[PATCH v4 0/2] rtc: efi: defer to the ACPI TAD if it can drive the RTC

[PATCH v4 0/2] rtc: efi: defer to the ACPI TAD if it can drive the RTC

[Hamza Mahfooz]

On Microsoft C2141 and C4143 systems, the EFI GetTime() runtime service
faults when called by rtc-efi, which causes all EFI runtime services to
be disabled, even though the RTC can be driven via the ACPI Time and
Alarm Device (TAD) instead, which acpi-tad registers an RTC class device
for since commit 8a1e7f4b1764 ("ACPI: TAD: Add RTC class device
interface").

As suggested by Ard, add acpi_has_rtc() to the TAD driver (patch 1),
and make rtc-efi decline to probe if it returns true (patch 2).

v2:
- Rename acpi_tad_rtc_available() to acpi_has_rtc(), move it to within
  acpi-tad (which allows us to simplify it quite a bit) and declare it
  in linux/acpi.h. (Ard)
- Add acpi_disabled check. (Ard)
- Make rtc-efi decline to probe instead of not registering the device in
  efisubsys_init(). (Ard)
- Make RTC_DRV_EFI depend on !ACPI || ACPI_TAD. (Ard)
v3:
- Make RTC_DRV_EFI depend on ACPI_TAD || !ACPI_TAD. (Sashiko)
v4:
- Move acpi-tad code to existing CONFIG_RTC_CLASS #ifdef. (Rafael)
- Add IS_ENABLED(CONFIG_RTC_CLASS) case to linux/acpi.h.

Hamza Mahfooz (2):
  ACPI: TAD: introduce acpi_has_rtc()
  rtc: efi: defer to the ACPI TAD if it can drive the RTC

 drivers/acpi/acpi_tad.c | 48 ++++++++++++++++++++++++++++++++++++-----
 drivers/rtc/Kconfig     |  5 +++++
 drivers/rtc/rtc-efi.c   |  4 ++++
 include/linux/acpi.h    |  6 ++++++
 4 files changed, 58 insertions(+), 5 deletions(-)

-- 
2.56.0

[PATCH v4 1/2] ACPI: TAD: introduce acpi_has_rtc()

[Hamza Mahfooz]

The same RTC may also be accessible by other means, like the EFI
GetTime() and SetTime() runtime services used by rtc-efi, which are best
avoided if there is a native ACPI interface to the RTC.

To allow other RTC drivers to stand down in that case, add an exported
helper, acpi_has_rtc(), that returns true if there is a TAD the driver
will register an RTC class device for.

The helper looks for a platform device matching the driver's ID table
whose _GCP indicates real time support, like acpi_tad_probe() checks,
instead of checking whether the driver has bound to such a device, so
that callers need not depend on the order in which drivers are probed.

Suggested-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org> (v3)
Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>
---
 drivers/acpi/acpi_tad.c | 48 ++++++++++++++++++++++++++++++++++++-----
 include/linux/acpi.h    |  6 ++++++
 2 files changed, 49 insertions(+), 5 deletions(-)

diff --git a/drivers/acpi/acpi_tad.c b/drivers/acpi/acpi_tad.c
index 45459a4ca..a07f4f4fd 100644
--- a/drivers/acpi/acpi_tad.c
+++ b/drivers/acpi/acpi_tad.c
@@ -622,9 +622,52 @@ static const struct attribute_group acpi_tad_group = {
 
 __ATTRIBUTE_GROUPS(acpi_tad);
 
+static const struct acpi_device_id acpi_tad_ids[] = {
+	{"ACPI000E", 0},
+	{}
+};
+
 #ifdef CONFIG_RTC_CLASS
 /* RTC class device interface */
 
+static int acpi_tad_rtc_match(struct device *dev, const void *data)
+{
+	unsigned long long caps;
+	acpi_status status;
+
+	if (!acpi_match_device(acpi_tad_ids, dev))
+		return 0;
+
+	status = acpi_evaluate_integer(ACPI_HANDLE(dev), "_GCP", NULL, &caps);
+
+	return ACPI_SUCCESS(status) && (caps & ACPI_TAD_RT);
+}
+
+/**
+ * acpi_has_rtc - Check if the RTC should be driven via the ACPI TAD
+ *
+ * Check if there is a TAD implementing the real time methods that this driver
+ * will register an RTC class device for.
+ *
+ * Return: %true if the RTC should be driven via the ACPI TAD, %false otherwise.
+ */
+bool acpi_has_rtc(void)
+{
+	struct device *dev;
+
+	if (acpi_disabled)
+		return false;
+
+	dev = bus_find_device(&platform_bus_type, NULL, NULL,
+			      acpi_tad_rtc_match);
+	if (!dev)
+		return false;
+
+	put_device(dev);
+	return true;
+}
+EXPORT_SYMBOL_GPL(acpi_has_rtc);
+
 static void acpi_tad_rt_to_tm(struct acpi_tad_rt *rt, struct rtc_time *tm)
 {
 	tm->tm_year = rt->year - 1900;
@@ -894,11 +937,6 @@ static int acpi_tad_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static const struct acpi_device_id acpi_tad_ids[] = {
-	{"ACPI000E", 0},
-	{}
-};
-
 static struct platform_driver acpi_tad_driver = {
 	.driver = {
 		.name = "acpi-tad",
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index ddacac812..74c60b8e8 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -1553,6 +1553,12 @@ extern bool acpi_has_watchdog(void);
 static inline bool acpi_has_watchdog(void) { return false; }
 #endif
 
