Re: [RFC][PATCH] misc: Introduce reboot_reason driver

On Tue, Dec 8, 2015 at 2:26 PM, Rob Herring [off-list ref] wrote:

On Tue, Dec 8, 2015 at 3:29 PM, John Stultz [off-list ref] wrote:

quoted

This patch adds a basic driver to allow for commands like
"reboot bootloader" and "reboot recovery" to communicate this
reboot-reason to the bootloader.

This is commonly done on Android devices, in order to reboot
the device into fastboot or recovery mode. It also supports
custom OEM specific commands, via "reboot oem-<value>".

What are some examples of OEM specific commands?

Actually, I don't know. That bit was preserved from the 3.4 based
logic in a vendor tree.

I can drop it for now if you'd rather.

quoted

This driver pulls the phys memory address from DT as well as
the magic reason values that are written to the address for
each mode.

Starting with what does the h/w look like, this is typically
implemented with some sort of persistent register(s) either in the SOC
or PMIC. So I think persistent memory/registers is what we need to
describe in DT. Perhaps this could be tied into pstore (an overkill
for a single register, but useful if you already have pstore support
for persistent memory)?

Hrm. Yea. I'm hesitant to tie it into pstore, but that's partly due to
not wanting the bootloader to have to parse the pstore area (ideally
the bootloader doesn't touch it).

To me having the bootloader reserve a page of memory and having the
kernel map and write that reserved page makes the most sense to me.
It then being special memory mapped registers or just a reserved page
can be somewhat equivalent.

But maybe I'm being naive?

The 2nd part is which register to use and the
mapping of values to reboot reason. Do these vary within SOC families?
If not, then we should just hardcode them in the reboot drivers which
are already vendor specific.

I'm not aware of differences between SOC families for the values,
though the addresses might change.

Also, while trying to standardize the values for reboot reason
probably won't work short term, we should define something so people
will start using it.

Ack. Though if the values are mostly custom/magic, is having them
defined in the DT problematic? Or do you just want  macros for
something like something like
   reason,bootloader = <GENERIC_REASON_BOOTLOADER>; ?

We also should consider any implications with
PSCI.

Sorry. I'm ignorant here. What would those implications possibly be?

thanks
-john
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html