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 --- arch/powerpc/boot/libfdt-wrapper.c | 185 +++++++++++++++++++++++++++++++++++++ 1 file changed, 185 insertions(+) create mode 100644 arch/powerpc/boot/libfdt-wrapper.c (limited to 'arch/powerpc/boot/libfdt-wrapper.c') diff --git a/arch/powerpc/boot/libfdt-wrapper.c b/arch/powerpc/boot/libfdt-wrapper.c new file mode 100644 index 000000000..217d0d7a6 --- /dev/null +++ b/arch/powerpc/boot/libfdt-wrapper.c @@ -0,0 +1,185 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * This file does the necessary interface mapping between the bootwrapper + * device tree operations and the interface provided by shared source + * files flatdevicetree.[ch]. + * + * Copyright 2007 David Gibson, IBM Corporation. + */ + +#include +#include +#include +#include +#include "ops.h" + +#define DEBUG 0 +#define BAD_ERROR(err) (((err) < 0) \ + && ((err) != -FDT_ERR_NOTFOUND) \ + && ((err) != -FDT_ERR_EXISTS)) + +#define check_err(err) \ + ({ \ + if (BAD_ERROR(err) || ((err < 0) && DEBUG)) \ + printf("%s():%d %s\n\r", __func__, __LINE__, \ + fdt_strerror(err)); \ + if (BAD_ERROR(err)) \ + exit(); \ + (err < 0) ? -1 : 0; \ + }) + +#define offset_devp(off) \ + ({ \ + unsigned long _offset = (off); \ + check_err(_offset) ? NULL : (void *)(_offset+1); \ + }) + +#define devp_offset_find(devp) (((unsigned long)(devp))-1) +#define devp_offset(devp) (devp ? ((unsigned long)(devp))-1 : 0) + +static void *fdt; +static void *buf; /* = NULL */ + +#define EXPAND_GRANULARITY 1024 + +static void expand_buf(int minexpand) +{ + int size = fdt_totalsize(fdt); + int rc; + + size = _ALIGN(size + minexpand, EXPAND_GRANULARITY); + buf = platform_ops.realloc(buf, size); + if (!buf) + fatal("Couldn't find %d bytes to expand device tree\n\r", size); + rc = fdt_open_into(fdt, buf, size); + if (rc != 0) + fatal("Couldn't expand fdt into new buffer: %s\n\r", + fdt_strerror(rc)); + + fdt = buf; +} + +static void *fdt_wrapper_finddevice(const char *path) +{ + return offset_devp(fdt_path_offset(fdt, path)); +} + +static int fdt_wrapper_getprop(const void *devp, const char *name, + void *buf, const int buflen) +{ + const void *p; + int len; + + p = fdt_getprop(fdt, devp_offset(devp), name, &len); + if (!p) + return check_err(len); + memcpy(buf, p, min(len, buflen)); + return len; +} + +static int fdt_wrapper_setprop(const void *devp, const char *name, + const void *buf, const int len) +{ + int rc; + + rc = fdt_setprop(fdt, devp_offset(devp), name, buf, len); + if (rc == -FDT_ERR_NOSPACE) { + expand_buf(len + 16); + rc = fdt_setprop(fdt, devp_offset(devp), name, buf, len); + } + + return check_err(rc); +} + +static int fdt_wrapper_del_node(const void *devp) +{ + return fdt_del_node(fdt, devp_offset(devp)); +} + +static void *fdt_wrapper_get_parent(const void *devp) +{ + return offset_devp(fdt_parent_offset(fdt, devp_offset(devp))); +} + +static void *fdt_wrapper_create_node(const void *devp, const char *name) +{ + int offset; + + offset = fdt_add_subnode(fdt, devp_offset(devp), name); + if (offset == -FDT_ERR_NOSPACE) { + expand_buf(strlen(name) + 16); + offset = fdt_add_subnode(fdt, devp_offset(devp), name); + } + + return offset_devp(offset); +} + +static void *fdt_wrapper_find_node_by_prop_value(const void *prev, + const char *name, + const char *val, + int len) +{ + int offset = fdt_node_offset_by_prop_value(fdt, devp_offset_find(prev), + name, val, len); + return offset_devp(offset); +} + +static void *fdt_wrapper_find_node_by_compatible(const void *prev, + const char *val) +{ + int offset = fdt_node_offset_by_compatible(fdt, devp_offset_find(prev), + val); + return offset_devp(offset); +} + +static char *fdt_wrapper_get_path(const void *devp, char *buf, int len) +{ + int rc; + + rc = fdt_get_path(fdt, devp_offset(devp), buf, len); + if (check_err(rc)) + return NULL; + return buf; +} + +static unsigned long fdt_wrapper_finalize(void) +{ + int rc; + + rc = fdt_pack(fdt); + if (rc != 0) + fatal("Couldn't pack flat tree: %s\n\r", + fdt_strerror(rc)); + return (unsigned long)fdt; +} + +void fdt_init(void *blob) +{ + int err; + int bufsize; + + dt_ops.finddevice = fdt_wrapper_finddevice; + dt_ops.getprop = fdt_wrapper_getprop; + dt_ops.setprop = fdt_wrapper_setprop; + dt_ops.get_parent = fdt_wrapper_get_parent; + dt_ops.create_node = fdt_wrapper_create_node; + dt_ops.find_node_by_prop_value = fdt_wrapper_find_node_by_prop_value; + dt_ops.find_node_by_compatible = fdt_wrapper_find_node_by_compatible; + dt_ops.del_node = fdt_wrapper_del_node; + dt_ops.get_path = fdt_wrapper_get_path; + dt_ops.finalize = fdt_wrapper_finalize; + + /* Make sure the dt blob is the right version and so forth */ + fdt = blob; + bufsize = fdt_totalsize(fdt) + EXPAND_GRANULARITY; + buf = malloc(bufsize); + if(!buf) + fatal("malloc failed. can't relocate the device tree\n\r"); + + err = fdt_open_into(fdt, buf, bufsize); + + if (err != 0) + fatal("fdt_init(): %s\n\r", fdt_strerror(err)); + + fdt = buf; +} -- cgit v1.3.1