Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit ef4a293a authored by Linus Torvalds's avatar Linus Torvalds
Browse files

Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 

Pull perf fixes from Ingo Molnar:
 "Mostly tooling fixes, but also a lockdep annotation fix, a PMU event
  list fix and a new model addition"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tools/liblockdep: Fix compilation error
  tools/liblockdep: Fix linker error in case of cross compile
  perf tools: Use getconf to determine number of online CPUs
  tools: Fix tools/vm build
  perf/x86/rapl: Enable Broadwell-U RAPL support
  perf/x86/intel: Fix SLM cache event list
  perf: Annotate inherited event ctx->mutex recursion
parents 214e9f72 60d5ddea

arch/x86/kernel/cpu/perf_event_intel.c

+3 −4
Original line number Original line Diff line number Diff line
@@ -1134,7 +1134,7 @@ static __initconst const u64 slm_hw_cache_extra_regs 
 [ C(LL  ) ] = {

 [ C(LL  ) ] = {

	[ C(OP_READ) ] = {

	[ C(OP_READ) ] = {

		[ C(RESULT_ACCESS) ] = SLM_DMND_READ|SLM_LLC_ACCESS,

		[ C(RESULT_ACCESS) ] = SLM_DMND_READ|SLM_LLC_ACCESS,

		[ C(RESULT_MISS)   ] = SLM_DMND_READ|SLM_LLC_MISS,

		[ C(RESULT_MISS)   ] = 0,

	},

	},

	[ C(OP_WRITE) ] = {

	[ C(OP_WRITE) ] = {

		[ C(RESULT_ACCESS) ] = SLM_DMND_WRITE|SLM_LLC_ACCESS,

		[ C(RESULT_ACCESS) ] = SLM_DMND_WRITE|SLM_LLC_ACCESS,
@@ -1184,8 +1184,7 @@ static __initconst const u64 slm_hw_cache_event_ids 
	[ C(OP_READ) ] = {

	[ C(OP_READ) ] = {

		/* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */

		/* OFFCORE_RESPONSE.ANY_DATA.LOCAL_CACHE */

		[ C(RESULT_ACCESS) ] = 0x01b7,

		[ C(RESULT_ACCESS) ] = 0x01b7,

		/* OFFCORE_RESPONSE.ANY_DATA.ANY_LLC_MISS */

		[ C(RESULT_MISS)   ] = 0,

		[ C(RESULT_MISS)   ] = 0x01b7,

	},

	},

	[ C(OP_WRITE) ] = {

	[ C(OP_WRITE) ] = {

		/* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */

		/* OFFCORE_RESPONSE.ANY_RFO.LOCAL_CACHE */
@@ -1217,7 +1216,7 @@ static __initconst const u64 slm_hw_cache_event_ids 
 [ C(ITLB) ] = {

 [ C(ITLB) ] = {

	[ C(OP_READ) ] = {

	[ C(OP_READ) ] = {

		[ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P */

		[ C(RESULT_ACCESS) ] = 0x00c0, /* INST_RETIRED.ANY_P */

		[ C(RESULT_MISS)   ] = 0x0282, /* ITLB.MISSES */

		[ C(RESULT_MISS)   ] = 0x40205, /* PAGE_WALKS.I_SIDE_WALKS */

	},

	},

	[ C(OP_WRITE) ] = {

	[ C(OP_WRITE) ] = {

		[ C(RESULT_ACCESS) ] = -1,

		[ C(RESULT_ACCESS) ] = -1,

arch/x86/kernel/cpu/perf_event_intel_rapl.c

+1 −0
Original line number Original line Diff line number Diff line
@@ -722,6 +722,7 @@ static int __init rapl_pmu_init(void) 
		break;

		break;

	case 60: /* Haswell */

	case 60: /* Haswell */

	case 69: /* Haswell-Celeron */

	case 69: /* Haswell-Celeron */

	case 61: /* Broadwell */

		rapl_cntr_mask = RAPL_IDX_HSW;

		rapl_cntr_mask = RAPL_IDX_HSW;

		rapl_pmu_events_group.attrs = rapl_events_hsw_attr;

		rapl_pmu_events_group.attrs = rapl_events_hsw_attr;

		break;

		break;

kernel/events/core.c

+33 −8
Original line number Original line Diff line number Diff line
@@ -913,10 +913,30 @@ static void put_ctx(struct perf_event_context *ctx) 
 * Those places that change perf_event::ctx will hold both

 * Those places that change perf_event::ctx will hold both

 * perf_event_ctx::mutex of the 'old' and 'new' ctx value.

