On Wed, Sep 23, 2020 at 04:00:32PM +0200, Arnd Bergmann wrote:
> On Sun, Sep 20, 2020 at 8:37 PM Corentin Labbe <[email protected]> wrote:
> >
> > Ciphers produce invalid results on BE.
> > Key and IV need to be written in LE.
> >
> > Fixes: 6298e948215f2 ("crypto: sunxi-ss - Add Allwinner Security System crypto accelerator")
> > Cc: <[email protected]>
> > Signed-off-by: Corentin Labbe <[email protected]>
> > ---
> > drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c | 12 ++++++------
> > 1 file changed, 6 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c b/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > index c6c25204780d..a05889745097 100644
> > --- a/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > +++ b/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > @@ -52,13 +52,13 @@ static int noinline_for_stack sun4i_ss_opti_poll(struct skcipher_request *areq)
> >
> > spin_lock_irqsave(&ss->slock, flags);
> >
> > - for (i = 0; i < op->keylen; i += 4)
> > - writel(*(op->key + i / 4), ss->base + SS_KEY0 + i);
> > + for (i = 0; i < op->keylen / 4; i++)
> > + writel(cpu_to_le32(op->key[i]), ss->base + SS_KEY0 + i * 4);
>
> I suspect what you actually want here is writesl() in place of the
> loop. This skips the byteswap on big-endian, rather than swapping
> each word twice.
>
> The point is that this register seems to act as a FIFO for a byte-stream
> rather than a 32-bit fixed-endian register.
>
> Arnd
Thanks, using writesl() fixes the warning, but I need to keep the loop since the register is different each time.
Or does it is better to use directly __raw_writel() ?
On Wed, Sep 23, 2020 at 8:08 PM LABBE Corentin <[email protected]> wrote:
> On Wed, Sep 23, 2020 at 04:00:32PM +0200, Arnd Bergmann wrote:
> > On Sun, Sep 20, 2020 at 8:37 PM Corentin Labbe <[email protected]> wrote:
> > > diff --git a/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c b/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > > index c6c25204780d..a05889745097 100644
> > > --- a/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > > +++ b/drivers/crypto/allwinner/sun4i-ss/sun4i-ss-cipher.c
> > > @@ -52,13 +52,13 @@ static int noinline_for_stack sun4i_ss_opti_poll(struct skcipher_request *areq)
> > >
> > > spin_lock_irqsave(&ss->slock, flags);
> > >
> > > - for (i = 0; i < op->keylen; i += 4)
> > > - writel(*(op->key + i / 4), ss->base + SS_KEY0 + i);
> > > + for (i = 0; i < op->keylen / 4; i++)
> > > + writel(cpu_to_le32(op->key[i]), ss->base + SS_KEY0 + i * 4);
> >
> > I suspect what you actually want here is writesl() in place of the
> > loop. This skips the byteswap on big-endian, rather than swapping
> > each word twice.
> >
> > The point is that this register seems to act as a FIFO for a byte-stream
> > rather than a 32-bit fixed-endian register.
>
> Thanks, using writesl() fixes the warning, but I need to keep the loop
> since the register is different each time.
Ah, I see. I thought we had an interface for that as well, but I can't
find it now. I see memcpy_toio32() in one driver, but that implementation
appears to be wrong here (and probably also wrong for the machine
it was meant for)
There is the regular memcpy_toio(), but on big-endian Arm that
turns into a per-byte copy, which might either not work on your
hardware or be too slow.
There is also __iowrite32_copy(), which is not what I had remembered
but does seem to do what you want here.
> Or does it is better to use directly __raw_writel() ?
__raw_writel() is not very portable, so I would avoid that in normal
device drivers even when you only run them on specific hardware.
Arnd