Re: [PATCH] make 'user_access_begin()' do 'access_ok()'

[Stafford Horne]

On Mon, Jan 07, 2019 at 10:05:12AM -0800, Linus Torvalds wrote:
> Gaah. Re-sending this for the kernel mailing list just for posterity.
> I keep replying to emails that had the mailing list address wrong, and
> then my reply will have it wrong too.
> 
> I will learn to fix up the address just in time for this thread to die
> out, I suspect.
> 
>                 Linus
> 
> On Mon, Jan 7, 2019 at 10:02 AM Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Sun, Jan 6, 2019 at 8:05 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
> > >
> > > Of the above, my test system boots images for the following architectures
> > > in qemu.
> > >
> > > - mips (32/64 bit, big/little endian)
> > > - nios2
> > > - openrisc
> > > - xtensa (mmu and nommu)
> >
> > So most of those are "only" the "macro arguments used twice" problem
> > (although openrisc also does the "arguments not quoted right"). That
> > doesn't cause problems with the new commit, it's an independent issue
> > that could cause random problems elsewhere

Linus,  Thanks for pointing this out and Guenter thanks for testing.

I ack the issue on OpenRISC and will send a patch for review in a day or so.
I'll also do an audit of openrisc to see if we have similar issues.

Next to access_ok() is __addr_ok() which seems to have the similar quoting
issue, but I don't see it being used anywhere, it's also defined and not
used on: arm, x86 and csky.   It is used on SH.

OK to remove __addr_ok()?

-Stafford