Re: [PATCH 11/18] maccess: remove strncpy_from_unsafe

[Christoph Hellwig <hch@lst.de>]

On Wed, May 13, 2020 at 12:11:27PM -0700, Linus Torvalds wrote:
> On Wed, May 13, 2020 at 9:01 AM Christoph Hellwig <hch@lst.de> wrote:
> >
> > +static void bpf_strncpy(char *buf, long unsafe_addr)
> > +{
> > +       buf[0] = 0;
> > +       if (strncpy_from_kernel_nofault(buf, (void *)unsafe_addr,
> > +                       BPF_STRNCPY_LEN))
> > +               strncpy_from_user_nofault(buf, (void __user *)unsafe_addr,
> > +                               BPF_STRNCPY_LEN);
> > +}
> 
> This seems buggy when I look at it.
> 
> It seems to think that strncpy_from_kernel_nofault() returns an error code.
> 
> Not so, unless I missed where you changed the rules.

I didn't change the rules, so yes, this is wrong.

> Also, I do wonder if we shouldn't gate this on TASK_SIZE, and do the
> user trial first. On architectures where this thing is valid in the
> first place (ie kernel and user addresses are separate), the test for
> address size would allow us to avoid a pointless fault due to an
> invalid kernel access to user space.
> 
> So I think this function should look something like
> 
>   static void bpf_strncpy(char *buf, long unsafe_addr)
>   {
>           /* Try user address */
>           if (unsafe_addr < TASK_SIZE) {
>                   void __user *ptr = (void __user *)unsafe_addr;
>                   if (strncpy_from_user_nofault(buf, ptr, BPF_STRNCPY_LEN) >= 0)
>                           return;
>           }
> 
>           /* .. fall back on trying kernel access */
>           buf[0] = 0;
>           strncpy_from_kernel_nofault(buf, (void *)unsafe_addr,
> BPF_STRNCPY_LEN);
>   }
> 
> or similar. No?

So on say s390 TASK_SIZE_USUALLy is (-PAGE_SIZE), which means we'd alway
try the user copy first, which seems odd.

I'd really like to here from the bpf folks what the expected use case
is here, and if the typical argument is kernel or user memory.