On Mon, Apr 17, 2017 at 3:55 AM, Mauro Carvalho Chehab <mche...@s-opensource.com> wrote: > Em Sat, 15 Apr 2017 20:28:20 +0200 > Anders Eriksson <aerikss...@gmail.com> escreveu: > >> Hi Mauro, >> >> I've two devices using this driver, and whenever I have them both in >> use I eventually (between 10K and 100K secs uptime) i2c write errors >> such as in the log below. If only have one of the devices in use, the >> machine is stable. >> >> The machine never recovers from the error. >> >> All help apreciated. >> -Anders >> >> >> >> [ 0.000000] Booting Linux on physical CPU 0xf00 >> [ 0.000000] Initializing cgroup subsys cpuset >> [ 0.000000] Initializing cgroup subsys cpu >> [ 0.000000] Initializing cgroup subsys cpuacct >> [ 0.000000] Linux version 4.4.15-v7+ (dc4@dc4-XPS13-9333) (gcc >> version 4.9.3 (crosstool-NG crosstool-ng-1.22.0-88-g8460611) ) #897 >> SMP Tue Jul 12 18:42:55 BST 2016 >> [ 0.000000] CPU: ARMv7 Processor [410fc075] revision 5 (ARMv7), >> cr=10c5387d >> [ 0.000000] CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing >> instruction cache >> [ 0.000000] Machine model: Raspberry Pi 2 Model B Rev 1.1 > > Hmm.. RPi2... that explains a lot ;) > > I've seen similar behaviors on some arm devices with just one device. > > That's likely due to some problem with isochronous transfers at the > USB host driver. > > The thing is that ISOC transfers are heavily used by USB cameras: > they require that the USB chip would provide a steady throughput > that can eat up to 60% of the USB maximum bitrate, with just one > video stream. > > My experience says that several USB drivers can't sustain such > bit rates for a long time. > > The RPi tree uses an out-of-tree driver for the USB host driver > (otgdwc - I guess). Upstream uses a different driver (dwc2). > My recent experiences with upstream(dwc2) and USB cameras > is even worse: it doesn't work, if the camera supports only > ISOC frames. > > I'll eventually try to fix the upstream driver if I find > spare time for it, but I won't touch at the proprietary > driver that is shipped with the downstream Kernel. >
Hi, I'd appreciate any attempt to fix this. My experience is that it (the rpi2 with otgdwc) bugs out after 10k-100k of uptime, and can sustain parallel recordings (using both tv-receivers), so the rest of the systems seems to be working ok. I'd be happy to try out any patches you create for the upstream driver. Br, Anders
