| Message ID | 20240325145646.1044760-5-bhe@redhat.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | mm/mm_init.c: refactor free_area_init_core() | expand |
On Mon, Mar 25, 2024 at 10:56:44PM +0800, Baoquan He wrote: > Currently, in free_area_init_core(), when initialize zone's field, a > rough value is set to zone->managed_pages. That value is calculated by > (zone->present_pages - memmap_pages). > > In the meantime, add the value to nr_all_pages and nr_kernel_pages which > represent all free pages of system (only low memory or including HIGHMEM > memory separately). Both of them are gonna be used in > alloc_large_system_hash(). > > However, the rough calculation and setting of zone->managed_pages is > meaningless because > a) memmap pages are allocated on units of node in sparse_init() or > alloc_node_mem_map(pgdat); The simple (zone->present_pages - > memmap_pages) is too rough to make sense for zone; > b) the set zone->managed_pages will be zeroed out and reset with > acutal value in mem_init() via memblock_free_all(). Before the > resetting, no buddy allocation request is issued. > > Here, remove the meaningless and complicated calculation of > (zone->present_pages - memmap_pages), initialize zone->managed_pages as 0 > which reflect its actual value because no any page is added into buddy > system right now. It will be reset in mem_init(). > > And also remove the assignment of nr_all_pages and nr_kernel_pages in > free_area_init_core(). Instead, call the newly added calc_nr_kernel_pages() > to count up all free but not reserved memory in memblock and assign to > nr_all_pages and nr_kernel_pages. The counting excludes memmap_pages, > and other kernel used data, which is more accurate than old way and > simpler, and can also cover the ppc required arch_reserved_kernel_pages() > case. > > And also clean up the outdated code comment above free_area_init_core(). > And free_area_init_core() is easy to understand now, no need to add > words to explain. > > Signed-off-by: Baoquan He <bhe@redhat.com> Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org> > --- > mm/mm_init.c | 46 +++++----------------------------------------- > 1 file changed, 5 insertions(+), 41 deletions(-)
On 03/25/24 at 10:56pm, Baoquan He wrote: > Currently, in free_area_init_core(), when initialize zone's field, a > rough value is set to zone->managed_pages. That value is calculated by > (zone->present_pages - memmap_pages). > > In the meantime, add the value to nr_all_pages and nr_kernel_pages which > represent all free pages of system (only low memory or including HIGHMEM > memory separately). Both of them are gonna be used in > alloc_large_system_hash(). > > However, the rough calculation and setting of zone->managed_pages is > meaningless because > a) memmap pages are allocated on units of node in sparse_init() or > alloc_node_mem_map(pgdat); The simple (zone->present_pages - > memmap_pages) is too rough to make sense for zone; > b) the set zone->managed_pages will be zeroed out and reset with > acutal value in mem_init() via memblock_free_all(). Before the > resetting, no buddy allocation request is issued. > > Here, remove the meaningless and complicated calculation of > (zone->present_pages - memmap_pages), initialize zone->managed_pages as 0 > which reflect its actual value because no any page is added into buddy > system right now. It will be reset in mem_init(). > > And also remove the assignment of nr_all_pages and nr_kernel_pages in > free_area_init_core(). Instead, call the newly added calc_nr_kernel_pages() > to count up all free but not reserved memory in memblock and assign to > nr_all_pages and nr_kernel_pages. The counting excludes memmap_pages, > and other kernel used data, which is more accurate than old way and > simpler, and can also cover the ppc required arch_reserved_kernel_pages() > case. > > And also clean up the outdated code comment above free_area_init_core(). > And free_area_init_core() is easy to understand now, no need to add > words to explain. > > Signed-off-by: Baoquan He <bhe@redhat.com> > --- > mm/mm_init.c | 46 +++++----------------------------------------- > 1 file changed, 5 insertions(+), 41 deletions(-) > > diff --git a/mm/mm_init.c b/mm/mm_init.c > index c57a7fc97a16..7f71e56e83f3 100644 > --- a/mm/mm_init.c > +++ b/mm/mm_init.c > @@ -1565,15 +1565,6 @@ void __ref free_area_init_core_hotplug(struct pglist_data *pgdat) > } > #endif > > -/* > - * Set up the zone data structures: > - * - mark