src/hotspot/os/aix/os_aix.cpp - master - githubmirror/panama-vector

src/hotspot/os/aix/os_aix.cpp
2 677 строк90 KB
Thomas Stuefe
8388017: [aix] Narrow klass encoding protection region should be protected in mmap mode
11 авг 2026, 11:43
11 авг 2026, 11:43cc939a5
/*
 * Copyright (c) 1999, 2026, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2012, 2026 SAP SE. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 *
 */

#include "classfile/vmSymbols.hpp"
#include "code/vtableStubs.hpp"
#include "compiler/compileBroker.hpp"
#include "interpreter/interpreter.hpp"
#include "jvm.h"
#include "jvmtifiles/jvmti.h"
#include "libperfstat_aix.hpp"
#include "libodm_aix.hpp"
#include "loadlib_aix.hpp"
#include "logging/log.hpp"
#include "logging/logStream.hpp"
#include "memory/allocation.inline.hpp"
#include "misc_aix.hpp"
#include "oops/oop.inline.hpp"
#include "os_aix.inline.hpp"
#include "os_posix.hpp"
#include "porting_aix.hpp"
#include "prims/jniFastGetField.hpp"
#include "prims/jvm_misc.hpp"
#include "runtime/arguments.hpp"
#include "runtime/atomicAccess.hpp"
#include "runtime/globals.hpp"
#include "runtime/globals_extension.hpp"
#include "runtime/interfaceSupport.inline.hpp"
#include "runtime/java.hpp"
#include "runtime/javaCalls.hpp"
#include "runtime/javaThread.hpp"
#include "runtime/mutexLocker.hpp"
#include "runtime/objectMonitor.hpp"
#include "runtime/os.hpp"
#include "runtime/osInfo.hpp"
#include "runtime/osThread.hpp"
#include "runtime/perfMemory.hpp"
#include "runtime/safefetch.hpp"
#include "runtime/sharedRuntime.hpp"
#include "runtime/threads.hpp"
#include "runtime/timer.hpp"
#include "runtime/vm_version.hpp"
#include "services/attachListener.hpp"
#include "services/runtimeService.hpp"
#include "signals_posix.hpp"
#include "utilities/align.hpp"
#include "utilities/checkedCast.hpp"
#include "utilities/debug.hpp"
#include "utilities/decoder.hpp"
#include "utilities/defaultStream.hpp"
#include "utilities/events.hpp"
#include "utilities/growableArray.hpp"
#include "utilities/permitForbiddenFunctions.hpp"
#include "utilities/vmError.hpp"
#if INCLUDE_JFR
#include "jfr/support/jfrNativeLibraryLoadEvent.hpp"
#endif

// put OS-includes here (sorted alphabetically)
#include <alloca.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h>
#include <poll.h>
#include <procinfo.h>
#include <pthread.h>
#include <pwd.h>
#include <semaphore.h>
#include <signal.h>
#include <stdint.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <sys/ioctl.h>
#include <sys/ipc.h>
#include <sys/mman.h>
// sys/mman.h defines MAP_ANON_64K beginning with AIX7.3 TL1
#ifndef MAP_ANON_64K
  #define MAP_ANON_64K 0x400
#else
  STATIC_ASSERT(MAP_ANON_64K == 0x400);
#endif
#include <sys/resource.h>
#include <sys/select.h>
#include <sys/shm.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/sysinfo.h>
#include <sys/systemcfg.h>
#include <sys/time.h>
#include <sys/times.h>
#include <sys/types.h>
#include <sys/utsname.h>
#include <sys/vminfo.h>

#ifndef _LARGE_FILES
#error Hotspot on AIX must be compiled with -D_LARGE_FILES
#endif

// Missing prototypes for various system APIs.
extern "C"
int mread_real_time(timebasestruct_t *t, size_t size_of_timebasestruct_t);

#define MAX_PATH (2 * K)

// for multipage initialization error analysis (in 'g_multipage_error')
#define ERROR_MP_OS_TOO_OLD                          100
#define ERROR_MP_VMGETINFO_FAILED                    102
#define ERROR_MP_VMGETINFO_CLAIMS_NO_SUPPORT_FOR_64K 103

// excerpts from sys/systemcfg.h that might be missing on older os levels
#ifndef PV_8
  #define PV_8 0x300000          /* Power PC 8 */
#endif
#ifndef PV_8_Compat
  #define PV_8_Compat 0x308000   /* Power PC 8 */
#endif
#ifndef PV_9
  #define PV_9 0x400000          /* Power PC 9 */
#endif
#ifndef PV_9_Compat
  #define PV_9_Compat  0x408000  /* Power PC 9 */
#endif
#ifndef PV_10
  #define PV_10 0x500000          /* Power PC 10 */
#endif
#ifndef PV_10_Compat
  #define PV_10_Compat  0x508000  /* Power PC 10 */
#endif
#ifndef PV_11
  #define PV_11           0x600000        /* Power PC 11 */
#endif
#ifndef PV_11_Compat
  #define PV_11_Compat    0x608000        /* Power PC 11 */
#endif

static address resolve_function_descriptor_to_code_pointer(address p);

static void vmembk_print_on(outputStream* os);

////////////////////////////////////////////////////////////////////////////////
// global variables (for a description see os_aix.hpp)

physical_memory_size_type    os::Aix::_physical_memory = 0;

pthread_t os::Aix::_main_thread = ((pthread_t)0);

