On Sun, Mar 20, 2011 at 12:26:57PM +0200, Amir Goldstein wrote: > If I am not mistaken, while 'bit' was already converted to cluster units, > 'count' is still in block units. > > I think ext4_free_blocks() need to do 2 things: > 1. convert 'count' to clusters (after issuing journal_forget()) > 2. round 'bit' up (and round 'count' down) if start block is not > on cluster boundary, so truncate/punch hole, will not free a > cluster when it's 'base' block is still allocated. Good catch! It's actually more complicated than that, though. Whether or not we can free the first and the last cluster in an unaligned extent is going to depend on whether or not there are any blocks still in use --- which is something that has to be communicated by extents.c, since mballoc has no way of knowing this. (Consider what happens with sparse blocks.) What I think I'm going to have to do is to teach truncate to check to see whether this is an unaligned truncate, and if so, whether there are any blocks still left used. If so, it will need to pass a flag to ext4_free_blocks() to tell it to skip the first incomplete cluster. Similarly, when the punch code lands, we'll need to worry about this at the end of the region which gets "punched" out, and we'll need a similar flag telling ext4_free_blocks() whether or not to release the last incomplete cluster. Thanks for catching this; I have a bit more coding work to do to deal with this case. - Ted -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