all pages reserved > - * - mark all memory queues empty > - * - clear the memory bitmaps > - * > - * NOTE: pgdat should get zeroed by caller. > - * NOTE: this function is only called during early init. > - */ > static void __init free_area_init_core(struct pglist_data *pgdat) > { > enum zone_type j; > @@ -1584,41 +1575,13 @@ static void __init free_area_init_core(struct pglist_data *pgdat) > > for (j = 0; j < MAX_NR_ZONES; j++) { > struct zone *zone = pgdat->node_zones + j; > - unsigned long size, freesize, memmap_pages; > - > - size = zone->spanned_pages; > - freesize = zone->present_pages; > - > - /* > - * Adjust freesize so that it accounts for how much memory > - * is used by this zone for memmap. This affects the watermark > - * and per-cpu initialisations > - */ > - memmap_pages = calc_memmap_size(size, freesize); > - if (!is_highmem_idx(j)) { > - if (freesize >= memmap_pages) { > - freesize -= memmap_pages; > - if (memmap_pages) > - pr_debug(" %s zone: %lu pages used for memmap\n", > - zone_names[j], memmap_pages); > - } else > - pr_warn(" %s zone: %lu memmap pages exceeds freesize %lu\n", > - zone_names[j], memmap_pages, freesize); > - } > - > - if (!is_highmem_idx(j)) > - nr_kernel_pages += freesize; > - /* Charge for highmem memmap if there are enough kernel pages */ > - else if (nr_kernel_pages > memmap_pages * 2) > - nr_kernel_pages -= memmap_pages; > - nr_all_pages += freesize; > + unsigned long size = zone->spanned_pages; > > /* > - * Set an approximate value for lowmem here, it will be adjusted > - * when the bootmem allocator frees pages into the buddy system. > - * And all highmem pages will be managed by the buddy system. > + * Initialize zone->managed_pages as 0 , it will be reset > + * when memblock allocator frees pages into buddy system. > */ > - zone_init_internals(zone, j, nid, freesize); > + zone_init_internals(zone, j, nid, 0); Here, we should initialize zone->managed_pages as zone->present_pages because later page_group_by_mobility_disabled need be set according to zone->managed_pages. Otherwise page_group_by_mobility_disabled will be set to 1 always. I will sent out v3. From a17b0921b4bd00596330f61ee9ea4b82386a9fed Mon Sep 17 00:00:00 2001 From: Baoquan He <bhe@redhat.com> Date: Thu, 28 Mar 2024 16:20:15 +0800 Subject: [PATCH] mm/mm_init.c: set zone's ->managed_pages as ->present_pages for now Content-type: text/plain Because page_group_by_mobility_disabled need be set according to zone's managed_pages later. Signed-off-by: Baoquan He <bhe@redhat.com> --- mm/mm_init.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mm/mm_init.c b/mm/mm_init.c index cc24e7958c0c..dd875f943cbb 100644 --- a/mm/mm_init.c +++ b/mm/mm_init.c @@ -1561,7 +1561,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat) * Initialize zone->managed_pages as 0 , it will be reset * when memblock allocator frees pages into buddy system. */ - zone_init_internals(zone, j, nid, 0); + zone_init_internals(zone, j, nid, zone->present_pages); if (!size) continue;
On Thu, Mar 28, 2024 at 04:32:38PM +0800, Baoquan He wrote: > On 03/25/24 at 10:56pm, Baoquan He wrote: > > > > /* > > - * Set an approximate value for lowmem here, it will be adjusted > > - * when the bootmem allocator frees pages into the buddy system. > > - * And all highmem pages will be managed by the buddy system. > > + * Initialize zone->managed_pages as 0 , it will be reset > > + * when memblock allocator frees pages into buddy system. > > */ > > - zone_init_internals(zone, j, nid, freesize); > > + zone_init_internals(zone, j, nid, 0); > > Here, we should initialize zone->managed_pages as zone->present_pages > because later page_group_by_mobility_disabled need be set according to > zone->managed_pages. Otherwise page_group_by_mobility_disabled will be > set to 1 always. I will sent out v3. With zone->managed_pages set to zone->present_pages we won't account for the reserved memory for initialization of page_group_by_mobility_disabled. As watermarks are still not initialized at the time build_all_zonelists() is called, we may use nr_all_pages - nr_kernel_pages instead of nr_free_zone_pages(), IMO. > From a17b0921b4bd00596330f61ee9ea4b82386a9fed Mon Sep 17 00:00:00 2001 > From: Baoquan He <bhe@redhat.com> > Date: Thu, 28 Mar 2024 16:20:15 +0800 > Subject: [PATCH] mm/mm_init.c: set zone's ->managed_pages as ->present_pages > for now > Content-type: text/plain > > Because page_group_by_mobility_disabled need be set according to zone's > managed_pages later. > > Signed-off-by: Baoquan He <bhe@redhat.com> > --- > mm/mm_init.