK, so this, I think, is the one remaining point of disagreement here; >unfortunately it's a biggish one. > >Users, I believe, don't really care which underlying framework the driver >is using; they just want a camera implementing the V4L2 spec. So, this >particular option does not have any real value for them. But it has a >real cost in terms of duplicated code, added complexity, and namespace >pollution. If you believe I'm wrong, please tell me why, but I think that >this option is not worth the cost. > >The reason for the soc_camera conversion is because that's how your >drivers do it — not necessarily the strongest of reasons. Of course, the >reason for keeping things as they are is because that's how the in-tree >drivers does it; not necessarily a whole lot stronger. > >I'm not sold on the soc_camera conversion, but neither am I implacably >opposed to it. But I *really* dislike the idea of having both, I don't >see that leading to good things in the long run. So can I ask one more >time: if soc_camera is important to you, could you please just convert the >driver over 100% and drop the other mode entirely? It seems that should >be easier than trying to support both, and it should certainly be easier >to maintain in the future. > [Albert Wang] So if we add B_DMA_SG and B_VMALLOC support and OLPC XO 1.0 support in soc_camera mode. Then we can just remove the original mode and only support soc_camera mode in marvell-ccic? >I'm sorry to be obnoxious about this. > >Meanwhile, the bulk of this last patch series seems good; most of them >have my acks, and I saw acks from Guennadi on some as well. I would >recommend that you separate those out into a different series and submit >them for merging, presumably for 3.9. That will give you a bit less code >to carry going forward as this last part gets worked out. > >Thanks again for doing this work and persevering with it! > >jon Thanks Albert Wang 86-21-61092656 ��.n��������+%������w��{.n�����{��g����^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
