From 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 Mon Sep 17 00:00:00 2001 From: Tim Chen Date: Mon, 21 Sep 2026 17:37:22 -0700 Subject: sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling bug alb_break_llc() decides whether to break LLC preference during active load balance. It does so by testing that every runnable fair task on the source rq prefers its LLC: env->src_rq->nr_pref_llc_running == env->src_rq->cfs.h_nr_runnable But the two counters cover different sets. nr_pref_llc_running is updated in account_llc_enqueue()/account_llc_dequeue(), next to cfs_rq->nr_queued, so it follows queued tasks. h_nr_runnable is updated in set_delayed()/ clear_delayed() and drops delay-dequeued tasks. So under DELAY_DEQUEUE, a preferring task that goes to sleep stays counted in nr_pref_llc_running while h_nr_runnable falls. The equality then breaks, alb_break_llc() returns false, and active balance is free to pull a task off its preferred LLC. Active balance only moves runnable tasks, and this is the only LLC check it consults: once the stopper runs, LBF_ACTIVE_LB skips the per-task test in can_migrate_task(). The runnable set is the one we want. Fix it on the counter side. A task should be counted in nr_pref_llc_running exactly while it is both queued on its preferred LLC (pref_llc_queued) and runnable (!sched_delayed). Define that membership once in task_pref_llc_runnable(), and adjust the counter only through pref_llc_running_inc()/pref_llc_running_dec() from the four sites that change either input: account_llc_enqueue(), account_llc_dequeue(), set_delayed() and clear_delayed(). Gating every update on the same predicate keeps the delay, wake and dequeue paths from double-counting or underflowing; see the comments at those sites for the ordering. nr_llc_running and sd->llc_counts are not touched and stay on queued semantics. Fixes: 714059f79ff0 ("sched/cache: Handle moving single tasks to/from their preferred LLC") Closes: https://lore.kernel.org/lkml/20260827135000.735138-1-zhanxusheng@xiaomi.com/ Reported-by: Zhan Xusheng Suggested-by: Chen Yu Signed-off-by: Tim Chen Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Ingo Molnar Reviewed-by: Kayra Cizmeci Cc: # v7.2.x Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com --- Documentation/userspace-api/gpio/chardev.rst | 116 +++++++++++++++++++++++++++ 1 file changed, 116 insertions(+) create mode 100644 Documentation/userspace-api/gpio/chardev.rst (limited to 'Documentation/userspace-api/gpio/chardev.rst') diff --git a/Documentation/userspace-api/gpio/chardev.rst b/Documentation/userspace-api/gpio/chardev.rst new file mode 100644 index 000000000..c58dd9771 --- /dev/null +++ b/Documentation/userspace-api/gpio/chardev.rst @@ -0,0 +1,116 @@ +.. SPDX-License-Identifier: GPL-2.0 + +=================================== +GPIO Character Device Userspace API +=================================== + +This is latest version (v2) of the character device API, as defined in +``include/uapi/linux/gpio.h.`` + +First added in 5.10. + +.. note:: + Do NOT abuse userspace APIs to control hardware that has proper kernel + drivers. There may already be a driver for your use case, and an existing + kernel driver is sure to provide a superior solution to bitbashing + from userspace. + + Read Documentation/driver-api/gpio/drivers-on-gpio.rst to avoid reinventing + kernel wheels in userspace. + + Similarly, for multi-function lines there may be other subsystems, such as + Documentation/spi/index.rst, Documentation/i2c/index.rst, + Documentation/driver-api/pwm.rst, Documentation/w1/index.rst etc, that + provide suitable drivers and APIs for your hardware. + +Basic examples using the character device API can be found in ``tools/gpio/*``. + +The API is based around two major objects, the :ref:`gpio-v2-chip` and the +:ref:`gpio-v2-line-request`. + +.. _gpio-v2-chip: + +Chip +==== + +The Chip represents a single GPIO chip and is exposed to userspace using device +files of the form ``/dev/gpiochipX``. + +Each chip supports a number of GPIO lines, +:c:type:`chip.lines`. Lines on the chip are identified by an +``offset`` in the range from 0 to ``chip.lines - 1``, i.e. `[0,chip.lines)`. + +Lines are requested from the chip using gpio-v2-get-line-ioctl.rst +and the resulting line request is used to access the GPIO chip's lines or +monitor the lines for edge events. + +Within this documentation, the file descriptor returned by calling `open()` +on the GPIO device file is referred to as ``chip_fd``. + +Operations +---------- + +The following operations may be performed on the chip: + +.. toctree:: + :titlesonly: + + Get Line + Get Chip Info + Get Line Info + Watch Line Info + Unwatch Line Info + Read Line Info Changed Events + +.. _gpio-v2-line-request: + +Line Request +============ + +Line requests are created by gpio-v2-get-line-ioctl.rst and provide +access to a set of requested lines. The line request is exposed to userspace +via the anonymous file descriptor returned in +:c:type:`request.fd` by gpio-v2-get-line-ioctl.rst. + +Within this documentation, the line request file descriptor is referred to +as ``req_fd``. + +Operations +---------- + +The following operations may be performed on the line request: + +.. toctree:: + :titlesonly: + + Get Line Values + Set Line Values + Read Line Edge Events + Reconfigure Lines + +Types +===== + +This section contains the structs and enums that are referenced by the API v2, +as defined in ``include/uapi/linux/gpio.h``. + +.. kernel-doc:: include/uapi/linux/gpio.h + :identifiers: + gpio_v2_line_attr_id + gpio_v2_line_attribute + gpio_v2_line_changed_type + gpio_v2_line_config + gpio_v2_line_config_attribute + gpio_v2_line_event + gpio_v2_line_event_id + gpio_v2_line_flag + gpio_v2_line_info + gpio_v2_line_info_changed + gpio_v2_line_request + gpio_v2_line_values + gpiochip_info + +.. toctree:: + :hidden: + + error-codes -- cgit v1.3.1