vmstat: don't auto expand the sysfs files

Message ID	20231211154644.4103495-1-pasha.tatashin@soleen.com (mailing list archive)
State	New
Headers	show Return-Path: <owner-linux-mm@kvack.org> From: Pasha Tatashin <pasha.tatashin@soleen.com> To: gregkh@linuxfoundation.org, rafael@kernel.org, akpm@linux-foundation.org, surenb@google.com, pasha.tatashin@soleen.com, linux-kernel@vger.kernel.org, linux-mm@kvack.org, souravpanda@google.com Subject: [PATCH] vmstat: don't auto expand the sysfs files Date: Mon, 11 Dec 2023 15:46:44 +0000 Message-ID: <20231211154644.4103495-1-pasha.tatashin@soleen.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Sender: owner-linux-mm@kvack.org Precedence: bulk
Series	vmstat: don't auto expand the sysfs files \| expand vmstat: don't auto expand the sysfs files

Pasha Tatashin Dec. 11, 2023, 3:46 p.m. UTC

Whenever a new fields are added one of the following: node_stat_item
numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
files are auto expanded.

This is a problem, as sysfs files should be only one value per file.
Also, once a field is exported via vmstat it is hard to remove it as
there could be user applications that rely on this field. This is why
we still cary "nr_unstable 0" in /proc/vmstat that is not used.

Also, since vmstat is auto-expanded the fields are not documented, so
users do not know whether they are counted in bytes/kilobytes/pages,
and the exact meaning of these fields.

Modify the code that the new fields do not auto-expand the vmstat in
sysfs. The new fields can still be exported via their own files in
sysfs, and be properly documents.

vmstat values are named using vmstat_text[] array, which contains names
for zone_stat, numa_stat, node_stat, lru_list, writeback_stat,
vm_event.

Change vmstat_text[] to be an array of structs that contain two values:
name, and flags.  The new flags field contains information whether to
show stat value in vmstat files in sysfs (VMSTAT_SHOW_SYSFS), and in
procfs (VMSTAT_SHOW_PROCFS). The comment above VMSTAT_SHOW_SYSFS
documents that this flag should not be used for new stat values when
they are added.

Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
---
 drivers/base/node.c    |  34 ++--
 include/linux/vmstat.h |  48 ++++--
 mm/vmstat.c            | 377 +++++++++++++++++++++--------------------
 3 files changed, 244 insertions(+), 215 deletions(-)

Examples of the projects that are currently under review and that add new
fields to the one of the vmstat items:

[1] mm: report per-page metadata information
https://lore.kernel.org/all/20231205223118.3575485-1-souravpanda@google.com

[2] IOMMU memory observability
https://lore.kernel.org/all/20231130201504.2322355-1-pasha.tatashin@soleen.com

Greg KH has been requesting for these files not to grow:
https://lore.kernel.org/all/2023120731-deception-handmade-8d49@gregkh

Matthew Wilcox (Oracle) Dec. 11, 2023, 5 p.m. UTC | #1

On Mon, Dec 11, 2023 at 03:46:44PM +0000, Pasha Tatashin wrote:
> +++ b/drivers/base/node.c
> @@ -520,26 +520,34 @@ static ssize_t node_read_vmstat(struct device *dev,
>  	int i;
>  	int len = 0;
>  
> -	for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++)
> -		len += sysfs_emit_at(buf, len, "%s %lu\n",
> -				     zone_stat_name(i),
> -				     sum_zone_node_page_state(nid, i));
> +	for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) {
> +		if (vmstat_text[ZONE_STAT_NAME_IDX(i)].flags & VMSTAT_SHOW_SYSFS) {
> +			len += sysfs_emit_at(buf, len, "%s %lu\n",
> +					     zone_stat_name(i),
> +					     sum_zone_node_page_state(nid, i));
> +		}
> +	}

This seems overly complicated.  Why not do:

	for (i = 0; i < NR_VM_ZONE_STAT_SYSFS_ITEMS; i++)

and have assertions that this number doesn't change (and require people
to add to the list after that point)?

Pasha Tatashin Dec. 11, 2023, 5:09 p.m. UTC | #2

On Mon, Dec 11, 2023 at 12:00 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Mon, Dec 11, 2023 at 03:46:44PM +0000, Pasha Tatashin wrote:
> > +++ b/drivers/base/node.c
> > @@ -520,26 +520,34 @@ static ssize_t node_read_vmstat(struct device *dev,
> >       int i;
> >       int len = 0;
> >
> > -     for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++)
> > -             len += sysfs_emit_at(buf, len, "%s %lu\n",
> > -                                  zone_stat_name(i),
> > -                                  sum_zone_node_page_state(nid, i));
> > +     for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) {
> > +             if (vmstat_text[ZONE_STAT_NAME_IDX(i)].flags & VMSTAT_SHOW_SYSFS) {
> > +                     len += sysfs_emit_at(buf, len, "%s %lu\n",
> > +                                          zone_stat_name(i),
> > +                                          sum_zone_node_page_state(nid, i));
> > +             }
> > +     }
>
> This seems overly complicated.  Why not do:
>
>         for (i = 0; i < NR_VM_ZONE_STAT_SYSFS_ITEMS; i++)
>
> and have assertions that this number doesn't change (and require people
> to add to the list after that point)?

This is what I wanted to do at the beginning. But, the problem is that
vmstat_text[] contains names from a number of different structs of
stats:  zone_stat, numa_stat, node_stat, lru_list, writeback_stat,
vm_even. Therefore, we can't simply count the number of
NR_VM_ZONE_STAT_SYSFS_ITEMS, as the new items can be added in the
middle of vmstat_text[]  when for example numa_stat is expanded.

Pasha

David Rientjes Dec. 14, 2023, 5:52 p.m. UTC | #3

On Mon, 11 Dec 2023, Pasha Tatashin wrote:

> Whenever a new fields are added one of the following: node_stat_item
> numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
> files are auto expanded.
> 
> This is a problem, as sysfs files should be only one value per file.

Does this patch address the one-value-per-file issue?  (I think that ship 
has sailed for vmstat.)

/sys/devices/system/node/nodeX/vmstat has been documented as a stable ABI 
in Linux for 13 years.

That said, the contents of the file has not been documented so I assume 
it's "whatever stats make sense for the current implementation of the 
Linux VM".

> Also, once a field is exported via vmstat it is hard to remove it as
> there could be user applications that rely on this field. This is why
> we still cary "nr_unstable 0" in /proc/vmstat that is not used.
> 

Implementations change over time, so this would be expected.

I'm assuming, but perhaps incorrectly, that userspace won't crash if 
nr_unstable just don't appear anymore.  That whoever is using it would 
just assume that it's zero if it doesn't appear.

So I think we need to answer the question of: are the *contents* of files 
like vmstat that are heavily dependent on implementation level details 
really part of a stable ABI that people can expect will carry on forever?  

> Also, since vmstat is auto-expanded the fields are not documented, so
> users do not know whether they are counted in bytes/kilobytes/pages,
> and the exact meaning of these fields.
> 

I think that's actually intended since there can also be ones that are 
event counters.  I don't think any fields in vmstat are intended to be 
long-term sacred stable ABIs.

