Skip to contentSkip to conversation

mm: page_alloc: dump migrate-failed pages

PATCH Superseded Minchan Kim 2 participants 3 messages Mar 11, 2021

Subsystems (2)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 mm: page_alloc: dump migrate-failed pages

Currently, debugging CMA allocation failures is quite limited. The most commong source of these failures seems to be page migration which doesn't provide any useful information on the reason of the failure by itself. alloc_contig_range can report those failures as it holds a list of migrate-failed pages.

page refcount, mapcount with page flags on dump_page are helpful information to deduce the culprit. Furthermore, dump_page_owner was super helpful to find long term pinner who initiated the page allocation.

The reason it approach with dynamic debug is the debug message could emit lots of noises as alloc_contig_range calls more frequently since it's a best effort allocator.

There are two ifdefery conditions to support common dyndbg options:

- CONFIG_DYNAMIC_DEBUG_CORE && DYNAMIC_DEBUG_MODULE It aims for supporting the feature with only specific file with adding ccflags.

- CONFIG_DYNAMIC_DEBUG
It aims for supporting the feature with system wide globally.

A simple example to enable the feature:

Admin could enable the dump like this(by default, disabled)

echo "func alloc_contig_dump_pages +p" > control

Admin could disable it.

echo "func alloc_contig_dump_pages =_" > control

Detail goes Documentation/admin-guide/dynamic-debug-howto.rst

A concern is utility functions in dump_page uses inconsistent loglevels.

__dump_page: KERN_WARNING
__dump_page_owner: KERN_ALERT
stack_trace_print: KERN_DEFAULT

There are bunch of places to use the inconsistent loglevel utility functions(e.g., just grep dump_page/strace_trace_print). It's unfortunate but here we are. It could be addressed different patchset.

---
* from v3 - https://lore.kernel.org/linux-mm/[email protected]
* add dyndgb usage comment - akpm
* use dumpstack instead of warn_on - david

* from v2 - https://lore.kernel.org/linux-mm/[email protected]/
* remove ratelimit - mhocko

* from v1 - https://lore.kernel.org/linux-mm/[email protected]/
* use dynamic debugging with system wide instead of per-call site - mhocko

 mm/page_alloc.c | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
mm/page_alloc.c +31 -0
index 3e4b29ee2b1e..76fc202cb105 100644
@@ -8453,6 +8453,36 @@ static unsigned long pfn_max_align_up(unsigned long pfn)
 				pageblock_nr_pages));
 }
 
+#if defined(CONFIG_DYNAMIC_DEBUG) || \
+	(defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
+/*
+ * usage)
+ * 	dyndbg_dir="/sys/kernel/debug/dynamic_debug"
+ * To enable:
+ * 	echo "func alloc_contig_dump_pages +p" > $dyndbg_dir/control
+ * To disable:
+ * 	echo "func alloc_contig_dump_pages =_" > $dyndbg_dir/control
+ * For detail, read dynamic-debug-howto.rst
+ */
+static void alloc_contig_dump_pages(struct list_head *page_list)
+{
+	DEFINE_DYNAMIC_DEBUG_METADATA(descriptor,
+			"migrate failure");
+
+	if (DYNAMIC_DEBUG_BRANCH(descriptor)) {
+		struct page *page;
+
+		dump_stack();
+		list_for_each_entry(page, page_list, lru)
+			dump_page(page, "migration failure");
+	}
+}
+#else
+static inline void alloc_contig_dump_pages(struct list_head *page_list)
+{
+}
+#endif
+
 /* [start, end) must belong to a single zone. */
 static int __alloc_contig_migrate_range(struct compact_control *cc,
 					unsigned long start, unsigned long end)
@@ -8496,6 +8526,7 @@ static int __alloc_contig_migrate_range(struct compact_control *cc,
 				NULL, (unsigned long)&mtc, cc->mode, MR_CONTIG_RANGE);
 	}
 	if (ret < 0) {
+		alloc_contig_dump_pages(&cc->migratepages);
 		putback_movable_pages(&cc->migratepages);
 		return ret;
 	}
Signature · 1 line
2.31.0.rc2.261.g7f71774620-goog
Reviewed-by on v4 1/1
#1 lore raw
Quoted 2 lines · Minchan Kim

Currently, debugging CMA allocation failures is quite limited. The most commong source of these failures seems to be page

s/commong/common/

Quoted 8 lines

migration which doesn't provide any useful information on the reason of the failure by itself. alloc_contig_range can report those failures as it holds a list of migrate-failed pages.

page refcount, mapcount with page flags on dump_page are helpful information to deduce the culprit. Furthermore, dump_page_owner was super helpful to find long term pinner who initiated the page allocation.

