I'm puzzled with the new behaviour in 2.6.7 kernel; it worked in 2.6.x (x<=6) and 2.4.x kernels. I'm writing an USB camera driver which streams frames using isochronous transfer mode.
The interrupt handler is called every 0.01 second normally (10 packets in each URB); but if the driver sleeps (by calling wait_for_event_interruptible) sometimes there is a 0.95 second pause in the isochronous interrupts, and data sent by the camera is lost. The wait_for_event_interruptible is called from V4L1 read()/ioctl() function implementation. Between the 0.95 second pause between calling schedule() (inside wait_for_event_interruptible) and the isochronous interrupt, there are no other kernel messages (I have "USB verbose debug messages" enabled) If the condition in wait_for_event() is true so that the driver doesn't really sleep, everything works well. Also if I call schedule() directly without setting the task state to TASK_INTERRUPTIBLE first, there is no problem. I tried changing all kernel configuration options that I thought could matter, in particular all debugging options, "Enforce USB bandwidth allocation", SMP support, and pre-emptible kernel. I also tried replacing uhci-hcd.c from the kernel 2.6.6; nothing helped. In 2.6.6 the problem occurs never; in 2.6.7 its frequency depends on whether the image decoding is done in kernel driver or in userspace program, and whether the userspace program has artificial delay (calling usleep(50000) or not). Below are some statistics (number of the 0.95s pauses/number of the schedule() calls): 267-mjpgdecshow-usleep.txt 10/21 (48%) 267-mjpgdecdrv-usleep.txt 13/13 (100%) 267-mjpgdecshow.txt 0/100 (0%) 267-mjpgdecdrv.txt 2/111 (2%) 267-bayerdecdrv-usleep.txt 2/737 (0.27%) 267uhci6-mjpgdecdrv-usleep.txt 5/5 (100%) 267uhci6-mjpgdecshow-usleep.txt 39/73 (53%) where 267uhci6 = uhci-hcd.c replaced from kernel 2.6.6 mjpg = using MJPG format (as opposed to Bayer) decdrv = decoding image into RGB in the driver decshow = decoding image into RGB in userspace usleep = userspace program contains usleep(50000) The problem doesn't appear in my laptop having OHCI (the problem happens in a computer having UHCI). Linux version 2.6.7 ([EMAIL PROTECTED]) (gcc version 2.95.4 20011002 (Debian prerelease)) #1 SMP Fri Jul 23 20: 18:05 EEST 2004 processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 8 model name : Pentium III (Coppermine) stepping : 6 cpu MHz : 927.479 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse bogomips : 1830.91 processor : 1 vendor_id : GenuineIntel cpu family : 6 model : 8 model name : Pentium III (Coppermine) stepping : 6 cpu MHz : 927.479 cache size : 256 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse bogomips : 1851.39 ------------------------------------------------------- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
