[PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Arnd Bergmann]

In randconfig builds with sysfs, procfs and numa all disabled,
but SMP enabled, we now get a link error in the newly introduced
vmstat_refresh function:

mm/built-in.o: In function `vmstat_refresh':
:(.text+0x15c78): undefined reference to `vmstat_text'

This modifes the already elaborate #ifdef to also cover that
configuration.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: mmotm ("mm: /proc/sys/vm/stat_refresh to force vmstat update")
---
 mm/vmstat.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/mm/vmstat.c b/mm/vmstat.c
index 57a24e919907..5367eb9b858b 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -678,7 +678,8 @@ int fragmentation_index(struct zone *zone, unsigned int order)
 }
 #endif
 
-#if defined(CONFIG_PROC_FS) || defined(CONFIG_SYSFS) || defined(CONFIG_NUMA)
+#if defined(CONFIG_PROC_FS) || defined(CONFIG_SYSFS) || \
+    defined(CONFIG_NUMA) || defined(CONFIG_SMP)
 #ifdef CONFIG_ZONE_DMA
 #define TEXT_FOR_DMA(xx) xx "_dma",
 #else
@@ -857,7 +858,7 @@ const char * const vmstat_text[] = {
 #endif
 #endif /* CONFIG_VM_EVENTS_COUNTERS */
 };
-#endif /* CONFIG_PROC_FS || CONFIG_SYSFS || CONFIG_NUMA */
+#endif /* CONFIG_PROC_FS || CONFIG_SYSFS || CONFIG_NUMA || CONFIG_SMP */
 
 
 #if (defined(CONFIG_DEBUG_FS) && defined(CONFIG_COMPACTION)) || \
-- 
2.7.0

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Christoph Lameter]

On Wed, 11 May 2016, Arnd Bergmann wrote:

> In randconfig builds with sysfs, procfs and numa all disabled,
> but SMP enabled, we now get a link error in the newly introduced
> vmstat_refresh function:
>
> mm/built-in.o: In function `vmstat_refresh':

Hmmm... vmstat_refresh should not be build if CONFIG_PROC_FS is not set
since there will be no way to trigger it. Lets not complicate this
further.



Subject: Do not build vmstat_refresh if there is no procfs support

It makes no sense to build functionality into the kernel that
cannot be used and causes build issues.

Signed-off-by: Christoph Lameter <cl@linux.com>

Index: linux/mm/vmstat.c
===================================================================
--- linux.orig/mm/vmstat.c
+++ linux/mm/vmstat.c
@@ -1358,7 +1358,6 @@ static const struct file_operations proc
 	.llseek		= seq_lseek,
 	.release	= seq_release,
 };
-#endif /* CONFIG_PROC_FS */

 #ifdef CONFIG_SMP
 static struct workqueue_struct *vmstat_wq;
@@ -1422,7 +1421,10 @@ int vmstat_refresh(struct ctl_table *tab
 		*lenp = 0;
 	return 0;
 }
+#endif /* CONFIG_SMP */
+#endif /* CONFIG_PROC_FS */

+#ifdef CONFIG_SMP
 static void vmstat_update(struct work_struct *w)
 {
 	if (refresh_cpu_vm_stats(true)) {

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Michal Hocko]

On Wed 11-05-16 10:32:11, Christoph Lameter wrote:
> Subject: Do not build vmstat_refresh if there is no procfs support
> 
> It makes no sense to build functionality into the kernel that
> cannot be used and causes build issues.
> 
> Signed-off-by: Christoph Lameter <cl@linux.com>
> 
> Index: linux/mm/vmstat.c
> ===================================================================
> --- linux.orig/mm/vmstat.c
> +++ linux/mm/vmstat.c
> @@ -1358,7 +1358,6 @@ static const struct file_operations proc
>  	.llseek		= seq_lseek,
>  	.release	= seq_release,
>  };
> -#endif /* CONFIG_PROC_FS */
> 
>  #ifdef CONFIG_SMP
>  static struct workqueue_struct *vmstat_wq;

This doesn't work because it makes the whole vmstat_wq depend on
CONFIG_PROC_FS. Which is obviously bad because we both rely on doing the
periodic sync even when counters are not exported to the userspace and
it wound't compile anyway...
-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Michal Hocko]

