On Wed, May 6, 2020 at 10:47 AM Christoph Hellwig <hch@xxxxxx> wrote:
>
> > The fact that we have "probe_kernel_read()" but then
> > "strncpy_from_user_unsafe()" for the _same_ conceptual difference
> > really tells me how inconsistent the naming for these kinds of "we
> > can't take page faults" is. No?
>
> True. If we wanted to do _nofaul, what would the basic read/write
> versions be?
I think "copy_to/from_kernel_nofault()" might be the most consistent
model, if we are looking to be kind of consistent with the user access
functions..
Unless we want to make "memcpy" be part of the name, but I think that
we really want to have that 'from/to' part anyway, which forces the
"copy_from/to_xyz" kind of naming.
I dunno. I don't want to be too nit-picky, I just would like us to be
more consistent and have the naming say what's up without having
multiple different versions of the same thing.
We've had this same discussion with the nvdimm case, but there the
issues are somewhat different (faulting is ok on user addresses - you
can sleep - but kernel address faults aren't about the _context_ any
more, they are about the data not being safe to access any more)
Anybody else?
Linus
