Re: [PATCH v2 1/1] mm: fix interrupt disabled long time inside deferred_init_memmap()

From: Shile Zhang
Date: Wed Mar 11 2020 - 08:02:56 EST



On 2020/3/11 19:42, Kirill Tkhai wrote:
On 11.03.2020 04:44, Shile Zhang wrote:
Hi Kirill,

Sorry for late to reply!
I'm not fully understood the whole thing about deferred page init, so I
just force on the jiffies update issue itself.

Maybe I'm in wrong path, it seems make no sense that deferred page init in 1 CPU system,
it cannot be initialize memory parallel.
Yes, it can't be initialized in parallel. But scheduler interprets deferred thread as a separate
task, so CPU time will be shared between that thread and the rest of tasks. This still should
make boot faster.

Thanks for your quickly reply!

Sorry, I don't think the CPU time can be shared to the rest of tasks since the CPU be blocked until
the deferred pages are all initialized, as following code:
8<------
#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT

ÂÂÂÂÂÂÂ /* There will be num_node_state(N_MEMORY) threads */
ÂÂÂÂÂÂÂ atomic_set(&pgdat_init_n_undone, num_node_state(N_MEMORY));
ÂÂÂÂÂÂÂ for_each_node_state(nid, N_MEMORY) {
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ kthread_run(deferred_init_memmap, NODE_DATA(nid), "pgdatinit%d", nid);
ÂÂÂÂÂÂÂ }

ÂÂÂÂÂÂÂ /* Block until all are initialised */
ÂÂÂÂÂÂÂ wait_for_completion(&pgdat_init_all_done_comp);
8<-------

Maybe I misunderstood this point.

@Pavel,
Could you please give any advice on this point? Thanks!

It might be better to disable deferred page init in 'deferred_init' in case of 1 CPU
(or only one memory node).

In other word, seems the better way to solve this issue is do not bind 'pgdatinit' thread
on boot CPU.

I also refactor the patch based on your comment, please help to check, thanks!


On 2020/3/4 18:47, Kirill Tkhai wrote:
On 04.03.2020 05:34, Shile Zhang wrote:
Hi Kirill,

Thanks for your quickly reply!

On 2020/3/4 00:52, Kirill Tkhai wrote:
On 03.03.2020 19:15, Shile Zhang wrote:
When 'CONFIG_DEFERRED_STRUCT_PAGE_INIT' is set, 'pgdatinit' kthread will
initialise the deferred pages with local interrupts disabled. It is
introduced by commit 3a2d7fa8a3d5 ("mm: disable interrupts while
initializing deferred pages").

The local interrupt will be disabled long time inside
deferred_init_memmap(), depends on memory size.
On machine with NCPUS <= 2, the 'pgdatinit' kthread could be pined on
boot CPU, then the tick timer will stuck long time, which caused the
system wall time inaccuracy.

For example, the dmesg shown that:

ÂÂÂ [ÂÂÂ 0.197975] node 0 initialised, 32170688 pages in 1ms

Obviously, 1ms is unreasonable.
Now, fix it by restore in the pending interrupts inside the while loop.
The reasonable demsg shown likes:

[ÂÂÂ 1.069306] node 0 initialised, 32203456 pages in 894ms
The way I understand the original problem, that Pavel fixed:

we need disable irqs in deferred_init_memmap() since this function may be called
in parallel with deferred_grow_zone() called from interrupt handler. So, Pavel
added lock to fix the race.

In case of we temporary unlock the lock, interrupt still be possible,
so my previous proposition returns the problem back.

Now thought again, I think we have to just add:

ÂÂÂÂÂpgdat_resize_unlock();
ÂÂÂÂÂpgdat_resize_lock();

instead of releasing interrupts, since in case of we just release them with lock held,
a call of interrupt->deferred_grow_zone() bring us to a deadlock.

So, unlock the lock is must.
Yes, you're right! I missed this point.
Thanks for your comment!

