On Fri, 12 Feb 2016, to...@tuxteam.de wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On Thu, Feb 11, 2016 at 10:49:43AM -0800, Patrick Bartek wrote: > > [...] > > > In 3 years of use, I've experienced no problems. Why does no one > > believe me? > > I *do* believe what you state above. I just *strongly* recommend > against the practice you propose. Anyone is, of course, free to > use my recommendations the way (s)he sees fit.
My comment was a theatrical aside. I should have put it in parentheses. > It's as simple as that: pulling out a still mounted medium is > risky, as the OS makes no guarantees about the state of the > cache. To be really effective, your udev rules would need the > ability to peek into the future, to unmount the USB stick a tad > before you're going to pull it out. Doesn't work like that. The system notes FIRST through D-Bus IIRC the drive has been removed, and then the specific rule to unmount it is run. Yes, that doesn't seem right, but that's the way it works. During my initial stress testing, I was able to create repeatable situations where the drive was not unmounted when pulled. The easiest way to do this was while writing to the drive was in progress. The "fix' was just plug the drive back in, wait a few seconds, and unplug it. It would unmount. However, what was being copied to the drive would not be completed, but the drive was still usable. One time, I was able to corrupt the drive to the point that reformatting was required as well as a "force" unmount on the system. > You can, of course, reduce the risk by waiting for the system > to be long enough in a "quiescent" state, by reducing the > complexity of the whole (Rube Goldbergian desktop systems > have lots of moving parts doing lots of things which tend to > make system behaviour less predictable), but your risk won't > disappear. If you want to get rid of this risk, unmount before > pulling. I may have an advantage: I don't use a desktop/environment, just X and a window manager. So, I don't have all the overhead that imposes. > I think we've stated our opinions clearly enough. I bow out. Hope your problem with GNOME is resolved quickly. B
