user/sven/linux.git/drivers/rtc/rtc-cmos.c, branch v5.4.36

rtc: cmos: Stop using shared IRQ

2020-02-14T21:34:12Z

commit b6da197a2e9670df6f07e6698629e9ce95ab614e upstream. As reported by Guilherme G. Piccoli: ---8<---8<---8<--- The rtc-cmos interrupt setting was changed in the commit 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch") in order to allow shared interrupts; according to that commit's description, some machine got kernel warnings due to the interrupt line being shared between rtc-cmos and other hardware, and rtc-cmos didn't allow IRQ sharing that time. After the aforementioned commit though it was observed a huge increase in lost HPET interrupts in some systems, observed through the following kernel message: [...] hpet1: lost 35 rtc interrupts After investigation, it was narrowed down to the shared interrupts usage when having the kernel option "irqpoll" enabled. In this case, all IRQ handlers are called for non-timer interrupts, if such handlers are setup in shared IRQ lines. The rtc-cmos IRQ handler could be set to hpet_rtc_interrupt(), which will produce the kernel "lost interrupts" message after doing work - lots of readl/writel to HPET registers, which are known to be slow. Although "irqpoll" is not a default kernel option, it's used in some contexts, one being the kdump kernel (which is an already "impaired" kernel usually running with 1 CPU available), so the performance burden could be considerable. Also, the same issue would happen (in a shorter extent though) when using "irqfixup" kernel option. In a quick experiment, a virtual machine with uptime of 2 minutes produced >300 calls to hpet_rtc_interrupt() when "irqpoll" was set, whereas without sharing interrupts this number reduced to 1 interrupt. Machines with more hardware than a VM should generate even more unnecessary HPET interrupts in this scenario. ---8<---8<---8<--- After looking into the rtc-cmos driver history and DSDT table from the Microsoft Surface 3, we may notice that Hans de Goede submitted a correct fix (see dependency below). Thus, we simply revert the culprit commit. Fixes: 079062b28fb4 ("rtc: cmos: prevent kernel warning on IRQ flags mismatch") Depends-on: a1e23a42f1bd ("rtc: cmos: Do not assume irq 8 for rtc when there are no legacy irqs") Reported-by: Guilherme G. Piccoli Cc: Hans de Goede Signed-off-by: Andy Shevchenko Tested-by: Guilherme G. Piccoli Reviewed-by: Hans de Goede Link: https://lore.kernel.org/r/20200123131437.28157-1-andriy.shevchenko@linux.intel.com Signed-off-by: Alexandre Belloni Signed-off-by: Greg Kroah-Hartman

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 152

2019-05-30T18:26:32Z

Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 3029 file(s). Signed-off-by: Thomas Gleixner Reviewed-by: Allison Randal Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070032.746973796@linutronix.de Signed-off-by: Greg Kroah-Hartman

rtc: cmos: Do not export alarm rtc_ops when we do not support alarms

2018-11-07T16:13:55Z

When there is no IRQ configured for the RTC, the rtc-cmos code does not support alarms, all alarm rtc_ops fail with -EIO / -EINVAL. The rtc-core expects a rtc driver which does not support rtc alarms to not have alarm ops at all. Otherwise the wakealarm sysfs attr will read as empty rather then returning an error, making it impossible for userspace to find out beforehand if alarms are supported. A system without an IRQ for the RTC before this patch: [root@localhost ~]# cat /sys/class/rtc/rtc0/wakealarm [root@localhost ~]# After this patch: [root@localhost ~]# cat /sys/class/rtc/rtc0/wakealarm cat: /sys/class/rtc/rtc0/wakealarm: No such file or directory [root@localhost ~]# This fixes gnome-session + systemd trying to use suspend-then-hibernate, which causes systemd to abort the suspend when writing the RTC alarm fails. BugLink: https://github.com/systemd/systemd/issues/9988 Signed-off-by: Hans de Goede Signed-off-by: Alexandre Belloni