Maybe simply "The information logged by dump_page() has already proven helpful for debugging allocation issues, like identifying long-term pinnings on ZONE_MOVABLE or MIGRATE_CMA."

Quoted 4 lines

The reason it approach with dynamic debug is the debug message could emit lots of noises as alloc_contig_range calls more frequently since it's a best effort allocator.

"Let's use the dynamic debugging infrastructure, such that we avoid
flooding the logs and creating a lot of noise on frequent alloc_contig_range() calls. This information is helpful for debugging only."

Quoted 16 lines
Quoted 1 line
Quoted 1 line

There are two ifdefery conditions to support common dyndbg options:

- CONFIG_DYNAMIC_DEBUG_CORE && DYNAMIC_DEBUG_MODULE It aims for supporting the feature with only specific file with adding ccflags.

- CONFIG_DYNAMIC_DEBUG
It aims for supporting the feature with system wide globally.

A simple example to enable the feature:

Admin could enable the dump like this(by default, disabled)

Quoted 7 lines

echo "func alloc_contig_dump_pages +p" > control

Admin could disable it.

echo "func alloc_contig_dump_pages =_" > control

Detail goes Documentation/admin-guide/dynamic-debug-howto.rst

<<< I'd drop that completely and only mention:

"For details on dynamic debugging, see Documentation/admin-guide/dynamic-debug-howto.rst."

As you have usage in the code itself, I think you don't have to be repetitive. The ifdeffery seems to be common (e.g., include/linux/netdevice.) for dynamic debugging users, so I don't see the need to describe that in detail.

Quoted 6 lines
Quoted 1 line
Quoted 1 line

A concern is utility functions in dump_page uses inconsistent loglevels.

Quoted 8 lines

__dump_page: KERN_WARNING
__dump_page_owner: KERN_ALERT
stack_trace_print: KERN_DEFAULT

There are bunch of places to use the inconsistent loglevel utility functions(e.g., just grep dump_page/strace_trace_print). It's unfortunate but here we are. It could be addressed different patchset.

<<< I'd drop that completely and mention

"In the future, we might want to make the loglevels used inside dump_page() consistent and eventually rework the way we log the information here. See [1]"

Where [1] is a link to the discussion.

Quoted 7 lines

---
* from v3 - https://lore.kernel.org/linux-mm/[email protected]
* add dyndgb usage comment - akpm
* use dumpstack instead of warn_on - david

Quoted 9 lines

* from v2 - https://lore.kernel.org/linux-mm/[email protected]/
* remove ratelimit - mhocko

* from v1 - https://lore.kernel.org/linux-mm/[email protected]/
* use dynamic debugging with system wide instead of per-call site - mhocko

  mm/page_alloc.c | 31 +++++++++++++++++++++++++++++++
  1 file changed, 31 insertions(+)

Minor nits:

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8459

Quoted 12 lines
mm/page_alloc.c +4 -0
index 3e4b29ee2b1e..76fc202cb105 100644
@@ -8453,6 +8453,36 @@ static unsigned long pfn_max_align_up(unsigned long pfn)
  				pageblock_nr_pages));
  }
 
+#if defined(CONFIG_DYNAMIC_DEBUG) || \
+	(defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
+/*
+ * usage)

"usage)" looks wrong here. Did you mean "Usage:"

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8465

Quoted 6 lines
mm/page_alloc.c +6 -0
+ * 	dyndbg_dir="/sys/kernel/debug/dynamic_debug"
+ * To enable:
+ * 	echo "func alloc_contig_dump_pages +p" > $dyndbg_dir/control
+ * To disable:
+ * 	echo "func alloc_contig_dump_pages =_" > $dyndbg_dir/control
+ * For detail, read dynamic-debug-howto.rst

Maybe simply

"See admin-guide/dynamic-debug-howto.rst"

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8470

Quoted 5 lines
mm/page_alloc.c +5 -0
+ */
+static void alloc_contig_dump_pages(struct list_head *page_list)
+{
+	DEFINE_DYNAMIC_DEBUG_METADATA(descriptor,
+			"migrate failure");

You can fit that into a single line.

