Am 11.04.2015 um 15:17 schrieb Bernhard Übelacker: > Package: udev > Version: 215-14 > Severity: normal > > Dear Maintainer, > when I insert a specific USB-stick I see following symptoms: > - cpu usage rises (/lib/systemd/systemd-udevd, /usr/lib/udisks2/udisksd, > /usr/bin/plasma-desktop) > - it takes up to a minute until the device notifier popups. > > Took then a closer look by calling "(udevadm monitor | ts & journalctl -f) > > systemd-udevd-008-hama.txt" > when doing following actions (see attached file systemd-udevd-006.txt): > > - 14:50:00 physically attached USB-stick > - 14:50:01 kernel: sd 12:0:0:0: [sdd] Attached SCSI removable disk > - 14:50:02 KERNEL[99837.550175] change > /devices/pci0000:00/0000:00:02.1/usb1/1-1/1-1:1.0/host15/target15:0:0/15:0:0:0/block/sde > (block) > - this "KERNEL.*change.*block" message is received another 741 times > - 14:50:05 rechner kernel: sd 15:0:0:0: [sde] 2004991 512-byte logical blocks: > (1.02 GB/978 MiB) > - 14:50:05 KERNEL[99840.590862] add > /devices/pci0000:00/0000:00:02.1/usb1/1-1/1-1:1.0/host15/target15:0:0/15:0:0:0/block/sde/sde1 > (block) > - 14:50:59 UDEV [99894.356135] add > /devices/pci0000:00/0000:00:02.1/usb1/1-1/1-1:1.0/host15/target15:0:0/15:0:0:0/block/sde/sde1 > (block) > - 14:50:59 now the device notifier was opened > > It seems that for this USB-stick the kernel sends too early/too many > of these "change block" events. > Unfortunately receiving this events is much cheaper than processing > them, so in these 3 seconds there are 9 times more received than processed. > > As all events are put into a queue the final "add block" event is only > processed when all "change block" before were finished. In this case > this took from 14:50:02 to 14:50:59. > > I am unsure if this is strictly a problem with systemd-udev, which should > be prepared for such a "event storm", or if it should be assigned to the > kernel, where I assume these events are originating from. > > Kind regards, > Bernhard > > > > root@rechner:~# uname -a > Linux rechner 3.16.0-4-amd64 #1 SMP Debian 3.16.7-ckt7-1 (2015-03-01) x86_64 > GNU/Linux Processing 1000 udev add events can be costly, because udev will likely spawn lots of (external) helpers, like blkid. Depending on how fast your system is, that can take a while. Smells like a kernel (or hardware bug), if it generates almost 1000 add events for the same device. Ben, what do you think? -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth?
Attachment:
signature.asc
Description: OpenPGP digital signature