> Modify the code that the new fields do not auto-expand the vmstat in
> sysfs. The new fields can still be exported via their own files in
> sysfs, and be properly documents.
> 
> vmstat values are named using vmstat_text[] array, which contains names
> for zone_stat, numa_stat, node_stat, lru_list, writeback_stat,
> vm_event.
> 
> Change vmstat_text[] to be an array of structs that contain two values:
> name, and flags.  The new flags field contains information whether to
> show stat value in vmstat files in sysfs (VMSTAT_SHOW_SYSFS), and in
> procfs (VMSTAT_SHOW_PROCFS). The comment above VMSTAT_SHOW_SYSFS
> documents that this flag should not be used for new stat values when
> they are added.
> 
> Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com>
> ---
>  drivers/base/node.c    |  34 ++--
>  include/linux/vmstat.h |  48 ++++--
>  mm/vmstat.c            | 377 +++++++++++++++++++++--------------------
>  3 files changed, 244 insertions(+), 215 deletions(-)
> 
> Examples of the projects that are currently under review and that add new
> fields to the one of the vmstat items:
> 
> [1] mm: report per-page metadata information
> https://lore.kernel.org/all/20231205223118.3575485-1-souravpanda@google.com
> 
> [2] IOMMU memory observability
> https://lore.kernel.org/all/20231130201504.2322355-1-pasha.tatashin@soleen.com
> 
> Greg KH has been requesting for these files not to grow:
> https://lore.kernel.org/all/2023120731-deception-handmade-8d49@gregkh
> 
> diff --git a/drivers/base/node.c b/drivers/base/node.c
> index 493d533f8375..f139d7ab58f5 100644
> --- a/drivers/base/node.c
> +++ b/drivers/base/node.c
> @@ -520,26 +520,34 @@ static ssize_t node_read_vmstat(struct device *dev,
>  	int i;
>  	int len = 0;
>  
> -	for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++)
> -		len += sysfs_emit_at(buf, len, "%s %lu\n",
> -				     zone_stat_name(i),
> -				     sum_zone_node_page_state(nid, i));
> +	for (i = 0; i < NR_VM_ZONE_STAT_ITEMS; i++) {
> +		if (vmstat_text[ZONE_STAT_NAME_IDX(i)].flags & VMSTAT_SHOW_SYSFS) {
> +			len += sysfs_emit_at(buf, len, "%s %lu\n",
> +					     zone_stat_name(i),
> +					     sum_zone_node_page_state(nid, i));
> +		}
> +	}
>  
>  #ifdef CONFIG_NUMA
>  	fold_vm_numa_events();
> -	for (i = 0; i < NR_VM_NUMA_EVENT_ITEMS; i++)
> -		len += sysfs_emit_at(buf, len, "%s %lu\n",
> -				     numa_stat_name(i),
> -				     sum_zone_numa_event_state(nid, i));
> +	for (i = 0; i < NR_VM_NUMA_EVENT_ITEMS; i++) {
> +		if (vmstat_text[NUMA_STAT_NAME_IDX(i)].flags & VMSTAT_SHOW_SYSFS) {
> +			len += sysfs_emit_at(buf, len, "%s %lu\n",
> +					     numa_stat_name(i),
> +					     sum_zone_numa_event_state(nid, i));
> +		}
> +	}
>  
>  #endif
>  	for (i = 0; i < NR_VM_NODE_STAT_ITEMS; i++) {
> -		unsigned long pages = node_page_state_pages(pgdat, i);
> +		if (vmstat_text[NODE_STAT_NAME_IDX(i)].flags & VMSTAT_SHOW_SYSFS) {
> +			unsigned long pages = node_page_state_pages(pgdat, i);
>  
> -		if (vmstat_item_print_in_thp(i))
> -			pages /= HPAGE_PMD_NR;
> -		len += sysfs_emit_at(buf, len, "%s %lu\n", node_stat_name(i),
> -				     pages);
> +			if (vmstat_item_print_in_thp(i))
> +				pages /= HPAGE_PMD_NR;
> +			len += sysfs_emit_at(buf, len, "%s %lu\n", node_stat_name(i),
> +					     pages);
> +		}
>  	}
>  
>  	return len;
> diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h
> index fed855bae6d8..2dd46daf69aa 100644
> --- a/include/linux/vmstat.h
> +++ b/include/linux/vmstat.h
> @@ -495,26 +495,44 @@ static inline void __mod_zone_freepage_state(struct zone *zone, int nr_pages,
>  		__mod_zone_page_state(zone, NR_FREE_CMA_PAGES, nr_pages);
>  }
>  
> -extern const char * const vmstat_text[];
>  
> +/*
> + * Show this stat in /sys/devices/system/node/nodeX/vmstat
> + * IMPORTANT: Don't use this flag for new stats, as the right way to output only
> + * one stat per file in sysfs. Instead, add new individual sysfs files for new
> + * stats, and document them in Documentation/ABI/TYPE/sysfs-new_field_name.
> + */
> +#define VMSTAT_SHOW_SYSFS	BIT(0)
> +
> +/* Show this stat in /proc/vmstat */
> +#define VMSTAT_SHOW_PROCFS	BIT(1)
> +
> +struct vmstat_text {
> +	const char *name;
> +	char flags;
> +};
> +
> +extern const struct vmstat_text vmstat_text[];
> +
> +#define ZONE_STAT_NAME_IDX(item)	((int)(item))
>  static inline const char *zone_stat_name(enum zone_stat_item item)
>  {
> -	return vmstat_text[item];
> +	return vmstat_text[ZONE_STAT_NAME_IDX(item)].name;
>  }
>  
>  #ifdef CONFIG_NUMA
> +#define NUMA_STAT_NAME_IDX(item)	(NR_VM_ZONE_STAT_ITEMS + (int)(item))
>  static inline const char *numa_stat_name(enum numa_stat_item item)
>  {
> -	return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
> -			   item];
> +	return vmstat_text[NUMA_STAT_NAME_IDX(item)].name;
>  }
>  #endif /* CONFIG_NUMA */
>  
> +#define NODE_STAT_NAME_IDX(item)	(NR_VM_NUMA_EVENT_ITEMS +	\
> +	NR_VM_ZONE_STAT_ITEMS + (int)(item))
>  static inline const char *node_stat_name(enum node_stat_item item)
>  {
> -	return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
> -			   NR_VM_NUMA_EVENT_ITEMS +
> -			   item];
> +	return vmstat_text[NODE_STAT_NAME_IDX(item)].name;
>  }
>  
>  static inline const char *lru_list_name(enum lru_list lru)
> @@ -522,22 +540,20 @@ static inline const char *lru_list_name(enum lru_list lru)
>  	return node_stat_name(NR_LRU_BASE + lru) + 3; // skip "nr_"
>  }
>  
> +#define WRITEBACK_STAT_NAME_IDX(item)	(NR_VM_NODE_STAT_ITEMS +	\
> +	NR_VM_NUMA_EVENT_ITEMS + NR_VM_ZONE_STAT_ITEMS + (int)(item))
>  static inline const char *writeback_stat_name(enum writeback_stat_item item)
>  {
> -	return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
> -			   NR_VM_NUMA_EVENT_ITEMS +
> -			   NR_VM_NODE_STAT_ITEMS +
> -			   item];
> +	return vmstat_text[WRITEBACK_STAT_NAME_IDX(item)].name;
>  }
>  
>  #if defined(CONFIG_VM_EVENT_COUNTERS) || defined(CONFIG_MEMCG)
> +#define VM_EVENT_NAME_IDX(item)	(NR_VM_WRITEBACK_STAT_ITEMS +		\
> +	NR_VM_NODE_STAT_ITEMS + NR_VM_NUMA_EVENT_ITEMS +		\
