Re: [PATCH] kernel: introduce prctl(PR_LOG_UACCESS)

[Peter Collingbourne]

On Tue, Sep 21, 2021 at 11:30 PM Cyrill Gorcunov <gorcunov@xxxxxxxxx> wrote:
>
> On Tue, Sep 21, 2021 at 11:18:09PM -0700, Peter Collingbourne wrote:
> > This patch introduces a kernel feature known as uaccess logging.
> > With uaccess logging, the userspace program passes the address and size
> > of a so-called uaccess buffer to the kernel via a prctl(). The prctl()
> > is a request for the kernel to log any uaccesses made during the next
> > syscall to the uaccess buffer. When the next syscall returns, the address
> > one past the end of the logged uaccess buffer entries is written to the
> > location specified by the third argument to the prctl(). In this way,
> > the userspace program may enumerate the uaccesses logged to the access
> > buffer to determine which accesses occurred.
> ...
> > diff --git a/include/linux/sched.h b/include/linux/sched.h
> > index e12b524426b0..3fecb0487b97 100644
> > --- a/include/linux/sched.h
> > +++ b/include/linux/sched.h
> > @@ -34,6 +34,7 @@
> >  #include <linux/rseq.h>
> >  #include <linux/seqlock.h>
> >  #include <linux/kcsan.h>
> > +#include <linux/uaccess_buffer_info.h>
> >  #include <asm/kmap_size.h>
> >
> >  /* task_struct member predeclarations (sorted alphabetically): */
> > @@ -1487,6 +1488,8 @@ struct task_struct {
> >       struct callback_head            l1d_flush_kill;
> >  #endif
> >
> > +     struct uaccess_buffer_info      uaccess_buffer;
> > +
>
> Hi, Peter! I didn't read the patch carefully yet (will do once time permit)
> but from a glance should not this member be under #ifdef CONFIG_UACCESS_BUFFER
> or something? task_struct is already bloated too much :(

Yes, I've now added an ifdef here (previously I had the ifdef inside
the struct uaccess_buffer_info, but I think this would still leave
some space due to C struct layout rules).

>
> > +     case PR_LOG_UACCESS:
> > +             if (arg5)
> > +                     return -EINVAL;
> > +             error = uaccess_buffer_set_logging(arg2, arg3, arg4);
> > +             break;
>
> Same here (if only I didn't miss something obvious). If there is no support
> for CONFIG_UACCESS_BUFFER we should return an error I guess.

The uaccess_buffer_set_logging (now
uaccess_buffer_set_descriptor_addr_addr) function is defined to return
-EINVAL if CONFIG_UACCESS_BUFFER is not defined.

Peter