bpf: avoid holding freeze_mutex during mmap operation
authorAndrii Nakryiko <andrii@kernel.org>
Wed, 29 Jan 2025 01:22:46 +0000 (17:22 -0800)
committerAlexei Starovoitov <ast@kernel.org>
Wed, 29 Jan 2025 17:49:50 +0000 (09:49 -0800)
commitbc27c52eea189e8f7492d40739b7746d67b65beb
tree3884a988f9f6f46626b75a8f2bd08c30dd11f590
parent98671a0fd1f14e4a518ee06b19037c20014900eb
bpf: avoid holding freeze_mutex during mmap operation

We use map->freeze_mutex to prevent races between map_freeze() and
memory mapping BPF map contents with writable permissions. The way we
naively do this means we'll hold freeze_mutex for entire duration of all
the mm and VMA manipulations, which is completely unnecessary. This can
potentially also lead to deadlocks, as reported by syzbot in [0].

So, instead, hold freeze_mutex only during writeability checks, bump
(proactively) "write active" count for the map, unlock the mutex and
proceed with mmap logic. And only if something went wrong during mmap
logic, then undo that "write active" counter increment.

  [0] https://lore.kernel.org/bpf/678dcbc9.050a0220.303755.0066.GAE@google.com/

Fixes: fc9702273e2e ("bpf: Add mmap() support for BPF_MAP_TYPE_ARRAY")
Reported-by: syzbot+4dc041c686b7c816a71e@syzkaller.appspotmail.com
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/r/20250129012246.1515826-2-andrii@kernel.org
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
kernel/bpf/syscall.c