Hi Axel, On 05/01/2014 01:46 AM, Axel Beckert wrote: >> On 04/29/2014 12:35 PM, Axel Beckert wrote: >>> # needrestart >>> Scanning processes............................................... >>> Scanning candidates.Use of uninitialized value in pattern match (m//) at >>> /usr/sbin/needrestart line 285, <HPIPE> line 1. >>> Use of uninitialized value in pattern match (m//) at /usr/sbin/needrestart >>> line 285, <HPIPE> line 1. >> >> this is very strange... it looks like the return of Proc::ProcessTable >> is broken - I could not reproduce it, yet. It looks like the >> /proc/<pid>/exe symlink is missing for some processes?! > > Doesn't seem so: > > # for i in /proc/[0-9]*; do if [ ! -e $i/exe ]; then ls -l $i/exe; fi; done > ls: cannot read symbolic link /proc/10/exe: No such file or directory > lrwxrwxrwx 1 root root 0 Apr 28 22:53 /proc/10/exe
> At least all of these pids I checked were kernel threads for which the > above behaviour is IIRC normal: > > # ps auxwww | fgrep --color '[' > root 1 0.0 0.0 13228 668 ? Ss 2013 2:21 init [2] > root 2 0.0 0.0 0 0 ? S 2013 0:00 [kthreadd] > But the /proc/$pid/exe symlink is always there, it just doesn't have a > target. Which should be ok for kernel threads. It was just to late at night... I was already aware of the kernel thread behavior: > # ignore kernel threads > next unless(defined($ptable->{$pid}->{exec})); The problem is the second scanning pass which looks for the parent processes. It seems to happen if a process is a child of a kernel thread... sadly I could not reproduce it, yet. > P.S.: Feel free to clone this report if you want to track the probably > at least two different issues in here separately. ACK, kernel image stuff goes into #746550. Thanks, Thomas -- :: WWW: http://fiasko-nw.net/~thomas/ :: ::: Jabber: xmpp:tho...@jabber.fiasko-nw.net ::: :: flickr: http://www.flickr.com/photos/laugufe/ :: -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org