Quoted 20 lines
mm/page_alloc.c +15 -0
+
+	if (DYNAMIC_DEBUG_BRANCH(descriptor)) {
+		struct page *page;
+
+		dump_stack();
+		list_for_each_entry(page, page_list, lru)
+			dump_page(page, "migration failure");
+	}
+}
+#else
+static inline void alloc_contig_dump_pages(struct list_head *page_list)
+{
+}
+#endif
+
  /* [start, end) must belong to a single zone. */
  static int __alloc_contig_migrate_range(struct compact_control *cc,
  					unsigned long start, unsigned long end)

mm/page_alloc.c static int __alloc_contig_migrate_range(struct compact_control *cc, · line 8532

Quoted 7 lines
mm/page_alloc.c +1 -0
@@ -8496,6 +8526,7 @@ static int __alloc_contig_migrate_range(struct compact_control *cc,
  				NULL, (unsigned long)&mtc, cc->mode, MR_CONTIG_RANGE);
  	}
  	if (ret < 0) {
+		alloc_contig_dump_pages(&cc->migratepages);
  		putback_movable_pages(&cc->migratepages);
  		return ret;
  	}

As I said, for my taste good enough for now. I would certainly preferred what Michal suggested (e.g., doing it via debug loglevels), but this gets the job done and is not too ugly.

Signature · 3 lines
Thanks,

David / dhildenb
#2 lore raw
Quoted 5 lines · David Hildenbrand
Quoted 2 lines · Minchan Kim

Currently, debugging CMA allocation failures is quite limited. The most commong source of these failures seems to be page

s/commong/common/

Oops.

Quoted 5 lines
Quoted 5 lines

migration which doesn't provide any useful information on the reason of the failure by itself. alloc_contig_range can report those failures as it holds a list of migrate-failed pages.

Quoted 8 lines
Quoted 4 lines

page refcount, mapcount with page flags on dump_page are helpful information to deduce the culprit. Furthermore, dump_page_owner was super helpful to find long term pinner who initiated the page allocation.

Maybe simply "The information logged by dump_page() has already proven helpful for debugging allocation issues, like identifying long-term pinnings on ZONE_MOVABLE or MIGRATE_CMA."

Will take it.

Quoted 9 lines
Quoted 5 lines

The reason it approach with dynamic debug is the debug message could emit lots of noises as alloc_contig_range calls more frequently since it's a best effort allocator.

"Let's use the dynamic debugging infrastructure, such that we avoid flooding the logs and creating a lot of noise on frequent alloc_contig_range() calls. This information is helpful for debugging only."

Sure.

Quoted 21 lines
Quoted 21 lines
Quoted 2 lines
Quoted 2 lines

There are two ifdefery conditions to support common dyndbg options:

- CONFIG_DYNAMIC_DEBUG_CORE && DYNAMIC_DEBUG_MODULE It aims for supporting the feature with only specific file with adding ccflags.

- CONFIG_DYNAMIC_DEBUG
It aims for supporting the feature with system wide globally.

A simple example to enable the feature:

Admin could enable the dump like this(by default, disabled)

echo "func alloc_contig_dump_pages +p" > control

Admin could disable it.

Quoted 8 lines
Quoted 3 lines

echo "func alloc_contig_dump_pages =_" > control

Detail goes Documentation/admin-guide/dynamic-debug-howto.rst

<<< I'd drop that completely and only mention:

"For details on dynamic debugging, see Documentation/admin-guide/dynamic-debug-howto.rst."

I'd like to keep it since many developer are not familiar with dynamic debug and the syntax was not straightforward to me when I read the doc. I believe that's why Andrew asked me to add simple comment on the comment. Including simple how to use in the description makes reviewer more comfortable to start the detail doc if they want to read and test.

Quoted 48 lines

As you have usage in the code itself, I think you don't have to be repetitive. The ifdeffery seems to be common (e.g., include/linux/netdevice.) for dynamic debugging users, so I don't see the need to describe that in detail.

