Patchwork [6/9,v5] powerpc/perf: Define big-endian version of perf_mem_data_src

login
register
mail settings
Submitter sukadev@linux.vnet.ibm.com
Date Oct. 5, 2013, 3:53 a.m.
Message ID <20131005035322.GA32354@us.ibm.com>
Download mbox | patch
Permalink /patch/280750/
State Superseded
Headers show

Comments

sukadev@linux.vnet.ibm.com - Oct. 5, 2013, 3:53 a.m.
Michael Ellerman [michael@ellerman.id.au] wrote:
| On Tue, Oct 01, 2013 at 05:15:07PM -0700, Sukadev Bhattiprolu wrote:
| > perf_mem_data_src is an union that is initialized via the ->val field
| > and accessed via the bitmap fields. For this to work on big endian
| > platforms, we also need a big-endian represenation of perf_mem_data_src.
| > 
| > diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
| > index ca1d90b..846f399 100644
| > --- a/include/uapi/linux/perf_event.h
| > +++ b/include/uapi/linux/perf_event.h
| > @@ -19,6 +19,50 @@
| >  #include <asm/byteorder.h>
| >  
| >  /*
| > + * Kernel and userspace check for endianness in incompatible ways.
| > + * In user space, <endian.h> defines both __BIG_ENDIAN and __LITTLE_ENDIAN
| > + * but sets __BYTE_ORDER to one or the other. So user space uses checks are:
| 
| 
| Why can't you use __BIG_ENDIAN_BITFIELD ?


So, the perf tool overrides the <asm/byteorder.h> with a local version.

And since this local version is arch neutral, we can't excplicitly include
the endian headers like <asm/byteorder.h> does.

How about we do something like this (both kernel and tool seem to build
on both x86 and power).

Sukadev.

---
 include/uapi/linux/perf_event.h         |   16 ++++++++++++++++
 tools/perf/util/include/asm/byteorder.h |   27 +++++++++++++++++++++++++++
 2 files changed, 43 insertions(+), 0 deletions(-)

Patch

diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index ca1d90b..dcfa74f 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -695,6 +695,7 @@  enum perf_callchain_context {
 #define PERF_FLAG_FD_OUTPUT		(1U << 1)
 #define PERF_FLAG_PID_CGROUP		(1U << 2) /* pid=cgroup id, per-cpu mode only */
 
+#if defined (__LITTLE_ENDIAN_BITFIELD)
 union perf_mem_data_src {
 	__u64 val;
 	struct {
@@ -706,6 +707,21 @@  union perf_mem_data_src {
 			mem_rsvd:31;
 	};
 };
+#elif defined(__BIG_ENDIAN_BITFIELD)
+union perf_mem_data_src {
+	__u64 val;
+	struct {
+		__u64	mem_rsvd:31,
+			mem_dtlb:7,	/* tlb access */
+			mem_lock:2,	/* lock instr */
+			mem_snoop:5,	/* snoop mode */
+			mem_lvl:14,	/* memory hierarchy level */
+			mem_op:5;	/* type of opcode */
+	};
+};
+#else
+#error "Unknown endianness"
+#endif
 
 /* type of opcode (load/store/prefetch,code) */
 #define PERF_MEM_OP_NA		0x01 /* not available */
diff --git a/tools/perf/util/include/asm/byteorder.h b/tools/perf/util/include/asm/byteorder.h
index 2a9bdc0..521a382 100644
--- a/tools/perf/util/include/asm/byteorder.h
+++ b/tools/perf/util/include/asm/byteorder.h
@@ -1,2 +1,29 @@ 
 #include <asm/types.h>
 #include "../../../../include/uapi/linux/swab.h"
+#include <endian.h>
+
+/*
+ * __LITTLE_ENDIAN_BITFIELD and __BIG_ENDIAN_BITFIELD are normally picked
+ * from <byteorder.h>. Since we override the default <byteorder.h>, define
+ * them explicitly here
+ */
+#if __BYTE_ORDER == __LITTLE_ENDIAN
+
+#ifndef __LITTLE_ENDIAN_BITFIELD
+#define __LITTLE_ENDIAN_BITFIELD
+#endif
+
+#elif __BYTE_ORDER == __BIG_ENDIAN
+
+#ifndef __BIG_ENDIAN_BITFIELD
+#define __BIG_ENDIAN_BITFIELD
+#endif
+
+#else
+#error "Unknown endianness"
+#endif
+
+#if defined(__LITTLE_ENDIAN_BITFIELD) && defined(__BIG_ENDIAN_BITFIELD)
+#error 	Both __LITTLE_ENDIAN_BITFIELD and __BIG_ENDIAN_BITFIELD defined! \
+	Some perf data structures (eg: perf_mem_data_src) will be wrong.
+#endif