> +	NR_VM_ZONE_STAT_ITEMS + (int)(item))
>  static inline const char *vm_event_name(enum vm_event_item item)
>  {
> -	return vmstat_text[NR_VM_ZONE_STAT_ITEMS +
> -			   NR_VM_NUMA_EVENT_ITEMS +
> -			   NR_VM_NODE_STAT_ITEMS +
> -			   NR_VM_WRITEBACK_STAT_ITEMS +
> -			   item];
> +	return vmstat_text[VM_EVENT_NAME_IDX(item)].name;
>  }
>  #endif /* CONFIG_VM_EVENT_COUNTERS || CONFIG_MEMCG */
>  
> diff --git a/mm/vmstat.c b/mm/vmstat.c
> index 359460deb377..691d8c90b4ac 100644
> --- a/mm/vmstat.c
> +++ b/mm/vmstat.c
> @@ -1142,278 +1142,281 @@ int fragmentation_index(struct zone *zone, unsigned int order)
>  #if defined(CONFIG_PROC_FS) || defined(CONFIG_SYSFS) || \
>      defined(CONFIG_NUMA) || defined(CONFIG_MEMCG)
>  #ifdef CONFIG_ZONE_DMA
> -#define TEXT_FOR_DMA(xx) xx "_dma",
> +#define TEXT_FOR_DMA(xx) {xx "_dma",	VMSTAT_SHOW_PROCFS},
>  #else
>  #define TEXT_FOR_DMA(xx)
>  #endif
>  
>  #ifdef CONFIG_ZONE_DMA32
> -#define TEXT_FOR_DMA32(xx) xx "_dma32",
> +#define TEXT_FOR_DMA32(xx) {xx "_dma32",	VMSTAT_SHOW_PROCFS},
>  #else
>  #define TEXT_FOR_DMA32(xx)
>  #endif
>  
>  #ifdef CONFIG_HIGHMEM
> -#define TEXT_FOR_HIGHMEM(xx) xx "_high",
> +#define TEXT_FOR_HIGHMEM(xx) {xx "_high",	VMSTAT_SHOW_PROCFS},
>  #else
>  #define TEXT_FOR_HIGHMEM(xx)
>  #endif
>  
>  #ifdef CONFIG_ZONE_DEVICE
> -#define TEXT_FOR_DEVICE(xx) xx "_device",
> +#define TEXT_FOR_DEVICE(xx) {xx "_device",	VMSTAT_SHOW_PROCFS},
>  #else
>  #define TEXT_FOR_DEVICE(xx)
>  #endif
>  
> -#define TEXTS_FOR_ZONES(xx) TEXT_FOR_DMA(xx) TEXT_FOR_DMA32(xx) xx "_normal", \
> -					TEXT_FOR_HIGHMEM(xx) xx "_movable", \
> -					TEXT_FOR_DEVICE(xx)
> +#define TEXT_FOR_NORMAL(xx) {xx "_normal", VMSTAT_SHOW_PROCFS},
> +#define TEXT_FOR_MOVABLE(xx) {xx "_movable", VMSTAT_SHOW_PROCFS},
>  
> -const char * const vmstat_text[] = {
> +#define TEXTS_FOR_ZONES(xx) TEXT_FOR_DMA(xx) TEXT_FOR_DMA32(xx)		\
> +	TEXT_FOR_NORMAL(xx) TEXT_FOR_HIGHMEM(xx) TEXT_FOR_MOVABLE(xx)	\
> +	TEXT_FOR_DEVICE(xx)
> +
> +const struct vmstat_text vmstat_text[] = {
>  	/* enum zone_stat_item counters */
> -	"nr_free_pages",
> -	"nr_zone_inactive_anon",
> -	"nr_zone_active_anon",
> -	"nr_zone_inactive_file",
> -	"nr_zone_active_file",
> -	"nr_zone_unevictable",
> -	"nr_zone_write_pending",
> -	"nr_mlock",
> -	"nr_bounce",
> +	{"nr_free_pages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_inactive_anon",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_active_anon",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_inactive_file",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_active_file",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_unevictable",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_zone_write_pending",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_mlock",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_bounce",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #if IS_ENABLED(CONFIG_ZSMALLOC)
> -	"nr_zspages",
> +	{"nr_zspages",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
> -	"nr_free_cma",
> +	{"nr_free_cma",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #ifdef CONFIG_UNACCEPTED_MEMORY
> -	"nr_unaccepted",
> +	{"nr_unaccepted",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
>  
>  	/* enum numa_stat_item counters */
>  #ifdef CONFIG_NUMA
> -	"numa_hit",
> -	"numa_miss",
> -	"numa_foreign",
> -	"numa_interleave",
> -	"numa_local",
> -	"numa_other",
> +	{"numa_hit",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"numa_miss",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"numa_foreign",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"numa_interleave",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"numa_local",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"numa_other",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
> -
>  	/* enum node_stat_item counters */
> -	"nr_inactive_anon",
> -	"nr_active_anon",
> -	"nr_inactive_file",
> -	"nr_active_file",
> -	"nr_unevictable",
> -	"nr_slab_reclaimable",
> -	"nr_slab_unreclaimable",
> -	"nr_isolated_anon",
> -	"nr_isolated_file",
> -	"workingset_nodes",
> -	"workingset_refault_anon",
> -	"workingset_refault_file",
> -	"workingset_activate_anon",
> -	"workingset_activate_file",
> -	"workingset_restore_anon",
> -	"workingset_restore_file",
> -	"workingset_nodereclaim",
> -	"nr_anon_pages",
> -	"nr_mapped",
> -	"nr_file_pages",
> -	"nr_dirty",
> -	"nr_writeback",
> -	"nr_writeback_temp",
> -	"nr_shmem",
> -	"nr_shmem_hugepages",
> -	"nr_shmem_pmdmapped",
> -	"nr_file_hugepages",
> -	"nr_file_pmdmapped",
> -	"nr_anon_transparent_hugepages",
> -	"nr_vmscan_write",
> -	"nr_vmscan_immediate_reclaim",
> -	"nr_dirtied",
> -	"nr_written",
> -	"nr_throttled_written",
> -	"nr_kernel_misc_reclaimable",
> -	"nr_foll_pin_acquired",
> -	"nr_foll_pin_released",
> -	"nr_kernel_stack",
> +	{"nr_inactive_anon",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_active_anon",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_inactive_file",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_active_file",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_unevictable",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_slab_reclaimable",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_slab_unreclaimable",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_isolated_anon",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_isolated_file",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_nodes",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_refault_anon",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_refault_file",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_activate_anon",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_activate_file",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_restore_anon",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_restore_file",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"workingset_nodereclaim",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