File 0001-Revert-remove-affinity_hint-infrastructure.patch of Package irqbalance.13273
From a43f7cdb820d4d81184b03c065f09ca7f27e8fce Mon Sep 17 00:00:00 2001
From: Olaf Hering <olaf@aepfle.de>
Date: Tue, 21 May 2019 16:18:55 +0200
Subject: Revert "remove affinity_hint infrastructure"
This reverts commit dcc411e7bfdd95bbdb7fd0af8699f8cafe686ff4.
---
activate.c | 28 ++++++++++++++++++++++++++--
classify.c | 35 ++++++++++++++++++++++++++++++++---
irqbalance.1 | 19 +++++++++++++++++++
irqbalance.c | 15 ++++++++++++++-
irqbalance.h | 7 +++++++
irqlist.c | 7 +++++++
placement.c | 14 ++++++++++++++
procinterrupts.c | 1 +
8 files changed, 120 insertions(+), 6 deletions(-)
--- a/activate.c
+++ b/activate.c
@@ -69,9 +69,33 @@ static void activate_mapping(struct irq_
if (!info->moved)
return;
- if (info->assigned_obj) {
+ if ((info->hint_policy == HINT_POLICY_EXACT) &&
+ (!cpus_empty(info->affinity_hint))) {
+ if (cpus_intersects(info->affinity_hint, banned_cpus))
+ log(TO_ALL, LOG_WARNING,
+ "irq %d affinity_hint and banned cpus confict\n",
+ info->irq);
+ else {
+ applied_mask = info->affinity_hint;
+ valid_mask = 1;
+ }
+ } else if (info->assigned_obj) {
applied_mask = info->assigned_obj->mask;
- valid_mask = 1;
+ if ((info->hint_policy == HINT_POLICY_SUBSET) &&
+ (!cpus_empty(info->affinity_hint))) {
+ cpus_and(applied_mask, applied_mask, info->affinity_hint);
+ if (!cpus_intersects(applied_mask, unbanned_cpus)) {
+ if (!info->warned) {
+ info->warned = 1;
+ log(TO_ALL, LOG_WARNING,
+ "irq %d affinity_hint subset empty\n",
+ info->irq);
+ }
+ } else
+ valid_mask = 1;
+ } else {
+ valid_mask = 1;
+ }
}
/*
--- a/classify.c
+++ b/classify.c
@@ -31,6 +31,7 @@ struct user_irq_policy {
int level;
int numa_node_set;
int numa_node;
+ enum hp_e hintpolicy;
};
static GList *interrupts_db = NULL;
@@ -282,6 +283,7 @@ static void add_banned_irq(int irq, GLis
new->irq = irq;
new->flags |= IRQ_FLAG_BANNED;
+ new->hint_policy = HINT_POLICY_EXACT;
*list = g_list_append(*list, new);
log(TO_CONSOLE, LOG_INFO, "IRQ %d was BANNED.\n", irq);
@@ -345,6 +347,7 @@ void add_cl_banned_module(char *modname)
static struct irq_info *add_one_irq_to_db(const char *devpath, int irq, struct user_irq_policy *pol)
{
int irq_class = IRQ_OTHER;
+ int rc;
struct irq_info *new, find;
int numa_node;
char path[PATH_MAX];
@@ -375,6 +378,7 @@ static struct irq_info *add_one_irq_to_d
new->irq = irq;
new->class = IRQ_OTHER;
+ new->hint_policy = pol->hintpolicy;
interrupts_db = g_list_append(interrupts_db, new);
@@ -398,7 +402,7 @@ get_numa_node:
sprintf(path, "%s/numa_node", devpath);
fd = fopen(path, "r");
if (fd) {
- fscanf(fd, "%d", &numa_node);
+ rc = fscanf(fd, "%d", &numa_node);
fclose(fd);
}
}
