>> I think [chdir] should be done _after_ the mount() call, > If you do it that way, there would be a race with an immediately > following umount, which would fail if attemted between the mount and > the chdir which follows.
That happens only if cwd is the mount-on directory (at least in 5.2); if a process's cwd is in a subdir of the mount-on directory, its cwd after the mount is still under the covered directory, not the covering directory. (I also would argue it's a bug that cwds equaling the covered directory are silently moved to the covering directory, but that's a separate issue.) mlelstv pointed out upthread that 10.1 (and presumably current) mount_mfs absolutifies relative mount-on paths before mounting, so all that's needed is that the chdir happen after converting the path; I haven't checked that (cvsweb seems to be handing out 403s at the moment), but would assume it's so. /~\ The ASCII Mouse \ / Ribbon Campaign X Against HTML mo...@rodents-montreal.org / \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
