reply: [PATCH] Revert "mm: skip CMA pages when they are not available"

["黄朝阳 (Zhaoyang Huang)" <zhaoyang.huang@unisoc.com>]

On Thu, Mar 14, 2024 at 10:15 AM <liuhailong@oppo.com> wrote:
>
> From: "Hailong.Liu" <liuhailong@oppo.com>
>
> This reverts commit 5da226dbfce3a2f44978c2c7cf88166e69a6788b.
>
> patch may cause system not responding. if cma pages is large in lru_list
> and system is in lowmemory, many tasks would enter direct reclaim and waste
> cpu time to isolate and return. Test this patch on android-5.15 device
> and tasks call stack as below.
>
> Task name: UsbFfs-worker [affinity: 0xff] pid: 3374 cpu: 7 prio: 120 start: ffffff8897a35c80
> state: 0x0[R] exit_state: 0x0 stack base: 0xffffffc01eaa0000
> Last_enqueued_ts:       0.000000000 Last_sleep_ts:       0.000000000
> Stack:
> [<ffffffd32ee7d910>] __switch_to+0x180
> [<ffffffd3302022fc>] __schedule+0x4dc
> [<ffffffd330201e08>] preempt_schedule+0x5c
> [<ffffffd33020a4d0>] _raw_spin_unlock_irq+0x54
> [<ffffffd32f14906c>] shrink_inactive_list+0x1d0
> [<ffffffd32f143998>] shrink_lruvec+0x1bc
> [<ffffffd32f147c0c>] shrink_node_memcgs+0x184
> [<ffffffd32f147414>] shrink_node+0x2d0
> [<ffffffd32f146d38>] shrink_zones+0x14c
> [<ffffffd32f142e84>] do_try_to_free_pages+0xe8
> [<ffffffd32f142b08>] try_to_free_pages+0x2e0
> [<ffffffd32f1a8e44>] __alloc_pages_direct_reclaim+0x84
> [<ffffffd32f1a2d58>] __alloc_pages_slowpath+0x4d0
> [<ffffffd32f1a23bc>] __alloc_pages_nodemask[jt]+0x124
> [<ffffffd32f19a220>] __vmalloc_area_node+0x188
> [<ffffffd32f19a540>] __vmalloc_node+0x148
> [<ffffffd32f19a60c>] vmalloc+0x4c
> [<ffffffd32f910218>] ffs_epfile_io+0x258
> [<ffffffd330033780>] kretprobe_trampoline[jt]+0x0
> [<ffffffd330033780>] kretprobe_trampoline[jt]+0x0
> [<ffffffd32f28129c>] __io_submit_one+0x1c0
> [<ffffffd32f280e38>] io_submit_one+0x88
> [<ffffffd32f280c88>] __do_sys_io_submit+0x178
> [<ffffffd32f27eac0>] __arm64_sys_io_submit+0x20
> [<ffffffd32eeabb74>] el0_svc_common.llvm.9961749221945255377+0xd0
> [<ffffffd32eeaba34>] do_el0_svc+0x28
> [<ffffffd32ff21be8>] el0_svc+0x14
> [<ffffffd32ff21b70>] el0_sync_handler+0x88
> [<ffffffd32ee128b8>] el0_sync+0x1b8
>
> Task name: kthreadd [affinity: 0xff] pid: 2 cpu: 7 prio: 120 start: ffffff87808c0000
> state: 0x0[R] exit_state: 0x0 stack base: 0xffffffc008078000
> Last_enqueued_ts:       0.000000000 Last_sleep_ts:       0.000000000
> Stack:
> [<ffffffd32ee7d910>] __switch_to+0x180
> [<ffffffd3302022fc>] __schedule+0x4dc
> [<ffffffd330201e08>] preempt_schedule+0x5c
> [<ffffffd33020a4d0>] _raw_spin_unlock_irq+0x54
> [<ffffffd32f149168>] shrink_inactive_list+0x2cc
> [<ffffffd32f143998>] shrink_lruvec+0x1bc
> [<ffffffd32f147c0c>] shrink_node_memcgs+0x184
> [<ffffffd32f147414>] shrink_node+0x2d0
> [<ffffffd32f146d38>] shrink_zones+0x14c
> [<ffffffd32f142e84>] do_try_to_free_pages+0xe8
> [<ffffffd32f142b08>] try_to_free_pages+0x2e0
> [<ffffffd32f1a8e44>] __alloc_pages_direct_reclaim+0x84
> [<ffffffd32f1a2d58>] __alloc_pages_slowpath+0x4d0
> [<ffffffd32f1a23bc>] __alloc_pages_nodemask[jt]+0x124
> [<ffffffd32f19a220>] __vmalloc_area_node+0x188
> [<ffffffd32f19a044>] __vmalloc_node_range+0x88
> [<ffffffd32f0fb430>] scs_alloc+0x1b8
> [<ffffffd32f0fb62c>] scs_prepare+0x20
> [<ffffffd32ef2ce04>] dup_task_struct+0xd4
> [<ffffffd32ef2a77c>] copy_process+0x144
> [<ffffffd32ef2bae4>] kernel_clone+0xb4
> [<ffffffd32ef2c040>] kernel_thread+0x5c
> [<ffffffd32ef618d0>] kthreadd+0x184
>
> without this patch, the tasks will reclaim cma pages and wakeup
> oom-killer or not spin on cpus.
>
> Signed-off-by: Hailong.Liu <liuhailong@oppo.com>
> ---
>  mm/vmscan.c | 22 +---------------------
>  1 file changed, 1 insertion(+), 21 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 2fe4a11d63f4..197ddf62019f 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -2261,25 +2261,6 @@ static __always_inline void update_lru_sizes(struct lruvec *lruvec,
>
>  }
>
> -#ifdef CONFIG_CMA
> -/*
> - * It is waste of effort to scan and reclaim CMA pages if it is not available
> - * for current allocation context. Kswapd can not be enrolled as it can not
> - * distinguish this scenario by using sc->gfp_mask = GFP_KERNEL
> - */
> -static bool skip_cma(struct folio *folio, struct scan_control *sc)
> -{
> -       return !current_is_kswapd() &&
> -                       gfp_migratetype(sc->gfp_mask) != MIGRATE_MOVABLE &&
> -                       get_pageblock_migratetype(&folio->page) == MIGRATE_CMA;
> -}
> -#else
> -static bool skip_cma(struct folio *folio, struct scan_control *sc)
> -{
> -       return false;
> -}
> -#endif
> -

