Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20191009192105.GC26530@ZenIV.linux.org.uk> <CAMo8BfKUOmExGRMaUPmcRsy=iyRrguLF6JOLUMegNnzkF9vcvQ@mail.gmail.com>
 <20191010015606.GD26530@ZenIV.linux.org.uk> <CAMo8BfLo4hy+WGA7p+7iZaLmmgFOyzMRAtG5dzNj=JEU04GoKA@mail.gmail.com>
 <20191010142956.GG26530@ZenIV.linux.org.uk>
In-Reply-To: <20191010142956.GG26530@ZenIV.linux.org.uk>
From:   Max Filippov <jcmvbkbc@gmail.com>
Date:   Thu, 10 Oct 2019 20:35:39 -0700
Message-ID: <CAMo8BfL3j4odehiR8KzHwjohfoBADOhjSjeXFG7AwgBXoXQRTw@mail.gmail.com>
Subject: Re: [PATCH] xtensa: fix {get,put}_user() for 64bit values
To:     Al Viro <viro@zeniv.linux.org.uk>
Cc:     "open list:TENSILICA XTENSA PORT (xtensa)" 
        <linux-xtensa@linux-xtensa.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Oct 10, 2019 at 7:29 AM Al Viro <viro@zeniv.linux.org.uk> wrote:
> Hmm...   Looking at __get_user_size(), we have retval = 0; very early
> in it.  So I wonder if it should simply be
> #define __get_user_size(x, ptr, size, retval)                           \
> do {                                                                    \
>         int __cb;                                                       \
>         retval = 0;                                                     \
>         switch (size) {                                                 \
>         case 1: __get_user_asm(x, ptr, retval, 1, "l8ui", __cb);  break;\
>         case 2: __get_user_asm(x, ptr, retval, 2, "l16ui", __cb); break;\
>         case 4: __get_user_asm(x, ptr, retval, 4, "l32i", __cb);  break;\
>         case 8: if (unlikely(__copy_from_user(&x, ptr, 8)) {            \
>                         retval = -EFAULT;                               \
>                         x = 0;                                          \
>                 }                                                       \
>                 break;                                                  \
>         default: (x) = __get_user_bad();                                \
>         }                                                               \
> } while (0)
> so that 64bit case is closer to the others in that respect (i.e. zeroing
> done on failure and out of line).  No?

Ok, I agree.
The intermediate __gu_val in __get_user_[no]check doesn't work well
with some data types used in the kernel, unfortunately. I'll post a series
with what's close to your initial patch on top of rearranged
__get_user_[no]check.

-- 
Thanks.
-- Max