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 --- drivers/dma/qcom/Kconfig | 55 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 55 insertions(+) create mode 100644 drivers/dma/qcom/Kconfig (limited to 'drivers/dma/qcom/Kconfig') diff --git a/drivers/dma/qcom/Kconfig b/drivers/dma/qcom/Kconfig new file mode 100644 index 000000000..c71b0b5d5 --- /dev/null +++ b/drivers/dma/qcom/Kconfig @@ -0,0 +1,55 @@ +# SPDX-License-Identifier: GPL-2.0-only +config QCOM_ADM + tristate "Qualcomm ADM support" + depends on (ARCH_QCOM || COMPILE_TEST) && !PHYS_ADDR_T_64BIT + select DMA_ENGINE + select DMA_VIRTUAL_CHANNELS + help + Enable support for the Qualcomm Application Data Mover (ADM) DMA + controller, as present on MSM8x60, APQ8064, and IPQ8064 devices. + This controller provides DMA capabilities for both general purpose + and on-chip peripheral devices. + +config QCOM_BAM_DMA + tristate "Qualcomm BAM DMA support" + depends on ARCH_QCOM || (COMPILE_TEST && OF && ARM) + select DMA_ENGINE + select DMA_VIRTUAL_CHANNELS + help + Enable support for the QCOM BAM DMA controller. This controller + provides DMA capabilities for a variety of on-chip devices. + +config QCOM_GPI_DMA + tristate "Qualcomm GPI DMA support" + depends on ARCH_QCOM + select DMA_ENGINE + select DMA_VIRTUAL_CHANNELS + help + Enable support for the QCOM GPI DMA controller. This controller + provides DMA capabilities for a variety of peripheral buses such + as I2C, UART, and SPI. By using GPI dmaengine driver, bus drivers + can use a standardize interface that is protocol independent to + transfer data between DDR and peripheral. + +config QCOM_HIDMA_MGMT + tristate "Qualcomm HIDMA Management support" + depends on HAS_IOMEM + select DMA_ENGINE + help + Enable support for the Qualcomm Technologies HIDMA Management. + Each DMA device requires one management interface driver + for basic initialization before QCOM_HIDMA channel driver can + start managing the channels. In a virtualized environment, + the guest OS would run QCOM_HIDMA channel driver and the + host would run the QCOM_HIDMA_MGMT management driver. + +config QCOM_HIDMA + tristate "Qualcomm HIDMA Channel support" + depends on HAS_IOMEM + select DMA_ENGINE + help + Enable support for the Qualcomm Technologies HIDMA controller. + The HIDMA controller supports optimized buffer copies + (user to kernel, kernel to kernel, etc.). It only supports + memcpy interface. The core is not intended for general + purpose slave DMA. -- cgit v1.3.1