Signed-off-by: Shile Zhang<shile.zhang@xxxxxxxxxxxxxxxxx>
---
ÂÂ mm/page_alloc.c | 6 +++++-
ÂÂ 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 3c4eb750a199..d3f337f2e089 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1809,8 +1809,12 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂ * that we can avoid introducing any issues with the buddy
ÂÂÂÂÂÂÂ * allocator.
ÂÂÂÂÂÂÂ */
-ÂÂÂ while (spfn < epfn)
+ÂÂÂ while (spfn < epfn) {
ÂÂÂÂÂÂÂÂÂÂ nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+ÂÂÂÂÂÂÂ /* let in any pending interrupts */
+ÂÂÂÂÂÂÂ local_irq_restore(flags);
+ÂÂÂÂÂÂÂ local_irq_save(flags);
+ÂÂÂ }
ÂÂ zone_empty:
ÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
I think we need here something like below (untested):

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 79e950d76ffc..323afa9a4db5 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1828,7 +1828,7 @@ static int __init deferred_init_memmap(void *data)
ÂÂ {
ÂÂÂÂÂÂ pg_data_t *pgdat = data;
ÂÂÂÂÂÂ const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
-ÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0;
+ÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0, prev_nr_pages = 0;
ÂÂÂÂÂÂ unsigned long first_init_pfn, flags;
ÂÂÂÂÂÂ unsigned long start = jiffies;
ÂÂÂÂÂÂ struct zone *zone;
@@ -1869,8 +1869,18 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂ * that we can avoid introducing any issues with the buddy
ÂÂÂÂÂÂÂ * allocator.
ÂÂÂÂÂÂÂ */
-ÂÂÂ while (spfn < epfn)
+ÂÂÂ while (spfn < epfn) {
ÂÂÂÂÂÂÂÂÂÂ nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+ÂÂÂÂÂÂÂ /*
+ÂÂÂÂÂÂÂÂ * Release interrupts every 1Gb to give a possibility
+ÂÂÂÂÂÂÂÂ * a timer to advance jiffies.
+ÂÂÂÂÂÂÂÂ */
+ÂÂÂÂÂÂÂ if (nr_pages - prev_nr_pages > (1UL << (30 - PAGE_SHIFT))) {
+ÂÂÂÂÂÂÂÂÂÂÂ prev_nr_pages = nr_pages;
+ÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
+ÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_lock(pgdat, &flags);
+ÂÂÂÂÂÂÂ }
+ÂÂÂ }
ÂÂ zone_empty:
ÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
 (I believe the comment may be improved more).
Yeah, your patch is better!
I test your code and it works!
But it seems that 1G is still hold the interrupts too long, about 40ms in my env
with Intel(R) Xeon(R) 2.5GHz). I tried other size, it is OK to use 1024 pages (4MB),
which suggested by Andrew's before.

Could you please help to review it again?

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 3c4eb750a199..5def66d3ffcd 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1768,7 +1768,7 @@ static int __init deferred_init_memmap(void *data)
ÂÂ{
ÂÂÂÂÂÂÂÂ pg_data_t *pgdat = data;
ÂÂÂÂÂÂÂÂ const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
-ÂÂÂÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0;
+ÂÂÂÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0, prev_nr_pages = 0;
ÂÂÂÂÂÂÂÂ unsigned long first_init_pfn, flags;
ÂÂÂÂÂÂÂÂ unsigned long start = jiffies;
ÂÂÂÂÂÂÂÂ struct zone *zone;
@@ -1809,8 +1809,17 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂÂÂ * that we can avoid introducing any issues with the buddy
ÂÂÂÂÂÂÂÂÂ * allocator.
ÂÂÂÂÂÂÂÂÂ */
-ÂÂÂÂÂÂ while (spfn < epfn)
+ÂÂÂÂÂÂ while (spfn < epfn) {
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ /*
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ * Restore pending interrupts every 1024 pages to give
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ * the chance tick timer to advance jiffies.
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ */
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ if (nr_pages - prev_nr_pages > 1024) {
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_unlock(&flags);
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_lock(&flags);
Here is problem: prev_nr_pages must be updated.