On Wed 11-05-16 16:54:55, Arnd Bergmann wrote:
> In randconfig builds with sysfs, procfs and numa all disabled,
> but SMP enabled, we now get a link error in the newly introduced
> vmstat_refresh function:
> 
> mm/built-in.o: In function `vmstat_refresh':
> :(.text+0x15c78): undefined reference to `vmstat_text'
> 
> This modifes the already elaborate #ifdef to also cover that
> configuration.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> Fixes: mmotm ("mm: /proc/sys/vm/stat_refresh to force vmstat update")

I agree with Christoph that vmstat_refresh is PROC_FS only so we should
fix it there. It is not like this would be generally reusable helper...
Why don't we just do:
---
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 57a24e919907..c759b526287b 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1370,6 +1370,7 @@ static void refresh_vm_stats(struct work_struct *work)
 	refresh_cpu_vm_stats(true);
 }
 
+#ifdef CONFIG_PROC_FS
 int vmstat_refresh(struct ctl_table *table, int write,
 		   void __user *buffer, size_t *lenp, loff_t *ppos)
 {
@@ -1422,6 +1423,7 @@ int vmstat_refresh(struct ctl_table *table, int write,
 		*lenp = 0;
 	return 0;
 }
+#endif
 
 static void vmstat_update(struct work_struct *w)
 {
-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Christoph Lameter]

On Mon, 16 May 2016, Michal Hocko wrote:

> I agree with Christoph that vmstat_refresh is PROC_FS only so we should
> fix it there. It is not like this would be generally reusable helper...
> Why don't we just do:

Looks good.

Acked-by: Christoph Lameter <cl@linux.com>

> ---
> diff --git a/mm/vmstat.c b/mm/vmstat.c
> index 57a24e919907..c759b526287b 100644
> --- a/mm/vmstat.c
> +++ b/mm/vmstat.c
> @@ -1370,6 +1370,7 @@ static void refresh_vm_stats(struct work_struct *work)
>  	refresh_cpu_vm_stats(true);
>  }
>
> +#ifdef CONFIG_PROC_FS
>  int vmstat_refresh(struct ctl_table *table, int write,
>  		   void __user *buffer, size_t *lenp, loff_t *ppos)
>  {
> @@ -1422,6 +1423,7 @@ int vmstat_refresh(struct ctl_table *table, int write,
>  		*lenp = 0;
>  	return 0;
>  }
> +#endif
>
>  static void vmstat_update(struct work_struct *w)
>  {
>

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Michal Hocko]

Andrew, I think that the following is more straightforward fix and
should be folded in to the patch which has introduced vmstat_refresh.
---
>From b8dd18fb7df040e1bfe61aadde1d903589de15e4 Mon Sep 17 00:00:00 2001
From: Michal Hocko <mhocko@suse.com>
Date: Mon, 16 May 2016 16:19:53 +0200
Subject: [PATCH] mmotm: mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix

Arnd has reported:
In randconfig builds with sysfs, procfs and numa all disabled,
but SMP enabled, we now get a link error in the newly introduced
vmstat_refresh function:

mm/built-in.o: In function `vmstat_refresh':
:(.text+0x15c78): undefined reference to `vmstat_text'

vmstat_refresh is proc_fs specific so there is no reason to define it
when !CONFIG_PROC_FS.

Reported-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
---
 mm/vmstat.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/mm/vmstat.c b/mm/vmstat.c
index 57a24e919907..c759b526287b 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1370,6 +1370,7 @@ static void refresh_vm_stats(struct work_struct *work)
 	refresh_cpu_vm_stats(true);
 }
 
+#ifdef CONFIG_PROC_FS
 int vmstat_refresh(struct ctl_table *table, int write,
 		   void __user *buffer, size_t *lenp, loff_t *ppos)
 {
@@ -1422,6 +1423,7 @@ int vmstat_refresh(struct ctl_table *table, int write,
 		*lenp = 0;
 	return 0;
 }
+#endif
 
 static void vmstat_update(struct work_struct *w)
 {
-- 
2.8.1

-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Andrew Morton]

On Mon, 16 May 2016 16:23:33 +0200 Michal Hocko <mhocko@kernel.org> wrote:

> Andrew, I think that the following is more straightforward fix and
> should be folded in to the patch which has introduced vmstat_refresh.
> ---
> >From b8dd18fb7df040e1bfe61aadde1d903589de15e4 Mon Sep 17 00:00:00 2001
> From: Michal Hocko <mhocko@suse.com>
> Date: Mon, 16 May 2016 16:19:53 +0200
> Subject: [PATCH] mmotm: mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix
> 
> Arnd has reported:
> In randconfig builds with sysfs, procfs and numa all disabled,
> but SMP enabled, we now get a link error in the newly introduced
> vmstat_refresh function:
> 
> mm/built-in.o: In function `vmstat_refresh':
> :(.text+0x15c78): undefined reference to `vmstat_text'
> 
> vmstat_refresh is proc_fs specific so there is no reason to define it
> when !CONFIG_PROC_FS.