+#if IS_REACHABLE(CONFIG_ACPI_TAD) && IS_ENABLED(CONFIG_RTC_CLASS)
+bool acpi_has_rtc(void);
+#else
+static inline bool acpi_has_rtc(void) { return false; }
+#endif
+
 #ifdef CONFIG_ACPI_SPCR_TABLE
 extern bool qdf2400_e44_present;
 int acpi_parse_spcr(bool enable_earlycon, bool enable_console);
-- 
2.56.0

[PATCH v4 2/2] rtc: efi: defer to the ACPI TAD if it can drive the RTC

[Hamza Mahfooz]

EFI runtime services are best avoided where possible, and ACPI defines a
standard interface to the RTC in the form of the Time and Alarm Device
(TAD), which the acpi-tad driver has been exposing as an RTC class
device since commit 8a1e7f4b1764 ("ACPI: TAD: Add RTC class device
interface").

So, make rtc-efi decline to probe, before calling GetTime(), if
acpi_has_rtc() says that the RTC can be driven via the TAD.

Also, to ensure that acpi_has_rtc() is always reachable from rtc-efi
when ACPI_TAD is enabled, make RTC_DRV_EFI depend on
ACPI_TAD || !ACPI_TAD.

Suggested-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>
---
 drivers/rtc/Kconfig   | 5 +++++
 drivers/rtc/rtc-efi.c | 4 ++++
 2 files changed, 9 insertions(+)

diff --git a/drivers/rtc/Kconfig b/drivers/rtc/Kconfig
index 1e863f1d3..5b7e9d2a8 100644
--- a/drivers/rtc/Kconfig
+++ b/drivers/rtc/Kconfig
@@ -1262,10 +1262,15 @@ config RTC_DRV_DA9063
 config RTC_DRV_EFI
 	tristate "EFI RTC"
 	depends on EFI && !X86
+	depends on ACPI_TAD || !ACPI_TAD
 	help
 	  If you say yes here you will get support for the EFI
 	  Real Time Clock.
 
+	  On ACPI systems, the ACPI Time and Alarm Device (TAD) driver
+	  is used instead of this one if the TAD can drive the RTC (assuming
+	  it's enabled).
+
 	  This driver can also be built as a module. If so, the module
 	  will be called rtc-efi.
 
diff --git a/drivers/rtc/rtc-efi.c b/drivers/rtc/rtc-efi.c
index b4f44999e..545b21b17 100644
--- a/drivers/rtc/rtc-efi.c
+++ b/drivers/rtc/rtc-efi.c
@@ -10,6 +10,7 @@
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
+#include <linux/acpi.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/stringify.h>
@@ -192,6 +193,9 @@ static int __init efi_rtc_probe(struct platform_device *dev)
 	efi_time_t eft;
 	efi_time_cap_t cap;
 
+	if (acpi_has_rtc())
+		return -ENODEV;
+
 	/* First check if the RTC is usable */
 	if (efi.get_time(&eft, &cap) != EFI_SUCCESS)
 		return -ENODEV;
-- 
2.56.0

Re: [PATCH v4 2/2] rtc: efi: defer to the ACPI TAD if it can drive the RTC

[sashiko-bot]

> EFI runtime services are best avoided where possible, and ACPI defines a
> standard interface to the RTC in the form of the Time and Alarm Device
> (TAD), which the acpi-tad driver has been exposing as an RTC class
> device since commit 8a1e7f4b1764 ("ACPI: TAD: Add RTC class device
> interface").
> 
> So, make rtc-efi decline to probe, before calling GetTime(), if
> acpi_has_rtc() says that the RTC can be driven via the TAD.
> 
> Also, to ensure that acpi_has_rtc() is always reachable from rtc-efi
> when ACPI_TAD is enabled, make RTC_DRV_EFI depend on
> ACPI_TAD || !ACPI_TAD.
> 
> Suggested-by: Ard Biesheuvel <ardb@kernel.org>
> Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
> Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261027220537.3443104-1-hamzamahfooz@linux.microsoft.com?part=2

Re: [PATCH v4 1/2] ACPI: TAD: introduce acpi_has_rtc()

[sashiko-bot]

> The same RTC may also be accessible by other means, like the EFI
> GetTime() and SetTime() runtime services used by rtc-efi, which are best
> avoided if there is a native ACPI interface to the RTC.
> 
> To allow other RTC drivers to stand down in that case, add an exported
> helper, acpi_has_rtc(), that returns true if there is a TAD the driver
> will register an RTC class device for.
> 
> The helper looks for a platform device matching the driver's ID table
> whose _GCP indicates real time support, like acpi_tad_probe() checks,
> instead of checking whether the driver has bound to such a device, so
> that callers need not depend on the order in which drivers are probed.
> 
> Suggested-by: Ard Biesheuvel <ardb@kernel.org>
> Reviewed-by: Ard Biesheuvel <ardb@kernel.org> (v3)
> Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261027220537.3443104-1-hamzamahfooz@linux.microsoft.com?part=1