// 0 = uninitialized, otherwise 32 bit number:
//  0xVVRRTTSS
//  VV - major version
//  RR - minor version
//  TT - tech level, if known, 0 otherwise
//  SS - service pack, if known, 0 otherwise
uint32_t  os::Aix::_os_version = 0;

// -1 = uninitialized, 0 - no, 1 - yes
int       os::Aix::_xpg_sus_mode = -1;

////////////////////////////////////////////////////////////////////////////////
// local variables

static volatile jlong max_real_time = 0;

// Process break recorded at startup.
static address g_brk_at_startup = nullptr;

// This describes the state of multipage support of the underlying
// OS. Note that this is of no interest to the outsize world and
// therefore should not be defined in AIX class.
//
// AIX supports four different page sizes - 4K, 64K, 16MB, 16GB. The
// latter two (16M "large" resp. 16G "huge" pages) require special
// setup and are normally not available.
//
// AIX supports multiple page sizes per process, for:
//  - Stack (of the primordial thread, so not relevant for us)
//  - Data - data, bss, heap, for us also pthread stacks
//  - Text - text code
//  - shared memory
//
// Default page sizes can be set via linker options (-bdatapsize, -bstacksize, ...)
// and via environment variable LDR_CNTRL (DATAPSIZE, STACKPSIZE, ...).
//
// For shared memory, page size can be set dynamically via
// shmctl(). Different shared memory regions can have different page
// sizes.
//
// More information can be found at IBM info center:
//   http://publib.boulder.ibm.com/infocenter/aix/v6r1/index.jsp?topic=/com.ibm.aix.prftungd/doc/prftungd/multiple_page_size_app_support.htm
//
static struct {
  size_t pagesize;             // sysconf _SC_PAGESIZE (4K)
  size_t datapsize;            // default data page size (LDR_CNTRL DATAPSIZE)
  size_t shmpsize;             // default shared memory page size (LDR_CNTRL SHMPSIZE)
  size_t pthr_stack_pagesize;  // stack page size of pthread threads
  size_t textpsize;            // default text page size (LDR_CNTRL STACKPSIZE)
  bool can_use_64K_pages;      // True if we can alloc 64K pages dynamically with Sys V shm.
  bool can_use_16M_pages;      // True if we can alloc 16M pages dynamically with Sys V shm.
  bool can_use_64K_mmap_pages; // True if we can alloc 64K pages dynamically with mmap.
  int error;                   // Error describing if something went wrong at multipage init.
} g_multipage_support = {
  (size_t) -1,
  (size_t) -1,
  (size_t) -1,
  (size_t) -1,
  (size_t) -1,
  false, false, false,
  0
};

// We must not accidentally allocate memory close to the BRK - even if
// that would work - because then we prevent the BRK segment from
// growing which may result in a malloc OOM even though there is
// enough memory. The problem only arises if we shmat() or mmap() at
// a specific wish address, e.g. to place the heap in a
// compressed-oops-friendly way.
static bool is_close_to_brk(address a) {
  assert0(g_brk_at_startup != nullptr);
  if (a >= g_brk_at_startup &&
      a < (g_brk_at_startup + MaxExpectedDataSegmentSize)) {
    return true;
  }
  return false;
}

bool os::free_memory(physical_memory_size_type& value) {
  return Aix::available_memory(value);
}

bool os::Machine::free_memory(physical_memory_size_type& value) {
  return Aix::available_memory(value);
}

bool os::available_memory(physical_memory_size_type& value) {
  return Aix::available_memory(value);
}

bool os::Machine::available_memory(physical_memory_size_type& value) {
  return Aix::available_memory(value);
}

bool os::Aix::available_memory(physical_memory_size_type& value) {
  os::Aix::meminfo_t mi;
  if (os::Aix::get_meminfo(&mi)) {
    value = static_cast<physical_memory_size_type>(mi.real_free);
    return true;
  } else {
    return false;
  }
}

bool os::total_swap_space(physical_memory_size_type& value) {
  return Machine::total_swap_space(value);
}

bool os::Machine::total_swap_space(physical_memory_size_type& value) {
  perfstat_memory_total_t memory_info;
  if (libperfstat::perfstat_memory_total(nullptr, &memory_info, sizeof(perfstat_memory_total_t), 1) == -1) {
    return false;
  }
  value = static_cast<physical_memory_size_type>(memory_info.pgsp_total * 4 * K);
  return true;
}

bool os::free_swap_space(physical_memory_size_type& value) {
  return Machine::free_swap_space(value);
}

bool os::Machine::free_swap_space(physical_memory_size_type& value) {
  perfstat_memory_total_t memory_info;
  if (libperfstat::perfstat_memory_total(nullptr, &memory_info, sizeof(perfstat_memory_total_t), 1) == -1) {
    return false;
  }
  value = static_cast<physical_memory_size_type>(memory_info.pgsp_free * 4 * K);
  return true;
}

physical_memory_size_type os::physical_memory() {
  return Aix::physical_memory();
}

physical_memory_size_type os::Machine::physical_memory() {
  return Aix::physical_memory();
}

size_t os::rss() { return (size_t)0; }

// Cpu architecture string
#if defined(PPC32)
static char cpu_arch[] = "ppc";
#elif defined(PPC64)
static char cpu_arch[] = "ppc64";
#else
#error Add appropriate cpu_arch setting
#endif

