NEW DATA - Linux 1394 somewhat unfriendly during boot?
Brought to you by:
aeb,
bencollins
From: Mark K. <mk...@co...> - 2000-05-21 18:20:38
|
Hi all, =A0=A0 Contained herein is some data I've collected concerning how my = Linux system is acting during a warm boot. Basic status of the system is: =A0 1) Linux PC using CVS code tree for both ieee1394 & libraw1394 2) 1394 support compiled into kernel 3) Five 1394 devices on the bus, Linux PC is the only 400Mb device. (On purpose!) =A0 =A0=A0 When I reboot this PC, sometimes it boots 'friendly', meaning my = other PC, a Windoze box, does not especially react to the fact that the = reboot is happening. Many times, however, that PC reacts noticeably by slowing = down to a crawl for what appears to be 5-10 seconds. Even the mouse on that = system acts badly when the Linux box is rebooting. I think the data in the = attached files shows generally what is going on. =A0 =A0=A0 In all cases, the other four 1394 devices are powered up and = ready to go, and there is no 'intended' bus traffic taking place. No applications = are running. I don't think the camera device is transmitting anything. The = CR-RW & RDK aren't doing anything. =A0 =A0=A0 I'm placing the simple version of the text for you to look at in = the attachment 'e-mail-text.txt' and the more complete version of the boot sequences in 'boot-sequence.txt' so that email line length limiters = don't make it too hard to read. =A0 =A0=A0 In e-mail-text.txt, notice that in the friendly sequence we are = the only 400Mb device, and that we generated the bus reset. However, in the unfriendly sequence, we did not generate the bus reset, but device 0 = did. At the end of the unfriendly sequence in this file you see the = 'unsolicited response packet', which starts off the next set of bus resets and = Self-ID sequences. This goes on for quite awhile, and then eventually calms = down. (See boot-sequence.txt) (Notice that the two sequences are actually = quite different much earlier on...) =A0 =A0=A0 I am concerned about whether some of this might be caused by way = I built and installed the new kernel. LILO possibly not reading the right = system.map file, or the right module-info file. Comments on this are very welcome. = I have multiple kernels installed in lilo.conf, but it does not seem to = me that after a reboot the System.Map link is always pointing at the one = that is currently running, and the line in the message file doesn't seem to = me that it is reading the right one. I may rebuild this system with a = single kernel just to ensure that this is not the issue. =A0 =A0=A0 If anyone has ideas about other experiments I could do, I = welcome those also. =A0 With best regards, Mark |