On Wed, 10 Jul 2024, Yu Zhao wrote:
> Remove boilerplate by using a macro to choose the corresponding lock
> and handler for each folio_batch in cpu_fbatches.
>
> Signed-off-by: Yu Zhao <yuzhao@xxxxxxxxxx>
Andrew, please revert this "remove boilerplate" patch (and of course its
followup fix) from mm-unstable. From the title I presume it was intended
to make no functional change, but that's far from so.
Under tmpfs swapping load, on different runs I see various badnesses:
"Bad page" in __free_one_page(), Oops in __run_timer_base(),
WARNING at kernel/workqueue.c:790 in set_work_data(), PageBuddy BUG
at page-flags.h:1009 from __del_page_from_freelist(), something (I'd
given up taking better notes by this time) in __queue_work(), others.
All those were including the fix to Barry's report: without that fix,
the boot is drowned in warnings scrolling past too fast to be read.
(All the above were on the HP workstation, swapping to SSD; whereas on
this ThinkPad, swapping to NVMe, no problem seen at all - I mention the
swapping medium, but have no idea whether that's a relevant difference.
In each case, MGLRU compiled in but not enabled. THPs and 64kTHPs active.)
Sorry, but I've put no effort whatsoever into debugging this: "remove
boilerplate" didn't seem worth the effort, and my personal preference
is for readable boilerplate over hiding in a macro. If you prefer the
macro, I expect Yu can soon come up with a fix (which I could test here):
but for now please revert "remove boilerplate", since its issues get in
the way of further mm testing.
Thanks,
Hugh
> ---
> mm/swap.c | 107 +++++++++++++++++++-----------------------------------
> 1 file changed, 37 insertions(+), 70 deletions(-)
>
> diff --git a/mm/swap.c b/mm/swap.c
> index 4a66d2f87f26..342ff4e39ba4 100644
> --- a/mm/swap.c
> +++ b/mm/swap.c
> @@ -220,16 +220,45 @@ static void folio_batch_move_lru(struct folio_batch *fbatch, move_fn_t move_fn)
> folios_put(fbatch);
> }
>
> -static void folio_batch_add_and_move(struct folio_batch *fbatch,
> - struct folio *folio, move_fn_t move_fn)
> +static void __folio_batch_add_and_move(struct folio_batch *fbatch,
> + struct folio *folio, move_fn_t move_fn,
> + bool on_lru, bool disable_irq)
> {
> + unsigned long flags;
> +
> + folio_get(folio);
> +
> + if (on_lru && !folio_test_clear_lru(folio)) {
> + folio_put(folio);
> + return;
> + }
> +
> if (folio_batch_add(fbatch, folio) && !folio_test_large(folio) &&
> !lru_cache_disabled())
> return;
>
> + if (disable_irq)
> + local_lock_irqsave(&cpu_fbatches.lock_irq, flags);
> + else
> + local_lock(&cpu_fbatches.lock);
> +
> folio_batch_move_lru(fbatch, move_fn);
> +
> + if (disable_irq)
> + local_unlock_irqrestore(&cpu_fbatches.lock_irq, flags);
> + else
> + local_unlock(&cpu_fbatches.lock);
> }
>
> +#define folio_batch_add_and_move(folio, op, on_lru) \
> + __folio_batch_add_and_move( \
> + this_cpu_ptr(&cpu_fbatches.op), \
> + folio, \
> + op, \
> + on_lru, \
> + offsetof(struct cpu_fbatches, op) > offsetof(struct cpu_fbatches, lock_irq) \
> + )
> +
> static void lru_move_tail(struct lruvec *lruvec, struct folio *folio)
> {
> if (folio_test_unevictable(folio))
> @@ -250,23 +279,11 @@ static void lru_move_tail(struct lruvec *lruvec, struct folio *folio)
> */
> void folio_rotate_reclaimable(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> - unsigned long flags;
> -
> if (folio_test_locked(folio) || folio_test_dirty(folio) ||
> folio_test_unevictable(folio))
> return;
>
> - folio_get(folio);
> - if (!folio_test_clear_lru(folio)) {
> - folio_put(folio);
> - return;
> - }
> -
> - local_lock_irqsave(&cpu_fbatches.lock_irq, flags);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_move_tail);
