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

1 files changed, 54 insertions, 0 deletions

diff --git a/Documentation/gpu/drm-compute.rst b/Documentation/gpu/drm-compute.rst
new file mode 100644
index 000000000..f90c3e63a
--- /dev/null
+++ b/Documentation/gpu/drm-compute.rst
@@ -0,0 +1,54 @@
+==================================
+Long running workloads and compute
+==================================
+
+Long running workloads (compute) are workloads that will not complete in 10
+seconds. (The time let the user wait before he reaches for the power button).
+This means that other techniques need to be used to manage those workloads,
+that cannot use fences.
+
+Some hardware may schedule compute jobs, and have no way to pre-empt them, or
+have their memory swapped out from them. Or they simply want their workload
+not to be preempted or swapped out at all.
+
+This means that it differs from what is described in driver-api/dma-buf.rst.
+
+As with normal compute jobs, dma-fence may not be used at all. In this case,
+not even to force preemption. The driver with is simply forced to unmap a BO
+from the long compute job's address space on unbind immediately, not even
+waiting for the workload to complete. Effectively this terminates the workload
+when there is no hardware support to recover.
+
+Since this is undesirable, there need to be mitigations to prevent a workload
+from being terminated. There are several possible approach, all with their
+advantages and drawbacks.
+
+The first approach you will likely try is to pin all buffers used by compute.
+This guarantees that the job will run uninterrupted, but also allows a very
+denial of service attack by pinning as much memory as possible, hogging the
+all GPU memory, and possibly a huge chunk of CPU memory.
+
+A second approach that will work slightly better on its own is adding an option
+not to evict when creating a new job (any kind). If all of userspace opts in
+to this flag, it would prevent cooperating userspace from forced terminating
+older compute jobs to start a new one.
+
+If job preemption and recoverable pagefaults are not available, those are the
+only approaches possible. So even with those, you want a separate way of
+controlling resources. The standard kernel way of doing so is cgroups.
+
+This creates a third option, using cgroups to prevent eviction. Both GPU and
+driver-allocated CPU memory would be accounted to the correct cgroup, and
+eviction would be made cgroup aware. This allows the GPU to be partitioned
+into cgroups, that will allow jobs to run next to each other without
+interference.
+
+The interface to the cgroup would be similar to the current CPU memory
+interface, with similar semantics for min/low/high/max, if eviction can
+be made cgroup aware.
+
+What should be noted is that each memory region (tiled memory for example)
+should have its own accounting.
+
+The key is set to the regionid set by the driver, for example "tile0".
+For the value of $card, we use drmGetUnique().