[v4,bpf-next] bpf: select proper size for bpf_prog_pack

Commit Message

Song Liu March 11, 2022, 8:11 p.m. UTC

Using HPAGE_PMD_SIZE as the size for bpf_prog_pack is not ideal in some
cases. Specifically, for NUMA systems, __vmalloc_node_range requires
PMD_SIZE * num_online_nodes() to allocate huge pages. Also, if the system
does not support huge pages (i.e., with cmdline option nohugevmalloc), it
is better to use PAGE_SIZE packs.

Add logic to select proper size for bpf_prog_pack. This solution is not
ideal, as it makes assumption about the behavior of module_alloc and
__vmalloc_node_range. However, it appears to be the easiest solution as
it doesn't require changes in module_alloc and vmalloc code.

Fixes: 57631054fae6 ("bpf: Introduce bpf_prog_pack allocator")
Signed-off-by: Song Liu <[e-mail]>

---
Changes v3 => v4:
1. Fix a race condition reported by kernel test robot
   <[e-mail]>.

Changes v2 => v3:
1. Remove a leftover debug change.
2. Shuffle alloc_new_pack() for cleaner patch.

Changes v1 => v2:
1. Fix case with first program > PAGE_SIZE. (Daniel)
2. Add Fixes tag.
3. Remove a inline to avoid netdev/source_inline error.
---
 kernel/bpf/core.c | 70 +++++++++++++++++++++++++++++++----------------
 1 file changed, 47 insertions(+), 23 deletions(-)

Comments

[e-mail] March 21, 2022, 2:20 a.m. UTC | #1

Hello:

This patch was applied to bpf/bpf-next.git (master)
by Alexei Starovoitov <[e-mail]>:

On Fri, 11 Mar 2022 12:11:35 -0800 you wrote:
> Using HPAGE_PMD_SIZE as the size for bpf_prog_pack is not ideal in some
> cases. Specifically, for NUMA systems, __vmalloc_node_range requires
> PMD_SIZE * num_online_nodes() to allocate huge pages. Also, if the system
> does not support huge pages (i.e., with cmdline option nohugevmalloc), it
> is better to use PAGE_SIZE packs.
> 
> Add logic to select proper size for bpf_prog_pack. This solution is not
> ideal, as it makes assumption about the behavior of module_alloc and
> __vmalloc_node_range. However, it appears to be the easiest solution as
> it doesn't require changes in module_alloc and vmalloc code.
> 
> [...]

Here is the summary with links:
  - [v4,bpf-next] bpf: select proper size for bpf_prog_pack
    https://git.kernel.org/bpf/bpf-next/c/ef078600eec2

You are awesome, thank you!

Patch

diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index ab630f773ec1..9d0a56aa7c90 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -33,6 +33,7 @@ 
 #include <linux/extable.h>
 #include <linux/log2.h>
 #include <linux/bpf_verifier.h>
+#include <linux/nodemask.h>
 
 #include <asm/barrier.h>
 #include <asm/unaligned.h>
@@ -814,15 +815,9 @@  int bpf_jit_add_poke_descriptor(struct bpf_prog *prog,
  * allocator. The prog_pack allocator uses HPAGE_PMD_SIZE page (2MB on x86)
  * to host BPF programs.
  */
-#ifdef CONFIG_TRANSPARENT_HUGEPAGE
-#define BPF_PROG_PACK_SIZE	HPAGE_PMD_SIZE
-#else
-#define BPF_PROG_PACK_SIZE	PAGE_SIZE
-#endif
 #define BPF_PROG_CHUNK_SHIFT	6
 #define BPF_PROG_CHUNK_SIZE	(1 << BPF_PROG_CHUNK_SHIFT)
 #define BPF_PROG_CHUNK_MASK	(~(BPF_PROG_CHUNK_SIZE - 1))
-#define BPF_PROG_CHUNK_COUNT	(BPF_PROG_PACK_SIZE / BPF_PROG_CHUNK_SIZE)
 
 struct bpf_prog_pack {
 	struct list_head list;
@@ -830,30 +825,56 @@  struct bpf_prog_pack {
 	unsigned long bitmap[];
 };
 
-#define BPF_PROG_MAX_PACK_PROG_SIZE	BPF_PROG_PACK_SIZE
 #define BPF_PROG_SIZE_TO_NBITS(size)	(round_up(size, BPF_PROG_CHUNK_SIZE) / BPF_PROG_CHUNK_SIZE)
 
+static size_t bpf_prog_pack_size = -1;
+
+static int bpf_prog_chunk_count(void)
+{
+	WARN_ON_ONCE(bpf_prog_pack_size == -1);
+	return bpf_prog_pack_size / BPF_PROG_CHUNK_SIZE;
+}
+
 static DEFINE_MUTEX(pack_mutex);
 static LIST_HEAD(pack_list);
 
