From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4) Gecko/20030703 Description of problem: magicdev seems to chew up a lot of excess CPU cycles... I usually just end up killing it and removing it as I dont need its functionality, but was asked to file a bug :) Version-Release number of selected component (if applicable): magicdev-1.1.4-7 How reproducible: Always Steps to Reproduce: Start with a freshly installed Fedora box (severn-test2) 1. login to gnome 2. wait some time 3. check top output Actual Results: magicdev seems to be taking up lots of cpu cycles for its function. usually sitting around 10-20% on my system (athlon 2100+) Expected Results: If magicdev consistently appears at the top of your top, then that's a bug. (Please provide information as detailed in the last mail) What I'd expect is that it doesn't appear at all in top, except perhaps popping up occasionally with 1-2% cpu if the sampling gets lucky. Regards, Owen Additional info: my system is a dual athlon 2100+ running 2.4.22-1.2061.nptl (smp kernel broken in test2 release). I have a dual channel U160 SCSI controller onboard, and a 2940 for my "cdroms" ... I have a Toshiba 5x DVD, TEAC 32X CDROM (532S), and Yamaha CD burner.. (16x10x40) ... I cannot find the equivilant of /proc/bus/ide/hdc/model for scsi. [root@cookies aic7xxx]# magicdev --verbose magicdev: Checking for blacklisting of /dev/cdrom (/dev/scd0) magicdev: Could not retrieve model magicdev: Checking for blacklisting of /dev/cdrom1 (/dev/scd1) magicdev: Could not retrieve model magicdev: Checking for blacklisting of /dev/cdrom2 (/dev/scd2) magicdev: Could not retrieve model
Created attachment 94970 [details] strace -o magicdev.strace magicdev (30 seconds or so) here is a 30 second (approx) strace output. it was riding at 11% towards the end.. I think over time it takes more and more. Tommy
also... take a look at that strace... I wonder why it opens like every single png file in /usr/share/icons/Bluecurve/24x24/stock/ ? this application doesnt even have a human interface does it?
- The model information should be in /proc/scsi/scsi - With three CDROM drives, I'd expect usage to be a bit higher, but 7-11% sounds too high. - The strace, while not explaining why the CPU usage is high, is a bit strange; for all three drives, you are getting: ioctl(16, 0x5325, 0x7fffffff) = 1 MEDIA_CHANGED YES ioctl(16, 0x5326, 0x7fffffff) = 2 DRIVE_STATUS TRAY_OPEN - Could you rerun the strace with strace -ttT; that will giving timestamps and syscall times, and hopefully figure out where the CPU is going. - The bluecurve icons all getting loaded at startup is a consequence of doing icon themes in a bit of a hacky way. GTK+-2.4 will do better. When timed, it turns out that loading all those icons doesn't actually take that much time, though it certainly looks bad in a strace. Magicdev actually will display a dialog or two on occasion ... for example, the confirmation dialog for CD-ROMS with "autorun" scripts.
