mm: cma: add trace events for CMA allocations and freeings

PATCH In review Stefan Strogin 3 participants 4 messages Apr 1, 2015 – Apr 2, 2015

Subsystems (3)

Memory management Tracing MM misc

lore

Share & tools

lore

Stefan Strogin

author

2015-04-01 13:31 UTC #0 lore raw

PATCH 1/1 mm: cma: add trace events for CMA allocations and freeings

Add trace events for cma_alloc() and cma_release().

The cma_alloc tracepoint is used both for successful and failed allocations, in case of allocation failure pfn=-1UL is stored and printed.

Signed-off-by Stefan Strogin <[email protected]>

---

Took out from the patch set "mm: cma: add some debug information for CMA" v4
(http://thread.gmane.org/gmane.linux.kernel.mm/129903) because of probable uselessness of the rest of the patches.

Changes from the version from the patch set:
- Constify the struct page * parameter passed to the tracepoints.
- Pass both pfn and struct page * to the tracepoints to decrease unnecessary pfn_to_page() and page_to_pfn() calls and avoid using them in TP_printk.
- Store and print pfn=-1UL instead of pfn=0, because 0th pfn can truly exist on some architectures.

 include/trace/events/cma.h | 63 ++++++++++++++++++++++++++++++++++++++++++++++
 mm/cma.c                   |  5 ++++
 2 files changed, 68 insertions(+)
 create mode 100644 include/trace/events/cma.h

include/trace/events/cma.h new file +63 -0

new file mode 100644
index 0000000..e01b35d
@@ -0,0 +1,63 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM cma
+
+#if !defined(_TRACE_CMA_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_CMA_H
+
+#include <linux/types.h>
+#include <linux/tracepoint.h>
+
+TRACE_EVENT(cma_alloc,
+
+	TP_PROTO(unsigned long pfn, const struct page *page,
+		 unsigned int count),
+
+	TP_ARGS(pfn, page, count),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, pfn)
+		__field(const struct page *, page)
+		__field(unsigned int, count)
+	),
+
+	TP_fast_assign(
+		__entry->pfn = pfn;
+		__entry->page = page;
+		__entry->count = count;
+	),
+
+	TP_printk("pfn=%lx page=%p count=%u",
+		  __entry->pfn,
+		  __entry->page,
+		  __entry->count)
+);
+
+TRACE_EVENT(cma_release,
+
+	TP_PROTO(unsigned long pfn, const struct page *page,
+		 unsigned int count),
+
+	TP_ARGS(pfn, page, count),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, pfn)
+		__field(const struct page *, page)
+		__field(unsigned int, count)
+	),
+
+	TP_fast_assign(
+		__entry->pfn = pfn;
+		__entry->page = page;
+		__entry->count = count;
+	),
+
+	TP_printk("pfn=%lx page=%p count=%u",
+		  __entry->pfn,
+		  __entry->page,
+		  __entry->count)
+);
+
+#endif /* _TRACE_CMA_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>

mm/cma.c +5 -0

index 47203fa..e9410b7c 100644
@@ -23,6 +23,7 @@
 #  define DEBUG
 #endif
 #endif
+#define CREATE_TRACE_POINTS
 
 #include <linux/memblock.h>
 #include <linux/err.h>

@@ -34,6 +35,7 @@
 #include <linux/cma.h>
 #include <linux/highmem.h>
 #include <linux/io.h>
+#include <trace/events/cma.h>
 
 #include "cma.h"

@@ -414,6 +416,8 @@ struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align)
 		start = bitmap_no + mask + 1;
 	}
 
+	trace_cma_alloc(page ? pfn : -1UL, page, count);
+
 	pr_debug("%s(): returned %p\n", __func__, page);
 	return page;
 }

@@ -446,6 +450,7 @@ bool cma_release(struct cma *cma, const struct page *pages, unsigned int count)
 
 	free_contig_range(pfn, count);
 	cma_clear_bitmap(cma, pfn, count);
+	trace_cma_release(pfn, pages, count);
 
 	return true;
 }

Signature · 7 lines

2.1.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Joonsoo Kim

2015-04-02 07:33 UTC #1 lore raw

Hello,

Quoted 11 lines · Stefan Strogin

Add trace events for cma_alloc() and cma_release().
The cma_alloc tracepoint is used both for successful and failed allocations, in case of allocation failure pfn=-1UL is stored and printed.
Signed-off-by Stefan Strogin <[email protected]>
---
Took out from the patch set "mm: cma: add some debug information for CMA" v4
(http://thread.gmane.org/gmane.linux.kernel.mm/129903) because of probable uselessness of the rest of the patches.

