Merge branch 'Atomics for eBPF'

Brendan Jackman says:

====================

There's still one unresolved review comment from John[3] which I
will resolve with a followup patch.

Differences from v6->v7 [1]:

* Fixed riscv build error detected by 0-day robot.

Differences from v5->v6 [1]:

* Carried Björn Töpel's ack for RISC-V code, plus a couple more acks from
  Yonhgong.

* Doc fixups.

* Trivial cleanups.

Differences from v4->v5 [1]:

* Fixed bogus type casts in interpreter that led to warnings from
  the 0day robot.

* Dropped feature-detection for Clang per Andrii's suggestion in [4].
  The selftests will now fail to build unless you have llvm-project
  commit 286daafd6512. The ENABLE_ATOMICS_TEST macro is still needed
  to support the no_alu32 tests.

* Carried some Acks from John and Yonghong.

* Dropped confusing usage of __atomic_exchange from prog_test in
  favour of __sync_lock_test_and_set.

* [Really] got rid of all the forest of instruction macros
  (BPF_ATOMIC_FETCH_ADD and friends); now there's just BPF_ATOMIC_OP
  to define all the instructions as we use them in the verifier
  tests. This makes the atomic ops less special in that API, and I
  don't think the resulting usage is actually any harder to read.

Differences from v3->v4 [1]:

* Added one Ack from Yonghong. He acked some other patches but those
  have now changed non-trivally so I didn't add those acks.

* Fixups to commit messages.

* Fixed disassembly and comments: first arg to atomic_fetch_* is a
  pointer.

* Improved prog_test efficiency. BPF progs are now all loaded in a
  single call, then the skeleton is re-used for each subtest.

* Dropped use of tools/build/feature in favour of a one-liner in the
  Makefile.

* Dropped the commit that created an emit_neg helper in the x86
  JIT. It's not used any more (it wasn't used in v3 either).

* Combined all the different filter.h macros (used to be
  BPF_ATOMIC_ADD, BPF_ATOMIC_FETCH_ADD, BPF_ATOMIC_AND, etc) into
  just BPF_ATOMIC32 and BPF_ATOMIC64.

* Removed some references to BPF_STX_XADD from tools/, samples/ and
  lib/ that I missed before.

Differences from v2->v3 [1]:

* More minor fixes and naming/comment changes

* Dropped atomic subtract: compilers can implement this by preceding
  an atomic add with a NEG instruction (which is what the x86 JIT did
  under the hood anyway).

* Dropped the use of -mcpu=v4 in the Clang BPF command-line; there is
  no longer an architecture version bump. Instead a feature test is
  added to Kbuild - it builds a source file to check if Clang
  supports BPF atomics.

* Fixed the prog_test so it no longer breaks
  test_progs-no_alu32. This requires some ifdef acrobatics to avoid
  complicating the prog_tests model where the same userspace code
  exercises both the normal and no_alu32 BPF test objects, using the
  same skeleton header.

Differences from v1->v2 [1]:

* Fixed mistakes in the netronome driver

* Addd sub, add, or, xor operations

* The above led to some refactors to keep things readable. (Maybe I
  should have just waited until I'd implemented these before starting
  the review...)

* Replaced BPF_[CMP]SET | BPF_FETCH with just BPF_[CMP]XCHG, which
  include the BPF_FETCH flag

* Added a bit of documentation. Suggestions welcome for more places
  to dump this info...

The prog_test that's added depends on Clang/LLVM features added by
Yonghong in commit 286daafd6512 (was
https://reviews.llvm.org/D72184).

This only includes a JIT implementation for x86_64 - I don't plan to
implement JIT support myself for other architectures.

Operations
==========

This patchset adds atomic operations to the eBPF instruction set. The
use-case that motivated this work was a trivial and efficient way to
generate globally-unique cookies in BPF progs, but I think it's
obvious that these features are pretty widely applicable.  The
instructions that are added here can be summarised with this list of
kernel operations:

* atomic[64]_[fetch_]add
* atomic[64]_[fetch_]and
* atomic[64]_[fetch_]or
* atomic[64]_xchg
* atomic[64]_cmpxchg

The following are left out of scope for this effort:

* 16 and 8 bit operations
* Explicit memory barriers

Encoding
========

I originally planned to add new values for bpf_insn.opcode. This was
rather unpleasant: the opcode space has holes in it but no entire
instruction classes[2]. Yonghong Song had a better idea: use the
immediate field of the existing STX XADD instruction to encode the
operation. This works nicely, without breaking existing programs,
because the immediate field is currently reserved-must-be-zero, and
extra-nicely because BPF_ADD happens to be zero.

Note that this of course makes immediate-source atomic operations
impossible. It's hard to imagine a measurable speedup from such
instructions, and if it existed it would certainly not benefit x86,
which has no support for them.

The BPF_OP opcode fields are re-used in the immediate, and an
additional flag BPF_FETCH is used to mark instructions that should
fetch a pre-modification value from memory.

So, BPF_XADD is now called BPF_ATOMIC (the old name is kept to avoid
breaking userspace builds), and where we previously had .imm = 0, we
now have .imm = BPF_ADD (which is 0).

Operands
========

Reg-source eBPF instructions only have two operands, while these
atomic operations have up to four. To avoid needing to encode
additional operands, then:

- One of the input registers is re-used as an output register
  (e.g. atomic_fetch_add both reads from and writes to the source
  register).

- Where necessary (i.e. for cmpxchg) , R0 is "hard-coded" as one of
  the operands.

This approach also allows the new eBPF instructions to map directly
to single x86 instructions.

[1] Previous iterations:
    v1: https://lore.kernel.org/bpf/20201123173202.1335708-1-jackmanb@google.com/
    v2: https://lore.kernel.org/bpf/20201127175738.1085417-1-jackmanb@google.com/
    v3: https://lore.kernel.org/bpf/X8kN7NA7bJC7aLQI@google.com/
    v4: https://lore.kernel.org/bpf/20201207160734.2345502-1-jackmanb@google.com/
    v5: https://lore.kernel.org/bpf/20201215121816.1048557-1-jackmanb@google.com/
    v6: https://lore.kernel.org/bpf/20210112154235.2192781-1-jackmanb@google.com/

[2] Visualisation of eBPF opcode space:
    https://gist.github.com/bjackman/00fdad2d5dfff601c1918bc29b16e778

[3] Comment from John about propagating bounds in verifier:
    https://lore.kernel.org/bpf/5fcf0fbcc8aa8_9ab320853@john-XPS-13-9370.notmuch/

[4] Mail from Andrii about not supporting old Clang in selftests:
    https://lore.kernel.org/bpf/CAEf4BzYBddPaEzRUs=jaWSo5kbf=LZdb7geAUVj85GxLQztuAQ@mail.gmail.com/
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>