diff options
| author | Tim Chen <tim.c.chen@linux.intel.com> | 2026-09-21 17:37:22 -0700 |
|---|---|---|
| committer | Ingo Molnar <mingo@kernel.org> | 2026-09-22 10:49:53 +0200 |
| commit | 0d6526f82c3cdefcca47f73f5fc08dc6f335eac6 (patch) | |
| tree | 176dc2d685debcfb082c151ee94f2425ac96b364 /Documentation/input/uinput.rst | |
| download | linux-stable-0d6526f82c3cdefcca47f73f5fc08dc6f335eac6.tar.gz linux-stable-0d6526f82c3cdefcca47f73f5fc08dc6f335eac6.zip | |
sched/cache: Keep nr_pref_llc_running in the runnable domain, to fix LLC mis-scheduling buggrafted
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 <zhanxusheng@xiaomi.com>
Suggested-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kayra Cizmeci <kayracizmeci@gmail.com>
Cc: <stable@kernel.org> # v7.2.x
Link: https://patch.msgid.link/06af61afedac32e6477f57feb4d658f6c411c3af.1790035273.git.tim.c.chen@linux.intel.com
Diffstat (limited to 'Documentation/input/uinput.rst')
| -rw-r--r-- | Documentation/input/uinput.rst | 245 |
1 files changed, 245 insertions, 0 deletions
diff --git a/Documentation/input/uinput.rst b/Documentation/input/uinput.rst new file mode 100644 index 000000000..30fe80e32 --- /dev/null +++ b/Documentation/input/uinput.rst @@ -0,0 +1,245 @@ +============= +uinput module +============= + +Introduction +============ + +uinput is a kernel module that makes it possible to emulate input devices +from userspace. By writing to /dev/uinput (or /dev/input/uinput) device, a +process can create a virtual input device with specific capabilities. Once +this virtual device is created, the process can send events through it, +that will be delivered to userspace and in-kernel consumers. + +Interface +========= + +:: + + linux/uinput.h + +The uinput header defines ioctls to create, set up, and destroy virtual +devices. + +libevdev +======== + +libevdev is a wrapper library for evdev devices that provides interfaces to +create uinput devices and send events. libevdev is less error-prone than +accessing uinput directly, and should be considered for new software. + +For examples and more information about libevdev: +https://www.freedesktop.org/software/libevdev/doc/latest/ + +Examples +======== + +Keyboard events +--------------- + +This first example shows how to create a new virtual device, and how to +send a key event. All default imports and error handlers were removed for +the sake of simplicity. + +.. code-block:: c + + #include <linux/uinput.h> + + void emit(int fd, int type, int code, int val) + { + struct input_event ie; + + ie.type = type; + ie.code = code; + ie.value = val; + /* timestamp values below are ignored */ + ie.time.tv_sec = 0; + ie.time.tv_usec = 0; + + write(fd, &ie, sizeof(ie)); + } + + int main(void) + { + struct uinput_setup usetup; + + int fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK); + + + /* + * The ioctls below will enable the device that is about to be + * created, to pass key events, in this case the space key. + */ + ioctl(fd, UI_SET_EVBIT, EV_KEY); + ioctl(fd, UI_SET_KEYBIT, KEY_SPACE); + + memset(&usetup, 0, sizeof(usetup)); + usetup.id.bustype = BUS_USB; + usetup.id.vendor = 0x1234; /* sample vendor */ + usetup.id.product = 0x5678; /* sample product */ + strcpy(usetup.name, "Example device"); + + ioctl(fd, UI_DEV_SETUP, &usetup); + ioctl(fd, UI_DEV_CREATE); + + /* + * On UI_DEV_CREATE the kernel will create the device node for this + * device. We are inserting a pause here so that userspace has time + * to detect, initialize the new device, and can start listening to + * the event, otherwise it will not notice the event we are about + * to send. This pause is only needed in our example code! + */ + sleep(1); + + /* Key press, report the event, send key release, and report again */ + emit(fd, EV_KEY, KEY_SPACE, 1); + emit(fd, EV_SYN, SYN_REPORT, 