+static size_t select_bpf_prog_pack_size(void)
+{
+	size_t size;
+	void *ptr;
+
+	size = PMD_SIZE * num_online_nodes();
+	ptr = module_alloc(size);
+
+	/* Test whether we can get huge pages. If not just use PAGE_SIZE
+	 * packs.
+	 */
+	if (!ptr || !is_vm_area_hugepages(ptr))
+		size = PAGE_SIZE;
+
+	vfree(ptr);
+	return size;
+}
+
 static struct bpf_prog_pack *alloc_new_pack(void)
 {
 	struct bpf_prog_pack *pack;
 
-	pack = kzalloc(sizeof(*pack) + BITS_TO_BYTES(BPF_PROG_CHUNK_COUNT), GFP_KERNEL);
+	pack = kzalloc(struct_size(pack, bitmap, BITS_TO_LONGS(bpf_prog_chunk_count())),
+		       GFP_KERNEL);
 	if (!pack)
 		return NULL;
-	pack->ptr = module_alloc(BPF_PROG_PACK_SIZE);
+	pack->ptr = module_alloc(bpf_prog_pack_size);
 	if (!pack->ptr) {
 		kfree(pack);
 		return NULL;
 	}
-	bitmap_zero(pack->bitmap, BPF_PROG_PACK_SIZE / BPF_PROG_CHUNK_SIZE);
+	bitmap_zero(pack->bitmap, bpf_prog_pack_size / BPF_PROG_CHUNK_SIZE);
 	list_add_tail(&pack->list, &pack_list);
 
 	set_vm_flush_reset_perms(pack->ptr);
-	set_memory_ro((unsigned long)pack->ptr, BPF_PROG_PACK_SIZE / PAGE_SIZE);
-	set_memory_x((unsigned long)pack->ptr, BPF_PROG_PACK_SIZE / PAGE_SIZE);
+	set_memory_ro((unsigned long)pack->ptr, bpf_prog_pack_size / PAGE_SIZE);
+	set_memory_x((unsigned long)pack->ptr, bpf_prog_pack_size / PAGE_SIZE);
 	return pack;
 }
 
@@ -864,7 +885,11 @@  static void *bpf_prog_pack_alloc(u32 size)
 	unsigned long pos;
 	void *ptr = NULL;
 
-	if (size > BPF_PROG_MAX_PACK_PROG_SIZE) {
+	mutex_lock(&pack_mutex);
+	if (bpf_prog_pack_size == -1)
+		bpf_prog_pack_size = select_bpf_prog_pack_size();
+
+	if (size > bpf_prog_pack_size) {
 		size = round_up(size, PAGE_SIZE);
 		ptr = module_alloc(size);
 		if (ptr) {
@@ -872,13 +897,12 @@  static void *bpf_prog_pack_alloc(u32 size)
 			set_memory_ro((unsigned long)ptr, size / PAGE_SIZE);
 			set_memory_x((unsigned long)ptr, size / PAGE_SIZE);
 		}
-		return ptr;
+		goto out;
 	}
-	mutex_lock(&pack_mutex);
 	list_for_each_entry(pack, &pack_list, list) {
-		pos = bitmap_find_next_zero_area(pack->bitmap, BPF_PROG_CHUNK_COUNT, 0,
+		pos = bitmap_find_next_zero_area(pack->bitmap, bpf_prog_chunk_count(), 0,
 						 nbits, 0);
-		if (pos < BPF_PROG_CHUNK_COUNT)
+		if (pos < bpf_prog_chunk_count())
 			goto found_free_area;
 	}
 
@@ -904,13 +928,13 @@  static void bpf_prog_pack_free(struct bpf_binary_header *hdr)
 	unsigned long pos;
 	void *pack_ptr;
 
-	if (hdr->size > BPF_PROG_MAX_PACK_PROG_SIZE) {
+	mutex_lock(&pack_mutex);
+	if (hdr->size > bpf_prog_pack_size) {
 		module_memfree(hdr);
-		return;
+		goto out;
 	}
 
-	pack_ptr = (void *)((unsigned long)hdr & ~(BPF_PROG_PACK_SIZE - 1));
-	mutex_lock(&pack_mutex);
+	pack_ptr = (void *)((unsigned long)hdr & ~(bpf_prog_pack_size - 1));
 
 	list_for_each_entry(tmp, &pack_list, list) {
 		if (tmp->ptr == pack_ptr) {
@@ -926,8 +950,8 @@  static void bpf_prog_pack_free(struct bpf_binary_header *hdr)
 	pos = ((unsigned long)hdr - (unsigned long)pack_ptr) >> BPF_PROG_CHUNK_SHIFT;
 
 	bitmap_clear(pack->bitmap, pos, nbits);
-	if (bitmap_find_next_zero_area(pack->bitmap, BPF_PROG_CHUNK_COUNT, 0,
-				       BPF_PROG_CHUNK_COUNT, 0) == 0) {
+	if (bitmap_find_next_zero_area(pack->bitmap, bpf_prog_chunk_count(), 0,
+				       bpf_prog_chunk_count(), 0) == 0) {
 		list_del(&pack->list);
 		module_memfree(pack->ptr);
 		kfree(pack);