3 files changed, 119 insertions, 5 deletions
diff --git a/Documentation/bpf/bpf_design_QA.rst b/Documentation/bpf/bpf_design_QA.rst
index cb402c59eca5..12a246fcf6cb 100644
--- a/Documentation/bpf/bpf_design_QA.rst
+++ b/Documentation/bpf/bpf_design_QA.rst
@@ -172,11 +172,31 @@ registers which makes BPF inefficient virtual machine for 32-bit
 CPU architectures and 32-bit HW accelerators. Can true 32-bit registers
 be added to BPF in the future?
 
-A: NO. The first thing to improve performance on 32-bit archs is to teach
-LLVM to generate code that uses 32-bit subregisters. Then second step
-is to teach verifier to mark operations where zero-ing upper bits
-is unnecessary. Then JITs can take advantage of those markings and
-drastically reduce size of generated code and improve performance.
+A: NO.
+
+But some optimizations on zero-ing the upper 32 bits for BPF registers are
+available, and can be leveraged to improve the performance of JITed BPF
+programs for 32-bit architectures.
+
+Starting with version 7, LLVM is able to generate instructions that operate
+on 32-bit subregisters, provided the option -mattr=+alu32 is passed for
+compiling a program. Furthermore, the verifier can now mark the
+instructions for which zero-ing the upper bits of the destination register
+is required, and insert an explicit zero-extension (zext) instruction
+(a mov32 variant). This means that for architectures without zext hardware
+support, the JIT back-ends do not need to clear the upper bits for
+subregisters written by alu32 instructions or narrow loads. Instead, the
+back-ends simply need to support code generation for that mov32 variant,
+and to overwrite bpf_jit_needs_zext() to make it return "true" (in order to
+enable zext insertion in the verifier).
+
+Note that it is possible for a JIT back-end to have partial hardware
+support for zext. In that case, if verifier zext insertion is enabled,
+it could lead to the insertion of unnecessary zext instructions. Such
+instructions could be removed by creating a simple peephole inside the JIT
+back-end: if one instruction has hardware support for zext and if the next
+instruction is an explicit zext, then the latter can be skipped when doing
+the code generation.
 
 Q: Does BPF have a stable ABI?
 ------------------------------
diff --git a/Documentation/bpf/index.rst b/Documentation/bpf/index.rst
index d3fe4cac0c90..801a6ed3f2e5 100644
--- a/Documentation/bpf/index.rst
+++ b/Documentation/bpf/index.rst
@@ -42,6 +42,7 @@ Program types
 .. toctree::
    :maxdepth: 1
 
+   prog_cgroup_sockopt
    prog_cgroup_sysctl
    prog_flow_dissector
 
diff --git a/Documentation/bpf/prog_cgroup_sockopt.rst b/Documentation/bpf/prog_cgroup_sockopt.rst
new file mode 100644
index 000000000000..c47d974629ae
--- /dev/null
+++ b/Documentation/bpf/prog_cgroup_sockopt.rst
@@ -0,0 +1,93 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+============================
+BPF_PROG_TYPE_CGROUP_SOCKOPT
+============================
+
+``BPF_PROG_TYPE_CGROUP_SOCKOPT`` program type can be attached to two
+cgroup hooks:
+
+* ``BPF_CGROUP_GETSOCKOPT`` - called every time process executes ``getsockopt``
+  system call.
+* ``BPF_CGROUP_SETSOCKOPT`` - called every time process executes ``setsockopt``
+  system call.
+
+The context (``struct bpf_sockopt``) has associated socket (``sk``) and
+all input arguments: ``level``, ``optname``, ``optval`` and ``optlen``.
+
+BPF_CGROUP_SETSOCKOPT
+=====================
+
+``BPF_CGROUP_SETSOCKOPT`` is triggered *before* the kernel handling of
+sockopt and it has writable context: it can modify the supplied arguments
+before passing them down to the kernel. This hook has access to the cgroup
+and socket local storage.
+
+If BPF program sets ``optlen`` to -1, the control will be returned
+back to the userspace after all other BPF programs in the cgroup
+chain finish (i.e. kernel ``setsockopt`` handling will *not* be executed).
+
+Note, that ``optlen`` can not be increased beyond the user-supplied
+value. It can only be decreased or set to -1. Any other value will
+trigger ``EFAULT``.
+
+Return Type
+-----------
+
+* ``0`` - reject the syscall, ``EPERM`` will be returned to the userspace.
+* ``1`` - success, continue with next BPF program in the cgroup chain.
+
+BPF_CGROUP_GETSOCKOPT
+=====================
+
+``BPF_CGROUP_GETSOCKOPT`` is triggered *after* the kernel handing of
+sockopt. The BPF hook can observe ``optval``, ``optlen`` and ``retval``
+if it's interested in whatever kernel has returned. BPF hook can override
+the values above, adjust ``optlen`` and reset ``retval`` to 0. If ``optlen``
+has been increased above initial ``getsockopt`` value (i.e. userspace
+buffer is too small), ``EFAULT`` is returned.
+
+This hook has access to the cgroup and socket local storage.
+
+Note, that the only acceptable value to set to ``retval`` is 0 and the
+original value that the kernel returned. Any other value will trigger
+``EFAULT``.
+
+Return Type
+-----------
+
+* ``0`` - reject the syscall, ``EPERM`` will be returned to the userspace.
+* ``1`` - success: copy ``optval`` and ``optlen`` to userspace, return
+  ``retval`` from the syscall (note that this can be overwritten by
+  the BPF program from the parent cgroup).
+
+Cgroup Inheritance
+==================
+
+Suppose, there is the following cgroup hierarchy where each cgroup
+has ``BPF_CGROUP_GETSOCKOPT`` attached at each level with
+``BPF_F_ALLOW_MULTI`` flag::
+
+  A (root, parent)
+   \
+    B (child)
+
+When the application calls ``getsockopt`` syscall from the cgroup B,
+the programs are executed from the bottom up: B, A. First program
+(B) sees the result of kernel's ``getsockopt``. It can optionally
+adjust ``optval``, ``optlen`` and reset ``retval`` to 0. After that
+control will be passed to the second (A) program which will see the
+same context as B including any potential modifications.
+
+Same for ``BPF_CGROUP_SETSOCKOPT``: if the program is attached to
+A and B, the trigger order is B, then A. If B does any changes
+to the input arguments (``level``, ``optname``, ``optval``, ``optlen``),
+then the next program in the chain (A) will see those changes,
+*not* the original input ``setsockopt`` arguments. The potentially
+modified values will be then passed down to the kernel.
+
+Example
+=======
+
+See ``tools/testing/selftests/bpf/progs/sockopt_sk.c`` for an example
+of BPF program that handles socket options.