rtc: cmos: Remove the `use_acpi_alarm' module parameter for !ACPI

2018-10-04T09:15:57Z

Fix a problem with commit 311ee9c151ad ("rtc: cmos: allow using ACPI for RTC alarm instead of HPET") defining `use_acpi_alarm' module parameter even for non-ACPI platforms, which ignore it. Wrap the definition into #ifdef CONFIG_ACPI and use a static inline wrapper function, hardcoded to return 0 and consequently optimized away for !ACPI, following the existing pattern with HPET handling functions. Signed-off-by: Maciej W. Rozycki Fixes: 311ee9c151ad ("rtc: cmos: allow using ACPI for RTC alarm instead of HPET") Cc: stable@vger.kernel.org # 4.18+ Signed-off-by: Alexandre Belloni

rtc: cmos: Fix non-ACPI undefined reference to `hpet_rtc_interrupt'

2018-10-04T09:15:55Z

Fix a commit 311ee9c151ad ("rtc: cmos: allow using ACPI for RTC alarm instead of HPET") `rtc-cmos' regression causing a link error: drivers/rtc/rtc-cmos.o: In function `cmos_platform_probe': rtc-cmos.c:(.init.text+0x33c): undefined reference to `hpet_rtc_interrupt' rtc-cmos.c:(.init.text+0x3f4): undefined reference to `hpet_rtc_interrupt' with non-ACPI platforms using this driver. The cause is the change of the condition guarding the use of `hpet_rtc_interrupt'. Previously it was a call to `is_hpet_enabled'. That function is static inline and has a hardcoded 0 result for non-ACPI platforms, which imply !HPET_EMULATE_RTC. Consequently the compiler optimized the whole block away including the reference to `hpet_rtc_interrupt', which never made it to the link stage. Now the guarding condition is a call to `use_hpet_alarm', which is not static inline and therefore the compiler may not be able to prove that it actually always returns 0 for non-ACPI platforms. Consequently the build breaks with an unsatisfied reference, because `hpet_rtc_interrupt' is nowhere defined at link time. Fix the problem by marking `use_hpet_alarm' inline. As the `inline' keyword serves as an optimization hint rather than a requirement the compiler is still free to choose whether inlining will be beneficial or not for ACPI platforms. Signed-off-by: Maciej W. Rozycki Fixes: 311ee9c151ad ("rtc: cmos: allow using ACPI for RTC alarm instead of HPET") Cc: stable@vger.kernel.org # 4.18+ Signed-off-by: Alexandre Belloni

rtc: cmos: introduce quirks to enable use_acpi_alarm mode

2018-04-19T16:01:50Z

Use ACPI for RTC Alarm only for Intel platforms 1. with Low Power S0 support 2. with HPET RTC emulation enabled 3. no earlier than 2015 Note that, during the test, it is found that this patch 1. works in 4.15-rc kernel 2. hangs the platform after suspend-to-idle for 2 or 3 times, in 4.15.0 3. works again in 4.16-rc3 kernel. 4. works in the latest 4.15.12 stable kernel. Thus although this patch breaks 4.15.0 kernel for some unknown reason, still, it is safe for both upstream and backport. Signed-off-by: Zhang Rui Signed-off-by: Alexandre Belloni

rtc: cmos: acknowledge ACPI driven wake alarms upon resume

2018-04-19T16:01:50Z

Previously, the RTC alarm is acknowledged either by the cmos rtc irq handler, or by the hpet rtc irq handler. When using ACPI RTC Fixed event as the RTC alarm, the RTC alarm is acknowledged by the ACPI RTC event handler, as addressed in the previous patch. But, when resume from suspend-to-ram (ACPI S3), the ACPI SCI is cleared right after resume, thus the ACPI RTC event handler is not invoked at all, results in the RTC Alarm unacknowledged. Handle this by comparing the current time and the RTC Alarm time in the rtc_cmos driver .resume() callback 1. Assume the wakeup event has already been fired if the RTC Alarm time is earlier than/equal to the current time, and ACK the RTC Alarm. 2. Assume the wakeup event has not been fired if the RTC Alarm time is later than current time, and re-arm it if needed. Signed-off-by: Zhang Rui Signed-off-by: Alexandre Belloni