_anon_pages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_mapped",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_file_pages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_dirty",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_writeback",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_writeback_temp",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_shmem",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_shmem_hugepages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_shmem_pmdmapped",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_file_hugepages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_file_pmdmapped",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_anon_transparent_hugepages", VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_vmscan_write",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_vmscan_immediate_reclaim",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_dirtied",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_written",			VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_throttled_written",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_kernel_misc_reclaimable",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_foll_pin_acquired",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_foll_pin_released",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_kernel_stack",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #if IS_ENABLED(CONFIG_SHADOW_CALL_STACK)
> -	"nr_shadow_call_stack",
> +	{"nr_shadow_call_stack",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
> -	"nr_page_table_pages",
> -	"nr_sec_page_table_pages",
> +	{"nr_page_table_pages",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"nr_sec_page_table_pages",	VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #ifdef CONFIG_SWAP
> -	"nr_swapcached",
> +	{"nr_swapcached",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
>  #ifdef CONFIG_NUMA_BALANCING
> -	"pgpromote_success",
> -	"pgpromote_candidate",
> +	{"pgpromote_success",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
> +	{"pgpromote_candidate",		VMSTAT_SHOW_PROCFS | VMSTAT_SHOW_SYSFS},
>  #endif
>  
>  	/* enum writeback_stat_item counters */
> -	"nr_dirty_threshold",
> -	"nr_dirty_background_threshold",
> +	{"nr_dirty_threshold",		VMSTAT_SHOW_PROCFS},
> +	{"nr_dirty_background_threshold", VMSTAT_SHOW_PROCFS},
>  
>  #if defined(CONFIG_VM_EVENT_COUNTERS) || defined(CONFIG_MEMCG)
>  	/* enum vm_event_item counters */
> -	"pgpgin",
> -	"pgpgout",
> -	"pswpin",
> -	"pswpout",
> +	{"pgpgin",			VMSTAT_SHOW_PROCFS},
> +	{"pgpgout",			VMSTAT_SHOW_PROCFS},
> +	{"pswpin",			VMSTAT_SHOW_PROCFS},
> +	{"pswpout",			VMSTAT_SHOW_PROCFS},
>  
>  	TEXTS_FOR_ZONES("pgalloc")
>  	TEXTS_FOR_ZONES("allocstall")
>  	TEXTS_FOR_ZONES("pgskip")
>  
> -	"pgfree",
> -	"pgactivate",
> -	"pgdeactivate",
> -	"pglazyfree",
> -
> -	"pgfault",
> -	"pgmajfault",
> -	"pglazyfreed",
> -
> -	"pgrefill",
> -	"pgreuse",
> -	"pgsteal_kswapd",
> -	"pgsteal_direct",
> -	"pgsteal_khugepaged",
> -	"pgdemote_kswapd",
> -	"pgdemote_direct",
> -	"pgdemote_khugepaged",
> -	"pgscan_kswapd",
> -	"pgscan_direct",
> -	"pgscan_khugepaged",
> -	"pgscan_direct_throttle",
> -	"pgscan_anon",
> -	"pgscan_file",
> -	"pgsteal_anon",
> -	"pgsteal_file",
> +	{"pgfree",			VMSTAT_SHOW_PROCFS},
> +	{"pgactivate",			VMSTAT_SHOW_PROCFS},
> +	{"pgdeactivate",		VMSTAT_SHOW_PROCFS},
> +	{"pglazyfree",			VMSTAT_SHOW_PROCFS},
> +
> +	{"pgfault",			VMSTAT_SHOW_PROCFS},
> +	{"pgmajfault",			VMSTAT_SHOW_PROCFS},
> +	{"pglazyfreed",			VMSTAT_SHOW_PROCFS},
> +
> +	{"pgrefill",			VMSTAT_SHOW_PROCFS},
> +	{"pgreuse",			VMSTAT_SHOW_PROCFS},
> +	{"pgsteal_kswapd",		VMSTAT_SHOW_PROCFS},
> +	{"pgsteal_direct",		VMSTAT_SHOW_PROCFS},
> +	{"pgsteal_khugepaged",		VMSTAT_SHOW_PROCFS},
> +	{"pgdemote_kswapd",		VMSTAT_SHOW_PROCFS},
> +	{"pgdemote_direct",		VMSTAT_SHOW_PROCFS},
> +	{"pgdemote_khugepaged",		VMSTAT_SHOW_PROCFS},
> +	{"pgscan_kswapd",		VMSTAT_SHOW_PROCFS},
> +	{"pgscan_direct",		VMSTAT_SHOW_PROCFS},
> +	{"pgscan_khugepaged",		VMSTAT_SHOW_PROCFS},
> +	{"pgscan_direct_throttle",	VMSTAT_SHOW_PROCFS},
> +	{"pgscan_anon",			VMSTAT_SHOW_PROCFS},
> +	{"pgscan_file",			VMSTAT_SHOW_PROCFS},
> +	{"pgsteal_anon",		VMSTAT_SHOW_PROCFS},
> +	{"pgsteal_file",		VMSTAT_SHOW_PROCFS},
>  
>  #ifdef CONFIG_NUMA
> -	"zone_reclaim_failed",
> +	{"zone_reclaim_failed",		VMSTAT_SHOW_PROCFS},
>  #endif
> -	"pginodesteal",
> -	"slabs_scanned",
> -	"kswapd_inodesteal",
> -	"kswapd_low_wmark_hit_quickly",
> -	"kswapd_high_wmark_hit_quickly",
> -	"pageoutrun",
> +	{"pginodesteal",		VMSTAT_SHOW_PROCFS},
> +	{"slabs_scanned",		VMSTAT_SHOW_PROCFS},
> +	{"kswapd_inodesteal",		VMSTAT_SHOW_PROCFS},
> +	{"kswapd_low_wmark_hit_quickly", VMSTAT_SHOW_PROCFS},
> +	{"kswapd_high_wmark_hit_quickly", VMSTAT_SHOW_PROCFS},
> +	{"pageoutrun",			VMSTAT_SHOW_PROCFS},
>  
> -	"pgrotated",
> +	{"pgrotated",			VMSTAT_SHOW_PROCFS},
>  
> -	"drop_pagecache",
> -	"drop_slab",
> -	"oom_kill",
> +	{"drop_pagecache",		VMSTAT_SHOW_PROCFS},
> +	{"drop_slab",			VMSTAT_SHOW_PROCFS},
> +	{"oom_kill",			VMSTAT_SHOW_PROCFS},
>  
>  #ifdef CONFIG_NUMA_BALANCING
> -	"numa_pte_updates",
> -	"numa_huge_pte_updates",
> -	"numa_hint_faults",
> -	"numa_hint_faults_local",
> -	"numa_pages_migrated",
> +	{"numa_pte_updates",		VMSTAT_SHOW_PROCFS},
> +	{"numa_huge_pte_updates",	VMSTAT_SHOW_PROCFS},
> +	{"numa_hint_faults",		VMSTAT_SHOW_PROCFS},
> +	{"numa_hint_faults_local",	VMSTAT_SHOW_PROCFS},
> +	{"numa_pages_migrated",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_MIGRATION
> -	"pgmigrate_success",
> -	"pgmigrate_fail",
> -	"thp_migration_success",
> -	"thp_migration_fail",
> -	"thp_migration_split",
> +	{"pgmigrate_success",		VMSTAT_SHOW_PROCFS},
> +	{"pgmigrate_fail",		VMSTAT_SHOW_PROCFS},
> +	{"thp_migration_success",	VMSTAT_SHOW_PROCFS},
> +	{"thp_migration_fail",		VMSTAT_SHOW_PROCFS},
> +	{"thp_migration_split",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_COMPACTION
