linux-kernel - [PATCH 24/31] perf, x86: Add the Haswell implementation of the generic transaction events

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <1348806696-31170-25-git-send-email-andi@firstfloor.org>
Date:	Thu, 27 Sep 2012 21:31:29 -0700
From:	Andi Kleen <andi@...stfloor.org>
To:	linux-kernel@...r.kernel.org
Cc:	x86@...nel.org, a.p.zijlstra@...llo.nl, eranian@...gle.com,
	acme@...hat.com, Andi Kleen <ak@...ux.intel.com>
Subject: [PATCH 24/31] perf, x86: Add the Haswell implementation of the generic transaction events

From: Andi Kleen <ak@...ux.intel.com>

Straight forward table mapping from the generic transactional memory events
to the Haswell TSX events.

One special case is that the abort-all events force PEBS with precise level
two. Without using eventingrip abort IPs are generally useless (you get
something after the abort). So we really want PEBS here for any
sampling. Since it was very unintuitive for users to do this manually
I just made this default.

To do this the mapping table sets a magic flag, that is later checked in
the event setup and forces the precise event. For counting events
PEBS is not forced.

Signed-off-by: Andi Kleen <ak@...ux.intel.com>
---
 arch/x86/kernel/cpu/perf_event_intel.c |   47 ++++++++++++++++++++++++++++++++
 1 files changed, 47 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 2c4cbf3..be0d3c8 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -820,6 +820,38 @@ static __initconst const u64 atom_hw_cache_event_ids
  },
 };
 
+#define FORCE_PEBS (1ULL << 63)
+
+static u64 __initdata hsw_transaction_event_ids
+				[PERF_COUNT_HW_TRANSACTION_MAX]
+				[PERF_COUNT_HW_ABORT_MAX] =
+{
+	/* RTM_RETIRED.START */
+	[ PERF_COUNT_HW_TRANSACTION_START ]  = { 0x1c9 },
+	/* RTM_RETIRED.COMMIT */
+	[ PERF_COUNT_HW_TRANSACTION_COMMIT ] = { 0x2c9 },
+	[ PERF_COUNT_HW_TRANSACTION_ABORT ]  = {
+		/* RTM_RETIRED.ABORTED with pebs */
+		[ PERF_COUNT_HW_ABORT_ALL ]      = 0x4c9|FORCE_PEBS,
+		/* TX_MEM.ABORT_CONFLICT */
+		[ PERF_COUNT_HW_ABORT_CONFLICT ] = 0x154,
+		/* TX_MEM.ABORT_CAPACITY */
+		[ PERF_COUNT_HW_ABORT_CAPACITY ] = 0x254,
+	},
+	/* HLE_RETIRED.START */
+	[ PERF_COUNT_HW_ELISION_START ]  =     { 0x1c8 },
+	/* HLE_RETIRED.COMMIT */
+	[ PERF_COUNT_HW_ELISION_COMMIT ] =     { 0x2c8 },
+	[ PERF_COUNT_HW_ELISION_ABORT ]  = {
+		/* HLE_RETIRED.ABORTED with pebs */
+		[ PERF_COUNT_HW_ABORT_ALL ]      = 0x4c8|FORCE_PEBS,
+		/* TX_MEM.ABORT_CONFLICT */
+		[ PERF_COUNT_HW_ABORT_CONFLICT ] = 0x154,
+		/* TX_MEM.ABORT_CAPACITY */
+		[ PERF_COUNT_HW_ABORT_CAPACITY ] = 0x254,
+	}
+};
+
 static inline bool intel_pmu_needs_lbr_smpl(struct perf_event *event)
 {
 	/* user explicitly requested branch sampling */
@@ -1655,6 +1687,18 @@ static int hsw_hw_config(struct perf_event *event)
 			return -EIO;
 		event->hw.config |= HSW_INTX_CHECKPOINTED;
 	}
+
+	/*
+	 * Sampling transaction abort events work very poorly without
+	 * PEBS. So force it.
+	 */
+	if (event->attr.type == PERF_TYPE_HW_TRANSACTION &&
+	    (event->hw.config & FORCE_PEBS)) {
+		event->hw.config &= ~FORCE_PEBS;
+		if (is_sampling_event(event))
+			event->attr.precise_ip = 2;
+	}
+
 	return 0;
 }
 
@@ -2165,6 +2209,9 @@ __init int intel_pmu_init(void)
 		x86_pmu.hw_config = hsw_hw_config;
 		x86_pmu.get_event_constraints = hsw_get_event_constraints;
 		x86_pmu.memory_lat_events = intel_hsw_memory_latency_events;
+		memcpy(hw_transaction_event_ids, hsw_transaction_event_ids,
+		       sizeof(hsw_transaction_event_ids));
+
 		pr_cont("Haswell events, ");
 		break;
 
-- 
1.7.7.6

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/