// Given an address, returns the size of the page backing that address.
size_t os::Aix::query_pagesize(void* addr) {
  vm_page_info pi;
  pi.addr = (uint64_t)addr;
  if (::vmgetinfo(&pi, VM_PAGE_INFO, sizeof(pi)) == 0) {
    return pi.pagesize;
  } else {
    log_warning(pagesize)("vmgetinfo(VM_PAGE_INFO) failed (errno: %d)", errno);
    assert(false, "vmgetinfo failed to retrieve page size");
    return 4*K;
  }
}

void os::Aix::initialize_system_info() {

  // Get the number of online(logical) cpus instead of configured.
  os::_processor_count = sysconf(_SC_NPROCESSORS_ONLN);
  assert(_processor_count > 0, "_processor_count must be > 0");

  // Retrieve total physical storage.
  os::Aix::meminfo_t mi;
  if (!os::Aix::get_meminfo(&mi)) {
    assert(false, "os::Aix::get_meminfo failed.");
  }
  _physical_memory = static_cast<physical_memory_size_type>(mi.real_total);
}

// Helper function for tracing page sizes.
static const char* describe_pagesize(size_t pagesize) {
  switch (pagesize) {
    case 4*K : return "4K";
    case 64*K: return "64K";
    case 16*M: return "16M";
    case 16*G: return "16G";
    default:
      assert(false, "surprise");
      return "??";
  }
}

// Probe OS for multipage support.
// Will fill the global g_multipage_support structure.
// Must be called before calling os::large_page_init().
static void query_multipage_support() {

  guarantee(g_multipage_support.pagesize == (size_t)-1,
            "do not call twice");

  g_multipage_support.pagesize = ::sysconf(_SC_PAGESIZE);

  // This really would surprise me.
  assert(g_multipage_support.pagesize == 4*K, "surprise!");

  // Query default data page size (default page size for C-Heap, pthread stacks and .bss).
  // Default data page size is defined either by linker options (-bdatapsize)
  // or by environment variable LDR_CNTRL (suboption DATAPSIZE). If none is given,
  // default should be 4K.
  {
    void* p = permit_forbidden_function::malloc(16*M);
    g_multipage_support.datapsize = os::Aix::query_pagesize(p);
    permit_forbidden_function::free(p);
  }

  // Query default shm page size (LDR_CNTRL SHMPSIZE).
  // Note that this is pure curiosity. We do not rely on default page size but set
  // our own page size after allocated.
  {
    const int shmid = ::shmget(IPC_PRIVATE, 1, IPC_CREAT | S_IRUSR | S_IWUSR);
    assert(shmid != -1, "shmget failed");
    if (shmid != -1) {
      void* p = ::shmat(shmid, nullptr, 0);
      ::shmctl(shmid, IPC_RMID, nullptr);
      assert(p != (void*) -1, "shmat failed");
      if (p != (void*) -1) {
        g_multipage_support.shmpsize = os::Aix::query_pagesize(p);
        ::shmdt(p);
      }
    }
  }

  // Before querying the stack page size, make sure we are not running as primordial
  // thread (because primordial thread's stack may have different page size than
  // pthread thread stacks). Running a VM on the primordial thread won't work for a
  // number of reasons so we may just as well guarantee it here.
  guarantee0(!os::is_primordial_thread());

  // Query pthread stack page size. Should be the same as data page size because
  // pthread stacks are allocated from C-Heap.
  {
    int dummy = 0;
    g_multipage_support.pthr_stack_pagesize = os::Aix::query_pagesize(&dummy);
  }

  // Query default text page size (LDR_CNTRL TEXTPSIZE).
  {
    address any_function =
      resolve_function_descriptor_to_code_pointer((address)describe_pagesize);
    g_multipage_support.textpsize = os::Aix::query_pagesize(any_function);
  }

  // Now check which page sizes the OS claims it supports, and of those, which actually can be used.
  {
    const int MAX_PAGE_SIZES = 4;
    psize_t sizes[MAX_PAGE_SIZES];
    const int num_psizes = ::vmgetinfo(sizes, VMINFO_GETPSIZES, MAX_PAGE_SIZES);
    if (num_psizes == -1) {
      log_warning(pagesize)("vmgetinfo(VMINFO_GETPSIZES) failed (errno: %d), disabling multipage support.", errno);
      g_multipage_support.error = ERROR_MP_VMGETINFO_FAILED;
      goto query_multipage_support_end;
    }
    guarantee(num_psizes > 0, "vmgetinfo(.., VMINFO_GETPSIZES, ...) failed.");
    assert(num_psizes <= MAX_PAGE_SIZES, "Surprise! more than 4 page sizes?");
    log_info(pagesize)("vmgetinfo(.., VMINFO_GETPSIZES, ...) returns %d supported page sizes: ", num_psizes);
    for (int i = 0; i < num_psizes; i ++) {
      trcVerbose(" %s ", describe_pagesize(sizes[i]));
    }