> -	"compact_migrate_scanned",
> -	"compact_free_scanned",
> -	"compact_isolated",
> -	"compact_stall",
> -	"compact_fail",
> -	"compact_success",
> -	"compact_daemon_wake",
> -	"compact_daemon_migrate_scanned",
> -	"compact_daemon_free_scanned",
> +	{"compact_migrate_scanned",	VMSTAT_SHOW_PROCFS},
> +	{"compact_free_scanned",	VMSTAT_SHOW_PROCFS},
> +	{"compact_isolated",		VMSTAT_SHOW_PROCFS},
> +	{"compact_stall",		VMSTAT_SHOW_PROCFS},
> +	{"compact_fail",		VMSTAT_SHOW_PROCFS},
> +	{"compact_success",		VMSTAT_SHOW_PROCFS},
> +	{"compact_daemon_wake",		VMSTAT_SHOW_PROCFS},
> +	{"compact_daemon_migrate_scanned", VMSTAT_SHOW_PROCFS},
> +	{"compact_daemon_free_scanned",	VMSTAT_SHOW_PROCFS},
>  #endif
>  
>  #ifdef CONFIG_HUGETLB_PAGE
> -	"htlb_buddy_alloc_success",
> -	"htlb_buddy_alloc_fail",
> +	{"htlb_buddy_alloc_success",	VMSTAT_SHOW_PROCFS},
> +	{"htlb_buddy_alloc_fail",	VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_CMA
> -	"cma_alloc_success",
> -	"cma_alloc_fail",
> +	{"cma_alloc_success",		VMSTAT_SHOW_PROCFS},
> +	{"cma_alloc_fail",		VMSTAT_SHOW_PROCFS},
>  #endif
> -	"unevictable_pgs_culled",
> -	"unevictable_pgs_scanned",
> -	"unevictable_pgs_rescued",
> -	"unevictable_pgs_mlocked",
> -	"unevictable_pgs_munlocked",
> -	"unevictable_pgs_cleared",
> -	"unevictable_pgs_stranded",
> +	{"unevictable_pgs_culled",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_scanned",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_rescued",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_mlocked",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_munlocked",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_cleared",	VMSTAT_SHOW_PROCFS},
> +	{"unevictable_pgs_stranded",	VMSTAT_SHOW_PROCFS},
>  
>  #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> -	"thp_fault_alloc",
> -	"thp_fault_fallback",
> -	"thp_fault_fallback_charge",
> -	"thp_collapse_alloc",
> -	"thp_collapse_alloc_failed",
> -	"thp_file_alloc",
> -	"thp_file_fallback",
> -	"thp_file_fallback_charge",
> -	"thp_file_mapped",
> -	"thp_split_page",
> -	"thp_split_page_failed",
> -	"thp_deferred_split_page",
> -	"thp_split_pmd",
> -	"thp_scan_exceed_none_pte",
> -	"thp_scan_exceed_swap_pte",
> -	"thp_scan_exceed_share_pte",
> +	{"thp_fault_alloc",		VMSTAT_SHOW_PROCFS},
> +	{"thp_fault_fallback",		VMSTAT_SHOW_PROCFS},
> +	{"thp_fault_fallback_charge",	VMSTAT_SHOW_PROCFS},
> +	{"thp_collapse_alloc",		VMSTAT_SHOW_PROCFS},
> +	{"thp_collapse_alloc_failed",	VMSTAT_SHOW_PROCFS},
> +	{"thp_file_alloc",		VMSTAT_SHOW_PROCFS},
> +	{"thp_file_fallback",		VMSTAT_SHOW_PROCFS},
> +	{"thp_file_fallback_charge",	VMSTAT_SHOW_PROCFS},
> +	{"thp_file_mapped",		VMSTAT_SHOW_PROCFS},
> +	{"thp_split_page",		VMSTAT_SHOW_PROCFS},
> +	{"thp_split_page_failed",	VMSTAT_SHOW_PROCFS},
> +	{"thp_deferred_split_page",	VMSTAT_SHOW_PROCFS},
> +	{"thp_split_pmd",		VMSTAT_SHOW_PROCFS},
> +	{"thp_scan_exceed_none_pte",	VMSTAT_SHOW_PROCFS},
> +	{"thp_scan_exceed_swap_pte",	VMSTAT_SHOW_PROCFS},
> +	{"thp_scan_exceed_share_pte",	VMSTAT_SHOW_PROCFS},
>  #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> -	"thp_split_pud",
> +	{"thp_split_pud",		VMSTAT_SHOW_PROCFS},
>  #endif
> -	"thp_zero_page_alloc",
> -	"thp_zero_page_alloc_failed",
> -	"thp_swpout",
> -	"thp_swpout_fallback",
> +	{"thp_zero_page_alloc",		VMSTAT_SHOW_PROCFS},
> +	{"thp_zero_page_alloc_failed",	VMSTAT_SHOW_PROCFS},
> +	{"thp_swpout",			VMSTAT_SHOW_PROCFS},
> +	{"thp_swpout_fallback",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_MEMORY_BALLOON
> -	"balloon_inflate",
> -	"balloon_deflate",
> +	{"balloon_inflate",		VMSTAT_SHOW_PROCFS},
> +	{"balloon_deflate",		VMSTAT_SHOW_PROCFS},
>  #ifdef CONFIG_BALLOON_COMPACTION
> -	"balloon_migrate",
> +	{"balloon_migrate",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #endif /* CONFIG_MEMORY_BALLOON */
>  #ifdef CONFIG_DEBUG_TLBFLUSH
> -	"nr_tlb_remote_flush",
> -	"nr_tlb_remote_flush_received",
> -	"nr_tlb_local_flush_all",
> -	"nr_tlb_local_flush_one",
> +	{"nr_tlb_remote_flush",		VMSTAT_SHOW_PROCFS},
> +	{"nr_tlb_remote_flush_received", VMSTAT_SHOW_PROCFS},
> +	{"nr_tlb_local_flush_all",	VMSTAT_SHOW_PROCFS},
> +	{"nr_tlb_local_flush_one",	VMSTAT_SHOW_PROCFS},
>  #endif /* CONFIG_DEBUG_TLBFLUSH */
>  
>  #ifdef CONFIG_SWAP
> -	"swap_ra",
> -	"swap_ra_hit",
> +	{"swap_ra",			VMSTAT_SHOW_PROCFS},
> +	{"swap_ra_hit",			VMSTAT_SHOW_PROCFS},
>  #ifdef CONFIG_KSM
> -	"ksm_swpin_copy",
> +	{"ksm_swpin_copy",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #endif
>  #ifdef CONFIG_KSM
> -	"cow_ksm",
> +	{"cow_ksm",			VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_ZSWAP
> -	"zswpin",
> -	"zswpout",
> +	{"zswpin",			VMSTAT_SHOW_PROCFS},
> +	{"zswpout",			VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_X86
> -	"direct_map_level2_splits",
> -	"direct_map_level3_splits",
> +	{"direct_map_level2_splits",	VMSTAT_SHOW_PROCFS},
> +	{"direct_map_level3_splits",	VMSTAT_SHOW_PROCFS},
>  #endif
>  #ifdef CONFIG_PER_VMA_LOCK_STATS
> -	"vma_lock_success",
> -	"vma_lock_abort",
> -	"vma_lock_retry",
> -	"vma_lock_miss",
> +	{"vma_lock_success",		VMSTAT_SHOW_PROCFS},
> +	{"vma_lock_abort",		VMSTAT_SHOW_PROCFS},
> +	{"vma_lock_retry",		VMSTAT_SHOW_PROCFS},
> +	{"vma_lock_miss",		VMSTAT_SHOW_PROCFS},
>  #endif
>  #endif /* CONFIG_VM_EVENT_COUNTERS || CONFIG_MEMCG */
>  };
> +
>  #endif /* CONFIG_PROC_FS || CONFIG_SYSFS || CONFIG_NUMA || CONFIG_MEMCG */
>  
>  #if (defined(CONFIG_DEBUG_FS) && defined(CONFIG_COMPACTION)) || \
> @@ -1845,9 +1848,11 @@ static int vmstat_show(struct seq_file *m, void *arg)
>  	unsigned long *l = arg;
>  	unsigned long off = l - (unsigned long *)m->private;
>  
> -	seq_puts(m, vmstat_text[off]);
> -	seq_put_decimal_ull(m, " ", *l);
> -	seq_putc(m, '\n');
> +	if (vmstat_text[off].flags & VMSTAT_SHOW_PROCFS) {
> +		seq_puts(m, vmstat_text[off].name);
> +		seq_put_decimal_ull(m, " ", *l);
> +		seq_putc(m, '\n');
> +	}
>  
>  	if (off == NR_VMSTAT_ITEMS - 1) {
>  		/*
> -- 
> 2.43.0.472.g3155946c3a-goog
> 
> 
>

