On 11/10/22 11:00 AM, Kevin Hilman wrote:
jerome Neanne <jneanne@xxxxxxxxxxxx> writes:
On 09/11/2022 22:59, Andrew Davis wrote:
On 11/7/22 3:14 PM, Kevin Hilman wrote:
Nishanth Menon <nm@xxxxxx> writes:
On 13:58-20221104, jerome Neanne wrote:
[...]
Can you try an compile with W=1 please.
This raise one warning on mfd:
drivers/mfd/tps65219.c:28:12: warning: ‘tps65219_soft_shutdown’
defined but
not used [-Wunused-function]
28 | static int tps65219_soft_shutdown(struct tps65219 *tps)
| ^~~~~~~~~~~~~~~~~~~~~~
soft_shutdown has been validated and is used in TI baseline even if not
hooked in upstream version further to this review:
https://lore.kernel.org/lkml/20220825150224.826258-5-msp@xxxxxxxxxxxx/
It was a TI requirement to implement it...
Let me know if you want me to remove this function or if we can keep
it like
this.
There are platforms without psci, correct? I think the comment was to
drop the force override with system-power-controller property,
if (!pm_power_off) {
tps65219_i2c_client = client;
pm_power_off = &tps65219_pm_power_off;
}
Could still be valid for such platforms, no? I do see that the
capability that the PMIC has - which is software shutdown is a valid
feature that we support in many different PMIC drivers. Is'nt the job of
the driver to introduce the functionality in a manner that is
appropriate to the OS framework?
Yeah, I think Nishanth is right here.
We should probably keep the `if (!pm_power_off)` part so the PMIC will
be used if PSCI is not, but it also allows an easy way to test/use the
PMIC
shutdown functionality downstream if needed.
Then should be using the sys-off handler API[0] so it doesn't block PSCI
which is also switching over[1].
Andrew
[0] https://lwn.net/Articles/894511/
[1] https://www.spinics.net/lists/arm-kernel/msg1024127.html
Can we go for upstream with v7 without tps65219_soft_shutdown. Then if
everyone agrees with Andrew proposal, I'll submit a separate patch which
adds implementation of tps65219_soft_shutdown support through sys-off
handler.
So that we are not blocking upstream in case further
discussions/alignment are required.
Seems OK to me. Nishanth? Andrew?
But I think you'll need to at least submit a v8 without the unused
code/dead code that Lee pointed out.
If you need the v8 anyway, then add support through sys-off in
that spin, should only be a couple lines of change.
Andrew