I already had this:

From: Christoph Lameter <cl@linux.com>
Subject: Do not build vmstat_refresh if there is no procfs support

It makes no sense to build functionality into the kernel that
cannot be used and causes build issues.

Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1605111011260.9351@east.gentwo.org
Signed-off-by: Christoph Lameter <cl@linux.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 mm/vmstat.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff -puN mm/vmstat.c~mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix mm/vmstat.c
--- a/mm/vmstat.c~mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix
+++ a/mm/vmstat.c
@@ -1371,7 +1371,6 @@ static const struct file_operations proc
 	.llseek		= seq_lseek,
 	.release	= seq_release,
 };
-#endif /* CONFIG_PROC_FS */
 
 #ifdef CONFIG_SMP
 static struct workqueue_struct *vmstat_wq;
@@ -1436,7 +1435,10 @@ int vmstat_refresh(struct ctl_table *tab
 		*lenp = 0;
 	return 0;
 }
+#endif /* CONFIG_SMP */
+#endif /* CONFIG_PROC_FS */
 
+#ifdef CONFIG_SMP
 static void vmstat_update(struct work_struct *w)
 {
 	if (refresh_cpu_vm_stats(true)) {
_

Re: Do not build vmstat_refresh if there is no procfs support

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 14871 bytes --]

Hi,

[auto build test ERROR on next-20160511]
[cannot apply to v4.6-rc7 v4.6-rc6 v4.6-rc5 v4.6]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Christoph-Lameter/Do-not-build-vmstat_refresh-if-there-is-no-procfs-support/20160511-233405
config: arm64-allnoconfig (attached as .config)
compiler: aarch64-linux-gnu-gcc (Debian 5.3.1-8) 5.3.1 20160205
reproduce:
        wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=arm64 

All error/warnings (new ones prefixed by >>):

