bpf: Make non-preallocated allocation low priority

GFP_ATOMIC doesn't cooperate well with memcg pressure so far, especially if we allocate too much GFP_ATOMIC memory. For example, when we set the memcg limit to limit a non-preallocated bpf memory, the GFP_ATOMIC can easily break the memcg limit by force charge. So it is very dangerous to use GFP_ATOMIC in non-preallocated case. One way to make it safe is to remove __GFP_HIGH from GFP_ATOMIC, IOW, use (__GFP_ATOMIC | __GFP_KSWAPD_RECLAIM) instead, then it will be limited if we allocate too much memory. There's a plan to completely remove __GFP_ATOMIC in the mm side[1], so let's use GFP_NOWAIT instead. We introduced BPF_F_NO_PREALLOC is because full map pre-allocation is too memory expensive for some cases. That means removing __GFP_HIGH doesn't break the rule of BPF_F_NO_PREALLOC, but has the same goal with it-avoiding issues caused by too much memory. So let's remove it. This fix can also apply to other run-time allocations, for example, the allocation in lpm trie, local storage and devmap. So let fix it consistently over the bpf code It also fixes a typo in the comment. [1]. https://lore.kernel.org/linux-mm/163712397076.13692.4727608274002939094@noble.neil.brown.name/

Yafang Shao committed 4 years ago

Commit ace2bee1 parent 1d5f82d4 files · +6 −6Browse files

Unified · Split

4 files

kernel/bpf/devmap.c +1 −1
kernel/bpf/hashtab.c +3 −3
kernel/bpf/local_storage.c +1 −1
kernel/bpf/lpm_trie.c +1 −1

kernel/bpf/devmap.c+1−1

@@ -845,7 +845,7 @@ static struct bpf_dtab_netdev __dev_map_alloc_node(struct net net, · 844 unmodified lines
845	845	struct bpf_dtab_netdev *dev;
846	846
847	847	dev = bpf_map_kmalloc_node(&dtab->map, sizeof(*dev),
848		− GFP_ATOMIC \| __GFP_NOWARN,
	848	+ GFP_NOWAIT \| __GFP_NOWARN,
849	849	dtab->map.numa_node);
850	850	if (!dev)
851	851	return ERR_PTR(-ENOMEM);

kernel/bpf/hashtab.c+3−3

@@ -61,7 +61,7 @@ · 60 unmodified lines
61	61	*
62	62	* As regular device interrupt handlers and soft interrupts are forced into
63	63	* thread context, the existing code which does
64		− * spin_lock(); alloc(GPF_ATOMIC); spin_unlock();
	64	+ * spin_lock(); alloc(GFP_ATOMIC); spin_unlock();
65	65	* just works.
66	66	*
67	67	* In theory the BPF locks could be converted to regular spinlocks as well,
@@ -978,7 +978,7 @@ static struct htab_elem alloc_htab_elem(struct bpf_htab htab, void *key, · 910 unmodified lines
978	978	goto dec_count;
979	979	}
980	980	l_new = bpf_map_kmalloc_node(&htab->map, htab->elem_size,
981		− GFP_ATOMIC \| __GFP_NOWARN,
	981	+ GFP_NOWAIT \| __GFP_NOWARN,
982	982	htab->map.numa_node);
983	983	if (!l_new) {
984	984	l_new = ERR_PTR(-ENOMEM);
@@ -996,7 +996,7 @@ static struct htab_elem alloc_htab_elem(struct bpf_htab htab, void *key, · 11 unmodified lines
996	996	} else {
997	997	/* alloc_percpu zero-fills */
998	998	pptr = bpf_map_alloc_percpu(&htab->map, size, 8,
999		− GFP_ATOMIC \| __GFP_NOWARN);
	999	+ GFP_NOWAIT \| __GFP_NOWARN);
1000	1000	if (!pptr) {
1001	1001	kfree(l_new);
1002	1002	l_new = ERR_PTR(-ENOMEM);

kernel/bpf/local_storage.c+1−1

@@ -165,7 +165,7 @@ static int cgroup_storage_update_elem(struct bpf_map map, void key, · 164 unmodified lines
165	165	}
166	166
167	167	new = bpf_map_kmalloc_node(map, struct_size(new, data, map->value_size),
168		− __GFP_ZERO \| GFP_ATOMIC \| __GFP_NOWARN,
	168	+ __GFP_ZERO \| GFP_NOWAIT \| __GFP_NOWARN,
169	169	map->numa_node);
170	170	if (!new)
171	171	return -ENOMEM;

kernel/bpf/lpm_trie.c+1−1

@@ -285,7 +285,7 @@ static struct lpm_trie_node lpm_trie_node_alloc(const struct lpm_trie trie, · 284 unmodified lines
285	285	if (value)
286	286	size += trie->map.value_size;
287	287
288		− node = bpf_map_kmalloc_node(&trie->map, size, GFP_ATOMIC \| __GFP_NOWARN,
	288	+ node = bpf_map_kmalloc_node(&trie->map, size, GFP_NOWAIT \| __GFP_NOWARN,
289	289	trie->map.numa_node);
290	290	if (!node)
291	291	return NULL;

bpf: Make non-preallocated allocation low priority

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code