Pasha Tatashin Dec. 14, 2023, 6:57 p.m. UTC | #4

Hi David,

Thank you for taking a look at this patch, my replies below.

On Thu, Dec 14, 2023 at 12:52 PM David Rientjes <rientjes@google.com> wrote:
>
> On Mon, 11 Dec 2023, Pasha Tatashin wrote:
>
> > Whenever a new fields are added one of the following: node_stat_item
> > numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
> > files are auto expanded.
> >
> > This is a problem, as sysfs files should be only one value per file.
>
> Does this patch address the one-value-per-file issue?  (I think that ship
> has sailed for vmstat.)

That ship has sailed for vmstat, this patch addresses what was asked
by GregKH: not to add new values to vmstat, as not to make the
existing problem even worse. The sysfs file system has a one page
limit per file. The developers will decide how to export the new items
added to node_stat, numa_stat, zone_stat individually. Each new item
can be exported in its own files, and must have its own documentation
about interface stability, value meaning, and expectations when the
stat file is absent.

> /sys/devices/system/node/nodeX/vmstat has been documented as a stable ABI
> in Linux for 13 years.
>
> That said, the contents of the file has not been documented so I assume
> it's "whatever stats make sense for the current implementation of the
> Linux VM".
>
> > Also, once a field is exported via vmstat it is hard to remove it as
> > there could be user applications that rely on this field. This is why
> > we still cary "nr_unstable 0" in /proc/vmstat that is not used.
> >
>
> Implementations change over time, so this would be expected.
>
> I'm assuming, but perhaps incorrectly, that userspace won't crash if
> nr_unstable just don't appear anymore.  That whoever is using it would
> just assume that it's zero if it doesn't appear.
>
> So I think we need to answer the question of: are the *contents* of files
> like vmstat that are heavily dependent on implementation level details
> really part of a stable ABI that people can expect will carry on forever?

I agree, but that is outside of the scope of this patch. The intent of
this patch is to keep the existing interfaces, and only prevents
future auto expansion of vmstat files. In the future, we work on
documenting the existing vmstat interfaces, and perhaps cleaning-up
them when possible.

> > Also, since vmstat is auto-expanded the fields are not documented, so
> > users do not know whether they are counted in bytes/kilobytes/pages,
> > and the exact meaning of these fields.
> >
>
> I think that's actually intended since there can also be ones that are
> event counters.  I don't think any fields in vmstat are intended to be
> long-term sacred stable ABIs.

Right, but we already carry fields i.e nr_unstable that are hardcoded,
but were removed from the kernel otherwise.

Thank you,
Pasha

David Rientjes Dec. 24, 2023, 9:26 p.m. UTC | #5

On Thu, 14 Dec 2023, Pasha Tatashin wrote:

> > > Whenever a new fields are added one of the following: node_stat_item
> > > numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
> > > files are auto expanded.
> > >
> > > This is a problem, as sysfs files should be only one value per file.
> >
> > Does this patch address the one-value-per-file issue?  (I think that ship
> > has sailed for vmstat.)
> 
> That ship has sailed for vmstat, this patch addresses what was asked
> by GregKH: not to add new values to vmstat, as not to make the
> existing problem even worse. The sysfs file system has a one page
> limit per file. The developers will decide how to export the new items
> added to node_stat, numa_stat, zone_stat individually. Each new item
> can be exported in its own files, and must have its own documentation
> about interface stability, value meaning, and expectations when the
> stat file is absent.
> 

As of at least 6.5, /proc/vmstat is a strict superset of the per-node 
vmstat.  Why is that a problem?

There's great benefit to being able to use the sample implementations to 
parse either /proc/vmstat *or* the per-node vmstat and without needing to 
read the per-node vmstat plus some new set of sysfs files that are 
one-value-per-file.  The per-node vmstat will always be multiple values, 
in fact it's a key value pair.

I have to think that doing anything else for vmstat is just adding 
complexity (like this patch) and actually making it *harder* on userspace 
to read the data it needs.

Yes, the per-node vmstat likely shouldn't be in sysfs at all but it 
appears to have been added there 13+ years ago because it was a convenient 
place to add a per-node variant.  That's not ideal, but owell.

Pasha Tatashin Dec. 26, 2023, 5:26 p.m. UTC | #6

On Sun, Dec 24, 2023 at 4:26 PM David Rientjes <rientjes@google.com> wrote:
>
> On Thu, 14 Dec 2023, Pasha Tatashin wrote:
>
> > > > Whenever a new fields are added one of the following: node_stat_item
> > > > numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
> > > > files are auto expanded.
> > > >
> > > > This is a problem, as sysfs files should be only one value per file.
> > >
> > > Does this patch address the one-value-per-file issue?  (I think that ship
> > > has sailed for vmstat.)
> >
> > That ship has sailed for vmstat, this patch addresses what was asked
> > by GregKH: not to add new values to vmstat, as not to make the
> > existing problem even worse. The sysfs file system has a one page
> > limit per file. The developers will decide how to export the new items
> > added to node_stat, numa_stat, zone_stat individually. Each new item
> > can be exported in its own files, and must have its own documentation
> > about interface stability, value meaning, and expectations when the
> > stat file is absent.
> >
>
> As of at least 6.5, /proc/vmstat is a strict superset of the per-node
> vmstat.  Why is that a problem?

The intent of this series is to stop auto expanding
/sys/devices/system/node/nodeX/vmstat as sysfs should only be one
value per file, and the task is not to make things worse. /proc/vmstat
is mostly ok, however we might not need to auto expand it as well, to
avoid situations where removing a field becomes a problem, and we have
to keep it in the file forever, like what we do with nr_unstable.

> There's great benefit to being able to use the sample implementations to
> parse either /proc/vmstat *or* the per-node vmstat and without needing to
> read the per-node vmstat plus some new set of sysfs files that are
> one-value-per-file.  The per-node vmstat will always be multiple values,
> in fact it's a key value pair.

