clang: loader: Allow user to override kernel version (iovisor#1895)

BCC currently requires exactly matching headers. Sometimes this is quite
inconvenient especially if the kernel version is only very slightly
different such as updates in a stable kernel. This patch gives the user
the flexibility to override the the LINUX_VERSION_CODE provided in the
linux kernel headers, so that the eBPF program may load. We also print a
message when this is done, so that the user is warned about the override
happening and that results may be unpredictable.

Also updated the docs.

Signed-off-by: Joel Fernandes <[email protected]>

docs/reference_guide.md

@@ -90,6 +90,10 @@ This guide is incomplete. If something feels missing, check the bcc and kernel s
  - [1. Invalid mem access](#1-invalid-mem-access)
  - [2. Cannot call GPL only function from proprietary program](#2-cannot-call-gpl-only-function-from-proprietary-program)
 
+- [Environment Variables](#envvars)
+ - [1. kernel source directory](#1-kernel-source-directory)
+ - [2. kernel version overriding](#2-kernel-version-overriding)
+
 # BPF C
 
 This section describes the C part of a bcc program.
@@ -1558,3 +1562,26 @@ bpf: Failed to load program: Invalid argument
 8: (85) call bpf_get_stackid#27
 cannot call GPL only function from proprietary program
 ```
+
+# Environment Variables
+
+## 1. Kernel source directory
+
+eBPF program compilation needs kernel sources or kernel headers with headers
+compiled. In case your kernel sources are at a non-standard location where BCC
+cannot find then, its possible to provide BCC the absolute path of the location
+by setting `BCC_KERNEL_SOURCE` to it.
+
+## 2. Kernel version overriding
+
+By default, BCC stores the `LINUX_VERSION_CODE` in the generated eBPF object
+which is then passed along to the kernel when the eBPF program is loaded.
+Sometimes this is quite inconvenient especially when the kernel is slightly
+updated such as an LTS kernel release. Its extremely unlikely the slight
+mismatch would cause any issues with the loaded eBPF program. By setting
+`BCC_LINUX_VERSION_CODE` to the version of the kernel that's running, the check
+for verifying the kernel version can be bypassed. This is needed for programs
+that use kprobes. This needs to be encoded in the format: `(VERSION * 65536) +
+(PATCHLEVEL * 256) + SUBLEVEL`. For example, if the running kernel is `4.9.10`,
+then can set `export BCC_LINUX_VERSION_CODE=264458` to override the kernel
+version check successfully.

src/cc/export/helpers.h

@@ -222,7 +222,11 @@ struct _name##_table_t _name = { .max_entries = (_max_entries) }
 #define cursor_advance(_cursor, _len) \
  ({ void *_tmp = _cursor; _cursor += _len; _tmp; })
 
+#ifdef LINUX_VERSION_CODE_OVERRIDE
+unsigned _version SEC("version") = LINUX_VERSION_CODE_OVERRIDE;
+#else
 unsigned _version SEC("version") = LINUX_VERSION_CODE;
+#endif
 
 /* helper functions called from eBPF programs written in C */
 static void *(*bpf_map_lookup_elem)(void *map, void *key) =

src/cc/frontends/clang/loader.cc

@@ -114,7 +114,9 @@ int ClangLoader::parse(unique_ptr<llvm::Module> *mod, TableStorage &ts,
  uname(&un);
  string kdir, kpath;
  const char *kpath_env = ::getenv("BCC_KERNEL_SOURCE");
+ const char *version_override = ::getenv("BCC_LINUX_VERSION_CODE");
  bool has_kpath_source = false;
+ string vmacro;
 
  if (kpath_env) {
  kpath = string(kpath_env);
@@ -175,6 +177,15 @@ int ClangLoader::parse(unique_ptr<llvm::Module> *mod, TableStorage &ts,
  flags_cstr.push_back(it->c_str());
 
  vector<const char *> flags_cstr_rem;
+
+ if (version_override) {
+ vmacro = "-DLINUX_VERSION_CODE_OVERRIDE=" + string(version_override);
+
+ std::cout << "WARNING: Linux version for eBPF program is being overridden with: " << version_override << "\n";
+ std::cout << "WARNING: Due to this, the results of the program may be unpredictable\n";
+ flags_cstr_rem.push_back(vmacro.c_str());
+ }
+
  flags_cstr_rem.push_back("-include");
  flags_cstr_rem.push_back("/virtual/include/bcc/helpers.h");
  flags_cstr_rem.push_back("-isystem");