   mm/vmstat.c: In function 'vmstat_update':
>> mm/vmstat.c:1436:45: error: 'vmstat_wq' undeclared (first use in this function)
      queue_delayed_work_on(smp_processor_id(), vmstat_wq,
                                                ^
   mm/vmstat.c:1436:45: note: each undeclared identifier is reported only once for each function it appears in
   In file included from include/asm-generic/percpu.h:6:0,
                    from arch/arm64/include/asm/percpu.h:276,
                    from include/linux/percpu.h:12,
                    from include/linux/percpu-rwsem.h:6,
                    from include/linux/fs.h:30,
                    from mm/vmstat.c:12:
>> mm/vmstat.c:1437:19: error: 'vmstat_work' undeclared (first use in this function)
        this_cpu_ptr(&vmstat_work),
                      ^
   include/linux/percpu-defs.h:206:47: note: in definition of macro '__verify_pcpu_ptr'
     const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
                                                  ^
   include/linux/percpu-defs.h:239:27: note: in expansion of macro 'raw_cpu_ptr'
    #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr)
                              ^
>> mm/vmstat.c:1437:5: note: in expansion of macro 'this_cpu_ptr'
        this_cpu_ptr(&vmstat_work),
        ^
   In file included from include/linux/fs.h:32:0,
                    from mm/vmstat.c:12:
   mm/vmstat.c: In function 'quiet_vmstat':
   mm/vmstat.c:1480:42: error: 'vmstat_work' undeclared (first use in this function)
     if (!delayed_work_pending(this_cpu_ptr(&vmstat_work)))
                                             ^
   include/linux/workqueue.h:26:51: note: in definition of macro 'work_data_bits'
    #define work_data_bits(work) ((unsigned long *)(&(work)->data))
                                                      ^
   include/linux/workqueue.h:271:2: note: in expansion of macro 'work_pending'
     work_pending(&(w)->work)
     ^
>> mm/vmstat.c:1480:7: note: in expansion of macro 'delayed_work_pending'
     if (!delayed_work_pending(this_cpu_ptr(&vmstat_work)))
          ^
   include/linux/percpu-defs.h:228:2: note: in expansion of macro '__verify_pcpu_ptr'
     __verify_pcpu_ptr(ptr);      \
     ^
   include/linux/percpu-defs.h:239:27: note: in expansion of macro 'raw_cpu_ptr'
    #define this_cpu_ptr(ptr) raw_cpu_ptr(ptr)
                              ^
   mm/vmstat.c:1480:28: note: in expansion of macro 'this_cpu_ptr'
     if (!delayed_work_pending(this_cpu_ptr(&vmstat_work)))
                               ^
   In file included from include/asm-generic/percpu.h:6:0,
                    from arch/arm64/include/asm/percpu.h:276,
                    from include/linux/percpu.h:12,
                    from include/linux/percpu-rwsem.h:6,
                    from include/linux/fs.h:30,
                    from mm/vmstat.c:12:
   mm/vmstat.c: In function 'vmstat_shepherd':
   mm/vmstat.c:1512:38: error: 'vmstat_work' undeclared (first use in this function)
      struct delayed_work *dw = &per_cpu(vmstat_work, cpu);
                                         ^
   include/linux/percpu-defs.h:206:47: note: in definition of macro '__verify_pcpu_ptr'
     const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
                                                  ^
   include/linux/percpu-defs.h:256:29: note: in expansion of macro 'per_cpu_ptr'
    #define per_cpu(var, cpu) (*per_cpu_ptr(&(var), cpu))
                                ^
>> mm/vmstat.c:1512:30: note: in expansion of macro 'per_cpu'
      struct delayed_work *dw = &per_cpu(vmstat_work, cpu);
                                 ^
   mm/vmstat.c:1515:32: error: 'vmstat_wq' undeclared (first use in this function)
        queue_delayed_work_on(cpu, vmstat_wq, dw, 0);
                                   ^
   In file included from include/linux/fs.h:32:0,
                    from mm/vmstat.c:12:
   mm/vmstat.c: In function 'start_shepherd_timer':
   mm/vmstat.c:1528:37: error: 'vmstat_work' undeclared (first use in this function)
      INIT_DEFERRABLE_WORK(per_cpu_ptr(&vmstat_work, cpu),
                                        ^
   include/linux/workqueue.h:216:16: note: in definition of macro '__INIT_WORK'
      __init_work((_work), _onstack);    \
                   ^
   include/linux/workqueue.h:231:3: note: in expansion of macro 'INIT_WORK'
      INIT_WORK(&(_work)->work, (_func));   \
      ^
   include/linux/workqueue.h:253:2: note: in expansion of macro '__INIT_DELAYED_WORK'
     __INIT_DELAYED_WORK(_work, _func, TIMER_DEFERRABLE)
     ^
>> mm/vmstat.c:1528:3: note: in expansion of macro 'INIT_DEFERRABLE_WORK'
      INIT_DEFERRABLE_WORK(per_cpu_ptr(&vmstat_work, cpu),
      ^
   include/linux/percpu-defs.h:222:2: note: in expansion of macro '__verify_pcpu_ptr'
     __verify_pcpu_ptr(ptr);      \
     ^
>> mm/vmstat.c:1528:24: note: in expansion of macro 'per_cpu_ptr'
      INIT_DEFERRABLE_WORK(per_cpu_ptr(&vmstat_work, cpu),
                           ^
   mm/vmstat.c:1531:2: error: 'vmstat_wq' undeclared (first use in this function)
     vmstat_wq = alloc_workqueue("vmstat", WQ_FREEZABLE|WQ_MEM_RECLAIM, 0);
     ^
   In file included from include/asm-generic/percpu.h:6:0,
                    from arch/arm64/include/asm/percpu.h:276,
                    from include/linux/percpu.h:12,
                    from include/linux/percpu-rwsem.h:6,
                    from include/linux/fs.h:30,
                    from mm/vmstat.c:12:
   mm/vmstat.c: In function 'vmstat_cpuup_callback':
   mm/vmstat.c:1568:37: error: 'vmstat_work' undeclared (first use in this function)
      cancel_delayed_work_sync(&per_cpu(vmstat_work, cpu));
                                        ^
   include/linux/percpu-defs.h:206:47: note: in definition of macro '__verify_pcpu_ptr'
     const void __percpu *__vpp_verify = (typeof((ptr) + 0))NULL; \
                                                  ^
   include/linux/percpu-defs.h:256:29: note: in expansion of macro 'per_cpu_ptr'
    #define per_cpu(var, cpu) (*per_cpu_ptr(&(var), cpu))
                                ^
   mm/vmstat.c:1568:29: note: in expansion of macro 'per_cpu'
      cancel_delayed_work_sync(&per_cpu(vmstat_work, cpu));
                                ^