    // Can we use 64K, 16M pages?
    for (int i = 0; i < num_psizes; i ++) {
      const size_t pagesize = sizes[i];
      if (pagesize != 64*K && pagesize != 16*M) {
        continue;
      }
      bool can_use = false;
      trcVerbose("Probing support for %s pages...", describe_pagesize(pagesize));
      const int shmid = ::shmget(IPC_PRIVATE, pagesize,
        IPC_CREAT | S_IRUSR | S_IWUSR);
      assert(shmid != -1, "shmget failed");
      if (shmid != -1) {
        // Try to set pagesize.
        struct shmid_ds shm_buf = { };
        shm_buf.shm_pagesize = pagesize;
        if (::shmctl(shmid, SHM_PAGESIZE, &shm_buf) != 0) {
          const int en = errno;
          ::shmctl(shmid, IPC_RMID, nullptr); // As early as possible!
          log_warning(pagesize)("shmctl(SHM_PAGESIZE) failed with errno=%d", errno);
        } else {
          // Attach and double check pageisze.
          void* p = ::shmat(shmid, nullptr, 0);
          ::shmctl(shmid, IPC_RMID, nullptr); // As early as possible!
          assert(p != (void*) -1, "shmat failed");
          if (p != (void*) -1) {
            const size_t real_pagesize = os::Aix::query_pagesize(p);
            if (real_pagesize != pagesize) {
              log_warning(pagesize)("real page size (0x%zx) differs.", real_pagesize);
            } else {
              can_use = true;
            }
            ::shmdt(p);
          }
        }
      }
      trcVerbose("Can use: %s", (can_use ? "yes" : "no"));
      if (pagesize == 64*K) {
        g_multipage_support.can_use_64K_pages = can_use;
      } else if (pagesize == 16*M) {
        g_multipage_support.can_use_16M_pages = can_use;
      }
    }

    // Can we use mmap with 64K pages? (Should be available with AIX7.3 TL1)
    {
      void* p = mmap(nullptr, 64*K, PROT_READ | PROT_WRITE, MAP_ANON_64K | MAP_ANONYMOUS | MAP_SHARED, -1, 0);
      assert(p != (void*) -1, "mmap failed");
      if (p != (void*) -1) {
        g_multipage_support.can_use_64K_mmap_pages = (64*K == os::Aix::query_pagesize(p));
        munmap(p, 64*K);
      }
    }

  } // end: check which pages can be used for shared memory

query_multipage_support_end:

  trcVerbose("base page size (sysconf _SC_PAGESIZE): %s",
      describe_pagesize(g_multipage_support.pagesize));
  trcVerbose("Data page size (C-Heap, bss, etc): %s",
      describe_pagesize(g_multipage_support.datapsize));
  trcVerbose("Text page size: %s",
      describe_pagesize(g_multipage_support.textpsize));
  trcVerbose("Thread stack page size (pthread): %s",
      describe_pagesize(g_multipage_support.pthr_stack_pagesize));
  trcVerbose("Can use 64K pages with mmap memory: %s",
      (g_multipage_support.can_use_64K_mmap_pages ? "yes" :"no"));
  trcVerbose("Default shared memory page size: %s",
      describe_pagesize(g_multipage_support.shmpsize));
  trcVerbose("Can use 64K pages dynamically with shared memory: %s",
      (g_multipage_support.can_use_64K_pages ? "yes" :"no"));
  trcVerbose("Can use 16M pages dynamically with shared memory: %s",
      (g_multipage_support.can_use_16M_pages ? "yes" :"no"));
  trcVerbose("Multipage error details: %d",
      g_multipage_support.error);

  // sanity checks
  assert0(g_multipage_support.pagesize == 4*K);
  assert0(g_multipage_support.datapsize == 4*K || g_multipage_support.datapsize == 64*K);
  assert0(g_multipage_support.textpsize == 4*K || g_multipage_support.textpsize == 64*K);
  assert0(g_multipage_support.pthr_stack_pagesize == g_multipage_support.datapsize);
  assert0(g_multipage_support.shmpsize == 4*K || g_multipage_support.shmpsize == 64*K);

}

void os::init_system_properties_values() {

#ifndef OVERRIDE_LIBPATH
  #define DEFAULT_LIBPATH "/lib:/usr/lib"
#else
  #define DEFAULT_LIBPATH OVERRIDE_LIBPATH
#endif
#define EXTENSIONS_DIR  "/lib/ext"

  // Buffer that fits several snprintfs.
  // Note that the space for the trailing null is provided
  // by the nulls included by the sizeof operator.
  const size_t bufsize =
    MAX2((size_t)MAXPATHLEN,  // For dll_dir & friends.
         (size_t)MAXPATHLEN + sizeof(EXTENSIONS_DIR)); // extensions dir
  char *buf = NEW_C_HEAP_ARRAY(char, bufsize, mtInternal);

  // sysclasspath, java_home, dll_dir
  {
    char *pslash;
    os::jvm_path(buf, bufsize);

    // Found the full path to libjvm.so.
    // Now cut the path to <java_home>/jre if we can.
    pslash = strrchr(buf, '/');
    if (pslash != nullptr) {
      *pslash = '\0';            // Get rid of /libjvm.so.
    }
    pslash = strrchr(buf, '/');
    if (pslash != nullptr) {
      *pslash = '\0';            // Get rid of /{client|server|hotspot}.
    }
    Arguments::set_dll_dir(buf);

    if (pslash != nullptr) {
      pslash = strrchr(buf, '/');
      if (pslash != nullptr) {
        *pslash = '\0';        // Get rid of /lib.
      }
    }
    Arguments::set_java_home(buf);
    if (!set_boot_path('/', ':')) {
      vm_exit_during_initialization("Failed setting boot class path.", nullptr);
    }
  }

  // Where to look for native libraries.