I think that patch 5/5 in previous submission is handy and simple to merge. Although we can calculate it by using bitmap, it would be good to get that information(used size and maxchunk size) directly.

mm/cma.c struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align) · line 420

Quoted 6 lines

mm/cma.c +2 -0

@@ -414,6 +416,8 @@ struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align)
 		start = bitmap_no + mask + 1;
 	}
 
+	trace_cma_alloc(page ? pfn : -1UL, page, count);
+

I think that tracing align is also useful.
Is there any reason not to include it?

Thanks.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Stefan Strogin

author

2015-04-02 13:04 UTC #2 lore raw

Hello Joonsoo,

Quoted 11 lines · Joonsoo Kim

Hello,
Quoted 8 lines · Stefan Strogin
Add trace events for cma_alloc() and cma_release().
The cma_alloc tracepoint is used both for successful and failed allocations, in case of allocation failure pfn=-1UL is stored and printed.
Signed-off-by Stefan Strogin <[email protected]>
---

Quoted 8 lines

Quoted 3 lines
Took out from the patch set "mm: cma: add some debug information for CMA" v4
(http://thread.gmane.org/gmane.linux.kernel.mm/129903) because of probable uselessness of the rest of the patches.
I think that patch 5/5 in previous submission is handy and simple to merge. Although we can calculate it by using bitmap, it would be good to get that information(used size and maxchunk size) directly.

Well, then I can send the patch 5/5 once more, this time singly.

Quoted 10 lines

mm/cma.c struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align) · line 420
Quoted 7 lines
mm/cma.c +2 -0
@@ -414,6 +416,8 @@ struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align)
 		start = bitmap_no + mask + 1;
 	}
 
+	trace_cma_alloc(page ? pfn : -1UL, page, count);
+
I think that tracing align is also useful.
Is there any reason not to include it?

In our case (DMA) alignment is easily calculated from the allocation size and CONFIG_CMA_ALIGNMENT. But I think you're right, e.g. it may be not so obvious on powerpc kvm? Anyway it won't be a shortcoming if we trace 'align' too.

Quoted 2 lines

Thanks.

Thank you for the reply.

Stefan Strogin

2015-04-02 13:14 UTC #3 lore raw

PATCH 1/1 mm-cma-add-trace-events-for-cma-allocations-and-freeings-fix

Trace 'align' too in cma_alloc trace event.

Signed-off-by Stefan Strogin <[email protected]>

---
 include/trace/events/cma.h | 11 +++++++----
 mm/cma.c                   |  2 +-
 2 files changed, 8 insertions(+), 5 deletions(-)

include/trace/events/cma.h +7 -4

index e01b35d..d7cd961 100644
@@ -10,26 +10,29 @@
 TRACE_EVENT(cma_alloc,
 
 	TP_PROTO(unsigned long pfn, const struct page *page,
-		 unsigned int count),
+		 unsigned int count, unsigned int align),
 
-	TP_ARGS(pfn, page, count),
+	TP_ARGS(pfn, page, count, align),
 
 	TP_STRUCT__entry(
 		__field(unsigned long, pfn)
 		__field(const struct page *, page)
 		__field(unsigned int, count)
+		__field(unsigned int, align)
 	),
 
 	TP_fast_assign(
 		__entry->pfn = pfn;
 		__entry->page = page;
 		__entry->count = count;
+		__entry->align = align;
 	),
 
-	TP_printk("pfn=%lx page=%p count=%u",
+	TP_printk("pfn=%lx page=%p count=%u align=%u",
 		  __entry->pfn,
 		  __entry->page,
-		  __entry->count)
+		  __entry->count,
+		  __entry->align)
 );
 
 TRACE_EVENT(cma_release,

mm/cma.c +1 -1

index e9410b7c..3a7a67b 100644
@@ -416,7 +416,7 @@ struct page *cma_alloc(struct cma *cma, unsigned int count, unsigned int align)
 		start = bitmap_no + mask + 1;
 	}
 
-	trace_cma_alloc(page ? pfn : -1UL, page, count);
+	trace_cma_alloc(page ? pfn : -1UL, page, count, align);
 
 	pr_debug("%s(): returned %p\n", __func__, page);
 	return page;

Signature · 7 lines

2.1.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>