Yes, but that file is already large, and soon can overflow a page
size, instead of converting it to a binary format, let's add new items
as one item per-file.

> I have to think that doing anything else for vmstat is just adding
> complexity (like this patch) and actually making it *harder* on userspace
> to read the data it needs.
>
> Yes, the per-node vmstat likely shouldn't be in sysfs at all but it
> appears to have been added there 13+ years ago because it was a convenient
> place to add a per-node variant.  That's not ideal, but owell.

It is up-to GregKH who requested this change. Greg, specifically
requested not to add new fields into per-node vmstat, and we are
adding new fields with per-page metadata series, and IOMMU accounting
series as these files are auto-expanded without this series.

Pasha

David Rientjes Dec. 27, 2023, 12:53 a.m. UTC | #7

On Tue, 26 Dec 2023, Pasha Tatashin wrote:

> > > > > Whenever a new fields are added one of the following: node_stat_item
> > > > > numa_stat_item zone_stat_item, the /sys/devices/system/node/nodeX/vmstat
> > > > > files are auto expanded.
> > > > >
> > > > > This is a problem, as sysfs files should be only one value per file.
> > > >
> > > > Does this patch address the one-value-per-file issue?  (I think that ship
> > > > has sailed for vmstat.)
> > >
> > > That ship has sailed for vmstat, this patch addresses what was asked
> > > by GregKH: not to add new values to vmstat, as not to make the
> > > existing problem even worse. The sysfs file system has a one page
> > > limit per file. The developers will decide how to export the new items
> > > added to node_stat, numa_stat, zone_stat individually. Each new item
> > > can be exported in its own files, and must have its own documentation
> > > about interface stability, value meaning, and expectations when the
> > > stat file is absent.
> > >
> >
> > As of at least 6.5, /proc/vmstat is a strict superset of the per-node
> > vmstat.  Why is that a problem?
> 
> The intent of this series is to stop auto expanding
> /sys/devices/system/node/nodeX/vmstat as sysfs should only be one
> value per file, and the task is not to make things worse. /proc/vmstat
> is mostly ok, however we might not need to auto expand it as well, to
> avoid situations where removing a field becomes a problem, and we have
> to keep it in the file forever, like what we do with nr_unstable.
> 
> > There's great benefit to being able to use the sample implementations to
> > parse either /proc/vmstat *or* the per-node vmstat and without needing to
> > read the per-node vmstat plus some new set of sysfs files that are
> > one-value-per-file.  The per-node vmstat will always be multiple values,
> > in fact it's a key value pair.
> 
> Yes, but that file is already large, and soon can overflow a page
> size, instead of converting it to a binary format, let's add new items
> as one item per-file.
> 
> > I have to think that doing anything else for vmstat is just adding
> > complexity (like this patch) and actually making it *harder* on userspace
> > to read the data it needs.
> >
> > Yes, the per-node vmstat likely shouldn't be in sysfs at all but it
> > appears to have been added there 13+ years ago because it was a convenient
> > place to add a per-node variant.  That's not ideal, but owell.
> 
> It is up-to GregKH who requested this change. Greg, specifically
> requested not to add new fields into per-node vmstat, and we are
> adding new fields with per-page metadata series, and IOMMU accounting
> series as these files are auto-expanded without this series.
> 

Thanks, let's get clarification on this once and for all from Andrew and 
Greg.

I'd argue that the ship on the "sysfs one-value-per-file rule" has sailed 
for long-standing use cases where either (1) switching is just not 
possible or (2) switching would be an undue burden to the user.

An example of (1) would be THP enablement and defrag options:

$ grep . /sys/kernel/mm/transparent_hugepage/{defrag,enabled,shmem_enabled}
/sys/kernel/mm/transparent_hugepage/defrag:always defer defer+madvise [madvise] never
/sys/kernel/mm/transparent_hugepage/enabled:[always] madvise never
/sys/kernel/mm/transparent_hugepage/shmem_enabled:always within_size advise [never] deny force

This convention isn't going to change.  We're not going to suddenly add a 
new enablement or defrag option that can only be set in a newly added 
file that is one-value-per-file.

THP was obviously introduced before any sysfs "one-value-per-file rule" 
and, in retrospect, I think most people would agree that these files would 
be much better off implemented returning a single word.  But, choices 
where made in the "before times", and it was implemented in a way that 
shows all the possible choices and which one is in effect.  Owell.  We 
deal with it, and we move on.

Imagine if I add a new choice of "lightweight" to the "defrag" file.  The 
only rational way to do that would be to extend the current interface, not 
create a new /sys/kernel/mm/transparent_hugepage/defrag/lightweight file 
that is one-value-per-file that unsets all the other options in "defrag."
Please.

I use this an an extreme example, but a very real one: we can't break 
userspace and that will always supercede the desire to maintain 
one-value-per-file.

Don't get me wrong, I am all *for* one-value-per-file, it has many 
benefits.  But I disagree we should use it a a northstar for *existing* 
use cases if it actually means we'll be breaking userspace or making it 
much harder on userspace.

Under discussion in this thread would be an example of (2).  /proc/vmstat 
is a strict superset of its per-node 
/sys/devices/system/node/node*/vmstat counterparts.

For ZVCs, we *want* to ensure new stats are added to both.  For the same 
reason that we want to ensure old stats are removed from both.  Starting 
to maintain completely different implementations in the kernel comes with 
complexity, but also with risk of bugs.  Especially for something that 
works exceedigly well for the user today.

Saying that we can extend /proc/vmstat for more entries but then *not* 
extend /sys/devices/system/node/node*/vmstat for the exact same entries, 
and forcing those a new files with a single value, is doing a disservice 
to the user who wants to collect the data comprehensively with the same 
implementation for the entire system and each node.

The *only* time I think that makes sense is if all new vmstats must be 
added in their own files in both procfs (system-wide) and sysfs 
(per-node).  That very well may be the path forward, but get ready for an 
explosion in new top-level /proc files.  (It'll also be a shame for 
anybody who just dumps /proc/meminfo and /proc/vmstat in testing for 
capturing the state of the system.)

The discussion about needing to make vmstat be a binary file because it 
doesn't fit into a single page is not very realistic.  It will result in 
userspace breakage for everything parsing it as it is today, so let's just 
set that aside.  If a sysfs file needs to support returning more than 
PAGE_SIZE of contents, which is just an implementation detail, so be it.

For brand new interfaces, I'm all for it: add one-value-per-file and be 
done with it.

But for existing files and conventions, I think we should settle it as 
"keep doing what you've been doing for 13+ years" and don't force this 
argument every time a kernel developer wants to just add one more stat.

Over to Andrew and Greg.

Andrew Morton Dec. 27, 2023, 6:42 p.m. UTC | #8

On Tue, 26 Dec 2023 16:53:31 -0800 (PST) David Rientjes <rientjes@google.com> wrote:

> But for existing files and conventions, I think we should settle it as 
> "keep doing what you've been doing for 13+ years" and don't force this 
> argument every time a kernel developer wants to just add one more stat.

Absolutely.  Let's do what makes most sense.  For new things, one value
per file.  For stats which logically group with other existing stats,
do whatever the existing other stats are currently doing.

