openwrt/target/linux/bcm27xx/patches-6.6/950-0359-mm-page_alloc-cma-introduce-a-customisable-threshold.patch
John Audia 69dd5a788f kernel: bump 6.6 to 6.6.37
Changelog: https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.6.37

Added:
	generic/hack-6.6/900-fix-build-to-handle-return-value.patch[1]

Manually rebased:
	generic/pending-6.6/834-ledtrig-libata.patch

Removed upstreamed:
	bcm27xx/patches-6.6/950-0398-drm-panel-panel-ilitek9881c-Use-cansleep-methods.patch[2]

All other patches automatically rebased.

1. Patch suggested by @DragonBluep to circumvent upstream breakage of kernel 6.6.37 compilation. See comments in https://github.com/openwrt/openwrt/pull/15879 for additional discussion.
2. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v6.6.37&id=1618f7a875ffd916596392fd29880c0429b8af60

Build system: x86/64
Build-tested: x86/64/AMD Cezanne, flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3
Run-tested: x86/64/AMD Cezanne, flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3

Signed-off-by: John Audia <therealgraysky@proton.me>
Link: https://github.com/openwrt/openwrt/pull/15879
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
2024-07-10 23:59:37 +02:00

68 lines
2.5 KiB
Diff

From 7b14f3444cf6b54b9905f2e7e1507e68fd92a9c0 Mon Sep 17 00:00:00 2001
From: David Plowman <david.plowman@raspberrypi.com>
Date: Tue, 29 Mar 2022 16:10:06 +0100
Subject: [PATCH 0359/1085] mm,page_alloc,cma: introduce a customisable
threshold for allocating pages in cma
On some platforms the cma area can be half the entire system memory,
meaning that allocations start happening in the cma area immediately.
This leads to fragmentation and subsequent fatal cma_alloc failures.
We introduce an "alloc_in_cma_threshold" parameter which requires that
this many sixteenths of the free pages must be in cma before it will
try to use them. By default this is set to 12, but the previous
behaviour can be restored by setting it to 8 on startup.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
---
mm/page_alloc.c | 28 +++++++++++++++++++++++++---
1 file changed, 25 insertions(+), 3 deletions(-)
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -204,6 +204,27 @@ EXPORT_SYMBOL(node_states);
gfp_t gfp_allowed_mask __read_mostly = GFP_BOOT_MASK;
+#define ALLOC_IN_CMA_THRESHOLD_MAX 16
+#define ALLOC_IN_CMA_THRESHOLD_DEFAULT 12
+
+static unsigned long _alloc_in_cma_threshold __read_mostly
+ = ALLOC_IN_CMA_THRESHOLD_DEFAULT;
+
+static int __init alloc_in_cma_threshold_setup(char *buf)
+{
+ unsigned long res;
+
+ if (kstrtoul(buf, 10, &res) < 0 ||
+ res > ALLOC_IN_CMA_THRESHOLD_MAX) {
+ pr_err("Bad alloc_cma_threshold value\n");
+ return 0;
+ }
+ _alloc_in_cma_threshold = res;
+ pr_info("Setting alloc_in_cma_threshold to %lu\n", res);
+ return 0;
+}
+early_param("alloc_in_cma_threshold", alloc_in_cma_threshold_setup);
+
/*
* A cached value of the page's pageblock's migratetype, used when the page is
* put on a pcplist. Used to avoid the pageblock migratetype lookup when
@@ -2095,12 +2116,13 @@ __rmqueue(struct zone *zone, unsigned in
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 when over more than a given proportion of
+ * the zone's free memory is in the CMA area.
*/
if (alloc_flags & ALLOC_CMA &&
zone_page_state(zone, NR_FREE_CMA_PAGES) >
- zone_page_state(zone, NR_FREE_PAGES) / 2) {
+ zone_page_state(zone, NR_FREE_PAGES) / ALLOC_IN_CMA_THRESHOLD_MAX
+ * _alloc_in_cma_threshold) {
page = __rmqueue_cma_fallback(zone, order);
if (page)
return page;