[PATCH bpf-next v3 1/6] mm, bpf: Introduce try_alloc_pages() for opportunistic page allocation

[alexei.starovoitov@gmail.com]

From: Alexei Starovoitov <ast@kernel.org>

Tracing BPF programs execute from tracepoints and kprobes where
running context is unknown, but they need to request additional
memory. The prior workarounds were using pre-allocated memory and
BPF specific freelists to satisfy such allocation requests.
Instead, introduce internal __GFP_TRYLOCK flag that makes page
allocator accessible from any context. It relies on percpu free
list of pages that rmqueue_pcplist() should be able to pop the
page from. If it fails (due to IRQ re-entrancy or list being
empty) then try_alloc_pages() attempts to spin_trylock zone->lock
and refill percpu freelist as normal.
BPF program may execute with IRQs disabled and zone->lock is
sleeping in RT, so trylock is the only option. In theory we can
introduce percpu reentrance counter and increment it every time
spin_lock_irqsave(&zone->lock, flags) is used, but we cannot rely
on it. Even if this cpu is not in page_alloc path the
spin_lock_irqsave() is not safe, since BPF prog might be called
from tracepoint where preemption is disabled. So trylock only.

Note, free_page and memcg are not taught about __GFP_TRYLOCK yet.
The support comes in the next patches.

This is a first step towards supporting BPF requirements in SLUB
and getting rid of bpf_mem_alloc.
That goal was discussed at LSFMM: https://lwn.net/Articles/974138/

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 include/linux/gfp.h       |  3 ++
 include/linux/gfp_types.h |  1 +
 mm/internal.h             |  2 ++
 mm/page_alloc.c           | 69 ++++++++++++++++++++++++++++++++++++---
 4 files changed, 71 insertions(+), 4 deletions(-)

diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index b0fe9f62d15b..65b8df1db26a 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -347,6 +347,9 @@ static inline struct page *alloc_page_vma_noprof(gfp_t gfp,
 }
 #define alloc_page_vma(...)			alloc_hooks(alloc_page_vma_noprof(__VA_ARGS__))
 
+struct page *try_alloc_pages_noprof(int nid, unsigned int order);
+#define try_alloc_pages(...)			alloc_hooks(try_alloc_pages_noprof(__VA_ARGS__))
+
 extern unsigned long get_free_pages_noprof(gfp_t gfp_mask, unsigned int order);
 #define __get_free_pages(...)			alloc_hooks(get_free_pages_noprof(__VA_ARGS__))
 
diff --git a/include/linux/gfp_types.h b/include/linux/gfp_types.h
index 65db9349f905..65b148ec86eb 100644
--- a/include/linux/gfp_types.h
+++ b/include/linux/gfp_types.h
@@ -48,6 +48,7 @@ enum {
 	___GFP_THISNODE_BIT,
 	___GFP_ACCOUNT_BIT,
 	___GFP_ZEROTAGS_BIT,
+	___GFP_TRYLOCK_BIT,
 #ifdef CONFIG_KASAN_HW_TAGS
 	___GFP_SKIP_ZERO_BIT,
 	___GFP_SKIP_KASAN_BIT,
diff --git a/mm/internal.h b/mm/internal.h
index cb8d8e8e3ffa..122fce7e1a9e 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -1175,6 +1175,8 @@ unsigned int reclaim_clean_pages_from_list(struct zone *zone,
 #endif
 #define ALLOC_HIGHATOMIC	0x200 /* Allows access to MIGRATE_HIGHATOMIC */
 #define ALLOC_KSWAPD		0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
+#define __GFP_TRYLOCK		((__force gfp_t)BIT(___GFP_TRYLOCK_BIT))
+#define ALLOC_TRYLOCK		0x1000000 /* Only use spin_trylock in allocation path */
 
 /* Flags that allow allocations below the min watermark. */
 #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 1cb4b8c8886d..d23545057b6e 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2304,7 +2304,11 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 	unsigned long flags;
 	int i;
 
-	spin_lock_irqsave(&zone->lock, flags);
+	if (!spin_trylock_irqsave(&zone->lock, flags)) {
+		if (unlikely(alloc_flags & ALLOC_TRYLOCK))
+			return 0;
+		spin_lock_irqsave(&zone->lock, flags);
+	}
 	for (i = 0; i < count; ++i) {
 		struct page *page = __rmqueue(zone, order, migratetype,
 								alloc_flags);
@@ -2904,7 +2908,11 @@ struct page *rmqueue_buddy(struct zone *preferred_zone, struct zone *zone,
 
