Hi, On Thu, Aug 25, 2005 at 03:37:18PM +0200, Love Hörnquist Åstrand wrote: > Michael Banck <[EMAIL PROTECTED]> writes: > > There is no kernel AFS support in the Hurd yet (though this issue might > > have stirred up some interest to write a translator for it), so the > > syscalls in afssys.c fail. > > > > I am not sure what should be done in the generic case (I saw there is > > some code for various other operating systems), but I think it makes no > > sense to call these syscalls without kernel support? There was quite a > > discussion whether a bogus ioctl should be added to glibc on the Hurd > > for this issue, but this got (rightfully, IMHO) discarded by the > > maintainers. > > I agree that the ioctl should not be added to the glibc just for this, that > why we define the cpp symbols ourself in kafs.h/afssys.c. > > In other unixes _IOW creates something standalone, apprently Hurd is born > diffrently.
Well, syscalls are different on the Hurd. I am no Hurd kernel hacker, but (as I understand it) as the Hurd is implemented in user space, there are no syscalls per se. ioctl's are merely provided for backwards compatibility with Unix/POSIX, and are implemented in glibc to do magic and notify the Hurd servers of the respective thing the syscall stands for. > In case of hurd we can use something other then _IOW and things will work > jus fine, the resulting numbers from _IOW are used in pioctl calls and can > overlap with ioctl numbers. > > So defineing _VICEIOCTL to > > #define _VICEIOCTL(v) ((unsigned int ) (0x56000000 | v) > #define VIOC_SYSCALL ((unsigned int ) (0x43000000 | 1) > > for Hurd should work just fine. I'll try to modify heimdal accordingly to test that (but if you could provide a patch, that would probably be much quicker...) > > Probably the whole kafs directory should not need to be compiled if AFS > > is not found (could configure check for that somehow, perhaps?), and the > > code in the other parts be conditionalized appropriately. > > Just because the build machine doesn't have AFS support doesn't mean > the binary should support it. AFS can be added later. Good point. However, I am not sure no modifications would be needed in that case. Probably a more hurdish solution would be provided for kafs in that case, using user-space libraries and file system translators. Again, I am not really a Hurd hacker though. Michael -- Michael Banck Debian Developer [EMAIL PROTECTED] http://www.advogato.org/person/mbanck/diary.html -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