rtc: cmos: allow using ACPI for RTC alarm instead of HPET

2018-04-19T16:01:50Z

It's found that the HPET timer prevents the platform from entering Low Power S0 on some new Intel platforms. This means that 1. users can still use RTC wake Alarm for suspend-to-idle, but the system never enters Low Power S0, which is a waste of power. or 2. if users want to put the system into Low Power S0, they can not use RTC as the wakeup source. To fix this, we need to stop using the HPET timer for wake alarm. But disabling CONFIG_HPET_EMULATE_RTC is not an option because HPET emulates PIT at the same time, and this is needed on some of these platforms. Thus, introduce a new mode (use_acpi_alarm) to the rtc_cmos driver, so that, even with CONFIG_HPET_EMULATE_RTC enabled, it's still possible to use ACPI SCI for RTC Alarm, including UIE/AIE/wkalrm, instead of HPET. Only necessary changes are made for the new "use_acpi_alarm" mode, including 1. drop all the calls to HPET emulation code, including the HPET irq handler for rtc interrupt. 2. enabling/disabling ACPI RTC Fixed event upon RTC UIE/AIE request. 3. acknowledge the RTC Alarm in ACPI RTC Fixed event handler. There is no functional change made in this patch if the new mode is not enabled. Note: this "use_acpi_alarm" mode is made based on the assumption that ACPI RTC Fixed event is reliable both at runtime and during system wakeup. And this has been verified on a couple of platforms I have, including a MS Surface Pro 4 (SKL), a Lenovo Yoga 900 (SKL), and a HP 9360 (KBL). Signed-off-by: Zhang Rui Signed-off-by: Alexandre Belloni

Merge tag 'rtc-4.17' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux

2018-04-10T17:22:27Z

Pull RTC updates from Alexandre Belloni: "This contains a few series that have been in preparation for a while and that will help systems with RTCs that will fail in 2038, 2069 or 2100. Subsystem: - Add tracepoints - Rework of the RTC/nvmem API to allow drivers to discard struct nvmem_config after registration - New range API, drivers can now expose the useful range of the RTC - New offset API the core is now able to add an offset to the RTC time, modifying the supported range. - Multiple rtc_time64_to_tm fixes - Handle time_t overflow on 32 bit platforms in the core instead of letting drivers do crazy things. - remove rtc_control API New driver: - Intersil ISL12026 Drivers: - Drivers exposing the RTC non volatile memory have been converted to use nvmem - Removed useless time and date validation - Removed an indirection pattern that was a cargo cult from ancient drivers - Removed VLA usage - Fixed a possible race condition in probe functions - AB8540 support is dropped from ab8500 - pcf85363 now has alarm support" * tag 'rtc-4.17' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux: (128 commits) rtc: snvs: Fix usage of snvs_rtc_enable rtc: mt7622: fix module autoloading for OF platform drivers rtc: isl12022: use true and false for boolean values rtc: ab8500: Drop AB8540 support rtc: remove a warning during scripts/kernel-doc step rtc: 88pm860x: remove artificial limitation rtc: 88pm80x: remove artificial limitation rtc: st-lpc: remove artificial limitation rtc: mrst: remove artificial limitation rtc: mv: remove artificial limitation rtc: hctosys: Ensure system time doesn't overflow time_t parisc: time: stop validating rtc_time in .read_time rtc: pcf85063: fix clearing bits in pcf85063_start_clock rtc: at91sam9: Set name of regmap_config rtc: s5m: Remove VLA usage rtc: s5m: Move enum from rtc.h to rtc-s5m.c rtc: remove VLA usage rtc: Add useful timestamp definitions rtc: Add one offset seconds to expand RTC range rtc: Factor out the RTC range validation into rtc_valid_range() ...

mn10300: Remove the architecture

2018-03-09T22:19:56Z

Remove the MN10300 arch as the hardware is defunct. Suggested-by: Arnd Bergmann Signed-off-by: David Howells cc: Masahiro Yamada cc: linux-am33-list@redhat.com Signed-off-by: Arnd Bergmann