All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ingo Molnar <mingo@kernel.org>
To: Adrian Hunter <adrian.hunter@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>,
	Vince Weaver <vince@deater.net>,
	hpa@zytor.com, linux-kernel@vger.kernel.org, tglx@linutronix.de,
	linux-tip-commits@vger.kernel.org, eranian@googlemail.com
Subject: [PATCH, v4] perf: Fix capabilities bitfield compatibility in 'struct perf_event_mmap_page'
Date: Thu, 19 Sep 2013 13:42:54 +0200	[thread overview]
Message-ID: <20130919114254.GA22807@gmail.com> (raw)
In-Reply-To: <523ADD8D.1030607@intel.com>


* Adrian Hunter <adrian.hunter@intel.com> wrote:

> >  		struct {
> > -			__u64	cap_usr_time		: 1,
> > -				cap_usr_rdpmc		: 1,
> > -				cap_usr_time_zero	: 1,
> > -				cap_____res		: 61;
> > +			__u64	cap_bit0		: 1, /* Always 0, deprecated, see commit 860f085b74e9 */
> > +				cap_bit0_is_deprecated	: 1, /* Always 1, signals that bit 0 is zero */
> > +
> > +				cap_user_rdpmc		: 1, /* The RDPMC instruction can be used to read counts */
> > +				cap_user_time		: 1, /* The time_* fields are used */
> > +				cap_user_time_zero	: 1, /* The time_zero field is used */
> > +				cap_____res		: 59;
> >  		};

> Please consider adding:

indeed - added. I also build-tested it all and added a changelog - see the 
updated patch below.

Thanks,

	Ingo

--------------------->
Subject: perf: Fix capabilities bitfield compatibility in 'struct perf_event_mmap_page'
From: Peter Zijlstra <peterz@infradead.org>
Date: Thu, 19 Sep 2013 10:16:42 +0200

Solve the problems around the broken definition of perf_event_mmap_page::
cap_usr_time and cap_usr_rdpmc fields which used to overlap, partially
fixed by:

  860f085b74e9 ("perf: Fix broken union in 'struct perf_event_mmap_page'")

The problem with the fix (merged in v3.12-rc1 and not yet released
officially), noticed by Vince Weaver is that the new behavior is
not detectable by new user-space, and that due to the reuse of the
field names it's easy to mis-compile a binary if old headers are used
on a new kernel or new headers are used on an old kernel.

To solve all that make this change explicit, detectable and self-contained,
by iterating the ABI the following way:

 - Always clear bit 0, and rename it to usrpage->cap_bit0, to at least not
   confuse old user-space binaries. RDPMC will be marked as unavailable
   to old binaries but that's within the ABI, this is a capability bit.

 - Rename bit 1 to ->cap_bit0_is_deprecated and always set it to 1, so new
   libraries can reliably detect that bit 0 is deprecated and perma-zero
   without having to check the kernel version.

 - Use bits 2, 3, 4 for the newly defined, correct functionality:

	cap_user_rdpmc		: 1, /* The RDPMC instruction can be used to read counts */
	cap_user_time		: 1, /* The time_* fields are used */
	cap_user_time_zero	: 1, /* The time_zero field is used */

 - Rename all the bitfield names in perf_event.h to be different from the
   old names, to make sure it's not possible to mis-compile it
   accidentally with old assumptions.

The 'size' field can then be used in the future to add new fields and it
will act as a natural ABI version indicator as well.

Also adjust tools/perf/ userspace for the new definitions, noticed by
Adrian Hunter.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Also-Fixed-by: Adrian Hunter <adrian.hunter@intel.com>
Link: http://lkml.kernel.org/n/tip-zr03yxjrpXesOzzupszqglbv@git.kernel.org
[ Restructured it, using Peter's original patches. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 arch/x86/kernel/cpu/perf_event.c |   10 +++++-----
 include/uapi/linux/perf_event.h  |   14 +++++++++-----
 kernel/events/core.c             |   21 +++++++++++++++++++++
 tools/perf/arch/x86/util/tsc.c   |    6 +++---
 4 files changed, 38 insertions(+), 13 deletions(-)