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/mm/mm_init.c b/mm/mm_init.c > index cc24e7958c0c..dd875f943cbb 100644 > --- a/mm/mm_init.c > +++ b/mm/mm_init.c > @@ -1561,7 +1561,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat) > * Initialize zone->managed_pages as 0 , it will be reset > * when memblock allocator frees pages into buddy system. > */ > - zone_init_internals(zone, j, nid, 0); > + zone_init_internals(zone, j, nid, zone->present_pages); > > if (!size) > continue; > -- > 2.41.0 > > > > > > if (!size) > > continue; > > @@ -1915,6 +1878,7 @@ void __init free_area_init(unsigned long *max_zone_pfn) > > check_for_memory(pgdat); > > } > > > > + calc_nr_kernel_pages(); > > memmap_init(); > > > > /* disable hash distribution for systems with a single node */ > > -- > > 2.41.0 > > >
On 03/28/24 at 11:53am, Mike Rapoport wrote: > On Thu, Mar 28, 2024 at 04:32:38PM +0800, Baoquan He wrote: > > On 03/25/24 at 10:56pm, Baoquan He wrote: > > > > > > /* > > > - * Set an approximate value for lowmem here, it will be adjusted > > > - * when the bootmem allocator frees pages into the buddy system. > > > - * And all highmem pages will be managed by the buddy system. > > > + * Initialize zone->managed_pages as 0 , it will be reset > > > + * when memblock allocator frees pages into buddy system. > > > */ > > > - zone_init_internals(zone, j, nid, freesize); > > > + zone_init_internals(zone, j, nid, 0); > > > > Here, we should initialize zone->managed_pages as zone->present_pages > > because later page_group_by_mobility_disabled need be set according to > > zone->managed_pages. Otherwise page_group_by_mobility_disabled will be > > set to 1 always. I will sent out v3. > > With zone->managed_pages set to zone->present_pages we won't account for > the reserved memory for initialization of page_group_by_mobility_disabled. The old zone->managed_pages didn't account for the reserved pages either. It's calculated by (zone->present_pages - memmap_pages). memmap pages only is only a very small portion, e.g on x86_64, 4K page size, assuming size of struct page is 64, then it's 1/64 of system memory. On arm64, 64K page size, it's 1/1024 of system memory. And about the setting of page_group_by_mobility_disabled, the compared value pageblock_nr_pages * MIGRATE_TYPES which is very small. On x86_64, it's 4M*6=24M; on arm64 with 64K size and 128M*6=768M which should be the biggest among ARCH-es. if (vm_total_pages < (pageblock_nr_pages * MIGRATE_TYPES)) page_group_by_mobility_disabled = 1; else page_group_by_mobility_disabled = 0; So page_group_by_mobility_disabled could be set to 1 only on system with very little memory which is very rarely seen. And setting zone->managed_pages as zone->present_pages is very close to its old value: (zone->present_pages - memmap_pages). Here we don't need be very accurate, just a rough value. > > As watermarks are still not initialized at the time build_all_zonelists() > is called, we may use nr_all_pages - nr_kernel_pages instead of > nr_free_zone_pages(), IMO. nr_all_pages should be fine if we take this way. nr_kernel_pages is a misleading name, it's all low memory pages excluding kernel reserved apges. nr_all_pages is all memory pages including highmema and exluding kernel reserved pages. Both is fine to me. The first one is easier, simply setting zone->managed_pages as zone->present_pages. The 2nd way is a little more accurate. > > > From a17b0921b4bd00596330f61ee9ea4b82386a9fed Mon Sep 17 00:00:00 2001 > > From: Baoquan He <bhe@redhat.com> > > Date: Thu, 28 Mar 2024 16:20:15 +0800 > > Subject: [PATCH] mm/mm_init.c: set zone's ->managed_pages as ->present_pages > > for now > > Content-type: text/plain > > > > Because page_group_by_mobility_disabled need be set according to zone's > > managed_pages later. > > > > Signed-off-by: Baoquan He <bhe@redhat.com> > > --- > > mm/mm_init.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/mm/mm_init.c b/mm/mm_init.c > > index cc24e7958c0c..dd875f943cbb 100644 > > --- a/mm/mm_init.c > > +++ b/mm/mm_init.c > > @@ -1561,7 +1561,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat) > > * Initialize zone->managed_pages as 0 , it will be reset > > * when memblock allocator frees pages into buddy system. > > */ > > - zone_init_internals(zone, j, nid, 0); > > + zone_init_internals(zone, j, nid, zone->present_pages); > > > > if (!size) > > continue; > > -- > > 2.41.0 > > > > > > > > > > if (!size) > > > continue; > > > @@ -1915,6 +1878,7 @@ void __init free_area_init(unsigned long *max_zone_pfn) > > > check_for_memory(pgdat); > > > } > > > > > > + calc_nr_kernel_pages(); > > > memmap_init(); > > > > > > /* disable hash distribution for systems with a single node */ > > > -- > > > 2.41.0 > > > > > > > -- > Sincerely yours, > Mike. >