 * perf_event_ctx::mutex of the 'old' and 'new' ctx value.

 *

 *

 * Lock ordering is by mutex address. There is one other site where

 * Lock ordering is by mutex address. There are two other sites where

 * perf_event_context::mutex nests and that is put_event(). But remember that

 * perf_event_context::mutex nests and those are:

 * that is a parent<->child context relation, and migration does not affect

 *

 * children, therefore these two orderings should not interact.

 *  - perf_event_exit_task_context()	[ child , 0 ]

 *      __perf_event_exit_task()

 *        sync_child_event()

 *          put_event()			[ parent, 1 ]

 *

 *  - perf_event_init_context()		[ parent, 0 ]

 *      inherit_task_group()

 *        inherit_group()

 *          inherit_event()

 *            perf_event_alloc()

 *              perf_init_event()

 *                perf_try_init_event()	[ child , 1 ]

 *

 * While it appears there is an obvious deadlock here -- the parent and child

 * nesting levels are inverted between the two. This is in fact safe because

 * life-time rules separate them. That is an exiting task cannot fork, and a

 * spawning task cannot (yet) exit.

 *

 * But remember that that these are parent<->child context relations, and

 * migration does not affect children, therefore these two orderings should not

 * interact.

 *

 *

 * The change in perf_event::ctx does not affect children (as claimed above)

 * The change in perf_event::ctx does not affect children (as claimed above)

 * because the sys_perf_event_open() case will install a new event and break

 * because the sys_perf_event_open() case will install a new event and break
@@ -3657,9 +3677,6 @@ static void perf_remove_from_owner(struct perf_event *event) 
	}

	}

}

}





/*

 * Called when the last reference to the file is gone.

 */

static void put_event(struct perf_event *event)

static void put_event(struct perf_event *event)

{

{

	struct perf_event_context *ctx;

	struct perf_event_context *ctx;
@@ -3697,6 +3714,9 @@ int perf_event_release_kernel(struct perf_event *event) 
}

}

EXPORT_SYMBOL_GPL(perf_event_release_kernel);

EXPORT_SYMBOL_GPL(perf_event_release_kernel);





/*

 * Called when the last reference to the file is gone.

 */

static int perf_release(struct inode *inode, struct file *file)

static int perf_release(struct inode *inode, struct file *file)

{

{

	put_event(file->private_data);

	put_event(file->private_data);
@@ -7364,7 +7384,12 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event) 
		return -ENODEV;

		return -ENODEV;





	if (event->group_leader != event) {

	if (event->group_leader != event) {

		ctx = perf_event_ctx_lock(event->group_leader);

		/*

		 * This ctx->mutex can nest when we're called through

		 * inheritance. See the perf_event_ctx_lock_nested() comment.

		 */

		ctx = perf_event_ctx_lock_nested(event->group_leader,

						 SINGLE_DEPTH_NESTING);

		BUG_ON(!ctx);

		BUG_ON(!ctx);

	}

	}

tools/lib/lockdep/Makefile

+2 −1
Original line number Original line Diff line number Diff line
@@ -14,9 +14,10 @@ define allow-override 
    $(eval $(1) = $(2)))

    $(eval $(1) = $(2)))

endef

endef





# Allow setting CC and AR, or setting CROSS_COMPILE as a prefix.

# Allow setting CC and AR and LD, or setting CROSS_COMPILE as a prefix.

$(call allow-override,CC,$(CROSS_COMPILE)gcc)

$(call allow-override,CC,$(CROSS_COMPILE)gcc)

$(call allow-override,AR,$(CROSS_COMPILE)ar)

$(call allow-override,AR,$(CROSS_COMPILE)ar)

$(call allow-override,LD,$(CROSS_COMPILE)ld)





INSTALL = install

INSTALL = install

tools/lib/lockdep/uinclude/linux/kernel.h

+3 −0
Original line number Original line Diff line number Diff line
@@ -28,6 +28,9 @@ 
#define __init

#define __init

#define noinline

#define noinline

#define list_add_tail_rcu list_add_tail

#define list_add_tail_rcu list_add_tail

#define list_for_each_entry_rcu list_for_each_entry

#define barrier() 

#define synchronize_sched()





#ifndef CALLER_ADDR0

#ifndef CALLER_ADDR0

#define CALLER_ADDR0 ((unsigned long)__builtin_return_address(0))

#define CALLER_ADDR0 ((unsigned long)__builtin_return_address(0))