@@ -412,7 +416,7 @@ get_numa_node:
fd = fopen(path, "r");
if (!fd) {
cpus_setall(new->cpumask);
- goto out;
+ goto assign_affinity_hint;
}
lcpu_mask = NULL;
ret = getline(&lcpu_mask, &blen, fd);
@@ -424,6 +428,19 @@ get_numa_node:
}
free(lcpu_mask);
+assign_affinity_hint:
+ cpus_clear(new->affinity_hint);
+ sprintf(path, "/proc/irq/%d/affinity_hint", irq);
+ fd = fopen(path, "r");
+ if (!fd)
+ goto out;
+ lcpu_mask = NULL;
+ ret = getline(&lcpu_mask, &blen, fd);
+ fclose(fd);
+ if (ret <= 0)
+ goto out;
+ cpumask_parse_user(lcpu_mask, ret, new->affinity_hint);
+ free(lcpu_mask);
out:
log(TO_CONSOLE, LOG_INFO, "Adding IRQ %d to database\n", irq);
return new;
@@ -482,6 +499,17 @@ static void parse_user_policy_key(char *
}
pol->numa_node = idx;
pol->numa_node_set = 1;
+ } else if (!strcasecmp("hintpolicy", key)) {
+ if (!strcasecmp("exact", value))
+ pol->hintpolicy = HINT_POLICY_EXACT;
+ else if (!strcasecmp("subset", value))
+ pol->hintpolicy = HINT_POLICY_SUBSET;
+ else if (!strcasecmp("ignore", value))
+ pol->hintpolicy = HINT_POLICY_IGNORE;
+ else {
+ key_set = 0;
+ log(TO_ALL, LOG_WARNING, "Unknown value for hitpolicy: %s\n", value);
+ }
} else {
key_set = 0;
log(TO_ALL, LOG_WARNING, "Unknown key returned, ignoring: %s\n", key);
@@ -506,6 +534,7 @@ static void get_irq_user_policy(char *pa
char *brc;
memset(pol, -1, sizeof(struct user_irq_policy));
+ pol->hintpolicy = global_hint_policy;
/* Return defaults if no script was given */
if (!polscript)
@@ -546,7 +575,7 @@ static int check_for_module_ban(char *na
return 0;
}
-static int check_for_irq_ban(char *path __attribute__((unused)), int irq, GList *proc_interrupts)
+static int check_for_irq_ban(char *path, int irq, GList *proc_interrupts)
{
struct irq_info find, *res;
GList *entry;
--- a/irqbalance.1
+++ b/irqbalance.1
@@ -50,6 +50,21 @@ Causes irqbalance to run in the foregrou
Enables log output optimized for systemd-journal.
.TP
+.B -h, --hintpolicy=[exact | subset | ignore]
+Set the policy for how IRQ kernel affinity hinting is treated. Can be one of:
+.P
+.I exact
+IRQ affinity hint is applied unilaterally and never violated.
+.P
+.I subset
+IRQ is balanced, but the assigned object will be a subset of the affinity hint.
+.P
+.I ignore
+IRQ affinity hint value is completely ignored.
+.P
+The default value for hintpolicy is ignore.
+
+.TP
.B -p, --powerthresh=<threshold>
Set the threshold at which we attempt to move a CPU into powersave mode
If more than <threshold> CPUs are more than 1 standard deviation below the
@@ -107,6 +122,10 @@ that irqbalance can bias IRQ affinity fo
node. Note that specifying a -1 here forces irqbalance to consider an interrupt
from a device to be equidistant from all nodes.