  // On Aix we get the user setting of LIBPATH.
  // Eventually, all the library path setting will be done here.
  // Get the user setting of LIBPATH.
  const char *v = ::getenv("LIBPATH");
  const char *v_colon = ":";
  if (v == nullptr) { v = ""; v_colon = ""; }

  // Concatenate user and invariant part of ld_library_path.
  // That's +1 for the colon and +1 for the trailing '\0'.
  size_t pathsize = strlen(v) + 1 + sizeof(DEFAULT_LIBPATH) + 1;
  char *ld_library_path = NEW_C_HEAP_ARRAY(char, pathsize, mtInternal);
  os::snprintf_checked(ld_library_path, pathsize, "%s%s" DEFAULT_LIBPATH, v, v_colon);
  Arguments::set_library_path(ld_library_path);
  FREE_C_HEAP_ARRAY(ld_library_path);

  // Extensions directories.
  os::snprintf_checked(buf, bufsize, "%s" EXTENSIONS_DIR, Arguments::get_java_home());
  Arguments::set_ext_dirs(buf);

  FREE_C_HEAP_ARRAY(buf);

#undef DEFAULT_LIBPATH
#undef EXTENSIONS_DIR
}

// retrieve memory information.
// Returns false if something went wrong;
// content of pmi undefined in this case.
bool os::Aix::get_meminfo(meminfo_t* pmi) {

  assert(pmi, "get_meminfo: invalid parameter");
  memset(pmi, 0, sizeof(meminfo_t));

  // dynamically loaded perfstat library is used to retrieve memory statistics
  perfstat_memory_total_t psmt;
  memset (&psmt, '\0', sizeof(psmt));
  const int rc = libperfstat::perfstat_memory_total(nullptr, &psmt, sizeof(psmt), 1);
  if (rc == -1) {
    log_warning(os)("perfstat_memory_total() failed (errno=%d)", errno);
    assert(0, "perfstat_memory_total() failed");
    return false;
  }

  assert(rc == 1, "perfstat_memory_total() - weird return code");

  // The fields of perfstat_memory_total_t:
  // u_longlong_t virt_total         Total virtual memory (in 4 KB pages).
  // u_longlong_t real_total         Total real memory (in 4 KB pages).
  // u_longlong_t real_free          Free real memory (in 4 KB pages).
  // u_longlong_t pgsp_total         Total paging space (in 4 KB pages).
  // u_longlong_t pgsp_free          Free paging space (in 4 KB pages).
  pmi->virt_total = psmt.virt_total * 4096;
  pmi->real_total = psmt.real_total * 4096;
  pmi->real_free = psmt.real_free * 4096;
  pmi->pgsp_total = psmt.pgsp_total * 4096;
  pmi->pgsp_free = psmt.pgsp_free * 4096;

  return true;
} // end os::Aix::get_meminfo

//////////////////////////////////////////////////////////////////////////////
// create new thread

// Thread start routine for all newly created threads
static void *thread_native_entry(Thread *thread) {

  thread->record_stack_base_and_size();

  const pthread_t pthread_id = ::pthread_self();
  const tid_t kernel_thread_id = ::thread_self();

  LogTarget(Info, os, thread) lt;
  if (lt.is_enabled()) {
    address low_address = thread->stack_end();
    address high_address = thread->stack_base();
    lt.print("Thread is alive (tid: %zu, kernel thread id: %zu"
             ", stack [" PTR_FORMAT " - " PTR_FORMAT " (%zuk using %luk pages)).",
             os::current_thread_id(), (uintx) kernel_thread_id, p2i(low_address), p2i(high_address),
             (high_address - low_address) / K, os::Aix::query_pagesize(low_address) / K);
  }

  // Normally, pthread stacks on AIX live in the data segment (are allocated with malloc()
  // by the pthread library). In rare cases, this may not be the case, e.g. when third-party
  // tools hook pthread_create(). In this case, we may run into problems establishing
  // guard pages on those stacks, because the stacks may reside in memory which is not
  // protectable (shmated).
  if (thread->stack_base() > ::sbrk(0)) {
    log_warning(os, thread)("Thread stack not in data segment.");
  }

  // Try to randomize the cache line index of hot stack frames.
  // This helps when threads of the same stack traces evict each other's
  // cache lines. The threads can be either from the same JVM instance, or
  // from different JVM instances. The benefit is especially true for
  // processors with hyperthreading technology.

  static int counter = 0;
  int pid = os::current_process_id();
  alloca(((pid ^ counter++) & 7) * 128);

  thread->initialize_thread_current();

  OSThread* osthread = thread->osthread();

  // Thread_id is pthread id.
  osthread->set_thread_id(pthread_id);

  // .. but keep kernel thread id too for diagnostics
  osthread->set_kernel_thread_id(kernel_thread_id);

  // Initialize signal mask for this thread.
  PosixSignals::hotspot_sigmask(thread);

  // Initialize floating point control register.
  os::Aix::init_thread_fpu_state();

  assert(osthread->get_state() == RUNNABLE, "invalid os thread state");

  // Call one more level start routine.
  thread->call_run();

