lib/Kconfig.debug: introduce CONFIG_CC_STACK_USAGE

This patch adds option for collecting stack usage statistics via gcc option '-fstack-usage' (needs gcc 4.6 or newer). For each .o file gcc dumps stack-frame sizes into .su file.

File format:
<file>:<line>:<column>:<function> <size> <qualifiers>

select.c:870:5:do_sys_poll      1040    static
rtnetlink.c:1880:12:rtnl_newlink        552     dynamic
random.c:1334:1:random_read     664     dynamic,bounded

Quote from the gcc manpage about qualifiers:

The qualifier "static" means that the function manipulates the stack statically: a fixed number of bytes are allocated for the frame on function entry and released on function exit; no stack adjustments are otherwise made in the function. The second field is this fixed number of bytes.

The qualifier "dynamic" means that the function manipulates the stack dynamically: in addition to the static allocation described above, stack adjustments are made in the body of the function, for example to push/pop arguments around function calls. If the qualifier "bounded" is also present, the amount of these adjustments is bounded at compile time and the second field is an upper bound of the total amount of stack used by the function. If it is not present, the amount of these adjustments is not bounded at compile time and the second field only represents the bounded part.

In comparison to the scripts/checkstack.pl this method is more accurate. It takes into account function arguments, push/pop pairs, frame pointers and saved return address. It gives information about all functions except purely assember (for example sha1_transform_avx2).

Also this patch adds make target 'stack-usage' which combines all .su files into one sorted 'stack-usage'. Also prints top 100 and all unbounded dynamic stack frames.

Signed-off-by Konstantin Khlebnikov <[email protected]>

---
 Makefile          |   22 +++++++++++++++++++++-
 lib/Kconfig.debug |    9 +++++++++
 2 files changed, 30 insertions(+), 1 deletion(-)

Makefile +21 -1

index 2167084..9ba61ee 100644
@@ -635,6 +635,10 @@ ifneq ($(CONFIG_FRAME_WARN),0)
 KBUILD_CFLAGS += $(call cc-option,-Wframe-larger-than=${CONFIG_FRAME_WARN})
 endif
 
+ifdef CONFIG_CC_STACK_USAGE
+KBUILD_CFLAGS += $(call cc-option,-fstack-usage)
+endif
+
 # Handle stack protector mode.
 ifdef CONFIG_CC_STACKPROTECTOR_REGULAR
   stackp-flag := -fstack-protector

@@ -1228,6 +1232,7 @@ help:
         echo  ''
        @echo  'Static analysers'
        @echo  '  checkstack      - Generate a list of stack hogs'
+       @echo  '  stack-usage     - Alternative list of stack hogs'
        @echo  '  namespacecheck  - Name space analysis on compiled kernel'
        @echo  '  versioncheck    - Sanity check on version.h usage'
        @echo  '  includecheck    - Check for duplicate included header files'

@@ -1376,7 +1381,7 @@ clean: $(clean-dirs)
        $(call cmd,rmfiles)
        @find $(if $(KBUILD_EXTMOD), $(KBUILD_EXTMOD), .) $(RCS_FIND_IGNORE) \
                \( -name '*.[oas]' -o -name '*.ko' -o -name '.*.cmd' \
-               -o -name '*.ko.*' \
+               -o -name '*.ko.*' -o -name '*.su' \
                -o -name '.*.d' -o -name '.*.tmp' -o -name '*.mod.c' \
                -o -name '*.symtypes' -o -name 'modules.order' \
                -o -name modules.builtin -o -name '.tmp_*.o.*' \

@@ -1432,6 +1437,21 @@ checkstack:
        $(OBJDUMP) -d vmlinux $$(find . -name '*.ko') | \
        $(PERL) $(src)/scripts/checkstack.pl $(CHECKSTACK_ARCH)
 
+quiet_cmd_cc_stack_usage = SORTING $@
+cmd_cc_stack_usage = find -name '*.su' | xargs cat | sort -k 2 -n -r > $@
+
+stack-usage: FORCE
+ifndef CONFIG_CC_STACK_USAGE
+       $(Q)echo >&2 'Please rebuild with CONFIG_CC_STACK_USAGE=y' && false
+endif
+       $(call cmd,cc_stack_usage)
+       $(Q)echo ''
+       $(Q)echo 'Unbounded dynamic stack frames:'
+       $(Q)grep '\bdynamic$$' $@
+       $(Q)echo ''
+       $(Q)echo 'Top 100 stack frames:'
+       $(Q)head -100 < $@
+
 kernelrelease:
        @echo "$(KERNELVERSION)$$($(CONFIG_SHELL) $(srctree)/scripts/setlocalversion $(srctree))"
 

lib/Kconfig.debug +9 -0

index 7a638aa..3b30853 100644
@@ -170,6 +170,15 @@ config FRAME_WARN
          Setting it to 0 disables the warning.
          Requires gcc 4.4
 
+config CC_STACK_USAGE
+       bool "Collect statistics about stack usage (needs gcc 4.6)"
+       help
+         This enables gcc option -fstack-usage which saves information
+         about stack frames in per-function basis into '.su' files.
+
+         Run 'make stack-usage' after building kernel and modules to
+         combine all information in one file.
+
 config STRIP_ASM_SYMS
        bool "Strip assembler-generated symbols during link"
        default n

This patch adds option for collecting stack usage statistics via gcc option '-fstack-usage' (needs gcc 4.6 or newer). For each .o file gcc dumps stack-frame sizes into .su file.

