Re: [RFC PATCH] asm/generic: introduce if_nospec and nospec_barrier

[Dan Williams]

On Wed, Jan 3, 2018 at 8:44 PM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
> On Thu, Jan 04, 2018 at 03:10:51AM +0000, Williams, Dan J wrote:
>
>> diff --git a/include/linux/fdtable.h b/include/linux/fdtable.h
>> index 1c65817673db..dbc12007da51 100644
>> --- a/include/linux/fdtable.h
>> +++ b/include/linux/fdtable.h
>> @@ -82,8 +82,10 @@ static inline struct file *__fcheck_files(struct files_struct *files, unsigned i
>>  {
>>       struct fdtable *fdt = rcu_dereference_raw(files->fdt);
>>
>> -     if (fd < fdt->max_fds)
>> +     if (fd < fdt->max_fds) {
>> +             osb();
>>               return rcu_dereference_raw(fdt->fd[fd]);
>> +     }
>>       return NULL;
>>  }
>
> ... and the point of that would be?  Possibly revealing the value of files->fdt?
> Why would that be a threat, assuming you manage to extract the information in
> question in the first place?

No, the concern is that an fd value >= fdt->max_fds may cause the cpu
to read arbitrary memory addresses relative to files->fdt and
userspace can observe that it got loaded. With the barrier the
speculation stops and never allows that speculative read to issue.
With the change, the cpu only issues a read for fdt->fd[fd] when fd is
valid.