Quoted 6 lines
Quoted 1 line
Quoted 1 line

A concern is utility functions in dump_page uses inconsistent loglevels.

Quoted 8 lines

__dump_page: KERN_WARNING
__dump_page_owner: KERN_ALERT
stack_trace_print: KERN_DEFAULT

There are bunch of places to use the inconsistent loglevel utility functions(e.g., just grep dump_page/strace_trace_print). It's unfortunate but here we are. It could be addressed different patchset.

<<< I'd drop that completely and mention

"In the future, we might want to make the loglevels used inside dump_page() consistent and eventually rework the way we log the information here. See [1]"

Where [1] is a link to the discussion.

Quoted 7 lines

---
* from v3 - https://lore.kernel.org/linux-mm/[email protected]
* add dyndgb usage comment - akpm
* use dumpstack instead of warn_on - david

Quoted 9 lines

* from v2 - https://lore.kernel.org/linux-mm/[email protected]/
* remove ratelimit - mhocko

* from v1 - https://lore.kernel.org/linux-mm/[email protected]/
* use dynamic debugging with system wide instead of per-call site - mhocko

  mm/page_alloc.c | 31 +++++++++++++++++++++++++++++++
  1 file changed, 31 insertions(+)

Minor nits:

Quoted 13 lines

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8458

Quoted 11 lines
mm/page_alloc.c +4 -0
index 3e4b29ee2b1e..76fc202cb105 100644
@@ -8453,6 +8453,36 @@ static unsigned long pfn_max_align_up(unsigned long pfn)
  				pageblock_nr_pages));
  }
+#if defined(CONFIG_DYNAMIC_DEBUG) || \
+	(defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
+/*
+ * usage)

"usage)" looks wrong here. Did you mean "Usage:"

Sure.

Quoted 11 lines

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8465

Quoted 7 lines
mm/page_alloc.c +6 -0
+ * 	dyndbg_dir="/sys/kernel/debug/dynamic_debug"
+ * To enable:
+ * 	echo "func alloc_contig_dump_pages +p" > $dyndbg_dir/control
+ * To disable:
+ * 	echo "func alloc_contig_dump_pages =_" > $dyndbg_dir/control
+ * For detail, read dynamic-debug-howto.rst

Maybe simply

"See admin-guide/dynamic-debug-howto.rst"

Yes, for code comment, it would be enough to point out where the detail is.

Quoted 8 lines

mm/page_alloc.c static unsigned long pfn_max_align_up(unsigned long pfn) · line 8470

Quoted 6 lines
mm/page_alloc.c +5 -0
+ */
+static void alloc_contig_dump_pages(struct list_head *page_list)
+{
+	DEFINE_DYNAMIC_DEBUG_METADATA(descriptor,
+			"migrate failure");

You can fit that into a single line.

Sure.

Quoted 23 lines
Quoted 23 lines
mm/page_alloc.c +15 -0
+
+	if (DYNAMIC_DEBUG_BRANCH(descriptor)) {
+		struct page *page;
+
+		dump_stack();
+		list_for_each_entry(page, page_list, lru)
+			dump_page(page, "migration failure");
+	}
+}
+#else
+static inline void alloc_contig_dump_pages(struct list_head *page_list)
+{
+}
+#endif
+
  /* [start, end) must belong to a single zone. */
  static int __alloc_contig_migrate_range(struct compact_control *cc,
  					unsigned long start, unsigned long end)
@@ -8496,6 +8526,7 @@ static int __alloc_contig_migrate_range(struct compact_control *cc,
  				NULL, (unsigned long)&mtc, cc->mode, MR_CONTIG_RANGE);
  	}
  	if (ret < 0) {
Quoted 8 lines

mm/page_alloc.c static int __alloc_contig_migrate_range(struct compact_control *cc, · line 8532

Quoted 5 lines
mm/page_alloc.c +1 -0
+		alloc_contig_dump_pages(&cc->migratepages);
  		putback_movable_pages(&cc->migratepages);
  		return ret;
  	}

As I said, for my taste good enough for now. I would certainly preferred what Michal suggested (e.g., doing it via debug loglevels), but this gets

Sure, that is definitely improvement. Someday, I will post the clean up.

Quoted 3 lines

the job done and is not too ugly.

Thanks.