Index: tip/arch/x86/kernel/cpu/perf_event.c
===================================================================
--- tip.orig/arch/x86/kernel/cpu/perf_event.c
+++ tip/arch/x86/kernel/cpu/perf_event.c
@@ -1883,9 +1883,9 @@ static struct pmu pmu = {
 
 void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
 {
-	userpg->cap_usr_time = 0;
-	userpg->cap_usr_time_zero = 0;
-	userpg->cap_usr_rdpmc = x86_pmu.attr_rdpmc;
+	userpg->cap_user_time = 0;
+	userpg->cap_user_time_zero = 0;
+	userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc;
 	userpg->pmc_width = x86_pmu.cntval_bits;
 
 	if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
@@ -1894,13 +1894,13 @@ void arch_perf_update_userpage(struct pe
 	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC))
 		return;
 
-	userpg->cap_usr_time = 1;
+	userpg->cap_user_time = 1;
 	userpg->time_mult = this_cpu_read(cyc2ns);
 	userpg->time_shift = CYC2NS_SCALE_FACTOR;
 	userpg->time_offset = this_cpu_read(cyc2ns_offset) - now;
 
 	if (sched_clock_stable && !check_tsc_disabled()) {
-		userpg->cap_usr_time_zero = 1;
+		userpg->cap_user_time_zero = 1;
 		userpg->time_zero = this_cpu_read(cyc2ns_offset);
 	}
 }
Index: tip/include/uapi/linux/perf_event.h
===================================================================
--- tip.orig/include/uapi/linux/perf_event.h
+++ tip/include/uapi/linux/perf_event.h
@@ -380,10 +380,13 @@ struct perf_event_mmap_page {
 	union {
 		__u64	capabilities;
 		struct {
-			__u64	cap_usr_time		: 1,
-				cap_usr_rdpmc		: 1,
-				cap_usr_time_zero	: 1,
-				cap_____res		: 61;
+			__u64	cap_bit0		: 1, /* Always 0, deprecated, see commit 860f085b74e9 */
+				cap_bit0_is_deprecated	: 1, /* Always 1, signals that bit 0 is zero */
+
+				cap_user_rdpmc		: 1, /* The RDPMC instruction can be used to read counts */
+				cap_user_time		: 1, /* The time_* fields are used */
+				cap_user_time_zero	: 1, /* The time_zero field is used */
+				cap_____res		: 59;
 		};
 	};
 
@@ -442,12 +445,13 @@ struct perf_event_mmap_page {
 	 *               ((rem * time_mult) >> time_shift);
 	 */
 	__u64	time_zero;
+	__u32	size;			/* Header size up to __reserved[] fields. */
 
 		/*
 		 * Hole for extension of the self monitor capabilities
 		 */
 
-	__u64	__reserved[119];	/* align to 1k */
+	__u8	__reserved[118*8+4];	/* align to 1k. */
 
 	/*
 	 * Control data for the mmap() data buffer.
Index: tip/kernel/events/core.c
===================================================================
--- tip.orig/kernel/events/core.c
+++ tip/kernel/events/core.c
@@ -3660,6 +3660,26 @@ static void calc_timer_values(struct per
 	*running = ctx_time - event->tstamp_running;
 }
 
+static void perf_event_init_userpage(struct perf_event *event)
+{
+	struct perf_event_mmap_page *userpg;
+	struct ring_buffer *rb;
+
+	rcu_read_lock();
+	rb = rcu_dereference(event->rb);
+	if (!rb)
+		goto unlock;
+
+	userpg = rb->user_page;
+
+	/* Allow new userspace to detect that bit 0 is deprecated */
+	userpg->cap_bit0_is_deprecated = 1;
+	userpg->size = offsetof(struct perf_event_mmap_page, __reserved);
+
+unlock:
+	rcu_read_unlock();
+}
+
 void __weak arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
 {
 }
@@ -4044,6 +4064,7 @@ again:
 	ring_buffer_attach(event, rb);
 	rcu_assign_pointer(event->rb, rb);
 
+	perf_event_init_userpage(event);
 	perf_event_update_userpage(event);
 
 unlock:
Index: tip/tools/perf/arch/x86/util/tsc.c
===================================================================
--- tip.orig/tools/perf/arch/x86/util/tsc.c
+++ tip/tools/perf/arch/x86/util/tsc.c
@@ -32,7 +32,7 @@ u64 tsc_to_perf_time(u64 cyc, struct per
 int perf_read_tsc_conversion(const struct perf_event_mmap_page *pc,
 			     struct perf_tsc_conversion *tc)
 {
-	bool cap_usr_time_zero;
+	bool cap_user_time_zero;
 	u32 seq;
 	int i = 0;
 
