Hello *, I've tested the latest linux kernel 2.6.36-rc3 and found that the $PWD variable in init scripts (/etc/init.d/*) during boot contains "(unreachable)/" only when passing init=/sbin/bootchartd to the kernel. This makes some init scripts to fail, for example the ones using start-stop-daemon (I'm using debian sid amd64). Maybe bootchartd is also guilty, but this is a change in behavior, a regression from older kernels. Just in case I'm using ext3 as /. Hope something can be done with it. Thanks, Santi
Another symptom of this bug is calls to ttyname() return "(unreachable)" in front of the ttyname when called inside a chroot environment, even though the tty actually IS reachable in the chroot. I have /dev bind-mounted to the /chroot/dev like this: mount --bind /dev /chroot/dev If I run "tty" in the chroot, I get "(unreachable)/dev/pts/1". This causes all manner of breakage in any program that calls ttyname() from the chroot. It looks like this was added to fs/dcache.c in 2.6.36. I don't see the real value of this feature. Shouldn't it be the problem of the userspace program to deal with the fact that a tty is unreachable?