Introduce getenv to LLVM libc

[rtenneti]

Summary

Add support for getenv as defined by the Open Group's "System Interface &
Header" in https://pubs.opengroup.org/onlinepubs/7908799/xsh/getenv.html

getenv requires a standard way of accessing the environment,
so a pointer to the environment is added to the startup in crt1.
Consquently, this function is not usable on top of other libcs.

Added starts_with method to StringView. getenv function uses it.

Co-authored-by: Jeff Bailey <[e-mail]>

[sivachandra]

Overall LGTM but I will make a fine grained pass in a day and get back to you. Thanks for working on this.

[rtenneti]

Added Stdlib to posix.td to fix integration tests.
Added few more test cases.

[sivachandra]

libc/loader/linux/x86_64/start.cpp

32

Nit: Now that you have declared app as an extern, we can remove this TODO.

libc/src/stdlib/getenv.cpp

33

Nit: I didn't understand this comment.

[jeffbailey]

libc/src/stdlib/getenv.cpp

33

This comment was my reminder to myself that starts_with could match stems as well as entire variables and that I needed to make sure that the equals sign came next.

[rtenneti]

Fixed comments.

[rtenneti]

Fixed comments.

[rtenneti]

Fixed comments.

[rtenneti]

Fixed comments

libc/config/linux/app.h

 
 // Data structure which captures properties of a linux application.
 struct AppProperties {
   // Page size used for the application.
   uintptr_t pageSize;
 
   // The properties of an application's TLS.
   TLS tls;
+
+  // Environment data.
+  uint64_t *envPtr;
 };
 
+extern AppProperties app;
+
 // Creates and initializes the TLS area for the current thread. Should not
 // be called before app.tls has been initialized.
 void initTLS();
 
 } // namespace __llvm_libc
 
 #endif // LLVM_LIBC_CONFIG_LINUX_APP_H

libc/config/linux/x86_64/entrypoints.txt

set(TARGET_LIBC_ENTRYPOINTS
     # stdlib.h entrypoints
     libc.src.stdlib.abs
     libc.src.stdlib.atoi
     libc.src.stdlib.atof
     libc.src.stdlib.atol
     libc.src.stdlib.atoll
     libc.src.stdlib.bsearch
     libc.src.stdlib.div
+    libc.src.stdlib.getenv
     libc.src.stdlib.labs
     libc.src.stdlib.ldiv
     libc.src.stdlib.llabs
     libc.src.stdlib.lldiv
     libc.src.stdlib.qsort
     libc.src.stdlib.strtod
     libc.src.stdlib.strtof
     libc.src.stdlib.strtol

libc/loader/linux/x86_64/start.cpp

 #ifdef SYS_mmap2
 static constexpr long mmapSyscallNumber = SYS_mmap2;
 #elif SYS_mmap
 static constexpr long mmapSyscallNumber = SYS_mmap;
 #else
 #error "Target platform does not have SYS_mmap or SYS_mmap2 defined"
 #endif
 
 // TODO: Declare var an extern var in config/linux/app.h so that other
 // libc functions can make use of the application wide information. For
 // example, mmap can pick up the page size from here.
 AppProperties app;
 
 // TODO: The function is x86_64 specific. Move it to config/linux/app.h
 // and generalize it. Also, dynamic loading is not handled currently.
 void initTLS() {
   if (app.tls.size == 0)
extern "C" void _start() {
   // "+ 1" below) to get to the args.
   Args *args = reinterpret_cast<Args *>(frame_ptr + 1);
 
   // After the argv array, is a 8-byte long NULL value before the array of env
   // values. The end of the env values is marked by another 8-byte long NULL
   // value. We step over it (the "+ 1" below) to get to the env values.
   uint64_t *env_ptr = args->argv + args->argc + 1;
   uint64_t *env_end_marker = env_ptr;
+  app.envPtr = env_ptr;
   while (*env_end_marker)
     ++env_end_marker;
 
   // After the env array, is the aux-vector. The end of the aux-vector is
   // denoted by an AT_NULL entry.
   Elf64_Phdr *programHdrTable = nullptr;
   uintptr_t programHdrCount;
   for (AuxEntry *aux_entry = reinterpret_cast<AuxEntry *>(env_end_marker + 1);

libc/spec/posix.td

[
         FunctionSpec<
           "write",
           RetValSpec<SSizeTType>,
           [ArgSpec<IntType>, ArgSpec<ConstVoidPtr>, ArgSpec<SizeTType>]
         >,
     ]
   >;
   
