[PATCH] bpf: fix a no-mmu build failure by providing a stub allocator

Commit fc9702273e2e ("bpf: Add mmap() support for BPF_MAP_TYPE_ARRAY")
added code that calls vmalloc_user_node_flags() and therefore requires
mm/vmalloc.c. However, that file is not built for the !CONFIG_MMU case.
This leads to a build failure when using ARM with the config provided
by at least one particular kbuild test robot report [1].

[1] https://lore/kernel.org/r/201911251639.UWS3hE3Y%lkp@intel.com

Fix the build by providing a stub function for __bpf_map_area_alloc().

Fixes: fc9702273e2e ("bpf: Add mmap() support for BPF_MAP_TYPE_ARRAY")
Reported-by: kbuild test robot <redacted>
Cc: Andrii Nakryiko <redacted>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Song Liu <redacted>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 kernel/bpf/syscall.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index e3461ec59570..cb3e13ee4123 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -128,6 +128,7 @@ static struct bpf_map *find_and_alloc_map(union bpf_attr *attr)
 	return map;
 }
 
+#ifdef CONFIG_MMU
 static void *__bpf_map_area_alloc(u64 size, int numa_node, bool mmapable)
 {
 	/* We really just want to fail instead of triggering OOM killer

@@ -162,6 +163,12 @@ static void *__bpf_map_area_alloc(u64 size, int numa_node, bool mmapable)
 					   GFP_KERNEL | __GFP_RETRY_MAYFAIL |
 					   flags, __builtin_return_address(0));
 }
+#else /* CONFIG_MMU */
+static void *__bpf_map_area_alloc(u64 size, int numa_node, bool mmapable)
+{
+	return NULL;
+}
+#endif /* !CONFIG_MMU */
 
 void *bpf_map_area_alloc(u64 size, int numa_node)
 {

-- 
2.24.0