Regression in next with "mfd: twl6040: The chip does not support bulk access"

[Tony Lindgren]

* Peter Ujfalusi <peter.ujfalusi at ti.com> [160923 00:21]:
> On 09/22/16 21:07, Tony Lindgren wrote:
> > Hi,
> > 
> > Looks like commit 7a17e47f6403 ("mfd: twl6040: The chip does not
> > support bulk access") breaks at least omap4-duovero.
> 
> That's odd. I see no such errors on omap4 PandaBoard-ES nor on omap5 uEVM.
> The IRQ status is one register so even in bulk access it is one read. So the
> use_single_rw should have no effect on the access.

Oh it's just one register.

> The only time when regmap would try to use bulk access to twl6040 is when we
> execute regcache_sync() on it after resuming the chip and this would fail at
> the first time when it tries to restore more than one consecutive registers.
> 
> I just tested things on next-20160915 and I see no errors at all.

OK interesting.

> > I now get tons of errors:
> > 
> > Skipping twl internal clock init and using bootloader value (unknown osc rate)
> > twl 0-0048: PIH (irq 332) nested IRQs
> > of_get_named_gpiod_flags: parsed 'ti,audpwron-gpio' property of node '/ocp/i2c at 48070000/t
> > wl at 4b[0]' - status (0)
> > omap_i2c 48070000.i2c: bus 0 rev0.10 at 400 kHz
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > twl6040 0-004b: Failed to read IRQ status: -16
> > 
> > It seems the regmap irqs don't work with use_single_rw?

So obviously that's a wrong conclusion then.

> > Also seems that twl6040 does support bulk access as things have been
> > working earlier?
> 
> Things kind of worked as we had single register reads and writes, but when I
> tried to power down/up the twl6040 the regcache_sync() would fail to restore
> the registers at the first place when we had two or more registers to update
> in bulk mode. In that case regmap gives an error, but I have not been checking
> the return of the regcache_sync() so the chip comes up, but not all settings
> were restored.
> 
> > Anyways, can you please revert?
> 
> Please don't.
> 
> On which linux-next version you are seeing this?

This was with next-20160922. But testing it again this is just another
regression caused by "softirq: fix tasklet_kill() and its users", so adding
Santosh to Cc.

So no need to revert $subject patch.

Regards,

Tony