On Tue 5 March 2013 16:19:52 Mauro Carvalho Chehab wrote: > Em Fri, 15 Feb 2013 09:46:59 +0100 > Hans Verkuil <hverkuil@xxxxxxxxx> escreveu: > > > This patch series cleans up the cx231xx driver based on v4l2-compliance > > reports. > > > > It is identical to the RFCv2 patch series I posted a week ago: > > > > http://www.mail-archive.com/linux-media@xxxxxxxxxxxxxxx/msg58508.html > > > > I have tested this on various cx231xx devices. However, I have no hardware > > that supports the radio tuner, so that's untested. To my knowledge, no such > > devices exist at the moment anyway. > > > > Also note that the MPEG encoder support does not seem to work. It didn't work > > before these patches are applied, and it doesn't work afterwards. At best it > > will stream for a bit and then hang the machine. While I did convert the 417 > > code to have it pass the compliance tests, I also disable 417 support in the > > single card that supports it (gracefully provided by Conexant for which I > > want to thank them!) until someone can find the time to dig into it and > > figure out what is wrong. Note that that board is an evaluation board and not > > a consumer product. > > > > In addition the vbi support is flaky as well. It was flaky before this patch > > series, and it is equally flaky afterwards. I have managed to get something > > to work only on rare occasions and only for NTSC, never for PAL. > > > > Finally I have tested this on a big-endian machine so there are a bunch of > > patches fixing a lot of endianness problems. > > > > A general note regarding this driver: I've found this to be a particularly > > fragile driver. Things like changing formats/standards, unplugging at > > unexpected times and vbi support all seem very prone to errors. I have > > serious doubts about the disconnect handling: this code really should use the > > core support for handling such events (in particular the v4l2_device release > > callback). > > > > Regards, > > > > Hans > > > > The following changes since commit ed72d37a33fdf43dc47787fe220532cdec9da528: > > > > [media] media: Add 0x3009 USB PID to ttusb2 driver (fixed diff) (2013-02-13 18:05:29 -0200) > > > > are available in the git repository at: > > > > git://linuxtv.org/hverkuil/media_tree.git cx231xx > > > > for you to fetch changes up to bb047a3fd001f7de36f94058c9c296332c494f83: > > > > cx231xx: fix gpio big-endian problems (2013-02-15 09:42:39 +0100) > > > > ---------------------------------------------------------------- > > Hans Verkuil (26): > > cx231xx: add device_caps support to QUERYCAP. > > cx231xx: add required VIDIOC_DBG_G_CHIP_IDENT support. > > cx231xx: clean up radio support. > > cx231xx: remove broken audio input support from the driver. > > cx231xx: fix tuner compliance issues. > > cx231xx: zero priv field and use right width in try_fmt > > cx231xx: fix frequency clamping. > > cx231xx: fix vbi compliance issues. > > cx231xx: convert to the control framework. > > cx231xx: add struct v4l2_fh to get prio and event support. > > cx231xx: remove current_norm usage. > > cx231xx: replace ioctl by unlocked_ioctl. > > cx231xx: get rid of a bunch of unused cx231xx_fh fields. > > cx231xx: improve std handling. > > cx231xx-417: remove empty functions. > > cx231xx-417: use one querycap for all device nodes. > > cx231xx-417: fix g/try_fmt compliance problems > > cx231xx-417: checkpatch cleanups. > > cx231xx-417: share ioctls with cx231xx-video. > > cx231xx-417: convert to the control framework. > > cx231xx: remove bogus driver prefix in log messages. > > cx231xx: disable 417 support from the Conexant video grabber > > Since when it was broken? That's impossible to say without going back a very long time. I doubt it ever worked reliably. Note that there were no consumer devices supporting this until now (the new OTG102 device), so nobody could ever test it unless they has the dev board from Conexant. > Did you c/c those patch series to Sri/Palash? Yes, to both of them. > What they said about the > issues with 417? Nothing :-) Anyway, this patch series does not affect the 417: it's just as broken before as it is after. I have been able to get MPEG out of it on occasion, but it crashed quickly afterwards. The only thing I did is disable it in the board config, waiting for someone with more time to dig into this and fix it. As I mentioned in my pull request this driver is quite fragile. Regards, Hans > > Palash/Sri: > > I remember Palash mentioned that he would be the "odd fixes" > maintainer for Conexant drivers. If you're willing to do so, could you > please send us a patch for MAINTAINERS, to avoid the risk of having a > series like this one without your review? > > Regards, > Mauro > > > cx231xx: don't reset width/height on first open. > > cx231xx: don't use port 3 on the Conexant video grabber. > > cx231xx: fix big-endian problems. > > cx231xx: fix gpio big-endian problems > > > > drivers/media/usb/cx231xx/cx231xx-417.c | 1178 +++++++++++++++++++++++++------------------------------------ > > drivers/media/usb/cx231xx/cx231xx-audio.c | 8 +- > > drivers/media/usb/cx231xx/cx231xx-avcore.c | 83 ++--- > > drivers/media/usb/cx231xx/cx231xx-cards.c | 24 +- > > drivers/media/usb/cx231xx/cx231xx-core.c | 2 +- > > drivers/media/usb/cx231xx/cx231xx-pcb-cfg.c | 2 +- > > drivers/media/usb/cx231xx/cx231xx-vbi.c | 25 +- > > drivers/media/usb/cx231xx/cx231xx-video.c | 589 ++++++++----------------------- > > drivers/media/usb/cx231xx/cx231xx.h | 54 ++- > > 9 files changed, 740 insertions(+), 1225 deletions(-) > > -- > > To unsubscribe from this list: send the line "unsubscribe linux-media" in > > the body of a message to majordomo@xxxxxxxxxxxxxxx > > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
