[PATCH 0/2] rust: sync: lock: allow fallible initializers in Lock::new

[Ivan Tverdokhleb <mbkchelofficialdev@gmail.com>]

GpuBuddyInner keeps the C gpu_buddy structure next to a separate
Mutex<()> and has a TODO to move the structure inside the mutex once
Mutex::new() accepts a fallible initializer.

Patch 1 makes Lock::new() generic over the error type of the initializer
for the protected data, so a fallible initializer such as
Opaque::try_ffi_init() can be wrapped in a Mutex or SpinLock.

Patch 2 uses it to put the gpu_buddy allocator directly inside the Mutex
in GPU buddy, so that the allocator can only be reached through the lock
guard, and resolves the TODO.

Tested by building the rust/kernel crate (x86_64 defconfig with
CONFIG_RUST=y and CONFIG_GPU_BUDDY=y, built with LLVM=1). Not tested at
runtime.

Ivan Tverdokhleb (2):
  rust: sync: lock: allow fallible initializers in Lock::new
  rust: gpu: buddy: store the allocator inside the Mutex

 rust/kernel/gpu/buddy.rs | 38 +++++++++++++++-----------------------
 rust/kernel/sync/lock.rs |  8 ++++----
 2 files changed, 19 insertions(+), 27 deletions(-)

-- 
2.56.1