The patch titled
Subject: mm: initialize pages on demand during boot
has been added to the -mm tree. Its filename is
mm-initialize-pages-on-demand-during-boot-v5.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/mm-initialize-pages-on-demand-during-boot-v5.patch
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/mm-initialize-pages-on-demand-during-boot-v5.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx>
Subject: mm: initialize pages on demand during boot
Link: http://lkml.kernel.org/r/20180309220807.24961-3-pasha.tatashin@xxxxxxxxxx
Signed-off-by: Pavel Tatashin <pasha.tatashin@xxxxxxxxxx>
Reviewed-by: Daniel Jordan <daniel.m.jordan@xxxxxxxxxx>
Reviewed-by: Steven Sistare <steven.sistare@xxxxxxxxxx>
Reviewed-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Tested-by: Masayoshi Mizuma <m.mizuma@xxxxxxxxxxxxxx>
Acked-by: Mel Gorman <mgorman@xxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
mm/page_alloc.c | 68 +++++++++++++++++++++++-----------------------
1 file changed, 35 insertions(+), 33 deletions(-)
diff -puN mm/page_alloc.c~mm-initialize-pages-on-demand-during-boot-v5 mm/page_alloc.c
--- a/mm/page_alloc.c~mm-initialize-pages-on-demand-during-boot-v5
+++ a/mm/page_alloc.c
@@ -1572,14 +1572,10 @@ static int __init deferred_init_memmap(v
}
/*
- * This lock guarantees that only one thread at a time is allowed to grow zones
- * (decrease number of deferred pages).
- * Protects first_deferred_pfn field in all zones during early boot before
- * deferred pages are initialized. Deferred pages are initialized in
- * page_alloc_init_late() soon after smp_init() is complete.
+ * During boot we initialize deferred pages on-demand, as needed, but once
+ * page_alloc_init_late() has finished, the deferred pages are all initialized,
+ * and we can permanently disable path.
*/
-static __initdata DEFINE_SPINLOCK(deferred_zone_grow_lock);
-static bool deferred_zone_grow __initdata = true;
static DEFINE_STATIC_KEY_TRUE(deferred_pages);
/*
@@ -1589,8 +1585,9 @@ static DEFINE_STATIC_KEY_TRUE(deferred_p
* of SECTION_SIZE bytes by initializing struct pages in increments of
* PAGES_PER_SECTION * sizeof(struct page) bytes.
*
- * Return true when zone was grown by at least number of pages specified by
- * order. Otherwise return false.
+ * Return true when zone was grown, otherwise return false. We return true even
+ * when we grow less than requested, let the caller decide if there are enough
+ * pages to satisfy allocation.
*
* Note: We use noinline because this function is needed only during boot, and
* it is called from a __ref function _deferred_grow_zone. This way we are
@@ -1604,7 +1601,8 @@ deferred_grow_zone(struct zone *zone, un
pg_data_t *pgdat = NODE_DATA(nid);
unsigned long nr_pages_needed = ALIGN(1 << order, PAGES_PER_SECTION);
unsigned long nr_pages = 0;
- unsigned long first_init_pfn, first_deferred_pfn, spfn, epfn, t, flags;
+ unsigned long first_init_pfn, spfn, epfn, t, flags;
+ unsigned long first_deferred_pfn = pgdat->first_deferred_pfn;
phys_addr_t spa, epa;
u64 i;
@@ -1612,21 +1610,32 @@ deferred_grow_zone(struct zone *zone, un
if (zone_end_pfn(zone) != pgdat_end_pfn(pgdat))
return false;
- spin_lock_irqsave(&deferred_zone_grow_lock, flags);
+ pgdat_resize_lock_irq(pgdat, &flags);
+
/*
- * Bail if we raced with another thread that disabled on demand
- * initialization.
+ * If deferred pages have been initialized while we were waiting for
+ * lock return true, as zone was grown. The caller will try again this
+ * zone. We won't return to this function again, since caller also has
+ * this static branch.
*/
- if (!static_branch_unlikely(&deferred_pages) || !deferred_zone_grow) {
- spin_unlock_irqrestore(&deferred_zone_grow_lock, flags);
- return false;
+ if (!static_branch_unlikely(&deferred_pages)) {
+ pgdat_resize_unlock_irq(pgdat, &flags);
+ return true;
+ }
+
+ /*
+ * If someone grew this zone while we were waiting for spinlock, return
+ * true, as there might be enough pages already.
+ */
+ if (first_deferred_pfn != pgdat->first_deferred_pfn) {
+ pgdat_resize_unlock_irq(pgdat, &flags);
+ return true;
}
- first_deferred_pfn = pgdat->first_deferred_pfn;
first_init_pfn = max(zone->zone_start_pfn, first_deferred_pfn);
if (first_init_pfn >= pgdat_end_pfn(pgdat)) {
- spin_unlock_irqrestore(&deferred_zone_grow_lock, flags);
+ pgdat_resize_unlock_irq(pgdat, &flags);
return false;
}
@@ -1655,9 +1664,9 @@ deferred_grow_zone(struct zone *zone, un
break;
}
pgdat->first_deferred_pfn = first_deferred_pfn;
- spin_unlock_irqrestore(&deferred_zone_grow_lock, flags);
+ pgdat_resize_unlock_irq(pgdat, &flags);
- return nr_pages >= nr_pages_needed;
+ return nr_pages > 0;
}
/*
@@ -1681,19 +1690,6 @@ void __init page_alloc_init_late(void)
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
int nid;
- /*
- * We are about to initialize the rest of deferred pages, permanently
- * disable on-demand struct page initialization.
- *
- * Note: it is prohibited to modify static branches in non-preemptible
- * context. Since, spin_lock() disables preemption, we must use an
- * extra boolean deferred_zone_grow.
- */
- spin_lock_irq(&deferred_zone_grow_lock);
- deferred_zone_grow = false;
- spin_unlock_irq(&deferred_zone_grow_lock);
- static_branch_disable(&deferred_pages);
-
/* There will be num_node_state(N_MEMORY) threads */
atomic_set(&pgdat_init_n_undone, num_node_state(N_MEMORY));
for_each_node_state(nid, N_MEMORY) {
@@ -1703,6 +1699,12 @@ void __init page_alloc_init_late(void)
/* Block until all are initialised */
wait_for_completion(&pgdat_init_all_done_comp);
+ /*
+ * We initialized the rest of deferred pages, permanently
+ * disable on-demand struct page initialization.
+ */
+ static_branch_disable(&deferred_pages);
+
/* Reinit limits that are based on free pages after the kernel is up */
files_maxfiles_init();
#endif
_
Patches currently in -mm which might be from pasha.tatashin@xxxxxxxxxx are
mm-disable-interrupts-while-initializing-deferred-pages.patch
mm-initialize-pages-on-demand-during-boot.patch
mm-initialize-pages-on-demand-during-boot-fix-3.patch
mm-initialize-pages-on-demand-during-boot-fix-4.patch
mm-initialize-pages-on-demand-during-boot-v5.patch
mm-memory_hotplug-enforce-block-size-aligned-range-check.patch
x86-mm-memory_hotplug-determine-block-size-based-on-the-end-of-boot-memory.patch
x86-mm-memory_hotplug-determine-block-size-based-on-the-end-of-boot-memory-v4.patch
mm-uninitialized-struct-page-poisoning-sanity-checking.patch
mm-uninitialized-struct-page-poisoning-sanity-checking-v4.patch
mm-memory_hotplug-optimize-probe-routine.patch
mm-memory_hotplug-dont-read-nid-from-struct-page-during-hotplug.patch
mm-memory_hotplug-dont-read-nid-from-struct-page-during-hotplug-v5.patch
mm-memory_hotplug-optimize-memory-hotplug.patch
mm-memory_hotplug-optimize-memory-hotplug-v5.patch
xen-mm-allow-deferred-page-initialization-for-xen-pv-domains.patch
sparc64-ng4-memset-32-bits-overflow.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