vim +/vmstat_wq +1436 mm/vmstat.c

0eb77e988 Christoph Lameter 2016-01-14  1430  	if (refresh_cpu_vm_stats(true)) {
7cc36bbdd Christoph Lameter 2014-10-09  1431  		/*
7cc36bbdd Christoph Lameter 2014-10-09  1432  		 * Counters were updated so we expect more updates
7cc36bbdd Christoph Lameter 2014-10-09  1433  		 * to occur in the future. Keep on running the
7cc36bbdd Christoph Lameter 2014-10-09  1434  		 * update worker thread.
7cc36bbdd Christoph Lameter 2014-10-09  1435  		 */
373ccbe59 Michal Hocko      2015-12-11 @1436  		queue_delayed_work_on(smp_processor_id(), vmstat_wq,
176bed1de Linus Torvalds    2015-10-15 @1437  				this_cpu_ptr(&vmstat_work),
98f4ebb29 Anton Blanchard   2009-04-02  1438  				round_jiffies_relative(sysctl_stat_interval));
f01f17d37 Michal Hocko      2016-02-05  1439  	}
7cc36bbdd Christoph Lameter 2014-10-09  1440  }
7cc36bbdd Christoph Lameter 2014-10-09  1441  
7cc36bbdd Christoph Lameter 2014-10-09  1442  /*
0eb77e988 Christoph Lameter 2016-01-14  1443   * Switch off vmstat processing and then fold all the remaining differentials
0eb77e988 Christoph Lameter 2016-01-14  1444   * until the diffs stay at zero. The function is used by NOHZ and can only be
0eb77e988 Christoph Lameter 2016-01-14  1445   * invoked when tick processing is not active.
0eb77e988 Christoph Lameter 2016-01-14  1446   */
0eb77e988 Christoph Lameter 2016-01-14  1447  /*
7cc36bbdd Christoph Lameter 2014-10-09  1448   * Check if the diffs for a certain cpu indicate that
7cc36bbdd Christoph Lameter 2014-10-09  1449   * an update is needed.
7cc36bbdd Christoph Lameter 2014-10-09  1450   */
7cc36bbdd Christoph Lameter 2014-10-09  1451  static bool need_update(int cpu)
7cc36bbdd Christoph Lameter 2014-10-09  1452  {
7cc36bbdd Christoph Lameter 2014-10-09  1453  	struct zone *zone;
7cc36bbdd Christoph Lameter 2014-10-09  1454  
7cc36bbdd Christoph Lameter 2014-10-09  1455  	for_each_populated_zone(zone) {
7cc36bbdd Christoph Lameter 2014-10-09  1456  		struct per_cpu_pageset *p = per_cpu_ptr(zone->pageset, cpu);
7cc36bbdd Christoph Lameter 2014-10-09  1457  
7cc36bbdd Christoph Lameter 2014-10-09  1458  		BUILD_BUG_ON(sizeof(p->vm_stat_diff[0]) != 1);
7cc36bbdd Christoph Lameter 2014-10-09  1459  		/*
7cc36bbdd Christoph Lameter 2014-10-09  1460  		 * The fast way of checking if there are any vmstat diffs.
7cc36bbdd Christoph Lameter 2014-10-09  1461  		 * This works because the diffs are byte sized items.
7cc36bbdd Christoph Lameter 2014-10-09  1462  		 */
7cc36bbdd Christoph Lameter 2014-10-09  1463  		if (memchr_inv(p->vm_stat_diff, 0, NR_VM_ZONE_STAT_ITEMS))
7cc36bbdd Christoph Lameter 2014-10-09  1464  			return true;
7cc36bbdd Christoph Lameter 2014-10-09  1465  
7cc36bbdd Christoph Lameter 2014-10-09  1466  	}
7cc36bbdd Christoph Lameter 2014-10-09  1467  	return false;
7cc36bbdd Christoph Lameter 2014-10-09  1468  }
7cc36bbdd Christoph Lameter 2014-10-09  1469  
043daba0d Christoph Lameter 2016-05-11  1470  /*
043daba0d Christoph Lameter 2016-05-11  1471   * Switch off vmstat processing and then fold all the remaining differentials
043daba0d Christoph Lameter 2016-05-11  1472   * until the diffs stay at zero. The function is used by NOHZ and can only be
043daba0d Christoph Lameter 2016-05-11  1473   * invoked when tick processing is not active.
043daba0d Christoph Lameter 2016-05-11  1474   */
f01f17d37 Michal Hocko      2016-02-05  1475  void quiet_vmstat(void)
f01f17d37 Michal Hocko      2016-02-05  1476  {
f01f17d37 Michal Hocko      2016-02-05  1477  	if (system_state != SYSTEM_RUNNING)
f01f17d37 Michal Hocko      2016-02-05  1478  		return;
f01f17d37 Michal Hocko      2016-02-05  1479  
043daba0d Christoph Lameter 2016-05-11 @1480  	if (!delayed_work_pending(this_cpu_ptr(&vmstat_work)))