> - folio_batch_add_and_move(fbatch, folio, lru_move_tail);
> - local_unlock_irqrestore(&cpu_fbatches.lock_irq, flags);
> + folio_batch_add_and_move(folio, lru_move_tail, true);
> }
>
> void lru_note_cost(struct lruvec *lruvec, bool file,
> @@ -355,21 +372,10 @@ static void folio_activate_drain(int cpu)
>
> void folio_activate(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> -
> if (folio_test_active(folio) || folio_test_unevictable(folio))
> return;
>
> - folio_get(folio);
> - if (!folio_test_clear_lru(folio)) {
> - folio_put(folio);
> - return;
> - }
> -
> - local_lock(&cpu_fbatches.lock);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_activate);
> - folio_batch_add_and_move(fbatch, folio, lru_activate);
> - local_unlock(&cpu_fbatches.lock);
> + folio_batch_add_and_move(folio, lru_activate, true);
> }
>
> #else
> @@ -513,8 +519,6 @@ EXPORT_SYMBOL(folio_mark_accessed);
> */
> void folio_add_lru(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> -
> VM_BUG_ON_FOLIO(folio_test_active(folio) &&
> folio_test_unevictable(folio), folio);
> VM_BUG_ON_FOLIO(folio_test_lru(folio), folio);
> @@ -524,11 +528,7 @@ void folio_add_lru(struct folio *folio)
> lru_gen_in_fault() && !(current->flags & PF_MEMALLOC))
> folio_set_active(folio);
>
> - folio_get(folio);
> - local_lock(&cpu_fbatches.lock);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_add);
> - folio_batch_add_and_move(fbatch, folio, lru_add);
> - local_unlock(&cpu_fbatches.lock);
> + folio_batch_add_and_move(folio, lru_add, false);
> }
> EXPORT_SYMBOL(folio_add_lru);
>
> @@ -702,22 +702,11 @@ void lru_add_drain_cpu(int cpu)
> */
> void deactivate_file_folio(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> -
> /* Deactivating an unevictable folio will not accelerate reclaim */
> if (folio_test_unevictable(folio))
> return;
>
> - folio_get(folio);
> - if (!folio_test_clear_lru(folio)) {
> - folio_put(folio);
> - return;
> - }
> -
> - local_lock(&cpu_fbatches.lock);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_deactivate_file);
> - folio_batch_add_and_move(fbatch, folio, lru_deactivate_file);
> - local_unlock(&cpu_fbatches.lock);
> + folio_batch_add_and_move(folio, lru_deactivate_file, true);
> }
>
> /*
> @@ -730,21 +719,10 @@ void deactivate_file_folio(struct folio *folio)
> */
> void folio_deactivate(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> -
> if (folio_test_unevictable(folio) || !(folio_test_active(folio) || lru_gen_enabled()))
> return;
>
> - folio_get(folio);
> - if (!folio_test_clear_lru(folio)) {
> - folio_put(folio);
> - return;
> - }
> -
> - local_lock(&cpu_fbatches.lock);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_deactivate);
> - folio_batch_add_and_move(fbatch, folio, lru_deactivate);
> - local_unlock(&cpu_fbatches.lock);
> + folio_batch_add_and_move(folio, lru_deactivate, true);
> }
>
> /**
> @@ -756,22 +734,11 @@ void folio_deactivate(struct folio *folio)
> */
> void folio_mark_lazyfree(struct folio *folio)
> {
> - struct folio_batch *fbatch;
> -
> if (!folio_test_anon(folio) || !folio_test_swapbacked(folio) ||
> folio_test_swapcache(folio) || folio_test_unevictable(folio))
> return;
>
> - folio_get(folio);
> - if (!folio_test_clear_lru(folio)) {
> - folio_put(folio);
> - return;
> - }
> -
> - local_lock(&cpu_fbatches.lock);
> - fbatch = this_cpu_ptr(&cpu_fbatches.lru_lazyfree);
> - folio_batch_add_and_move(fbatch, folio, lru_lazyfree);
> - local_unlock(&cpu_fbatches.lock);
> + folio_batch_add_and_move(folio, lru_lazyfree, true);
> }
>
> void lru_add_drain(void)
> --
> 2.45.2.803.g4e1b14247a-goog