Greg KH Dec. 28, 2023, 10:17 a.m. UTC | #9

On Tue, Dec 26, 2023 at 04:53:31PM -0800, David Rientjes wrote:
> I'd argue that the ship on the "sysfs one-value-per-file rule" has sailed 
> for long-standing use cases where either (1) switching is just not 
> possible or (2) switching would be an undue burden to the user.
> 
> An example of (1) would be THP enablement and defrag options:
> 
> $ grep . /sys/kernel/mm/transparent_hugepage/{defrag,enabled,shmem_enabled}
> /sys/kernel/mm/transparent_hugepage/defrag:always defer defer+madvise [madvise] never
> /sys/kernel/mm/transparent_hugepage/enabled:[always] madvise never
> /sys/kernel/mm/transparent_hugepage/shmem_enabled:always within_size advise [never] deny force
> 
> This convention isn't going to change.  We're not going to suddenly add a 
> new enablement or defrag option that can only be set in a newly added 
> file that is one-value-per-file.
> 
> THP was obviously introduced before any sysfs "one-value-per-file rule" 

No, the rule has been there since "day one" for sysfs, this file snuck
in much later with no one noticing it against the "rules" and I've been
complaining about it every time someone tries to add a new field to it
that I notice.

> and, in retrospect, I think most people would agree that these files would 
> be much better off implemented returning a single word.  But, choices 
> where made in the "before times", and it was implemented in a way that 
> shows all the possible choices and which one is in effect.  Owell.  We 
> deal with it, and we move on.
> 
> Imagine if I add a new choice of "lightweight" to the "defrag" file.  The 
> only rational way to do that would be to extend the current interface, not 
> create a new /sys/kernel/mm/transparent_hugepage/defrag/lightweight file 
> that is one-value-per-file that unsets all the other options in "defrag."
> Please.

Please remember that the sysfs rule is there for a good reason, it makes
it very hard to break existing userspace tools if you stick with it.  If
you decide to ignore that rule, then you are on your own and better make
sure that nothing breaks.

Again, please learn from our previous mistakes with proc files, that is
why the rule is there.

If you wish to ignore the rule, you all really are on your own, good
luck!

greg k-h

Pasha Tatashin Dec. 28, 2023, 2:48 p.m. UTC | #10

On Wed, Dec 27, 2023 at 1:42 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Tue, 26 Dec 2023 16:53:31 -0800 (PST) David Rientjes <rientjes@google.com> wrote:
>
> > But for existing files and conventions, I think we should settle it as
> > "keep doing what you've been doing for 13+ years" and don't force this
> > argument every time a kernel developer wants to just add one more stat.
>
> Absolutely.  Let's do what makes most sense.  For new things, one value
> per file.  For stats which logically group with other existing stats,
> do whatever the existing other stats are currently doing.

The intent of "[PATCH] vmstat: don't auto expand the sysfs files" is
to do exactly this: keep the current fields in
/sys/devices/system/node/node*/vmstat as-is, but prevent future fields
added to node_stat, numa_stat, zone_stat from showing in vmstat.

Potentially, we could also extend checkpatch.pl to warn when
VMSTAT_SHOW_SYSFS is used in the future patches.

Pasha

David Rientjes Dec. 28, 2023, 8:43 p.m. UTC | #11

On Thu, 28 Dec 2023, Pasha Tatashin wrote:

> > > But for existing files and conventions, I think we should settle it as
> > > "keep doing what you've been doing for 13+ years" and don't force this
> > > argument every time a kernel developer wants to just add one more stat.
> >
> > Absolutely.  Let's do what makes most sense.  For new things, one value
> > per file.  For stats which logically group with other existing stats,
> > do whatever the existing other stats are currently doing.
> 
> The intent of "[PATCH] vmstat: don't auto expand the sysfs files" is
> to do exactly this: keep the current fields in
> /sys/devices/system/node/node*/vmstat as-is, but prevent future fields
> added to node_stat, numa_stat, zone_stat from showing in vmstat.
> 

This is the opposite of what Andrew and I wrote above.  There should be no 
need to start preventing future stats from being grouped together in 
vmstat if it logically makes sense.

NAK.

Pasha Tatashin Dec. 28, 2023, 8:50 p.m. UTC | #12

On Thu, Dec 28, 2023 at 3:43 PM David Rientjes <rientjes@google.com> wrote:
>
> On Thu, 28 Dec 2023, Pasha Tatashin wrote:
>
> > > > But for existing files and conventions, I think we should settle it as
> > > > "keep doing what you've been doing for 13+ years" and don't force this
> > > > argument every time a kernel developer wants to just add one more stat.
> > >
> > > Absolutely.  Let's do what makes most sense.  For new things, one value
> > > per file.  For stats which logically group with other existing stats,
> > > do whatever the existing other stats are currently doing.
> >
> > The intent of "[PATCH] vmstat: don't auto expand the sysfs files" is
> > to do exactly this: keep the current fields in
> > /sys/devices/system/node/node*/vmstat as-is, but prevent future fields
> > added to node_stat, numa_stat, zone_stat from showing in vmstat.
> >
>
> This is the opposite of what Andrew and I wrote above.  There should be no
> need to start preventing future stats from being grouped together in
> vmstat if it logically makes sense.
>
> NAK.

Hm, I must have misunderstood what Andrew said.

I am ok with this patch being NAK, as long as the IOMMU observability
patch series and Per-Page Metadata patch can be merged upstream, and
extend vmstat without a NAK from GregKH for expanding a sysfs vmstat
file.

Pasha

David Rientjes Dec. 28, 2023, 8:52 p.m. UTC | #13

On Thu, 28 Dec 2023, Greg Kroah-Hartman wrote:

> > I'd argue that the ship on the "sysfs one-value-per-file rule" has sailed 
> > for long-standing use cases where either (1) switching is just not 
> > possible or (2) switching would be an undue burden to the user.
> > 
> > An example of (1) would be THP enablement and defrag options:
> > 
> > $ grep . /sys/kernel/mm/transparent_hugepage/{defrag,enabled,shmem_enabled}
> > /sys/kernel/mm/transparent_hugepage/defrag:always defer defer+madvise [madvise] never
> > /sys/kernel/mm/transparent_hugepage/enabled:[always] madvise never
> > /sys/kernel/mm/transparent_hugepage/shmem_enabled:always within_size advise [never] deny force
> > 
> > This convention isn't going to change.  We're not going to suddenly add a 
> > new enablement or defrag option that can only be set in a newly added 
> > file that is one-value-per-file.
> > 
> > THP was obviously introduced before any sysfs "one-value-per-file rule" 
> 
> No, the rule has been there since "day one" for sysfs, this file snuck
> in much later with no one noticing it against the "rules" and I've been
> complaining about it every time someone tries to add a new field to it
> that I notice.
> 

Ah, gotcha, thanks.  I had assumed that the push for one-value-per-file 
started after thp, and perhaps even because of thp :)  I have to admit 
that whenever I log into a new server type one of the first things I do is

$ cat /sys/devices/system/node/node*/distance

and that table just makes intuitive sense.  If we were to go back in time 
and reimplement that as one-value-per-file, I'd just assume that many 
userspace implementations would just need to read 64 different files to 
structure it into the same exact table.

On the other hand, I have wished countless times that the thp settings 
would have actually been one-value-per-file from the start.

vmstat: don't auto expand the sysfs files

Commit Message

Comments

Patch