File format:
<file>:<line>:<column>:<function> <size> <qualifiers>

select.c:870:5:do_sys_poll 1040 static
rtnetlink.c:1880:12:rtnl_newlink 552 dynamic
random.c:1334:1:random_read 664 dynamic,bounded

Quote from the gcc manpage about qualifiers:

The qualifier "static" means that the function manipulates the stack statically: a fixed number of bytes are allocated for the frame on function entry and released on function exit; no stack adjustments are otherwise made in the function. The second field is this fixed number of bytes.

The qualifier "dynamic" means that the function manipulates the stack dynamically: in addition to the static allocation described above, stack adjustments are made in the body of the function, for example to push/pop arguments around function calls. If the qualifier "bounded" is also present, the amount of these adjustments is bounded at compile time and the second field is an upper bound of the total amount of stack used by the function. If it is not present, the amount of these adjustments is not bounded at compile time and the second field only represents the bounded part.

In comparison to the scripts/checkstack.pl this method is more accurate. It takes into account function arguments, push/pop pairs, frame pointers and saved return address. It gives information about all functions except purely assember (for example sha1_transform_avx2).

Also this patch adds make target 'stack-usage' which combines all .su files into one sorted 'stack-usage'. Also prints top 100 and all unbounded dynamic stack frames.

Signed-off-by Konstantin Khlebnikov <[email protected]>

---
 .gitignore        |    3 +++
 Makefile          |   23 ++++++++++++++++++++++-
 lib/Kconfig.debug |    9 +++++++++
 3 files changed, 34 insertions(+), 1 deletion(-)

.gitignore +3 -0

index f4c0b09..96b8984 100644
@@ -85,6 +85,9 @@ GTAGS
 *~
 \#*#
 
+*.su
+/stack-usage
+
 #
 # Leavings from module signing
 #

Makefile +5 -0

index 2167084..29fe588 100644
@@ -635,6 +635,11 @@ ifneq ($(CONFIG_FRAME_WARN),0)
 KBUILD_CFLAGS += $(call cc-option,-Wframe-larger-than=${CONFIG_FRAME_WARN})
 endif
 
+ifdef CONFIG_CC_STACK_USAGE
+KBUILD_CFLAGS += $(call cc-option,-fstack-usage)
+endif
+CLEAN_FILES += stack-usage
+
 # Handle stack protector mode.
 ifdef CONFIG_CC_STACKPROTECTOR_REGULAR
   stackp-flag := -fstack-protector

Makefile +6 -1

@@ -1376,7 +1382,7 @@ clean: $(clean-dirs)
 	$(call cmd,rmfiles)
 	@find $(if $(KBUILD_EXTMOD), $(KBUILD_EXTMOD), .) $(RCS_FIND_IGNORE) \
 		\( -name '*.[oas]' -o -name '*.ko' -o -name '.*.cmd' \
-		-o -name '*.ko.*' \
+		-o -name '*.ko.*' -o -name '*.su' \
 		-o -name '.*.d' -o -name '.*.tmp' -o -name '*.mod.c' \
 		-o -name '*.symtypes' -o -name 'modules.order' \
 		-o -name modules.builtin -o -name '.tmp_*.o.*' \

@@ -1432,6 +1438,21 @@ checkstack:
 	$(OBJDUMP) -d vmlinux $$(find . -name '*.ko') | \
 	$(PERL) $(src)/scripts/checkstack.pl $(CHECKSTACK_ARCH)
 
+quiet_cmd_cc_stack_usage = SORTING $@
+cmd_cc_stack_usage = find -name '*.su' | xargs cat | sort -k 2 -n -r > $@
+
+stack-usage: FORCE
+ifndef CONFIG_CC_STACK_USAGE

Hi Konstantin.

Quoted 80 lines · Konstantin Khlebnikov

This patch adds option for collecting stack usage statistics via gcc option '-fstack-usage' (needs gcc 4.6 or newer). For each .o file gcc dumps stack-frame sizes into .su file.

