• Replaced the DEFINE_CLASS_0 -> struct driver_t
• removed pinctrl_single_pins_write / read
• update pinctrl_single_syscon_unlocked_write_4/read_4 to honor pinctrl-single,register-width. (even if the functionnames will be misleading and give the impression its an 4byte version of read/write)
• removed pinctrl_single_syscon_unlocked_modify_4
• update error handling after fdt_pinctrl_register() call to release syscon in case of failure.

Looks mostly good to me.

Thank you Manu. Followed your advice to move the bus_alloc_resource/syscon interface to the end of the attach function to avoid complicated error paths.

Replaced by D42124

I'm pretty sure that the last time I've tested my pandaboard (maybe 4-5 years ago) there were already problems.

In D41759#951401, @andrew wrote:

Why is the kernel getting into the undefined instruction handler? It should only use VFP instructions from within a specific section marked by the fpu_kern_* functions

In D41896#955137, @oskar.holmlund_ohdata.se wrote:

Mike, no you need more patches for it to boot. You will get it this week.
Beaglebone black boots with some u-boot issues regarding the ethernet and USB due to DTS missmatch to the one provided by Linux..
Pocketbeagle will not boot if you have a uboot version >= 2003.01.

So, apparently the goal of simple-pm-bus is only to enable runtime pm in the Linux kernel for the whole child tree, while simple-bus doesn't do that.
Apparently the bindings are wrong suggesting that either power-domain or clock is required and the Linux code seems to indicate this ...

Mike, no you need more patches for it to boot. You will get it this week.
Beaglebone black boots with some u-boot issues regarding the ethernet and USB due to DTS missmatch to the one provided by Linux..
Pocketbeagle will not boot if you have a uboot version >= 2003.01.

In D41896#954928, @manu wrote:

Since TI port is dead

Based on the bindings it looks like that clock is required if power-domain is missing (and vice-versa) but TI doesn't seems to do that, what's the point of TI using this compatible if they do not use it properly ?

Looks mostly ok, I guess TI loves to reimplement simple-bus like things ...
Same comments as Mike though.

Since TI port is dead I think that this driver should be made generic and not patching it this way.
Both pinctrl-single and pinconf-single aren't specific for TI.

Seems fine to me, I could actually approve this one but I'll wait for others to take a look.

What does this driver actually do? It looks like its only role is to show up in the namespace.

I'm not sure what this driver actually does, so I'm probably not the one to approve it. Just a few cosmetic issues.

Why is the kernel getting into the undefined instruction handler? It should only use VFP instructions from within a specific section marked by the fpu_kern_* functions

Approved.

Remove core-team, as we do not need to get involved.
Would still like to see someone commit this...

Fix styling error.

Will someone commit this? It's been approved for a while...

It looks correct, however I don't know much about the 32-bit arm busdma code.

Andrew: Do you have time to have a quick look here?

Updated patch a bit. Maybe this piece of code can be removed altogether ...

I've suggested some optimizations. Have a look at the PR.

Looks good.

What is the current state of this issue?

Any reason the clock are enabled after interrupts ? I suspect that some stuff will rely on the clock being activated at attach.
The bindings have this comment :

Hence, the bus controller's PM domain and/or
clock must be enabled for child devices connected to the bus (either
on-SoC or externally) to function.

D27889 are required to get AM335x to boot.

Update for latest DTS.
Previous I created clk_link in ti_prcm, but latest DTS introduce clkctrl in the very first ti,sysc nodes as a consequence the clk_link will be created in ti_sysc.c.
This patch alone will not boot the am335x, need a simplepmbus driver aswell.

• cleanup gpio gdbclk in ti_clk_clkctrl and use a TI_CLKCTRL_GATE subclock instead
• prepare for am572x subtypes, am335x only use TI_CLKCTRL_GATE
• Fix some wrong mappings like timer7 and so on

Note: I am a first time reviewer, and I hope that I do not something wrong.

I'll leave this for a few days in case someone have an objection.

In D29137#652543, @lwhsu wrote:

.dD field in manual page needs to be updated.

Remove share/man/man9/pwmbus.9
Fix space around &

Looks ok, could you split the updated man page change not related with this review ?

.dD field in manual page needs to be updated.

Update ti_prcm.c to add clk link in probe() instead of attach to avoid the need to declare syscon_generic.c attach as public.
Remove attach() function in ti_prcm.c and use inherited one from syscon_generic.c

Remove syscon_generic_handle from syscon_generic.c and add into ti_prcm.c and ti_scm_syscon.c

• Add parents for ti,clkctrl.
• since the node scm_conf@0 (compatible "syscon", "simple-bus") needs to be attached as soon as posible during boot and syscon_generic are in pass BUS_PASS_DEFAULT I readded ti_scm_syscon.c and removed more or less everything but the probe() function and "subclass" syscon_generic. Added checks in probe() to ensure it not connect to other SoCs syscon nodes.

Yes, this revision is nothing to care about anymore.

Fix cpu reset functions in ti_prcm.c

Not possible for 12.2; for 12.3 we need to make sure all of the required changes are merged to stable/12.