  // Note: at this point the thread object may already have deleted itself.
  // Prevent dereferencing it from here on out.
  thread = nullptr;

  log_info(os, thread)("Thread finished (tid: %zu, kernel thread id: %zu).",
    os::current_thread_id(), (uintx) kernel_thread_id);

  return nullptr;
}

bool os::create_thread(Thread* thread, ThreadType thr_type,
                       size_t req_stack_size) {

  assert(thread->osthread() == nullptr, "caller responsible");

  // Allocate the OSThread object.
  OSThread* osthread = new (std::nothrow) OSThread();
  if (osthread == nullptr) {
    return false;
  }

  // Initial state is ALLOCATED but not INITIALIZED
  osthread->set_state(ALLOCATED);

  thread->set_osthread(osthread);

  // Init thread attributes.
  pthread_attr_t attr;
  int rslt = pthread_attr_init(&attr);
  if (rslt != 0) {
    thread->set_osthread(nullptr);
    delete osthread;
    return false;
  }
  pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);

  // Make sure we run in 1:1 kernel-user-thread mode.
  guarantee(pthread_attr_setscope(&attr, PTHREAD_SCOPE_SYSTEM) == 0, "???");
  guarantee(pthread_attr_setinheritsched(&attr, PTHREAD_EXPLICIT_SCHED) == 0, "???");

  // Start in suspended state, and in os::thread_start, wake the thread up.
  guarantee(pthread_attr_setsuspendstate_np(&attr, PTHREAD_CREATE_SUSPENDED_NP) == 0, "???");

  // Calculate stack size if it's not specified by caller.
  size_t stack_size = os::Posix::get_initial_stack_size(thr_type, req_stack_size);

  // JDK-8187028: It was observed that on some configurations (4K backed thread stacks)
  // the real thread stack size may be smaller than the requested stack size, by as much as 64K.
  // This very much looks like a pthread lib error. As a workaround, increase the stack size
  // by 64K for small thread stacks (arbitrarily chosen to be < 4MB)
  if (stack_size < 4096 * K) {
    stack_size += 64 * K;
  }

  // On Aix, pthread_attr_setstacksize fails with huge values and leaves the
  // thread size in attr unchanged. If this is the minimal stack size as set
  // by pthread_attr_init this leads to crashes after thread creation. E.g. the
  // guard pages might not fit on the tiny stack created.
  int ret = pthread_attr_setstacksize(&attr, stack_size);
  if (ret != 0) {
    log_warning(os, thread)("The %sthread stack size specified is invalid: %zuk",
                            (thr_type == compiler_thread) ? "compiler " : ((thr_type == java_thread) ? "" : "VM "),
                            stack_size / K);
    thread->set_osthread(nullptr);
    delete osthread;
    pthread_attr_destroy(&attr);
    return false;
  }

  // Save some cycles and a page by disabling OS guard pages where we have our own
  // VM guard pages (in java threads). For other threads, keep system default guard
  // pages in place.
  if (thr_type == java_thread || thr_type == compiler_thread) {
    ret = pthread_attr_setguardsize(&attr, 0);
  }

  ResourceMark rm;
  pthread_t tid = 0;

  if (ret == 0) {
    int trials_remaining = 4;
    useconds_t next_delay = 1000;
    while (true) {
      ret = pthread_create(&tid, &attr, (void* (*)(void*)) thread_native_entry, thread);

      if (ret != EAGAIN) {
        break;
      }

      if (--trials_remaining <= 0) {
        break;
      }

      log_debug(os, thread)("Failed to start native thread (%s), retrying after %dus.", os::errno_name(ret), next_delay);
      ::usleep(next_delay);
      next_delay *= 2;
    }
  }

  if (ret == 0) {
    char buf[64];
    log_info(os, thread)("Thread \"%s\" started (pthread id: %zu, attributes: %s). ",
                         thread->name(), (uintx) tid, os::Posix::describe_pthread_attr(buf, sizeof(buf), &attr));
  } else {
    char buf[64];
    log_warning(os, thread)("Failed to start thread \"%s\" - pthread_create failed (%d=%s) for attributes: %s.",
                            thread->name(), ret, os::errno_name(ret), os::Posix::describe_pthread_attr(buf, sizeof(buf), &attr));
    // Log some OS information which might explain why creating the thread failed.
    log_warning(os, thread)("Number of threads approx. running in the VM: %d", Threads::number_of_threads());
    log_warning(os, thread)("Checking JVM parameter MaxExpectedDataSegmentSize (currently %zuk)  might be helpful", MaxExpectedDataSegmentSize/K);
    LogStream st(Log(os, thread)::info());
    os::Posix::print_rlimit_info(&st);
    os::print_memory_info(&st);
  }

  pthread_attr_destroy(&attr);

  if (ret != 0) {
    // Need to clean up stuff we've allocated so far.
    thread->set_osthread(nullptr);
    delete osthread;
    return false;
  }

  // OSThread::thread_id is the pthread id.
  osthread->set_thread_id(tid);

  // child thread synchronization is not done here on AIX, a thread is started in suspended state

  return true;
}

/////////////////////////////////////////////////////////////////////////////
// attach existing thread

// bootstrap the main thread
bool os::create_main_thread(JavaThread* thread) {
  assert(os::Aix::_main_thread == pthread_self(), "should be called inside main thread");
  return create_attached_thread(thread);
}

bool os::create_attached_thread(JavaThread* thread) {
#ifdef ASSERT
    thread->verify_not_published();
#endif

