On Wed, Feb 08, 2017 at 09:21:27AM -0500, Theodore Ts'o wrote: > On Tue, Feb 07, 2017 at 08:36:13PM -0800, Darrick J. Wong wrote: > > As one of 'those people', I think I'm well qualified to say that a fake > > xattr having side effects that doesn't behave in quite the same ways as > > regular xattrs is setting up other programs (the ones that listxattr all > > the attrs and getxattrs them) for all kinds of weird yuckiness. And by > > 'weird yuckiness' I mean "this one magic xattr returned EOPNOTSUPP and > > the whole backup program exploded". > > > > Now granted you can argue that we're just shifting that into an ioctl, > > but ioctls are already weird and yucky. :) > > If we were going to do anything at all, it would be to move it to a > syscall. But given that it took ***years*** for the glibc developers > to get around to adding getrandom(2) to glibc, we'd be stuck using the > syscall(3) interface and having to deal with different syscall numbers > for different architectures (in case we're compiling on a system which > hadn't update the kernel headers in /usr/include), and so at least in > the short term it would actually be worse than using ioctl's. > > For similar reasons I don't think there's going to be that much > interest in adding a syscall to replace XFS_IOC_GOINGDOWN. We'll > probably go with FS_IOC_GOINGDOWN, just as we have with > FS_IOC_SETFLAGS, etc. > > Speaking of XFS_IOC_GOINGDOWN, I'm sure the name was coined many, many > years agoo, back when concerns such as "avoiding a hostile working > environment" were much of an issue. Even so, I note that *someone* > decided that the name that should be exposed to customer (in the > xfs_io man page), was "shutdown", and not "goingdown". So is there > any objection if we use the name FS_IOC_SHUTDOWN moving forward? None here. I think Dave asked for GOINGDOWN -> SHUTDOWN too. > Some might accuse us of being overly concerned about political > correctness, but I have to admit I did have a slight twinge when I > checked in the ext4 shutdown changes into our internal kernel. I was > worried that there might be some folks who might have found the name > at least a tiny bit offensive. (Not that anyone complained, but I'd > much rather be conservative in what we send, and liberal in what we > accept.) Yes, and I further argue that io control command (ioctl) names should be imperative. One doesn't yell 'going down!' to turn off the light switches. :P --D > > - Ted > > > > > > > > > > >