f01f17d37 Michal Hocko      2016-02-05  1481  		return;
f01f17d37 Michal Hocko      2016-02-05  1482  
f01f17d37 Michal Hocko      2016-02-05  1483  	if (!need_update(smp_processor_id()))
f01f17d37 Michal Hocko      2016-02-05  1484  		return;
f01f17d37 Michal Hocko      2016-02-05  1485  
f01f17d37 Michal Hocko      2016-02-05  1486  	/*
f01f17d37 Michal Hocko      2016-02-05  1487  	 * Just refresh counters and do not care about the pending delayed
f01f17d37 Michal Hocko      2016-02-05  1488  	 * vmstat_update. It doesn't fire that often to matter and canceling
f01f17d37 Michal Hocko      2016-02-05  1489  	 * it would be too expensive from this path.
f01f17d37 Michal Hocko      2016-02-05  1490  	 * vmstat_shepherd will take care about that for us.
f01f17d37 Michal Hocko      2016-02-05  1491  	 */
f01f17d37 Michal Hocko      2016-02-05  1492  	refresh_cpu_vm_stats(false);
f01f17d37 Michal Hocko      2016-02-05  1493  }
f01f17d37 Michal Hocko      2016-02-05  1494  
7cc36bbdd Christoph Lameter 2014-10-09  1495  /*
7cc36bbdd Christoph Lameter 2014-10-09  1496   * Shepherd worker thread that checks the
7cc36bbdd Christoph Lameter 2014-10-09  1497   * differentials of processors that have their worker
7cc36bbdd Christoph Lameter 2014-10-09  1498   * threads for vm statistics updates disabled because of
7cc36bbdd Christoph Lameter 2014-10-09  1499   * inactivity.
7cc36bbdd Christoph Lameter 2014-10-09  1500   */
7cc36bbdd Christoph Lameter 2014-10-09  1501  static void vmstat_shepherd(struct work_struct *w);
7cc36bbdd Christoph Lameter 2014-10-09  1502  
0eb77e988 Christoph Lameter 2016-01-14  1503  static DECLARE_DEFERRABLE_WORK(shepherd, vmstat_shepherd);
7cc36bbdd Christoph Lameter 2014-10-09  1504  
7cc36bbdd Christoph Lameter 2014-10-09  1505  static void vmstat_shepherd(struct work_struct *w)
7cc36bbdd Christoph Lameter 2014-10-09  1506  {
7cc36bbdd Christoph Lameter 2014-10-09  1507  	int cpu;
7cc36bbdd Christoph Lameter 2014-10-09  1508  
7cc36bbdd Christoph Lameter 2014-10-09  1509  	get_online_cpus();
7cc36bbdd Christoph Lameter 2014-10-09  1510  	/* Check processors whose vmstat worker threads have been disabled */
043daba0d Christoph Lameter 2016-05-11  1511  	for_each_online_cpu(cpu) {
f01f17d37 Michal Hocko      2016-02-05 @1512  		struct delayed_work *dw = &per_cpu(vmstat_work, cpu);
7cc36bbdd Christoph Lameter 2014-10-09  1513  
043daba0d Christoph Lameter 2016-05-11  1514  		if (!delayed_work_pending(dw) && need_update(cpu))
f01f17d37 Michal Hocko      2016-02-05  1515  				queue_delayed_work_on(cpu, vmstat_wq, dw, 0);
f01f17d37 Michal Hocko      2016-02-05  1516  	}
7cc36bbdd Christoph Lameter 2014-10-09  1517  	put_online_cpus();
7cc36bbdd Christoph Lameter 2014-10-09  1518  
7cc36bbdd Christoph Lameter 2014-10-09  1519  	schedule_delayed_work(&shepherd,
7cc36bbdd Christoph Lameter 2014-10-09  1520  		round_jiffies_relative(sysctl_stat_interval));
d1187ed21 Christoph Lameter 2007-05-09  1521  }
d1187ed21 Christoph Lameter 2007-05-09  1522  
7cc36bbdd Christoph Lameter 2014-10-09  1523  static void __init start_shepherd_timer(void)
d1187ed21 Christoph Lameter 2007-05-09  1524  {
7cc36bbdd Christoph Lameter 2014-10-09  1525  	int cpu;
7cc36bbdd Christoph Lameter 2014-10-09  1526  
7cc36bbdd Christoph Lameter 2014-10-09  1527  	for_each_possible_cpu(cpu)
ccde8bd40 Michal Hocko      2016-02-05 @1528  		INIT_DEFERRABLE_WORK(per_cpu_ptr(&vmstat_work, cpu),
7cc36bbdd Christoph Lameter 2014-10-09  1529  			vmstat_update);
7cc36bbdd Christoph Lameter 2014-10-09  1530  
751e5f5c7 Michal Hocko      2016-01-08  1531  	vmstat_wq = alloc_workqueue("vmstat", WQ_FREEZABLE|WQ_MEM_RECLAIM, 0);

