Skip to content

L
linux
Commit e022e7eb authored by Len Brown's avatar Len Brown
Browse files
Options

intel_idle: remove assumption of one C-state per MWAIT flag 

Remove the assumption that cstate_tables are
indexed by MWAIT flag values.  Each entry
identifies itself via its own flags value.
This change is needed to support multiple states
that share the same MWAIT flags.

Note that this can have an effect on what state is described
by 'N' on cmdline intel_idle.max_cstate=N on some systems.

intel_idle.max_cstate=0 still disables the driver
intel_idle.max_cstate=1 still results in just C1(E)
However, "place holders" in the sparse C-state name-space
(eg. Atom) have been removed.
Signed-off-by: default avatarLen Brown <len.brown@intel.com>
parent 137ecc77
Show whitespace changes
Inline Side-by-side
Showing with 61 additions and 51 deletions
arch/x86/include/asm/mwait.h
View file @ e022e7eb
...@@ -4,6 +4,8 @@ ...@@ -4,6 +4,8 @@
#define MWAIT_SUBSTATE_MASK 0xf #define MWAIT_SUBSTATE_MASK 0xf
#define MWAIT_CSTATE_MASK 0xf #define MWAIT_CSTATE_MASK 0xf
#define MWAIT_SUBSTATE_SIZE 4 #define MWAIT_SUBSTATE_SIZE 4
#define MWAIT_HINT2CSTATE(hint) (((hint) >> MWAIT_SUBSTATE_SIZE) & MWAIT_CSTATE_MASK)
#define MWAIT_HINT2SUBSTATE(hint) ((hint) & MWAIT_CSTATE_MASK)
#define CPUID_MWAIT_LEAF 5 #define CPUID_MWAIT_LEAF 5
#define CPUID5_ECX_EXTENSIONS_SUPPORTED 0x1 #define CPUID5_ECX_EXTENSIONS_SUPPORTED 0x1
......
drivers/idle/intel_idle.c
View file @ e022e7eb
...@@ -124,158 +124,161 @@ static struct cpuidle_state *cpuidle_state_table; ...@@ -124,158 +124,161 @@ static struct cpuidle_state *cpuidle_state_table;
* Thus C0 is a dummy. * Thus C0 is a dummy.
*/ */
static struct cpuidle_state nehalem_cstates[CPUIDLE_STATE_MAX] = { static struct cpuidle_state nehalem_cstates[CPUIDLE_STATE_MAX] = {
{ /* MWAIT C0 */ }, {
{ /* MWAIT C1 */
.name = "C1-NHM", .name = "C1-NHM",
.desc = "MWAIT 0x00", .desc = "MWAIT 0x00",
.flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 3, .exit_latency = 3,
.target_residency = 6, .target_residency = 6,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C2 */ {
.name = "C3-NHM", .name = "C3-NHM",
.desc = "MWAIT 0x10", .desc = "MWAIT 0x10",
.flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 20, .exit_latency = 20,
.target_residency = 80, .target_residency = 80,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C3 */ {
.name = "C6-NHM", .name = "C6-NHM",
.desc = "MWAIT 0x20", .desc = "MWAIT 0x20",
.flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 200, .exit_latency = 200,
.target_residency = 800, .target_residency = 800,
.enter = &intel_idle }, .enter = &intel_idle },
{
.enter = NULL }
}; };
static struct cpuidle_state snb_cstates[CPUIDLE_STATE_MAX] = { static struct cpuidle_state snb_cstates[CPUIDLE_STATE_MAX] = {
{ /* MWAIT C0 */ }, {
{ /* MWAIT C1 */
.name = "C1-SNB", .name = "C1-SNB",
.desc = "MWAIT 0x00", .desc = "MWAIT 0x00",
.flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 1, .exit_latency = 1,
.target_residency = 1, .target_residency = 1,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C2 */ {
.name = "C3-SNB", .name = "C3-SNB",
.desc = "MWAIT 0x10", .desc = "MWAIT 0x10",
.flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 80, .exit_latency = 80,
.target_residency = 211, .target_residency = 211,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C3 */ {
.name = "C6-SNB", .name = "C6-SNB",
.desc = "MWAIT 0x20", .desc = "MWAIT 0x20",
.flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 104, .exit_latency = 104,
.target_residency = 345, .target_residency = 345,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C4 */ {
.name = "C7-SNB", .name = "C7-SNB",
.desc = "MWAIT 0x30", .desc = "MWAIT 0x30",
.flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 109, .exit_latency = 109,
