Hello, 2017-01-14 8:26 GMT-03:00 Jonathan de Boyne Pollard: > > The nosh package is now up to version 1.31 .
Some comments: 1) Programs from at least util-linux get confused by the way system-manager(8) mounts some of the filesystems from source/api_mounts.cpp. The problem seems to be those whose corresponding array of iovecs doesn't have a 'from' element, which IIUC results in a mount(2) call for them with an empty string as the 'source' argument in source/nmount.cpp. And it looks like other programs don't expect that and behave weirdly: $ ps -p 1 -o pid,tname,stat,time,args PID TTY STAT TIME COMMAND 1 ? Ss 00:00:00 /sbin/system-manager $ mount mount: /proc/self/mountinfo: parse error: ignore entry at line 2. mount: /proc/self/mountinfo: parse error: ignore entry at line 3. mount: /proc/self/mountinfo: parse error: ignore entry at line 4. mount: /proc/self/mountinfo: parse error: ignore entry at line 5. /dev/sda4 on / type ext4 (rw,relatime,data=ordered) none on /proc none on /sys none on /dev none on /dev/pts tmpfs on /run type tmpfs (rw,nosuid,nodev,size=206124k,mode=755) tmpfs on /run/shm type tmpfs (rw,nosuid,nodev,noexec,size=515312k) [...] $ findmnt --list findmnt: /proc/self/mountinfo: parse error at line 2 findmnt: /proc/self/mountinfo: parse error at line 3 findmnt: /proc/self/mountinfo: parse error at line 4 findmnt: /proc/self/mountinfo: parse error at line 5 TARGET SOURCE FSTYPE OPTIONS / /dev/sda4 ext4 rw,relatime,data=ordered /proc /sys /dev /dev/pts /run tmpfs tmpfs rw,nosuid,nodev,size=206124k,mode=755 /run/shm tmpfs tmpfs rw,nosuid,nodev,noexec,size=515312k [...] $ cat /proc/self/mountinfo 14 0 8:4 / / rw,relatime - ext4 /dev/sda4 rw,data=ordered 15 14 0:4 / /proc rw,nosuid,nodev,relatime - proc rw 16 14 0:14 / /sys rw,nosuid,nodev,noexec,relatime - sysfs rw 17 14 0:6 / /dev rw,nosuid - devtmpfs rw,size=10240k,nr_inodes=128775,mode=755 18 17 0:15 / /dev/pts rw,nosuid,noexec - devpts rw,gid=5,mode=620,ptmxmode=666 19 14 0:16 / /run rw,nosuid,nodev - tmpfs tmpfs rw,size=206124k,mode=755 20 19 0:17 / /run/shm rw,nosuid,nodev,noexec - tmpfs tmpfs rw,size=515312k [...] $ mountpoint /sys /sys is not a mountpoint (Uuuh... you're wrong) 2) This didn't work: 2017-01-13 10:01 GMT-03:00 Jonathan de Boyne Pollard: > >> Guillermo: >> >>> 2017-01-13 6:00 GMT-03:00 Jonathan de Boyne Pollard : >>> >>> Does your kernel support version 2 cgroupfs? >> >> Nope, it's a 4.4-series kernel. > > I've wangled a later kernel out of Debian 8 backports. (-: I'm going to > have to write v1 detection for you, then. /sys/fs/cgroup shows up empty even though the kernel has cgroups v1 support compiled in. And the catch-all logger reports this: $ tai64nlocal /run/system-manager/log/current 2017-01-19 21:36:21.771835085 system-manager: INFO: cg1: Not required by this kernel. 2017-01-19 21:36:21.771907685 system-manager: INFO: sc1: Not required by this kernel. 2017-01-19 21:36:21.771914950 system-manager: INFO: cg2: Not required by this kernel. 2017-01-19 21:36:21.771919422 system-manager: INFO: sc2: Not required by this kernel. 2017-01-19 21:36:21.771923829 system-manager: ERROR: symlink: /dev/ptmx: File exists 2017-01-19 21:36:21.771928172 system-manager: INFO: system-control (pid 61) started (init ) 2017-01-19 21:36:21.771932710 system-manager: INFO: cyclog (pid 62) started 2017-01-19 21:36:21.771936745 system-manager: INFO: service-manager (pid 63) started 2017-01-19 21:36:21.809549752 system-manager: INFO: system-control (pid 61) ended status 0 2017-01-19 21:36:21.810503507 system-manager: INFO: system-control (pid 64) started (start sysinit) [...] The lines about cg1 and sc1, the arrays of iovecs corresponding to /sys/fs/cgroup (fstype=tmpfs) and /sys/fs/cgroup/systemd (fstype=cgroup), shouldn't be there, since there is a 'cgroup' line in /proc/filesystems that query_control_group_level() in source/common-manager.cpp should have seen: $ grep cgroup /proc/filesystems nodev cgroup I couldn't figure this one out yet; I'll do more testing during the weekend. G.