diff --git a/mm/mm_init.c b/mm/mm_init.c index c57a7fc97a16..7f71e56e83f3 100644 --- a/mm/mm_init.c +++ b/mm/mm_init.c @@ -1565,15 +1565,6 @@ void __ref free_area_init_core_hotplug(struct pglist_data *pgdat) } #endif -/* - * Set up the zone data structures: - * - mark all pages reserved - * - mark all memory queues empty - * - clear the memory bitmaps - * - * NOTE: pgdat should get zeroed by caller. - * NOTE: this function is only called during early init. - */ static void __init free_area_init_core(struct pglist_data *pgdat) { enum zone_type j; @@ -1584,41 +1575,13 @@ static void __init free_area_init_core(struct pglist_data *pgdat) for (j = 0; j < MAX_NR_ZONES; j++) { struct zone *zone = pgdat->node_zones + j; - unsigned long size, freesize, memmap_pages; - - size = zone->spanned_pages; - freesize = zone->present_pages; - - /* - * Adjust freesize so that it accounts for how much memory - * is used by this zone for memmap. This affects the watermark - * and per-cpu initialisations - */ - memmap_pages = calc_memmap_size(size, freesize); - if (!is_highmem_idx(j)) { - if (freesize >= memmap_pages) { - freesize -= memmap_pages; - if (memmap_pages) - pr_debug(" %s zone: %lu pages used for memmap\n", - zone_names[j], memmap_pages); - } else - pr_warn(" %s zone: %lu memmap pages exceeds freesize %lu\n", - zone_names[j], memmap_pages, freesize); - } - - if (!is_highmem_idx(j)) - nr_kernel_pages += freesize; - /* Charge for highmem memmap if there are enough kernel pages */ - else if (nr_kernel_pages > memmap_pages * 2) - nr_kernel_pages -= memmap_pages; - nr_all_pages += freesize; + unsigned long size = zone->spanned_pages; /* - * Set an approximate value for lowmem here, it will be adjusted - * when the bootmem allocator frees pages into the buddy system. - * And all highmem pages will be managed by the buddy system. + * Initialize zone->managed_pages as 0 , it will be reset + * when memblock allocator frees pages into buddy system. */ - zone_init_internals(zone, j, nid, freesize); + zone_init_internals(zone, j, nid, 0); if (!size) continue; @@ -1915,6 +1878,7 @@ void __init free_area_init(unsigned long *max_zone_pfn) check_for_memory(pgdat); } + calc_nr_kernel_pages(); memmap_init(); /* disable hash distribution for systems with a single node */
Currently, in free_area_init_core(), when initialize zone's field, a rough value is set to zone->managed_pages. That value is calculated by (zone->present_pages - memmap_pages). In the meantime, add the value to nr_all_pages and nr_kernel_pages which represent all free pages of system (only low memory or including HIGHMEM memory separately). Both of them are gonna be used in alloc_large_system_hash(). However, the rough calculation and setting of zone->managed_pages is meaningless because a) memmap pages are allocated on units of node in sparse_init() or alloc_node_mem_map(pgdat); The simple (zone->present_pages - memmap_pages) is too rough to make sense for zone; b) the set zone->managed_pages will be zeroed out and reset with acutal value in mem_init() via memblock_free_all(). Before the resetting, no buddy allocation request is issued. Here, remove the meaningless and complicated calculation of (zone->present_pages - memmap_pages), initialize zone->managed_pages as 0 which reflect its actual value because no any page is added into buddy system right now. It will be reset in mem_init(). And also remove the assignment of nr_all_pages and nr_kernel_pages in free_area_init_core(). Instead, call the newly added calc_nr_kernel_pages() to count up all free but not reserved memory in memblock and assign to nr_all_pages and nr_kernel_pages. The counting excludes memmap_pages, and other kernel used data, which is more accurate than old way and simpler, and can also cover the ppc required arch_reserved_kernel_pages() case. And also clean up the outdated code comment above free_area_init_core(). And free_area_init_core() is easy to understand now, no need to add words to explain. Signed-off-by: Baoquan He <bhe@redhat.com> --- mm/mm_init.c | 46 +++++----------------------------------------- 1 file changed, 5 insertions(+), 41 deletions(-)