  // Allocate the OSThread object
  OSThread* osthread = new (std::nothrow) OSThread();

  if (osthread == nullptr) {
    return false;
  }

  const pthread_t pthread_id = ::pthread_self();
  const tid_t kernel_thread_id = ::thread_self();

  // OSThread::thread_id is the pthread id.
  osthread->set_thread_id(pthread_id);

  // .. but keep kernel thread id too for diagnostics
  osthread->set_kernel_thread_id(kernel_thread_id);

  // initialize floating point control register
  os::Aix::init_thread_fpu_state();

  // Initial thread state is RUNNABLE
  osthread->set_state(RUNNABLE);

  thread->set_osthread(osthread);

  // initialize signal mask for this thread
  // and save the caller's signal mask
  PosixSignals::hotspot_sigmask(thread);

  log_info(os, thread)("Thread attached (tid: %zu, kernel thread  id: %zu"
                       ", stack: " PTR_FORMAT " - " PTR_FORMAT " (%zuK) ).",
                       os::current_thread_id(), (uintx) kernel_thread_id,
                       p2i(thread->stack_base()), p2i(thread->stack_end()), thread->stack_size() / K);

  return true;
}

void os::pd_start_thread(Thread* thread) {
  int status = pthread_continue_np(thread->osthread()->pthread_id());
  assert(status == 0, "thr_continue failed");
}

// Free OS resources related to the OSThread
void os::free_thread(OSThread* osthread) {
  assert(osthread != nullptr, "osthread not set");

  // We are told to free resources of the argument thread, but we can only really operate
  // on the current thread. The current thread may be already detached at this point.
  assert(Thread::current_or_null() == nullptr || Thread::current()->osthread() == osthread,
         "os::free_thread but not current thread");

  // Restore caller's signal mask
  sigset_t sigmask = osthread->caller_sigmask();
  pthread_sigmask(SIG_SETMASK, &sigmask, nullptr);

  delete osthread;
}

////////////////////////////////////////////////////////////////////////////////
// time support

// We use mread_real_time here.
// On AIX: If the CPU has a time register, the result will be RTC_POWER and
// it has to be converted to real time. AIX documentations suggests to do
// this unconditionally, so we do it.
//
// See: https://www.ibm.com/support/knowledgecenter/ssw_aix_61/com.ibm.aix.basetrf2/read_real_time.htm
//
jlong os::javaTimeNanos() {
  timebasestruct_t time;
  int rc = mread_real_time(&time, TIMEBASE_SZ);

  if (rc != RTC_POWER) {
    rc = time_base_to_time(&time, TIMEBASE_SZ);
    assert(rc != -1, "error calling time_base_to_time()");
  }
  return jlong(time.tb_high) * NANOSECS_PER_SEC + jlong(time.tb_low);
}

void os::javaTimeNanos_info(jvmtiTimerInfo *info_ptr) {
  info_ptr->max_value = all_bits_jlong;
  // mread_real_time() is monotonic (see 'os::javaTimeNanos()')
  info_ptr->may_skip_backward = false;
  info_ptr->may_skip_forward = false;
  info_ptr->kind = JVMTI_TIMER_ELAPSED;    // elapsed not CPU time
}

intx os::current_thread_id() {
  return (intx)pthread_self();
}

int os::current_process_id() {
  return getpid();
}

// DLL functions

// This must be hard coded because it's the system's temporary
// directory not the java application's temp directory, ala java.io.tmpdir.
const char* os::get_temp_directory() { return "/tmp"; }

void os::prepare_native_symbols() {
  LoadedLibraries::reload();
}

// Check if addr is inside libjvm.so.
bool os::address_is_in_vm(address addr) {

  // Input could be a real pc or a function pointer literal. The latter
  // would be a function descriptor residing in the data segment of a module.
  loaded_module_t lm;
  if (LoadedLibraries::find_for_text_address(addr, &lm)) {
    return lm.is_in_vm;
  } else if (LoadedLibraries::find_for_data_address(addr, &lm)) {
    return lm.is_in_vm;
  } else {
    return false;
  }

}

// Resolve an AIX function descriptor literal to a code pointer.
// If the input is a valid code pointer to a text segment of a loaded module,
//   it is returned unchanged.
// If the input is a valid AIX function descriptor, it is resolved to the
//   code entry point.
// If the input is neither a valid function descriptor nor a valid code pointer,
//   null is returned.
static address resolve_function_descriptor_to_code_pointer(address p) {

  if (LoadedLibraries::find_for_text_address(p, nullptr)) {
    // It is a real code pointer.
    return p;
  } else if (LoadedLibraries::find_for_data_address(p, nullptr)) {
    // Pointer to data segment, potential function descriptor.
    address code_entry = (address)(((FunctionDescriptor*)p)->entry());
    if (LoadedLibraries::find_for_text_address(code_entry, nullptr)) {
      // It is a function descriptor.
      return code_entry;
    }
  }

  return nullptr;
}

bool os::dll_address_to_function_name(address addr, char *buf,
                                      int buflen, int *offset,
                                      bool demangle) {
  if (offset) {
    *offset = -1;
  }
  // Buf is not optional, but offset is optional.
  assert(buf != nullptr, "sanity check");
  buf[0] = '\0';

