From: SourceForge.net <no...@so...> - 2007-11-09 21:52:39
|
Bugs item #1829255, was opened at 2007-11-09 16:52 Message generated for change (Settings changed) made by cfriedt You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=109655&aid=1829255&group_id=9655 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: xine-lib Group: v1.1.8 Status: Open Resolution: None >Priority: 8 Private: No Submitted By: Christopher Friedt (cfriedt) Assigned to: Nobody/Anonymous (nobody) Summary: xine hangs, crashes linux, when media is removed Initial Comment: xine-lib-1.1.8 xine-ui-0.99.5 linux-2.6.23 (gentoo) 'file' executed on the media file: <filename>: RIFF (little-endian) data, AVI, 512 x 224, 23.98 fps, video: XviD, audio: Dolby AC3 (stereo, 48000 Hz) Hi, I own a fairly low-end USB hard-disk enclosure and this has happened to me at least twice because the USB enclosure randomly cycles its power. While watching video that is stored on the removable media, if the removable media is suddenly without warning removed, then xine will attempt to keep reading the open file descriptor infinitely. Since the I/O request fails practically instantaneously and xine has 0 wait time / or checking for such failures, the xine process enters a busy-loop, very quickly trying to consume 100% of the processor time. I've tried killing it with practically every signal number there is but all kill signals are ignored, even when from root. I've reniced it to -20 and it still uses ~49% of the cpu time. The other ~49% (aside from a rare process grabbing 2%) is used by the system logger. Examining the kernel log with dmesg, I see that the kernel is issuing the following warning extremely fast. ... Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device Nov 7 23:15:52 sith scsi 7:0:0:0: rejecting I/O to dead device ... The amount of similar messages indicate that this is happening perhaps 100 times per second or more. For some reason, `ps` hangs, but i was able to determine the process number of xine with `top`. I can't shut down linux or normally soft-reboot the machine otherwise. I have to do a hard power-off. X successfully shuts down, but the xine process still uses 48% of the CPU. If i kill the system logger, xine uses 98% of the CPU time. Even if I power off the device again, or if the device cycles power by itself, xine still does not let go of the file descriptor or check for errors. I wouldn't say so much that this would be a bug I would request fixed for myself, but rather that it poses a major problem in the software if the operating system is overwhelmed like this. Actually, it's hard to say whether this is a bug in Linux or xine, but since Linux is responding as it 'should' to all of the I/O requests, then I would assume that Xine should be fixed. Let me know if I can help out with any more information. Chris ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=109655&aid=1829255&group_id=9655 |