Anyway, releasing every 4M looks wrong for me, since you removes the fix that Pavel introduced.
He protected against big allocations made from interrupt content. But in case of we unlock
the lock after 4Mb, only 4Mb will be available for allocations from interrupts. pgdat->first_deferred_pfn
is updated at the start of function, so interrupt allocations won't be able to initialize
mode for themselve.
Yes, you're right. I missed this point since I'm not fully understood the code before.
Thanks for your advice!
In case of you want unlock interrupts very often, you should make some creativity with first_deferred_pfn.
We should update it sequentially. Something like below (untested):
I got your point now, thanks!
---
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 79e950d76ffc..be09d158baeb 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1828,7 +1828,7 @@ static int __init deferred_init_memmap(void *data)
 {
ÂÂÂÂÂ pg_data_t *pgdat = data;
ÂÂÂÂÂ const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
-ÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0;
+ÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages;
ÂÂÂÂÂ unsigned long first_init_pfn, flags;
ÂÂÂÂÂ unsigned long start = jiffies;
ÂÂÂÂÂ struct zone *zone;
@@ -1838,7 +1838,7 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂ /* Bind memory initialisation thread to a local node if possible */
ÂÂÂÂÂ if (!cpumask_empty(cpumask))
ÂÂÂÂÂÂÂÂÂ set_cpus_allowed_ptr(current, cpumask);
-
+again:
ÂÂÂÂÂ pgdat_resize_lock(pgdat, &flags);
ÂÂÂÂÂ first_init_pfn = pgdat->first_deferred_pfn;
ÂÂÂÂÂ if (first_init_pfn == ULONG_MAX) {
@@ -1850,7 +1850,6 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂ /* Sanity check boundaries */
ÂÂÂÂÂ BUG_ON(pgdat->first_deferred_pfn < pgdat->node_start_pfn);
ÂÂÂÂÂ BUG_ON(pgdat->first_deferred_pfn > pgdat_end_pfn(pgdat));
-ÂÂÂ pgdat->first_deferred_pfn = ULONG_MAX;
 Â /* Only the highest zone is deferred so find it */
ÂÂÂÂÂ for (zid = 0; zid < MAX_NR_ZONES; zid++) {
@@ -1864,14 +1863,30 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ first_init_pfn))
ÂÂÂÂÂÂÂÂÂ goto zone_empty;
 + nr_pages = 0;
'nr_pages' used to mark the total init pages before, so it cannot be zerolized each round.
seems we need one more to count the pages init each round.

+
ÂÂÂÂÂ /*
ÂÂÂÂÂÂ * Initialize and free pages in MAX_ORDER sized increments so
ÂÂÂÂÂÂ * that we can avoid introducing any issues with the buddy
ÂÂÂÂÂÂ * allocator.
+ÂÂÂÂ * Final iteration marker is: spfn=ULONG_MAX and epfn=0.
ÂÂÂÂÂÂ */
-ÂÂÂ while (spfn < epfn)
+ÂÂÂ while (spfn < epfn) {
ÂÂÂÂÂÂÂÂÂ nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+ÂÂÂÂÂÂÂ if (!epfn)
+ÂÂÂÂÂÂÂÂÂÂÂ break;
Seems 'epfn' never goes to 0 since it is "end page frame number", right?
So this is needless.
+ÂÂÂÂÂÂÂ pgdat->first_deferred_pfn = epfn;
I think first_deferred_pfn update wrong value here, it seems should be the spfn, the start pfn right?
+ÂÂÂÂÂÂÂ /*
+ÂÂÂÂÂÂÂÂ * Restore pending interrupts every 128Mb to give
+ÂÂÂÂÂÂÂÂ * the chance tick timer to advance jiffies.
+ÂÂÂÂÂÂÂÂ */
+ÂÂÂÂÂÂÂ if (nr_pages > (1UL << 27 - PAGE_SHIFT)) {
+ÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
+ÂÂÂÂÂÂÂÂÂÂÂ goto again;
+ÂÂÂÂÂÂÂ }
+ÂÂÂ }
 zone_empty:
+ÂÂÂ pgdat->first_deferred_pfn = ULONG_MAX;
ÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
 Â /* Sanity check that the next zone really is unpopulated */


I update the patch based on your comment, it passed the test.
Could you please help to review it again? Thanks!
The patch is OK for me. It may be sent into ml with an appropriate description.