:::::: The code at line 1436 was first introduced by commit
:::::: 373ccbe5927034b55bdc80b0f8b54d6e13fe8d12 mm, vmstat: allow WQ concurrency to discover memory reclaim doesn't make any progress

:::::: TO: Michal Hocko <mhocko@suse.com>
:::::: CC: Linus Torvalds <torvalds@linux-foundation.org>

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/octet-stream, Size: 6049 bytes --]

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Michal Hocko]

On Mon 16-05-16 15:36:56, Andrew Morton wrote:
> On Mon, 16 May 2016 16:23:33 +0200 Michal Hocko <mhocko@kernel.org> wrote:
> 
> > Andrew, I think that the following is more straightforward fix and
> > should be folded in to the patch which has introduced vmstat_refresh.
> > ---
> > >From b8dd18fb7df040e1bfe61aadde1d903589de15e4 Mon Sep 17 00:00:00 2001
> > From: Michal Hocko <mhocko@suse.com>
> > Date: Mon, 16 May 2016 16:19:53 +0200
> > Subject: [PATCH] mmotm: mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix
> > 
> > Arnd has reported:
> > In randconfig builds with sysfs, procfs and numa all disabled,
> > but SMP enabled, we now get a link error in the newly introduced
> > vmstat_refresh function:
> > 
> > mm/built-in.o: In function `vmstat_refresh':
> > :(.text+0x15c78): undefined reference to `vmstat_text'
> > 
> > vmstat_refresh is proc_fs specific so there is no reason to define it
> > when !CONFIG_PROC_FS.
> 
> I already had this:
> 
> From: Christoph Lameter <cl@linux.com>
> Subject: Do not build vmstat_refresh if there is no procfs support
> 
> It makes no sense to build functionality into the kernel that
> cannot be used and causes build issues.
> 
> Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1605111011260.9351@east.gentwo.org
> Signed-off-by: Christoph Lameter <cl@linux.com>
> Reported-by: Arnd Bergmann <arnd@arndb.de>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

