Hi Wolfram,
On Fri, Oct 2, 2020 at 5:44 PM Wolfram Sang <wsa@xxxxxxxxxxxxx> wrote:
> On Mon, Sep 28, 2020 at 05:59:50PM +0200, Ulrich Hecht wrote:
> > Implements atomic transfers to fix reboot/shutdown on r8a7790 Lager and
> > similar boards.
> >
> > Signed-off-by: Ulrich Hecht <uli+renesas@xxxxxxxx>
> > @@ -581,10 +585,12 @@ static void start_ch(struct sh_mobile_i2c_data *pd, struct i2c_msg *usr_msg,
> > + if (pd->atomic_xfer) {
> > + unsigned long j = jiffies + pd->adap.timeout;
> > +
> > + time_left = time_before_eq(jiffies, j);
> > + while (time_left &&
> > + !(pd->sr & (ICSR_TACK | SW_DONE))) {
> > + unsigned char sr = iic_rd(pd, ICSR);
> > +
> > + if (sr & (ICSR_AL | ICSR_TACK |
> > + ICSR_WAIT | ICSR_DTE)) {
> > + sh_mobile_i2c_isr(0, pd);
> > + udelay(150);
> > + } else {
> > + cpu_relax();
> > + }
>
> Is it 100% safe to call cpu_relax() that late? Aren't interrupts
> disabled? What is waking the CPU again? And where does the value 150us
> come from?
cpu_relax() does not sleep, usually it's merely a compiler directive.
On arm32/v7 (and most other platforms):
#define cpu_relax() barrier()
#define barrier() __asm__ __volatile__("": : :"memory")
On arm64:
static inline void cpu_relax(void)
{
asm volatile("yield" ::: "memory");
}
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