.TP
+.I hintpolicy=[exact | subset | ignore ]
+This allows a user to override the globally set hintpolicy for a given irq. Use
+is identical to the --hintpolicy setting, but applied per irq
+.TP
.B -s, --pid=<file>
Have irqbalance write its process id to the specified file. By default no
pidfile is written. The written pidfile is automatically unlinked when
--- a/irqbalance.c
+++ b/irqbalance.c
@@ -53,6 +53,7 @@ int journal_logging = 0;
int need_rescan;
unsigned int log_mask = TO_ALL;
const char *log_indent;
+enum hp_e global_hint_policy = HINT_POLICY_IGNORE;
unsigned long power_thresh = ULONG_MAX;
unsigned long deepest_cache = 2;
unsigned long long cycle_count = 0;
@@ -117,7 +118,7 @@ static void parse_command_line(int argc,
unsigned long val;
while ((opt = getopt_long(argc, argv,
- "odfji:p:s:c:b:l:m:t:V",
+ "odfjh:i:p:s:c:b:l:m:t:V",
lopts, &longind)) != -1) {
switch(opt) {
@@ -155,6 +156,18 @@ static void parse_command_line(int argc,
case 'f':
foreground_mode=1;
break;
+ case 'h':
+ if (!strncmp(optarg, "exact", strlen(optarg)))
+ global_hint_policy = HINT_POLICY_EXACT;
+ else if (!strncmp(optarg, "subset", strlen(optarg)))
+ global_hint_policy = HINT_POLICY_SUBSET;
+ else if (!strncmp(optarg, "ignore", strlen(optarg)))
+ global_hint_policy = HINT_POLICY_IGNORE;
+ else {
+ usage();
+ exit(1);
+ }
+ break;
case 'i':
val = strtoull(optarg, NULL, 10);
if (val == ULONG_MAX) {
--- a/irqbalance.h
+++ b/irqbalance.h
@@ -65,10 +65,17 @@ extern GList *cpus;
extern int numa_avail;
extern GList *cl_banned_irqs;
+enum hp_e {
+ HINT_POLICY_IGNORE,
+ HINT_POLICY_SUBSET,
+ HINT_POLICY_EXACT
+};
+
extern int debug_mode;
extern int journal_logging;
extern int one_shot_mode;
extern int need_rescan;
+extern enum hp_e global_hint_policy;
extern unsigned long long cycle_count;
extern unsigned long power_thresh;
extern unsigned long deepest_cache;
--- a/irqlist.c
+++ b/irqlist.c
@@ -77,6 +77,13 @@ static void move_candidate_irqs(struct i
{
struct load_balance_info *lb_info = data;
+ /* never move an irq that has an afinity hint when
+ * hint_policy is HINT_POLICY_EXACT
+ */
+ if (info->hint_policy == HINT_POLICY_EXACT)
+ if (!cpus_empty(info->affinity_hint))
+ return;
+
/* Don't rebalance irqs that don't want it */
if (info->level == BALANCE_NONE)
return;
--- a/placement.c
+++ b/placement.c
@@ -41,6 +41,7 @@ static void find_best_object(struct topo
{
struct obj_placement *best = (struct obj_placement *)data;
uint64_t newload;
+ cpumask_t subset;
/*
* Don't consider the unspecified numa node here
@@ -56,6 +57,19 @@ static void find_best_object(struct topo
(!cpus_intersects(d->mask, unbanned_cpus)))
return;
+ /*
+ * If the hint policy is subset, then we only want
+ * to consider objects that are within the irqs hint, but
+ * only if that irq in fact has published a hint
+ */
+ if (best->info->hint_policy == HINT_POLICY_SUBSET) {
+ if (!cpus_empty(best->info->affinity_hint)) {
+ cpus_and(subset, best->info->affinity_hint, d->mask);
+ if (cpus_empty(subset))
+ return;
+ }
+ }
+
if (d->powersave_mode)
return;
--- a/procinterrupts.c
+++ b/procinterrupts.c
@@ -221,6 +221,7 @@ GList* collect_full_irq_list()
info->class = IRQ_OTHER;
#endif
}
+ info->hint_policy = global_hint_policy;
info->name = strdupa(irq_mod);
tmp_list = g_list_append(tmp_list, info);
}