But this is broken:
http://lkml.kernel.org/r/20160516073144.GA23146@dhcp22.suse.cz and
kbuild robot agrees
http://lkml.kernel.org/r/201605171333.ANqJcwpy%fengguang.wu@intel.com
> ---
> 
>  mm/vmstat.c |    4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff -puN mm/vmstat.c~mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix mm/vmstat.c
> --- a/mm/vmstat.c~mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix
> +++ a/mm/vmstat.c
> @@ -1371,7 +1371,6 @@ static const struct file_operations proc
>  	.llseek		= seq_lseek,
>  	.release	= seq_release,
>  };
> -#endif /* CONFIG_PROC_FS */
>  
>  #ifdef CONFIG_SMP
>  static struct workqueue_struct *vmstat_wq;
> @@ -1436,7 +1435,10 @@ int vmstat_refresh(struct ctl_table *tab
>  		*lenp = 0;
>  	return 0;
>  }
> +#endif /* CONFIG_SMP */
> +#endif /* CONFIG_PROC_FS */
>  
> +#ifdef CONFIG_SMP
>  static void vmstat_update(struct work_struct *w)
>  {
>  	if (refresh_cpu_vm_stats(true)) {
> _

-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm: unhide vmstat_text definition for CONFIG_SMP

[Hugh Dickins]

On Tue, 17 May 2016, Michal Hocko wrote:
> On Mon 16-05-16 15:36:56, Andrew Morton wrote:
> > On Mon, 16 May 2016 16:23:33 +0200 Michal Hocko <mhocko@kernel.org> wrote:
> > 
> > > Andrew, I think that the following is more straightforward fix and
> > > should be folded in to the patch which has introduced vmstat_refresh.
> > > ---
> > > >From b8dd18fb7df040e1bfe61aadde1d903589de15e4 Mon Sep 17 00:00:00 2001
> > > From: Michal Hocko <mhocko@suse.com>
> > > Date: Mon, 16 May 2016 16:19:53 +0200
> > > Subject: [PATCH] mmotm: mm-proc-sys-vm-stat_refresh-to-force-vmstat-update-fix
> > > 
> > > Arnd has reported:
> > > In randconfig builds with sysfs, procfs and numa all disabled,
> > > but SMP enabled, we now get a link error in the newly introduced
> > > vmstat_refresh function:
> > > 
> > > mm/built-in.o: In function `vmstat_refresh':
> > > :(.text+0x15c78): undefined reference to `vmstat_text'
> > > 
> > > vmstat_refresh is proc_fs specific so there is no reason to define it
> > > when !CONFIG_PROC_FS.
> > 
> > I already had this:
> > 
> > From: Christoph Lameter <cl@linux.com>
> > Subject: Do not build vmstat_refresh if there is no procfs support
> > 
> > It makes no sense to build functionality into the kernel that
> > cannot be used and causes build issues.
> > 
> > Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1605111011260.9351@east.gentwo.org
> > Signed-off-by: Christoph Lameter <cl@linux.com>
> > Reported-by: Arnd Bergmann <arnd@arndb.de>
> > Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> 
> But this is broken:
> http://lkml.kernel.org/r/20160516073144.GA23146@dhcp22.suse.cz and
> kbuild robot agrees
> http://lkml.kernel.org/r/201605171333.ANqJcwpy%fengguang.wu@intel.com

Sorry for my noise, sorry for my silence, thanks to Arnd and everyone
for chipping in.  But now I try it, I find that even Michal's is not
quite right: if you build without CONFIG_PROC_FS, then it gives you
mm/vmstat.c:1381:13: warning: `refresh_vm_stats' defined but not used [-Wunused-function]
(well, that was on a tree with different line numbering).
So here's my attempt...

Signed-off-by: Hugh Dickins <hughd@google.com>
---
Fix to merge into mm-proc-sys-vm-stat_refresh-to-force-vmstat-update.patch

 mm/vmstat.c |    2 ++
 1 file changed, 2 insertions(+)

--- 4.6-rc7-mm1/mm/vmstat.c	2016-05-14 08:29:10.609386264 -0700
+++ linux/mm/vmstat.c	2016-05-17 17:43:02.861862648 -0700
@@ -1365,6 +1365,7 @@ static struct workqueue_struct *vmstat_w
 static DEFINE_PER_CPU(struct delayed_work, vmstat_work);
 int sysctl_stat_interval __read_mostly = HZ;
 
+#ifdef CONFIG_PROC_FS
 static void refresh_vm_stats(struct work_struct *work)
 {
 	refresh_cpu_vm_stats(true);
@@ -1422,6 +1423,7 @@ int vmstat_refresh(struct ctl_table *tab
 		*lenp = 0;
 	return 0;
 }
+#endif /* CONFIG_PROC_FS */
 
 static void vmstat_update(struct work_struct *w)
 {