+  HeaderSpec StdLib = HeaderSpec<
+    "stdlib.h",
+    [], // Macros
+    [], // Types
+    [], // Enumerations
+    [
+        FunctionSpec<
+          "getenv",
+          RetValSpec<CharPtr>,
+          [ArgSpec<ConstCharPtr>]
+        >,
+    ]
+  >;
+
   HeaderSpec String = HeaderSpec<
     "string.h",
     [
         Macro<"NULL">,
     ],
     [
         SizeTType,
     ],
HeaderSpec SysStat = HeaderSpec<
     ]
   >;
 
   let Headers = [
     CType,
     Errno,
     FCntl,
     Signal,
+    StdLib,
     SysMMan,
     SysStat,
     UniStd,
     String
   ];
 }

libc/src/__support/CPP/StringView.h

StringView trim(char C) const {
     for (; SuffixLen < Len; --NewEnd, ++SuffixLen) {
       if (*NewEnd != C)
         break;
     }
 
     return remove_prefix(PrefixLen).remove_suffix(SuffixLen);
   }
 
+  // Check if this string starts with the given \p Prefix.
+  bool starts_with(StringView Prefix) const {
+    if (Len < Prefix.Len)
+      return false;
+    for (size_t I = 0; I < Prefix.Len; ++I) {
+      if (Data[I] != Prefix.Data[I])
+        return false;
+    }
+    return true;
+  }
+
   // An equivalent method is not available in std::string_view.
   bool equals(StringView Other) const {
     if (Len != Other.Len)
       return false;
     for (size_t I = 0; I < Len; ++I) {
       if (Data[I] != Other.Data[I])
         return false;
     }
     return true;
   }
 };
 
 } // namespace cpp
 } // namespace __llvm_libc
 
 #endif // LLVM_LIBC_SRC_SUPPORT_CPP_STRINGVIEW_H

libc/src/stdlib/CMakeLists.txt

SRCS
     atoll.cpp
   HDRS
     atoll.h
   DEPENDS
     libc.src.__support.str_to_integer
 )
 
 add_entrypoint_object(
+  getenv
+  SRCS
+    getenv.cpp
+  HDRS
+    getenv.h
+  DEPENDS
+    libc.config.linux.app_h
+    libc.src.string.strncmp
+)
+
+add_entrypoint_object(
   strtof
   SRCS
     strtof.cpp
   HDRS
     strtof.h
   DEPENDS
     libc.src.__support.str_to_float
 )

libc/src/stdlib/getenv.h

+//===-- Implementation header for getenv --------------------------------*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_STDLIB_GETENV_H
+#define LLVM_LIBC_SRC_STDLIB_GETENV_H
+
+namespace __llvm_libc {
+
+char *getenv(const char *name);
+
+} // namespace __llvm_libc
+
+#endif // LLVM_LIBC_SRC_STDLIB_GETENV_H

libc/src/stdlib/getenv.cpp

+//===-- Implementation of getenv ------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/stdlib/getenv.h"
+#include "config/linux/app.h"
+#include "src/__support/CPP/StringView.h"
+#include "src/__support/common.h"
+
+#include <stddef.h> // For size_t.
+
+namespace __llvm_libc {
+
+LLVM_LIBC_FUNCTION(char *, getenv, (const char *name)) {
+  char **env_ptr = reinterpret_cast<char **>(__llvm_libc::app.envPtr);
+
+  if (name == nullptr || env_ptr == nullptr)
+    return nullptr;
+
+  __llvm_libc::cpp::StringView env_var_name(name);
+  if (env_var_name.size() == 0)
+    return nullptr;
+  for (char **env = env_ptr; *env != nullptr; env++) {
+    __llvm_libc::cpp::StringView cur(*env);
+    if (!cur.starts_with(env_var_name))
+      continue;
+
+    // Check for PATH=, vs PATH1=, PATH2=, etc...
+    if (cur[env_var_name.size()] != '=')
+      continue;
+
+    return const_cast<char *>(
+        cur.remove_prefix(env_var_name.size() + 1).data());
+  }
+
+  return nullptr;
+}
+
+} // namespace __llvm_libc

libc/test/loader/linux/CMakeLists.txt

add_loader_test(
   loader_no_args_test
   SRC
     main_without_args.cpp
   DEPENDS
     .loader_test
     libc.loader.linux.crt1
 )
 
+add_loader_test(
+  getenv_test
+  SRC
+    getenv_test.cpp
+  DEPENDS
+    .loader_test
+    libc.loader.linux.crt1
+    libc.src.stdlib.getenv
+  ENV
+    FRANCE=Paris
+    GERMANY=Berlin
+)
+
+
 # TODO: Disableing this test temporarily to investigate why gold fails to link
 # and produce an executable for this test. Test works all fine with ld.bfd.
 #add_loader_test(
 #  loader_tls_test
 #  SRC
 #    tls_test.cpp
 #  DEPENDS
 #    libc.config.linux.app_h
 #    libc.include.errno
 #    libc.include.sys_mman
 #    libc.loader.linux.crt1
 #    libc.src.assert.__assert_fail
 #    libc.src.errno.__errno_location
 #    libc.src.sys.mman.mmap
 #)