0); + emit(fd, EV_KEY, KEY_SPACE, 0); + emit(fd, EV_SYN, SYN_REPORT, 0); + + /* + * Give userspace some time to read the events before we destroy the + * device with UI_DEV_DESTROY. + */ + sleep(1); + + ioctl(fd, UI_DEV_DESTROY); + close(fd); + + return 0; + } + +Mouse movements +--------------- + +This example shows how to create a virtual device that behaves like a physical +mouse. + +.. code-block:: c + + #include <linux/uinput.h> + + /* emit function is identical to of the first example */ + + int main(void) + { + struct uinput_setup usetup; + int i = 50; + + int fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK); + + /* enable mouse button left and relative events */ + ioctl(fd, UI_SET_EVBIT, EV_KEY); + ioctl(fd, UI_SET_KEYBIT, BTN_LEFT); + + ioctl(fd, UI_SET_EVBIT, EV_REL); + ioctl(fd, UI_SET_RELBIT, REL_X); + ioctl(fd, UI_SET_RELBIT, REL_Y); + + memset(&usetup, 0, sizeof(usetup)); + usetup.id.bustype = BUS_USB; + usetup.id.vendor = 0x1234; /* sample vendor */ + usetup.id.product = 0x5678; /* sample product */ + strcpy(usetup.name, "Example device"); + + ioctl(fd, UI_DEV_SETUP, &usetup); + ioctl(fd, UI_DEV_CREATE); + + /* + * On UI_DEV_CREATE the kernel will create the device node for this + * device. We are inserting a pause here so that userspace has time + * to detect, initialize the new device, and can start listening to + * the event, otherwise it will not notice the event we are about + * to send. This pause is only needed in our example code! + */ + sleep(1); + + /* Move the mouse diagonally, 5 units per axis */ + while (i--) { + emit(fd, EV_REL, REL_X, 5); + emit(fd, EV_REL, REL_Y, 5); + emit(fd, EV_SYN, SYN_REPORT, 0); + usleep(15000); + } + + /* + * Give userspace some time to read the events before we destroy the + * device with UI_DEV_DESTROY. + */ + sleep(1); + + ioctl(fd, UI_DEV_DESTROY); + close(fd); + + return 0; + } + + +uinput old interface +-------------------- + +Before uinput version 5, there wasn't a dedicated ioctl to set up a virtual +device. Programs supporting older versions of uinput interface need to fill +a uinput_user_dev structure and write it to the uinput file descriptor to +configure the new uinput device. New code should not use the old interface +but interact with uinput via ioctl calls, or use libevdev. + +.. code-block:: c + + #include <linux/uinput.h> + + /* emit function is identical to of the first example */ + + int main(void) + { + struct uinput_user_dev uud; + int version, rc, fd; + + fd = open("/dev/uinput", O_WRONLY | O_NONBLOCK); + rc = ioctl(fd, UI_GET_VERSION, &version); + + if (rc == 0 && version >= 5) { + /* use UI_DEV_SETUP */ + return 0; + } + + /* + * The ioctls below will enable the device that is about to be + * created, to pass key events, in this case the space key. + */ + ioctl(fd, UI_SET_EVBIT, EV_KEY); + ioctl(fd, UI_SET_KEYBIT, KEY_SPACE); + + memset(&uud, 0, sizeof(uud)); + snprintf(uud.name, UINPUT_MAX_NAME_SIZE, "uinput old interface"); + write(fd, &uud, sizeof(uud)); + + ioctl(fd, UI_DEV_CREATE); + + /* + * On UI_DEV_CREATE the kernel will create the device node for this + * device. We are inserting a pause here so that userspace has time + * to detect, initialize the new device, and can start listening to + * the event, otherwise it will not notice the event we are about + * to send. This pause is only needed in our example code! + */ + sleep(1); + + /* Key press, report the event, send key release, and report again */ + emit(fd, EV_KEY, KEY_SPACE, 1); + emit(fd, EV_SYN, SYN_REPORT, 0); + emit(fd, EV_KEY, KEY_SPACE, 0); + emit(fd, EV_SYN, SYN_REPORT, 0); + + /* + * Give userspace some time to read the events before we destroy the + * device with UI_DEV_DESTROY. + */ + sleep(1); + + ioctl(fd, UI_DEV_DESTROY); + + close(fd); + return 0; + } + |