File format:
<file>:<line>:<column>:<function> <size> <qualifiers>

select.c:870:5:do_sys_poll 1040 static
rtnetlink.c:1880:12:rtnl_newlink 552 dynamic
random.c:1334:1:random_read 664 dynamic,bounded

Quote from the gcc manpage about qualifiers:

The qualifier "static" means that the function manipulates the stack statically: a fixed number of bytes are allocated for the frame on function entry and released on function exit; no stack adjustments are otherwise made in the function. The second field is this fixed number of bytes.

The qualifier "dynamic" means that the function manipulates the stack dynamically: in addition to the static allocation described above, stack adjustments are made in the body of the function, for example to push/pop arguments around function calls. If the qualifier "bounded" is also present, the amount of these adjustments is bounded at compile time and the second field is an upper bound of the total amount of stack used by the function. If it is not present, the amount of these adjustments is not bounded at compile time and the second field only represents the bounded part.

In comparison to the scripts/checkstack.pl this method is more accurate. It takes into account function arguments, push/pop pairs, frame pointers and saved return address. It gives information about all functions except purely assember (for example sha1_transform_avx2).

Also this patch adds make target 'stack-usage' which combines all .su files into one sorted 'stack-usage'. Also prints top 100 and all unbounded dynamic stack frames.

Signed-off-by Konstantin Khlebnikov <[email protected]>

---
 .gitignore        |    3 +++
 Makefile          |   23 ++++++++++++++++++++++-
 lib/Kconfig.debug |    9 +++++++++
 3 files changed, 34 insertions(+), 1 deletion(-)

.gitignore +3 -0

index f4c0b09..96b8984 100644
@@ -85,6 +85,9 @@ GTAGS
 *~
 \#*#
 
+*.su
+/stack-usage
+
 #
 # Leavings from module signing
 #

Makefile +5 -0

index 2167084..29fe588 100644
@@ -635,6 +635,11 @@ ifneq ($(CONFIG_FRAME_WARN),0)
 KBUILD_CFLAGS += $(call cc-option,-Wframe-larger-than=${CONFIG_FRAME_WARN})
 endif
 
+ifdef CONFIG_CC_STACK_USAGE
+KBUILD_CFLAGS += $(call cc-option,-fstack-usage)
+endif
+CLEAN_FILES += stack-usage
+
 # Handle stack protector mode.
 ifdef CONFIG_CC_STACKPROTECTOR_REGULAR
   stackp-flag := -fstack-protector

@@ -1228,6 +1233,7 @@ help:
       echo  ''

Makefile help: · line 1239

Quoted 6 lines

Makefile +1 -0

      @echo  'Static analysers'
      @echo  '  checkstack      - Generate a list of stack hogs'
+     @echo  '  stack-usage     - Alternative list of stack hogs'
      @echo  '  namespacecheck  - Name space analysis on compiled kernel'
      @echo  '  versioncheck    - Sanity check on version.h usage'
      @echo  '  includecheck    - Check for duplicate included header files'

The other "check" targets have check in their name.
Can we follow this naming convention and call it checkstackusage?

Quoted 20 lines

Makefile +6 -1

@@ -1376,7 +1382,7 @@ clean: $(clean-dirs)
      $(call cmd,rmfiles)
      @find $(if $(KBUILD_EXTMOD), $(KBUILD_EXTMOD), .) $(RCS_FIND_IGNORE) \
              \( -name '*.[oas]' -o -name '*.ko' -o -name '.*.cmd' \
-             -o -name '*.ko.*' \
+             -o -name '*.ko.*' -o -name '*.su' \
              -o -name '.*.d' -o -name '.*.tmp' -o -name '*.mod.c' \
              -o -name '*.symtypes' -o -name 'modules.order' \
              -o -name modules.builtin -o -name '.tmp_*.o.*' \

@@ -1432,6 +1438,21 @@ checkstack:
      $(OBJDUMP) -d vmlinux $$(find . -name '*.ko') | \
      $(PERL) $(src)/scripts/checkstack.pl $(CHECKSTACK_ARCH)
 
+quiet_cmd_cc_stack_usage = SORTING $@
+cmd_cc_stack_usage = find -name '*.su' | xargs cat | sort -k 2 -n -r > $@
+
+stack-usage: FORCE
+ifndef CONFIG_CC_STACK_USAGE

Makefile checkstack: · line 1455

Quoted 10 lines

Makefile +10 -0

+     $(Q)echo >&2 'Please rebuild with CONFIG_CC_STACK_USAGE=y' && false
+endif
+     $(call cmd,cc_stack_usage)
+     $(Q)echo ''
+     $(Q)echo 'Unbounded dynamic stack frames:'
+     $(Q)grep '\bdynamic$$' $@ || true
+     $(Q)echo ''
+     $(Q)echo 'Top 100 stack frames:'
+     $(Q)head -100 < $@
+

All this belongs to a script - for example named

scripts/checkstackusage.sh

The top-level Makefile shall be a thin driver and there are too much functionality in that file. but this is no excuse for adding more.

Sam