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 --- net/sunrpc/sysctl.c | 180 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 180 insertions(+) create mode 100644 net/sunrpc/sysctl.c (limited to 'net/sunrpc/sysctl.c') diff --git a/net/sunrpc/sysctl.c b/net/sunrpc/sysctl.c new file mode 100644 index 000000000..bdb587a72 --- /dev/null +++ b/net/sunrpc/sysctl.c @@ -0,0 +1,180 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * linux/net/sunrpc/sysctl.c + * + * Sysctl interface to sunrpc module. + * + * I would prefer to register the sunrpc table below sys/net, but that's + * impossible at the moment. + */ + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include + +#include "netns.h" + +/* + * Declare the debug flags here + */ +unsigned int rpc_debug; +EXPORT_SYMBOL_GPL(rpc_debug); + +unsigned int nfs_debug; +EXPORT_SYMBOL_GPL(nfs_debug); + +unsigned int nfsd_debug; +EXPORT_SYMBOL_GPL(nfsd_debug); + +unsigned int nlm_debug; +EXPORT_SYMBOL_GPL(nlm_debug); + +#if IS_ENABLED(CONFIG_SUNRPC_DEBUG) + +static int proc_do_xprt(const struct ctl_table *table, int write, + void *buffer, size_t *lenp, loff_t *ppos) +{ + char tmpbuf[256]; + ssize_t len; + + if (write || *ppos) { + *lenp = 0; + return 0; + } + len = svc_print_xprts(tmpbuf, sizeof(tmpbuf)); + len = memory_read_from_buffer(buffer, *lenp, ppos, tmpbuf, len); + + if (len < 0) { + *lenp = 0; + return -EINVAL; + } + *lenp = len; + return 0; +} + +static int +proc_dodebug(const struct ctl_table *table, int write, void *buffer, size_t *lenp, + loff_t *ppos) +{ + char tmpbuf[20], *s = NULL; + char *p; + unsigned int value; + size_t left, len; + + if ((*ppos && !write) || !*lenp) { + *lenp = 0; + return 0; + } + + left = *lenp; + + if (write) { + p = buffer; + while (left && isspace(*p)) { + left--; + p++; + } + if (!left) + goto done; + + if (left > sizeof(tmpbuf) - 1) + return -EINVAL; + memcpy(tmpbuf, p, left); + tmpbuf[left] = '\0'; + + value = simple_strtol(tmpbuf, &s, 0); + if (s) { + left -= (s - tmpbuf); + if (left && !isspace(*s)) + return -EINVAL; + while (left && isspace(*s)) { + left--; + s++; + } + } else + left = 0; + *(unsigned int *) table->data = value; + /* Display the RPC tasks on writing to rpc_debug */ + if (strcmp(table->procname, "rpc_debug") == 0) + rpc_show_tasks(&init_net); + } else { + len = sprintf(tmpbuf, "0x%04x", *(unsigned int *) table->data); + if (len > left) + len = left; + memcpy(buffer, tmpbuf, len); + if ((left -= len) > 0) { + *((char *)buffer + len) = '\n'; + left--; + } + } + +done: + *lenp -= left; + *ppos += *lenp; + return 0; +} + +static struct ctl_table_header *sunrpc_table_header; + +static struct ctl_table debug_table[] = { + { + .procname = "rpc_debug", + .data = &rpc_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dodebug + }, + { + .procname = "nfs_debug", + .data = &nfs_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dodebug + }, + { + .procname = "nfsd_debug", + .data = &nfsd_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dodebug + }, + { + .procname = "nlm_debug", + .data = &nlm_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dodebug + }, + { + .procname = "transports", + .maxlen = 256, + .mode = 0444, + .proc_handler = proc_do_xprt, + }, +}; + +void +rpc_register_sysctl(void) +{ + if (!sunrpc_table_header) + sunrpc_table_header = register_sysctl("sunrpc", debug_table); +} + +void +rpc_unregister_sysctl(void) +{ + if (sunrpc_table_header) { + unregister_sysctl_table(sunrpc_table_header); + sunrpc_table_header = NULL; + } +} +#endif -- cgit v1.3.1