[root@cookies root]# cat /proc/scsi/scsi Attached devices: Host: scsi0 Channel: 00 Id: 03 Lun: 00 Vendor: TOSHIBA Model: DVD-ROM SD-M1401 Rev: 1007 Type: CD-ROM ANSI SCSI revision: 02 Host: scsi0 Channel: 00 Id: 04 Lun: 00 Vendor: TEAC Model: CD-ROM CD-532S Rev: 3.0A Type: CD-ROM ANSI SCSI revision: 02 Host: scsi0 Channel: 00 Id: 06 Lun: 00 Vendor: YAMAHA Model: CRW2100S Rev: 1.0H Type: CD-ROM ANSI SCSI revision: 02 Host: scsi1 Channel: 00 Id: 00 Lun: 00 Vendor: IBM-PSG Model: DPSS-318350N F Rev: S94S Type: Direct-Access ANSI SCSI revision: 03 Host: scsi1 Channel: 00 Id: 02 Lun: 00 Vendor: IBM-PSG Model: DPSS-318350N F Rev: S94S Type: Direct-Access ANSI SCSI revision: 03 [root@cookies root]#
... ok .. for some reason it has decided to behave for right now? so, um, the only thing I did was watch a dvd on mplayer last night? I did cause gnome-terminal to hang the whole desktop by forgetting the -o filename option :) I suppose thats a bug too :) I am going to reboot and try it again (from a fresh state?) Tommy
Created attachment 95035 [details] strace -ttT -o magicdev.strace-ttT /usr/bin/magicdev here is the strace for approx 30 seconds... it was running from 3-8% the whole time... it seems to vary quite a bit, but when it behaves, it doesnt even hardly show up on top.. maybe .5% every 3rd or 4th refresh. Tommy
Created attachment 95143 [details] strace -ttT -p PID of magicdev using 20% PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 5102 tommy 16 0 1996 1208 1100 S 9.1 0.1 301:19 1 magicdev strace -ttT -o magicdev-20.strace -p 5102 Process 5102 attached - interrupt to quit Process 5102 detached its up to 25% now... if you need data, I have it in a "mis-behaved" state :) Tommy
damn I wish this was like a forum and I could edit previous posts :) PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 5102 tommy 16 0 1996 1212 1104 S 24.2 0.1 301:47 0 magicdev
it looks like the time is spent on whatever "irq" or "softirq" is? 17:56:05 up 3 days, 19:21, 3 users, load average: 0.59, 0.46, 0.36 99 processes: 96 sleeping, 3 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 0.0% 0.0% 0.0% 66.6% 66.8% 66.2% 0.0% cpu00 0.0% 0.0% 0.0% 33.3% 33.5% 33.1% 0.0% cpu01 0.0% 0.0% 0.0% 33.3% 33.4% 33.1% 0.0% Mem: 1032128k av, 1012828k used, 19300k free, 0k shrd, 40340k buff 475828k active, 488416k inactive Swap: 1831400k av, 202856k used, 1628544k free 677192k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 5102 tommy 16 0 1996 1212 1104 D 12.7 0.1 303:19 0 magicdev 24596 tommy 15 0 7256 7256 4776 S 6.6 0.7 8:33 0 xmms 5287 tommy 15 0 11912 10M 8304 R 1.9 1.0 4:05 0 gnome-terminal ... [tommy@cookies tommy]$ uname -a Linux cookies.local.redneck.nu 2.4.22-1.2087.nptlsmp #1 SMP Sat Oct 4 09:28:12 EDT 2003 i686 athlon i386 GNU/Linux
Bastien Nocera suggests that (if you don't actually need all three CDs once) it would be useful to try disconnecting some of them and see if there is a particular drive or combination of drives that triggers the problem.
well.. its kindof nice to be able to watch DVD's and copy CD's at the same time, but next time I can shut down my box, I will pull the cable on one of them and see how things go.. I have my suspects.. only one of them flashes its light for the check (TEAC 532S) so I will try unplugging it first :) Tommy
just noticed with just the Toshiba DVD and Yamaha CRW, its climbing again :) I will grab a strace and once my download (wget) gets done, I will reboot and take off the CDRW :) 97 processes: 96 sleeping, 1 running, 0 zombie, 0 stopped CPU states: cpu user nice system irq softirq iowait idle total 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% cpu00 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% cpu01 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% Mem: 1032128k av, 1017972k used, 14156k free, 0k shrd, 153176k buff 333336k active, 626264k inactive Swap: 1831400k av, 22876k used, 1808524k free 659880k cached PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 4680 tommy 16 0 5344 4848 4644 S 9.7 0.4 83:55 0 magicdev 16214 tommy 15 0 1624 1620 1320 S 6.7 0.1 4:19 1 wget 4561 root 15 0 297M 40M 9308 S 0.5 4.0 11:28 0 X
I was about to file a bug on this as well, but it would be a duplicate, so I'll just post a comment here. I've actually seen this behavior in magicdev since RH 7.3. This bug has been there for years. -Dave
I have also seen this behavior out of magicdev for years. I currently see it on Fedore Core 1.
Sorry if this is very obvious but here goes... why not just add a parameter for the number of seconds to sleep between checks (or something like that). I don't insert CD-ROM very often I won't mind waiting 10 seconds for it to be mounted.
Closing. Magicdev has been removed in favor of gnome-volume-manager in devel branch.
there is a longstanding bug (or performance issue) in the kernel, see https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=179573