Subject: Re: The demise of DEV_BSIZE
To: Bill Studenmund <firstname.lastname@example.org>
From: Chuck Silvers <email@example.com>
Date: 10/12/1999 10:04:04
On Tue, Oct 12, 1999 at 09:53:40AM -0700, Bill Studenmund wrote:
> On Tue, 12 Oct 1999, Chuck Silvers wrote:
> > if we're calling VFS_VGET(), then there must be a mounted filesystem in
> > there somewhere. mounting the filesystem would have opened the device
> > on which that filesystem is mounted.
> True, but the device on which the filesystem is mounted is not the one on
> which we want to call d_ioctl(). We want to call d_ioctl() on the device
> represented by the new vnode. :-)
> Take care,
oh, I was confused, you're talking about VFS_VGET() returning a device node...
hmm. it's starting to sound like checkalias() is the wrong place to be
dealing with the sector size, it's really more of an open-time thing rather
than vget-time. how about just requiring that d_open() deal with it
for disk devices? but d_open() doesn't have the right pointer for that.
it seems like it might be better to pass around a pointer to something
in many of the places where we currently pass around dev_t's, then we
can just get at the bsize via that pointer instead of needing to call
a function for it.