Feel free to not forget to add my:

Co-developed-by: Kirill Tkhai <ktkhai@xxxxxxxxxxxxx>

Yeah, sure!
Many thanks for your kindly help on this issue!
I'll send out v3 later for further review.
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 3c4eb750a199..841c902d4509 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1763,12 +1763,17 @@ deferred_init_maxorder(u64 *i, struct zone *zone, unsigned long *start_pfn,
ÂÂÂÂÂÂÂ return nr_pages;
Â}

+/*
+ * Release the tick timer interrupts for every TICK_PAGE_COUNT pages.
+ */
+#define TICK_PAGE_COUNTÂÂÂÂÂÂÂ (32 * 1024)
+
Â/* Initialise remaining memory on a node */
Âstatic int __init deferred_init_memmap(void *data)
Â{
ÂÂÂÂÂÂÂ pg_data_t *pgdat = data;
ÂÂÂÂÂÂÂ const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
-ÂÂÂÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0;
+ÂÂÂÂÂÂ unsigned long spfn = 0, epfn = 0, nr_pages = 0, prev_nr_pages = 0;
ÂÂÂÂÂÂÂ unsigned long first_init_pfn, flags;
ÂÂÂÂÂÂÂ unsigned long start = jiffies;
ÂÂÂÂÂÂÂ struct zone *zone;
@@ -1779,6 +1784,7 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂ if (!cpumask_empty(cpumask))
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ set_cpus_allowed_ptr(current, cpumask);

+again:
ÂÂÂÂÂÂÂ pgdat_resize_lock(pgdat, &flags);
ÂÂÂÂÂÂÂ first_init_pfn = pgdat->first_deferred_pfn;
ÂÂÂÂÂÂÂ if (first_init_pfn == ULONG_MAX) {
@@ -1790,7 +1796,6 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂ /* Sanity check boundaries */
ÂÂÂÂÂÂÂ BUG_ON(pgdat->first_deferred_pfn < pgdat->node_start_pfn);
ÂÂÂÂÂÂÂ BUG_ON(pgdat->first_deferred_pfn > pgdat_end_pfn(pgdat));
-ÂÂÂÂÂÂ pgdat->first_deferred_pfn = ULONG_MAX;

ÂÂÂÂÂÂÂ /* Only the highest zone is deferred so find it */
ÂÂÂÂÂÂÂ for (zid = 0; zid < MAX_NR_ZONES; zid++) {
@@ -1809,9 +1814,23 @@ static int __init deferred_init_memmap(void *data)
ÂÂÂÂÂÂÂÂ * that we can avoid introducing any issues with the buddy
ÂÂÂÂÂÂÂÂ * allocator.
ÂÂÂÂÂÂÂÂ */
-ÂÂÂÂÂÂ while (spfn < epfn)
+ÂÂÂÂÂÂ while (spfn < epfn) {
ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ /*
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ * Release the interrupts for every TICK_PAGE_COUNT pages
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ * (128MB) to give the chance that tick timer to advance
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ * the jiffies.
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ */
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ if ((nr_pages - prev_nr_pages) > TICK_PAGE_COUNT) {
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ prev_nr_pages = nr_pages;
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ pgdat->first_deferred_pfn = spfn;
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ goto again;
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ }
+ÂÂÂÂÂÂ }
+
Âzone_empty:
+ÂÂÂÂÂÂ pgdat->first_deferred_pfn = ULONG_MAX;
ÂÂÂÂÂÂÂ pgdat_resize_unlock(pgdat, &flags);

ÂÂÂÂÂÂÂ /* Sanity check that the next zone really is unpopulated */