> The attached patch should fix the floppy oops you reported. The problem
> was that the floppy_release operation can be called with a NULL file
> pointer in some cases, so it's not possible to pass filp->f_dentry to
> block_fsync. But in both of these cases the device has either been
> synced already or doesn't need syncing, so a test to avoid calling
> block_fsync should suffice.
>
> BTW, at some point we should reexamine what arguments are really needed
> by these low-level calls. The case in point, block_fsync, should only
> need a device number to call fsync_dev.
Is there any reason for not calling fsync_dev always without looking
at the flags?
Have a nice fortnight
-- Martin `MJ' Mares <mj@gts.cz> http://atrey.karlin.mff.cuni.cz/~mj/ Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth "What? DosShell wasn't supposed to be a joke?"