在 2025-10-09星期四的 11:55 -0400，Sasha Levin写道：
From: Icenowy Zheng uwu@icenowy.me
[ Upstream commit 64581f41f4c4aa1845edeee6bb0c8f2a7103d9aa ]
This commit seems to be not backportable -- it depends on the new driver for the auxiliary device.
I suggest not to backport it.
The reboot / power off operations require communication with the AON firmware too.
As the driver is already present, create an auxiliary device with name "reboot" to match that driver, and pass the AON channel by using platform_data.
Signed-off-by: Icenowy Zheng uwu@icenowy.me Signed-off-by: Ulf Hansson ulf.hansson@linaro.org Signed-off-by: Sasha Levin sashal@kernel.org
LLM Generated explanations, may be completely bogus:
YES
- User-visible bugfix: Without this, TH1520 systems cannot register
platform reboot/poweroff handlers via the AON firmware, so reboot/poweroff won’t work even though the AON protocol and the reboot driver exist. This commit wires them up by instantiating the auxiliary device that the reboot driver matches on.
- Small and contained: Adds a tiny helper and one call in the TH1520
PM domain driver; no architectural changes.
- Specific code changes
- Adds `th1520_pd_reboot_init()` which creates an auxiliary device named `reboot` and passes the AON channel through `platform_data` so the reboot driver can use it: drivers/pmdomain/thead/th1520-pm- domains.c:176 - `adev = devm_auxiliary_device_create(dev, "reboot", aon_chan);` - Hooks it into probe after setting up PM domains and the optional GPU pwrseq: drivers/pmdomain/thead/th1520-pm-domains.c:250 - `ret = th1520_pd_reboot_init(dev, aon_chan);` - Error path is correctly routed back to the provider cleanup.
- Correct driver pairing: The created device name matches the
existing driver’s ID table - Reboot driver expects `th1520_pm_domains.reboot` and consumes `adev->dev.platform_data` as the `th1520_aon_chan` to issue AON RPCs for poweroff/restart: drivers/power/reset/th1520-aon-reboot.c:51 and drivers/power/reset/th1520-aon-reboot.c:82
- Risk and scope:
- TH1520-specific; no impact on other platforms. - Only instantiates an auxiliary device; safe if the reboot driver isn’t present. - No ABI/uAPI changes.
- Important follow-up fix to include: The original change used
`PTR_ERR_OR_ZERO(adev)` with `devm_auxiliary_device_create()`, which returns NULL on failure (not an error pointer). That was fixed by “pmdomain: thead: Fix error pointer vs NULL bug in th1520_pd_reboot_init()” (bbc3110823eca), which changes the return to `-ENODEV` on NULL and returns 0 otherwise: drivers/pmdomain/thead/th1520-pm-domains.c:181. For stable backporting, include this fix alongside the main commit to avoid silently succeeding when the aux device creation fails.
- Stable policy fit:
- Fixes a real functionality gap (reboot/poweroff) for TH1520 users. - Minimal code, clear intent, and contained to the TH1520 PM domain driver. - No feature creep or architectural refactoring.
Recommendation: Backport this commit together with the follow-up fix bbc3110823eca to ensure correct error handling.
drivers/pmdomain/thead/th1520-pm-domains.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
diff --git a/drivers/pmdomain/thead/th1520-pm-domains.c b/drivers/pmdomain/thead/th1520-pm-domains.c index 9040b698e7f7f..5213994101a59 100644 --- a/drivers/pmdomain/thead/th1520-pm-domains.c +++ b/drivers/pmdomain/thead/th1520-pm-domains.c @@ -173,6 +173,16 @@ static int th1520_pd_pwrseq_gpu_init(struct device *dev) adev); } +static int th1520_pd_reboot_init(struct device *dev, + struct th1520_aon_chan *aon_chan) +{ + struct auxiliary_device *adev;
+ adev = devm_auxiliary_device_create(dev, "reboot", aon_chan);
+ return PTR_ERR_OR_ZERO(adev); +}
static int th1520_pd_probe(struct platform_device *pdev) { struct generic_pm_domain **domains; @@ -235,6 +245,10 @@ static int th1520_pd_probe(struct platform_device *pdev) if (ret) goto err_clean_provider; + ret = th1520_pd_reboot_init(dev, aon_chan); + if (ret) + goto err_clean_provider;
return 0; err_clean_provider: