Skip to content

Commit

Permalink
mm: disable interrupts while initializing deferred pages
Browse files Browse the repository at this point in the history 
Vlastimil Babka reported about a window issue during which when deferred
pages are initialized, and the current version of on-demand
initialization is finished, allocations may fail.  While this is highly
unlikely scenario, since this kind of allocation request must be large,
and must come from interrupt handler, we still want to cover it.

We solve this by initializing deferred pages with interrupts disabled,
and holding node_size_lock spin lock while pages in the node are being
initialized.  The on-demand deferred page initialization that comes
later will use the same lock, and thus synchronize with
deferred_init_memmap().

It is unlikely for threads that initialize deferred pages to be
interrupted.  They run soon after smp_init(), but before modules are
initialized, and long before user space programs.  This is why there is
no adverse effect of having these threads running with interrupts
disabled.

[[email protected]: v6]
  Link: http://lkml.kernel.org/r/[email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Pavel Tatashin <[email protected]>
Reviewed-by: Andrew Morton <[email protected]>
Cc: Steven Sistare <[email protected]>
Cc: Daniel Jordan <[email protected]>
Cc: Masayoshi Mizuma <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: AKASHI Takahiro <[email protected]>
Cc: Gioh Kim <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Yaowei Bai <[email protected]>
Cc: Wei Yang <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Miles Chen <[email protected]>
Cc: Vlastimil Babka <[email protected]>
Cc: Mel Gorman <[email protected]>
Cc: Johannes Weiner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
  • Loading branch information
@torvalds
Pavel Tatashin authored and torvalds committed Apr 6, 2018
1 parent 8e7a0c9 commit 3a2d7fa
Show file tree
Hide file tree
Showing 3 changed files with 42 additions and 35 deletions.
53 changes: 28 additions & 25 deletions include/linux/memory_hotplug.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -51,24 +51,6 @@ enum {
MMOP_ONLINE_MOVABLE,
};

/*
* pgdat resizing functions
*/
static inline
void pgdat_resize_lock(struct pglist_data *pgdat, unsigned long *flags)
{
spin_lock_irqsave(&pgdat->node_size_lock, *flags);
}
static inline
void pgdat_resize_unlock(struct pglist_data *pgdat, unsigned long *flags)
{
spin_unlock_irqrestore(&pgdat->node_size_lock, *flags);
}
static inline
void pgdat_resize_init(struct pglist_data *pgdat)
{
spin_lock_init(&pgdat->node_size_lock);
}
/*
* Zone resizing functions
*
Expand Down Expand Up @@ -246,13 +228,6 @@ extern void clear_zone_contiguous(struct zone *zone);
___page; \
})

/*
* Stub functions for when hotplug is off
*/
static inline void pgdat_resize_lock(struct pglist_data *p, unsigned long *f) {}
static inline void pgdat_resize_unlock(struct pglist_data *p, unsigned long *f) {}
static inline void pgdat_resize_init(struct pglist_data *pgdat) {}

static inline unsigned zone_span_seqbegin(struct zone *zone)
{
return 0;
Expand Down Expand Up @@ -293,6 +268,34 @@ static inline bool movable_node_is_enabled(void)
}
#endif /* ! CONFIG_MEMORY_HOTPLUG */

#if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
/*
* pgdat resizing functions
*/
static inline
void pgdat_resize_lock(struct pglist_data *pgdat, unsigned long *flags)
{
spin_lock_irqsave(&pgdat->node_size_lock, *flags);
}
static inline
void pgdat_resize_unlock(struct pglist_data *pgdat, unsigned long *flags)
{
spin_unlock_irqrestore(&pgdat->node_size_lock, *flags);
}
static inline
void pgdat_resize_init(struct pglist_data *pgdat)
{
spin_lock_init(&pgdat->node_size_lock);
}
#else /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */
/*
* Stub functions for when hotplug is off
*/
static inline void pgdat_resize_lock(struct pglist_data *p, unsigned long *f) {}
static inline void pgdat_resize_unlock(struct pglist_data *p, unsigned long *f) {}
static inline void pgdat_resize_init(struct pglist_data *pgdat) {}
#endif /* !(CONFIG_MEMORY_HOTPLUG || CONFIG_DEFERRED_STRUCT_PAGE_INIT) */

#ifdef CONFIG_MEMORY_HOTREMOVE

extern bool is_mem_section_removable(unsigned long pfn, unsigned long nr_pages);
Expand Down
5 changes: 3 additions & 2 deletions include/linux/mmzone.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -633,14 +633,15 @@ typedef struct pglist_data {
#ifndef CONFIG_NO_BOOTMEM
struct bootmem_data *bdata;
#endif
#ifdef CONFIG_MEMORY_HOTPLUG
#if defined(CONFIG_MEMORY_HOTPLUG) || defined(CONFIG_DEFERRED_STRUCT_PAGE_INIT)
/*
* Must be held any time you expect node_start_pfn, node_present_pages
* or node_spanned_pages stay constant. Holding this will also
* guarantee that any pfn_valid() stays that way.
*
* pgdat_resize_lock() and pgdat_resize_unlock() are provided to
* manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
* manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG
* or CONFIG_DEFERRED_STRUCT_PAGE_INIT.
*
* Nests above zone->lock and zone->span_seqlock
*/
Expand Down
19 changes: 11 additions & 8 deletions mm/page_alloc.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1506,7 +1506,7 @@ static void __init deferred_free_pages(int nid, int zid, unsigned long pfn,
} else if (!(pfn & nr_pgmask)) {
deferred_free_range(pfn - nr_free, nr_free);
nr_free = 1;
cond_resched();
touch_nmi_watchdog();
} else {
nr_free++;
}
Expand Down Expand Up @@ -1535,7 +1535,7 @@ static unsigned long __init deferred_init_pages(int nid, int zid,
continue;
} else if (!page || !(pfn & nr_pgmask)) {
page = pfn_to_page(pfn);
cond_resched();
touch_nmi_watchdog();
} else {
page++;
}
Expand All @@ -1552,23 +1552,25 @@ static int __init deferred_init_memmap(void *data)
int nid = pgdat->node_id;
unsigned long start = jiffies;
unsigned long nr_pages = 0;
unsigned long spfn, epfn;
unsigned long spfn, epfn, first_init_pfn, flags;
phys_addr_t spa, epa;
int zid;
struct zone *zone;
unsigned long first_init_pfn = pgdat->first_deferred_pfn;
const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
u64 i;

/* Bind memory initialisation thread to a local node if possible */
if (!cpumask_empty(cpumask))
set_cpus_allowed_ptr(current, cpumask);

pgdat_resize_lock(pgdat, &flags);
first_init_pfn = pgdat->first_deferred_pfn;
if (first_init_pfn == ULONG_MAX) {
pgdat_resize_unlock(pgdat, &flags);
pgdat_init_report_one_done();
return 0;
}

/* Bind memory initialisation thread to a local node if possible */
if (!cpumask_empty(cpumask))
set_cpus_allowed_ptr(current, cpumask);

/* Sanity check boundaries */
BUG_ON(pgdat->first_deferred_pfn < pgdat->node_start_pfn);
BUG_ON(pgdat->first_deferred_pfn > pgdat_end_pfn(pgdat));
Expand Down Expand Up @@ -1598,6 +1600,7 @@ static int __init deferred_init_memmap(void *data)
epfn = min_t(unsigned long, zone_end_pfn(zone), PFN_DOWN(epa));
deferred_free_pages(nid, zid, spfn, epfn);
}
pgdat_resize_unlock(pgdat, &flags);

/* Sanity check that the next zone really is unpopulated */
WARN_ON(++zid < MAX_NR_ZONES && populated_zone(++zone));
Expand Down

0 comments on commit 3a2d7fa

Please sign in to comment.