On Thursday 19 July 2007, M. Warner Losh wrote: > In message: <[EMAIL PROTECTED]> > > Jeremy Chadwick <[EMAIL PROTECTED]> writes: > : If someone wants to work on this and needs devices/toys (thumb > : drives, external enclosures + hard disks), let me know, I will be > : more than happy to buy them the hardware needed. > > Willing to fund the work on it too? This is a volunteer project, > and you have to motivate people to work on this. Tirades in > mailing lists has proven to be ineffective in the past. > > I've looked at the issue, and generically, if a device goes away, > it is *HARD* to not panic. The same thing happens if you eject a > CF card in a PC Card adapter in a PC Card slot. > > The best one can do without massive buffer cache work is what > firewire does: it has one attachment to handle all umass devices. > When the device goes away, it pauses all operations to that device. > If the device comes back, it resumes the I/O . If the device > never comes back, then the I/O never finishes. > > Warner >
Just curious, but what, if any, is the performance hit with this strategy? I could care less about performance on a usb stick, but if we are talking about changes that are going to affect all filesystems regardless of storage device implimentation then I'm sort of interested. eg: I wouldn't be happy trading filesystem performance for avoiding a panic that is trivial to avoid in the first place. -- Thanks, Josh Paetzel
pgpmLTsui8umW.pgp
Description: PGP signature