 	do {
 		page = NULL;
-		spin_lock_irqsave(&zone->lock, flags);
+		if (!spin_trylock_irqsave(&zone->lock, flags)) {
+			if (unlikely(alloc_flags & ALLOC_TRYLOCK))
+				return NULL;
+			spin_lock_irqsave(&zone->lock, flags);
+		}
 		if (alloc_flags & ALLOC_HIGHATOMIC)
 			page = __rmqueue_smallest(zone, order, MIGRATE_HIGHATOMIC);
 		if (!page) {
@@ -4001,6 +4009,7 @@ gfp_to_alloc_flags(gfp_t gfp_mask, unsigned int order)
 	 */
 	BUILD_BUG_ON(__GFP_HIGH != (__force gfp_t) ALLOC_MIN_RESERVE);
 	BUILD_BUG_ON(__GFP_KSWAPD_RECLAIM != (__force gfp_t) ALLOC_KSWAPD);
+	BUILD_BUG_ON(__GFP_TRYLOCK != (__force gfp_t) ALLOC_TRYLOCK);
 
 	/*
 	 * The caller may dip into page reserves a bit more if the caller
@@ -4009,7 +4018,7 @@ gfp_to_alloc_flags(gfp_t gfp_mask, unsigned int order)
 	 * set both ALLOC_NON_BLOCK and ALLOC_MIN_RESERVE(__GFP_HIGH).
 	 */
 	alloc_flags |= (__force int)
-		(gfp_mask & (__GFP_HIGH | __GFP_KSWAPD_RECLAIM));
+		(gfp_mask & (__GFP_HIGH | __GFP_KSWAPD_RECLAIM | __GFP_TRYLOCK));
 
 	if (!(gfp_mask & __GFP_DIRECT_RECLAIM)) {
 		/*
@@ -4509,7 +4518,8 @@ static inline bool prepare_alloc_pages(gfp_t gfp_mask, unsigned int order,
 
 	might_alloc(gfp_mask);
 
-	if (should_fail_alloc_page(gfp_mask, order))
+	if (!(*alloc_flags & ALLOC_TRYLOCK) &&
+	    should_fail_alloc_page(gfp_mask, order))
 		return false;
 
 	*alloc_flags = gfp_to_alloc_flags_cma(gfp_mask, *alloc_flags);
@@ -7023,3 +7033,54 @@ static bool __free_unaccepted(struct page *page)
 }
 
 #endif /* CONFIG_UNACCEPTED_MEMORY */
+
+struct page *try_alloc_pages_noprof(int nid, unsigned int order)
+{
+	gfp_t alloc_gfp = __GFP_NOWARN | __GFP_ZERO |
+			  __GFP_NOMEMALLOC | __GFP_TRYLOCK;
+	unsigned int alloc_flags = ALLOC_TRYLOCK;
+	struct alloc_context ac = { };
+	struct page *page;
+
+	/*
+	 * In RT spin_trylock() may call raw_spin_lock() which is unsafe in NMI.
+	 * If spin_trylock() is called from hard IRQ the current task may be
+	 * waiting for one rt_spin_lock, but rt_spin_trylock() will mark the
+	 * task as the owner of another rt_spin_lock which will confuse PI
+	 * logic, so return immediately if called form hard IRQ or NMI.
+	 *
+	 * Note, irqs_disabled() case is ok. This function can be called
+	 * from raw_spin_lock_irqsave region.
+	 */
+	if (IS_ENABLED(CONFIG_PREEMPT_RT) && (in_nmi() || in_hardirq()))
+		return NULL;
+	if (!pcp_allowed_order(order))
+		return NULL;
+
+#ifdef CONFIG_UNACCEPTED_MEMORY
+	if (has_unaccepted_memory() && !list_empty(&zone->unaccepted_pages))
+		return NULL;
+#endif
+
+	if (nid == NUMA_NO_NODE)
+		nid = numa_node_id();
+
+	prepare_alloc_pages(alloc_gfp, order, nid, NULL, &ac,
+			    &alloc_gfp, &alloc_flags);
+
+	/*
+	 * Best effort allocation from percpu free list.
+	 * If it's empty attempt to spin_trylock zone->lock.
+	 * Do not specify __GFP_KSWAPD_RECLAIM to avoid wakeup_kswapd
+	 * that may need to grab a lock.
+	 * Do not specify __GFP_ACCOUNT to avoid local_lock.
+	 * Do not warn either.
+	 */
+	page = get_page_from_freelist(alloc_gfp, order, alloc_flags, &ac);
+
+	/* Unlike regular alloc_pages() there is no __alloc_pages_slowpath(). */
+
+	trace_mm_page_alloc(page, order, alloc_gfp & ~__GFP_TRYLOCK, ac.migratetype);
+	kmsan_alloc_page(page, order, alloc_gfp);
+	return page;
+}
-- 
2.43.5