>NAK.

>+Charan Teja Kalla -- This can cause build errors when CONFIG_LRU_GEN=y.

>If you plan to post a v2, please include a reproducer. Thanks.

Could you please retest the case with bellow patch, which has not been in the aosp yet.

From: Zhaoyang Huang <zhaoyang.huang@unisoc.com>

According to current CMA utilization policy, an alloc_pages(GFP_USER)
could 'steal' UNMOVABLE & RECLAIMABLE page blocks via the help of
CMA(pass zone_watermark_ok by counting CMA in but use U&R in rmqueue),
which could lead to following alloc_pages(GFP_KERNEL) fail.
Solving this by introducing second watermark checking for GFP_MOVABLE,
which could have the allocation use CMA when proper.

-- Free_pages(30MB)
|
|
-- WMARK_LOW(25MB)
|
-- Free_CMA(12MB)
|
|
--

Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
---
v6: update comments
---
---
 mm/page_alloc.c | 44 ++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 40 insertions(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 452459836b71..5a146aa7c0aa 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2078,6 +2078,43 @@ __rmqueue_fallback(struct zone *zone, int order, int start_migratetype,

 }

+#ifdef CONFIG_CMA
+/*
+ * GFP_MOVABLE allocation could drain UNMOVABLE & RECLAIMABLE page blocks via
+ * the help of CMA which makes GFP_KERNEL failed. Checking if zone_watermark_ok
+ * again without ALLOC_CMA to see if to use CMA first.
+ */
+static bool use_cma_first(struct zone *zone, unsigned int order, unsigned int alloc_flags)
+{
+       unsigned long watermark;
+       bool cma_first = false;
+
+       watermark = wmark_pages(zone, alloc_flags & ALLOC_WMARK_MASK);
+       /* check if GFP_MOVABLE pass previous zone_watermark_ok via the help of CMA */
+       if (zone_watermark_ok(zone, order, watermark, 0, alloc_flags & (~ALLOC_CMA))) {
+               /*
+                * Balance movable allocations between regular and CMA areas by
+                * allocating from CMA when over half of the zone's free memory
+                * is in the CMA area.
+                */
+               cma_first = (zone_page_state(zone, NR_FREE_CMA_PAGES) >
+                               zone_page_state(zone, NR_FREE_PAGES) / 2);
+       } else {
+               /*
+                * watermark failed means UNMOVABLE & RECLAIMBLE is not enough
+                * now, we should use cma first to keep them stay around the
+                * corresponding watermark
+                */
+               cma_first = true;
+       }
+       return cma_first;
+}
+#else
+static bool use_cma_first(struct zone *zone, unsigned int order, unsigned int alloc_flags)
+{
+       return false;
+}
+#endif
 /*
  * Do the hard work of removing an element from the buddy allocator.
  * Call me with the zone->lock already held.
@@ -2091,12 +2128,11 @@ __rmqueue(struct zone *zone, unsigned int order, int migratetype,
        if (IS_ENABLED(CONFIG_CMA)) {
                /*
                 * Balance movable allocations between regular and CMA areas by
-                * allocating from CMA when over half of the zone's free memory
-                * is in the CMA area.
+                * allocating from CMA base on judging zone_watermark_ok again
+                * to see if the latest check got pass via the help of CMA
                 */
                if (alloc_flags & ALLOC_CMA &&
-                   zone_page_state(zone, NR_FREE_CMA_PAGES) >
-                   zone_page_state(zone, NR_FREE_PAGES) / 2) {
+                       use_cma_first(zone, order, alloc_flags)) {
                        page = __rmqueue_cma_fallback(zone, order);
                        if (page)
                                return page;
--


>  /*
>   * Isolating page from the lruvec to fill in @dst list by nr_to_scan times.
>   *
> @@ -2326,8 +2307,7 @@ static unsigned long isolate_lru_folios(unsigned long nr_to_scan,
>                 nr_pages = folio_nr_pages(folio);
>                 total_scan += nr_pages;
>
> -               if (folio_zonenum(folio) > sc->reclaim_idx ||
> -                               skip_cma(folio, sc)) {
> +               if (folio_zonenum(folio) > sc->reclaim_idx) {
>                         nr_skipped[folio_zonenum(folio)] += nr_pages;
>                         move_to = &folios_skipped;
>                         goto move;
> --
> 2.34.1
>
>