@@ -42,7 +42,7 @@ int perf_read_tsc_conversion(const struc
 		tc->time_mult = pc->time_mult;
 		tc->time_shift = pc->time_shift;
 		tc->time_zero = pc->time_zero;
-		cap_usr_time_zero = pc->cap_usr_time_zero;
+		cap_user_time_zero = pc->cap_user_time_zero;
 		rmb();
 		if (pc->lock == seq && !(seq & 1))
 			break;
@@ -52,7 +52,7 @@ int perf_read_tsc_conversion(const struc
 		}
 	}
 
-	if (!cap_usr_time_zero)
+	if (!cap_user_time_zero)
 		return -EOPNOTSUPP;
 
 	return 0;

  reply	other threads:[~2013-09-19 11:42 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-06-28 13:22 [PATCH 0/5] perf: add two new features Adrian Hunter
2013-06-28 13:22 ` [PATCH 1/5] perf: fix broken union in perf_event_mmap_page Adrian Hunter
2013-06-28 15:22   ` Peter Zijlstra
2013-07-16 11:51     ` H. Peter Anvin
2013-07-24  3:56   ` [tip:perf/core] perf: Fix broken union in ' struct perf_event_mmap_page' tip-bot for Adrian Hunter
2013-09-17 20:23     ` Vince Weaver
2013-09-17 20:35       ` Vince Weaver
2013-09-19  8:42         ` Ingo Molnar
2013-09-18  8:57       ` Peter Zijlstra
2013-09-18 14:19         ` Vince Weaver
2013-09-18 15:42           ` Peter Zijlstra
2013-09-18 18:33             ` Stephane Eranian
2013-09-19  8:43               ` Peter Zijlstra
2013-09-19  8:55                 ` Stephane Eranian
2013-09-19  9:16                   ` Ingo Molnar
2013-09-18 20:07             ` Vince Weaver
2013-09-19  8:16               ` Peter Zijlstra
2013-09-19  9:14                 ` [PATCH] perf: Always set bit 0 in the capabilities field of 'struct perf_event_mmap_page' to 0, to maintain the ABI Ingo Molnar
2013-09-19 10:12                   ` Peter Zijlstra
2013-09-19 10:28                     ` Ingo Molnar
2013-09-19 10:35                       ` Peter Zijlstra
2013-09-19 10:40                         ` [PATCH, v3] " Ingo Molnar
2013-09-19 11:18                           ` Adrian Hunter
2013-09-19 11:42                             ` Ingo Molnar [this message]
2013-09-19 17:40                               ` [PATCH, v4] perf: Fix capabilities bitfield compatibility in 'struct perf_event_mmap_page' Vince Weaver
2013-09-20  7:44                                 ` Ingo Molnar
2013-09-18  9:13       ` [tip:perf/core] perf: Fix broken union in ' struct perf_event_mmap_page' Adrian Hunter
2013-09-18 14:10         ` Vince Weaver
2013-06-28 13:22 ` [PATCH 2/5] x86: add ability to calculate TSC from perf sample timestamps Adrian Hunter
2013-07-24  3:56   ` [tip:perf/core] perf/x86: Add " tip-bot for Adrian Hunter
2013-06-28 13:22 ` [PATCH 3/5] perf tools: add test for converting perf time to/from TSC Adrian Hunter
2013-07-24  3:56   ` [tip:perf/core] perf tools: Add test for converting perf time to/ from TSC tip-bot for Adrian Hunter
2013-06-28 13:22 ` [PATCH 4/5] perf: add 'keep tracking' flag to PERF_EVENT_IOC_DISABLE Adrian Hunter
2013-06-28 13:22 ` [PATCH 5/5] perf tools: add 'keep tracking' test Adrian Hunter
2013-06-28 15:27 ` [PATCH 0/5] perf: add two new features Peter Zijlstra
2013-06-28 19:22   ` Adrian Hunter
2013-07-16  6:22     ` Adrian Hunter
2013-07-16 14:34       ` Peter Zijlstra
2013-07-17 11:28         ` Adrian Hunter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130919114254.GA22807@gmail.com \
    --to=mingo@kernel.org \
    --cc=adrian.hunter@intel.com \
    --cc=eranian@googlemail.com \
    --cc=hpa@zytor.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-tip-commits@vger.kernel.org \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    --cc=vince@deater.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.