| Message ID | 3578615.iBrKplh45Q@phil (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 01/20, Heiko Stuebner wrote: > Commit e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") removed > the special ops struct for read-only clocks and instead opted to handle > them regularly. > > On the rk3368 this results in breakage as aclkm now gets set a value. > While it is the same divider value, the A53 core still doesn't like it, > which can result in the cpu ending up in a hang. > The reason being that "ACLKENMasserts one clock cycle before the rising > edge of ACLKM" and the clock should only be touched when STANDBYWFIL2 > is asserted. > > So make sure read-only clocks don't touch the clock-register at all > even if only writing the same value, as even writing the same value > may not be safe in all cases. > > Fixes: e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") > Reported-by: Zhang Qing <zhangqing@rock-chips.com> > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > Reviewed-by: James Hogan <james.hogan@imgtec.com> > --- Hmph, the same patch[1] has been sitting in my inbox for months now but I never got a response to my last question and then forgot about it. I'll repeat the question here and hopefully we can finish the discussion. Maybe it would make more sense to have different ops for read only dividers? The clk_set_rate op would be empty, while we would have proper recalc_rate and round_rate ops. We can't just flat out revert commit e6d5e7d90be9 (clk-divider: Fix READ_ONLY when divider > 1, 2014-11-14) because that will introduce the problem that it was fixing, but as long as we implement round_rate in addition to recalc_rate it should work. [1] http://lkml.kernel.org/r/1428392806-14538-1-git-send-email-jy0922.shim@samsung.com
Am Mittwoch, 20. Januar 2016, 14:16:56 schrieb Stephen Boyd: > On 01/20, Heiko Stuebner wrote: > > Commit e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") > > removed the special ops struct for read-only clocks and instead opted > > to handle them regularly. > > > > On the rk3368 this results in breakage as aclkm now gets set a value. > > While it is the same divider value, the A53 core still doesn't like it, > > which can result in the cpu ending up in a hang. > > The reason being that "ACLKENMasserts one clock cycle before the rising > > edge of ACLKM" and the clock should only be touched when STANDBYWFIL2 > > is asserted. > > > > So make sure read-only clocks don't touch the clock-register at all > > even if only writing the same value, as even writing the same value > > may not be safe in all cases. > > > > Fixes: e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") > > Reported-by: Zhang Qing <zhangqing@rock-chips.com> > > Signed-off-by: Heiko Stuebner <heiko@sntech.de> > > Reviewed-by: James Hogan <james.hogan@imgtec.com> > > --- > > Hmph, the same patch[1] has been sitting in my inbox for months > now but I never got a response to my last question and then > forgot about it. I'll repeat the question here and hopefully we > can finish the discussion. > > Maybe it would make more sense to have different ops for read > only dividers? The clk_set_rate op would be empty, while we would > have proper recalc_rate and round_rate ops. We can't just flat > out revert commit e6d5e7d90be9 (clk-divider: Fix READ_ONLY when > divider > 1, 2014-11-14) because that will introduce the problem > that it was fixing, but as long as we implement round_rate in > addition to recalc_rate it should work. > > [1] > http://lkml.kernel.org/r/1428392806-14538-1-git-send-email-jy0922.shim@sa > msung.com Sorry for the second submission then. That really is from quite a while ago. Reintroducing a secondary ops-struct with a round-rate sounds fine for my side - I'll give that a shot tomorrow. Heiko
diff --git a/drivers/clk/clk-divider.c b/drivers/clk/clk-divider.c index ded3ff4..6dfe261 100644 --- a/drivers/clk/clk-divider.c +++ b/drivers/clk/clk-divider.c @@ -391,6 +391,10 @@ static int clk_divider_set_rate(struct clk_hw *hw, unsigned long rate, unsigned long flags = 0; u32 val; + /* don't do writes in read-only case, as it may be unsafe */ + if (divider->flags & CLK_DIVIDER_READ_ONLY) + return 0; + value = divider_get_val(rate, parent_rate, divider->table, divider->width, divider->flags);
Commit e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") removed the special ops struct for read-only clocks and instead opted to handle them regularly. On the rk3368 this results in breakage as aclkm now gets set a value. While it is the same divider value, the A53 core still doesn't like it, which can result in the cpu ending up in a hang. The reason being that "ACLKENMasserts one clock cycle before the rising edge of ACLKM" and the clock should only be touched when STANDBYWFIL2 is asserted. So make sure read-only clocks don't touch the clock-register at all even if only writing the same value, as even writing the same value may not be safe in all cases. Fixes: e6d5e7d90be9 ("clk-divider: Fix READ_ONLY when divider > 1") Reported-by: Zhang Qing <zhangqing@rock-chips.com> Signed-off-by: Heiko Stuebner <heiko@sntech.de> Reviewed-by: James Hogan <james.hogan@imgtec.com> --- changes in v2: - emphasize that even setting the same divider may be unsafe drivers/clk/clk-divider.c | 4 ++++ 1 file changed, 4 insertions(+)