.target_residency = 345, .target_residency = 345,
.enter = &intel_idle }, .enter = &intel_idle },
{
.enter = NULL }
}; };
static struct cpuidle_state ivb_cstates[CPUIDLE_STATE_MAX] = { static struct cpuidle_state ivb_cstates[CPUIDLE_STATE_MAX] = {
{ /* MWAIT C0 */ }, {
{ /* MWAIT C1 */
.name = "C1-IVB", .name = "C1-IVB",
.desc = "MWAIT 0x00", .desc = "MWAIT 0x00",
.flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 1, .exit_latency = 1,
.target_residency = 1, .target_residency = 1,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C2 */ {
.name = "C3-IVB", .name = "C3-IVB",
.desc = "MWAIT 0x10", .desc = "MWAIT 0x10",
.flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 59, .exit_latency = 59,
.target_residency = 156, .target_residency = 156,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C3 */ {
.name = "C6-IVB", .name = "C6-IVB",
.desc = "MWAIT 0x20", .desc = "MWAIT 0x20",
.flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 80, .exit_latency = 80,
.target_residency = 300, .target_residency = 300,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C4 */ {
.name = "C7-IVB", .name = "C7-IVB",
.desc = "MWAIT 0x30", .desc = "MWAIT 0x30",
.flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 87, .exit_latency = 87,
.target_residency = 300, .target_residency = 300,
.enter = &intel_idle }, .enter = &intel_idle },
{
.enter = NULL }
}; };
static struct cpuidle_state hsw_cstates[CPUIDLE_STATE_MAX] = { static struct cpuidle_state hsw_cstates[CPUIDLE_STATE_MAX] = {
{ /* MWAIT C0 */ }, {
{ /* MWAIT C1 */
.name = "C1-HSW", .name = "C1-HSW",
.desc = "MWAIT 0x00", .desc = "MWAIT 0x00",
.flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 2, .exit_latency = 2,
.target_residency = 2, .target_residency = 2,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C2 */ {
.name = "C3-HSW", .name = "C3-HSW",
.desc = "MWAIT 0x10", .desc = "MWAIT 0x10",
.flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 33, .exit_latency = 33,
.target_residency = 100, .target_residency = 100,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C3 */ {
.name = "C6-HSW", .name = "C6-HSW",
.desc = "MWAIT 0x20", .desc = "MWAIT 0x20",
.flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x20) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 133, .exit_latency = 133,
.target_residency = 400, .target_residency = 400,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C4 */ {
.name = "C7s-HSW", .name = "C7s-HSW",
.desc = "MWAIT 0x32", .desc = "MWAIT 0x32",
.flags = MWAIT2flg(0x32) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x32) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 166, .exit_latency = 166,
.target_residency = 500, .target_residency = 500,
.enter = &intel_idle }, .enter = &intel_idle },
{
.enter = NULL }
}; };
static struct cpuidle_state atom_cstates[CPUIDLE_STATE_MAX] = { static struct cpuidle_state atom_cstates[CPUIDLE_STATE_MAX] = {
{ /* MWAIT C0 */ }, {
{ /* MWAIT C1 */
.name = "C1-ATM", .name = "C1-ATM",
.desc = "MWAIT 0x00", .desc = "MWAIT 0x00",
.flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x00) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 1, .exit_latency = 1,
.target_residency = 4, .target_residency = 4,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C2 */ {
.name = "C2-ATM", .name = "C2-ATM",
.desc = "MWAIT 0x10", .desc = "MWAIT 0x10",
.flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID, .flags = MWAIT2flg(0x10) | CPUIDLE_FLAG_TIME_VALID,
.exit_latency = 20, .exit_latency = 20,
.target_residency = 80, .target_residency = 80,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C3 */ }, {
{ /* MWAIT C4 */
.name = "C4-ATM", .name = "C4-ATM",
.desc = "MWAIT 0x30", .desc = "MWAIT 0x30",
.flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x30) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 100, .exit_latency = 100,
.target_residency = 400, .target_residency = 400,
.enter = &intel_idle }, .enter = &intel_idle },
{ /* MWAIT C5 */ }, {
{ /* MWAIT C6 */
.name = "C6-ATM", .name = "C6-ATM",
.desc = "MWAIT 0x52", .desc = "MWAIT 0x52",
.flags = MWAIT2flg(0x52) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED, .flags = MWAIT2flg(0x52) | CPUIDLE_FLAG_TIME_VALID | CPUIDLE_FLAG_TLB_FLUSHED,
.exit_latency = 140, .exit_latency = 140,
.target_residency = 560, .target_residency = 560,
.enter = &intel_idle }, .enter = &intel_idle },
{
.enter = NULL }
}; };
/** /**
...@@ -503,32 +506,31 @@ static int intel_idle_cpuidle_driver_init(void) ...@@ -503,32 +506,31 @@ static int intel_idle_cpuidle_driver_init(void)
drv->state_count = 1; drv->state_count = 1;
for (cstate = 1; cstate < CPUIDLE_STATE_MAX; ++cstate) { for (cstate = 0; cstate < CPUIDLE_STATE_MAX; ++cstate) {
int num_substates; int num_substates, mwait_hint, mwait_cstate, mwait_substate;
if (cstate > max_cstate) { if (cpuidle_state_table[cstate].enter == NULL)
break;
if (cstate + 1 > max_cstate) {
printk(PREFIX "max_cstate %d reached\n", printk(PREFIX "max_cstate %d reached\n",
max_cstate); max_cstate);
break; break;
} }
mwait_hint = flg2MWAIT(cpuidle_state_table[cstate].flags);
mwait_cstate = MWAIT_HINT2CSTATE(mwait_hint);
mwait_substate = MWAIT_HINT2SUBSTATE(mwait_hint);
/* does the state exist in CPUID.MWAIT? */ /* does the state exist in CPUID.MWAIT? */
num_substates = (mwait_substates >> ((cstate) * 4)) num_substates = (mwait_substates >> ((mwait_cstate + 1) * 4))
& MWAIT_SUBSTATE_MASK; & MWAIT_SUBSTATE_MASK;
if (num_substates == 0)
continue; /* if sub-state in table is not enumerated by CPUID */
/* is the state not enabled? */ if ((mwait_substate + 1) > num_substates)
if (cpuidle_state_table[cstate].enter == NULL) {
/* does the driver not know about the state? */
if (*cpuidle_state_table[cstate].name == '\0')
pr_debug(PREFIX "unaware of model 0x%x"
" MWAIT %d please"
" contact lenb@kernel.org\n",
boot_cpu_data.x86_model, cstate);
continue; continue;
}
if ((cstate > 2) && if (((mwait_cstate + 1) > 2) &&
!boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) !boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
mark_tsc_unstable("TSC halts in idle" mark_tsc_unstable("TSC halts in idle"
" states deeper than C2"); " states deeper than C2");
...@@ -560,21 +562,27 @@ static int intel_idle_cpu_init(int cpu) ...@@ -560,21 +562,27 @@ static int intel_idle_cpu_init(int cpu)
dev->state_count = 1; dev->state_count = 1;
for (cstate = 1; cstate < CPUIDLE_STATE_MAX; ++cstate) { for (cstate = 0; cstate < CPUIDLE_STATE_MAX; ++cstate) {
int num_substates; int num_substates, mwait_hint, mwait_cstate, mwait_substate;
if (cstate > max_cstate) { if (cpuidle_state_table[cstate].enter == NULL)
continue;
if (cstate + 1 > max_cstate) {
printk(PREFIX "max_cstate %d reached\n", max_cstate); printk(PREFIX "max_cstate %d reached\n", max_cstate);
break; break;
} }
mwait_hint = flg2MWAIT(cpuidle_state_table[cstate].flags);
mwait_cstate = MWAIT_HINT2CSTATE(mwait_hint);
mwait_substate = MWAIT_HINT2SUBSTATE(mwait_hint);
/* does the state exist in CPUID.MWAIT? */ /* does the state exist in CPUID.MWAIT? */
num_substates = (mwait_substates >> ((cstate) * 4)) num_substates = (mwait_substates >> ((mwait_cstate + 1) * 4))
& MWAIT_SUBSTATE_MASK; & MWAIT_SUBSTATE_MASK;
if (num_substates == 0)
continue; /* if sub-state in table is not enumerated by CPUID */
/* is the state not enabled? */ if ((mwait_substate + 1) > num_substates)
if (cpuidle_state_table[cstate].enter == NULL)
continue; continue;
dev->state_count += 1; dev->state_count += 1;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7