libc/test/loader/linux/getenv_test.cpp

+//===-- Unittests for getenv ----------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "loader_test.h"
+#include "src/stdlib/getenv.h"
+
+static bool my_streq(const char *lhs, const char *rhs) {
+  if (lhs == rhs)
+    return true;
+  if (((lhs == static_cast<char *>(nullptr)) &&
+       (rhs != static_cast<char *>(nullptr))) ||
+      ((lhs != static_cast<char *>(nullptr)) &&
+       (rhs == static_cast<char *>(nullptr)))) {
+    return false;
+  }
+  const char *l, *r;
+  for (l = lhs, r = rhs; *l != '\0' && *r != '\0'; ++l, ++r)
+    if (*l != *r)
+      return false;
+
+  return *l == '\0' && *r == '\0';
+}
+
+int main(int argc, char **argv, char **envp) {
+  ASSERT_TRUE(my_streq(__llvm_libc::getenv(""), static_cast<char *>(nullptr)));
+  ASSERT_TRUE(my_streq(__llvm_libc::getenv("="), static_cast<char *>(nullptr)));
+  ASSERT_TRUE(my_streq(__llvm_libc::getenv("MISSING ENV VARIABLE"),
+                       static_cast<char *>(nullptr)));
+  ASSERT_FALSE(
+      my_streq(__llvm_libc::getenv("PATH"), static_cast<char *>(nullptr)));
+  ASSERT_TRUE(my_streq(__llvm_libc::getenv("FRANCE"), "Paris"));
+  ASSERT_FALSE(my_streq(__llvm_libc::getenv("FRANCE"), "Berlin"));
+  ASSERT_TRUE(my_streq(__llvm_libc::getenv("GERMANY"), "Berlin"));
+  ASSERT_TRUE(
+      my_streq(__llvm_libc::getenv("FRANC"), static_cast<char *>(nullptr)));
+  ASSERT_TRUE(
+      my_streq(__llvm_libc::getenv("FRANCE1"), static_cast<char *>(nullptr)));
+
+  return 0;
+}

libc/test/loader/linux/loader_test.h

 #define __CHECK(file, line, val, should_exit)                                  \
   if (!(val)) {                                                                \
     __llvm_libc::write_to_stderr(file ":" __AS_STRING(                         \
         line) ": Expected '" #val "' to be true, but is false\n");             \
     if (should_exit)                                                           \
       __llvm_libc::quick_exit(127);                                            \
   }
 
+#define __CHECK_NE(file, line, val, should_exit)                               \
+  if ((val)) {                                                                 \
+    __llvm_libc::write_to_stderr(file ":" __AS_STRING(                         \
+        line) ": Expected '" #val "' to be false, but is true\n");             \
+    if (should_exit)                                                           \
+      __llvm_libc::quick_exit(127);                                            \
+  }
+
 #define EXPECT_TRUE(val) __CHECK(__FILE__, __LINE__, val, false)
 #define ASSERT_TRUE(val) __CHECK(__FILE__, __LINE__, val, true)
+#define EXPECT_FALSE(val) __CHECK_NE(__FILE__, __LINE__, val, false)
+#define ASSERT_FALSE(val) __CHECK_NE(__FILE__, __LINE__, val, true)
 
 #endif // LLVM_LIBC_TEST_LOADER_LINUX_LOADER_TEST_H

libc/test/utils/CPP/stringview_test.cpp

TEST(LlvmLibcStringViewTest, Equals) {
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView()));
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView("")));
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView("123")));
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView("abd")));
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView("aaa")));
   ASSERT_FALSE(v.equals(__llvm_libc::cpp::StringView("abcde")));
 }
 
+TEST(LlvmLibcStringViewTest, startsWith) {
+  __llvm_libc::cpp::StringView v("abc");
+  ASSERT_TRUE(v.starts_with(__llvm_libc::cpp::StringView("a")));
+  ASSERT_TRUE(v.starts_with(__llvm_libc::cpp::StringView("ab")));
+  ASSERT_TRUE(v.starts_with(__llvm_libc::cpp::StringView("abc")));
+  ASSERT_TRUE(v.starts_with(__llvm_libc::cpp::StringView()));
+  ASSERT_TRUE(v.starts_with(__llvm_libc::cpp::StringView("")));
+  ASSERT_FALSE(v.starts_with(__llvm_libc::cpp::StringView("123")));
+  ASSERT_FALSE(v.starts_with(__llvm_libc::cpp::StringView("abd")));
+  ASSERT_FALSE(v.starts_with(__llvm_libc::cpp::StringView("aaa")));
+  ASSERT_FALSE(v.starts_with(__llvm_libc::cpp::StringView("abcde")));
+}
+
 TEST(LlvmLibcStringViewTest, RemovePrefix) {
   __llvm_libc::cpp::StringView v("123456789");
 
   auto p = v.remove_prefix(0);
   ASSERT_EQ(p.size(), size_t(9));
   ASSERT_TRUE(p.equals(__llvm_libc::cpp::StringView("123456789")));
 
   p = v.remove_prefix(4);