  // Resolve function ptr literals first.
  addr = resolve_function_descriptor_to_code_pointer(addr);
  if (!addr) {
    return false;
  }

  return AixSymbols::get_function_name(addr, buf, buflen, offset, nullptr, demangle);
}

bool os::dll_address_to_library_name(address addr, char* buf,
                                     int buflen, int* offset) {
  if (offset) {
    *offset = -1;
  }
  // Buf is not optional, but offset is optional.
  assert(buf != nullptr, "sanity check");
  buf[0] = '\0';

  // Resolve function ptr literals first.
  addr = resolve_function_descriptor_to_code_pointer(addr);
  if (!addr) {
    return false;
  }

  address  base = nullptr;
  if (!AixSymbols::get_module_name_and_base(addr, buf, buflen, &base)
      || base == nullptr) {
    return false;
  }
  assert(addr >= base && addr <= base + INT_MAX, "address not in library text range");
  if (offset != nullptr) {
    *offset = addr - base;
  }

  return true;
}

static void* dll_load_library(const char *filename, int *eno, char *ebuf, int ebuflen) {

  log_info(os)("attempting shared library load of %s", filename);
  if (ebuf && ebuflen > 0) {
    ebuf[0] = '\0';
    ebuf[ebuflen - 1] = '\0';
  }

  if (!filename || strlen(filename) == 0) {
    if (ebuf != nullptr && ebuflen > 0) {
      ::strncpy(ebuf, "dll_load: empty filename specified", ebuflen - 1);
    }
    return nullptr;
  }

  // RTLD_LAZY has currently the same behavior as RTLD_NOW
  // The dl is loaded immediately with all its dependants.
  int dflags = RTLD_LAZY;
  // check for filename ending with ')', it indicates we want to load
  // a MEMBER module that is a member of an archive.
  int flen = strlen(filename);
  if (flen > 0 && filename[flen - 1] == ')') {
    dflags |= RTLD_MEMBER;
  }

  Events::log_dll_message(nullptr, "Attempting to load shared library %s", filename);

  void* result;
  const char* error_report = nullptr;
  JFR_ONLY(NativeLibraryLoadEvent load_event(filename, &result);)
  result = Aix_dlopen(filename, dflags, eno, &error_report);
  if (result != nullptr) {
    Events::log_dll_message(nullptr, "Loaded shared library %s", filename);
    // Reload dll cache. Don't do this in signal handling.
    LoadedLibraries::reload();
    log_info(os)("shared library load of %s was successful", filename);
    return result;
  } else {
    // error analysis when dlopen fails
    if (error_report == nullptr) {
      error_report = "dlerror returned no error description";
    }
    if (ebuf != nullptr && ebuflen > 0) {
      os::snprintf_checked(ebuf, ebuflen, "%s, LIBPATH=%s, LD_LIBRARY_PATH=%s : %s",
                           filename, ::getenv("LIBPATH"), ::getenv("LD_LIBRARY_PATH"), error_report);
    }
    Events::log_dll_message(nullptr, "Loading shared library %s failed, %s", filename, error_report);
    log_info(os)("shared library load of %s failed, %s", filename, error_report);
    JFR_ONLY(load_event.set_error_msg(error_report);)
  }
  return nullptr;
}
// Load library named <filename>
// If filename matches <name>.so, and loading fails, repeat with <name>.a.
void *os::dll_load(const char *filename, char *ebuf, int ebuflen) {
  void* result = nullptr;
  const char old_extension[] = ".so";
  const char new_extension[] = ".a";
  // First try to load the existing file.
  int eno = 0;
  result = dll_load_library(filename, &eno, ebuf, ebuflen);
  // If the load fails, we try to reload by changing the extension to .a for .so files only.
  // Shared object in .so format dont have braces, hence they get removed for archives with members.
  if (result == nullptr && eno == ENOENT) {
    const char* pointer_to_dot = strrchr(filename, '.');
    if (pointer_to_dot != nullptr && strcmp(pointer_to_dot, old_extension) == 0) {
      STATIC_ASSERT(sizeof(old_extension) >= sizeof(new_extension));
      char* tmp_path = os::strdup(filename);
      size_t prefix_size = pointer_delta(pointer_to_dot, filename, 1);
      os::snprintf_checked(tmp_path + prefix_size, sizeof(old_extension), "%s", new_extension);
      result = dll_load_library(tmp_path, &eno, ebuf, ebuflen);
      os::free(tmp_path);
    }
  }
  return result;
}

void os::print_dll_info(outputStream *st) {
  st->print_cr("Dynamic libraries:");
  LoadedLibraries::print(st);
}

void os::get_summary_os_info(char* buf, size_t buflen) {
  // There might be something more readable than uname results for AIX.
  struct utsname name;
  uname(&name);
  os::snprintf_checked(buf, buflen, "%s %s", name.release, name.version);
}

int os::get_loaded_modules_info(os::LoadedModulesCallbackFunc callback, void *param) {

  if (!LoadedLibraries::for_each(callback, param)) {
    return -1;
  }

  return 0;
}

void os::print_os_info_brief(outputStream* st) {
  uint32_t ver = os::Aix::os_version();
  st->print_cr("AIX kernel version %u.%u.%u.%u",
               (ver >> 24) & 0xFF, (ver >> 16) & 0xFF, (ver >> 8) & 0xFF, ver & 0xFF);

  os::Posix::print_uname_info(st);

  // Linux uses print_libversion_info(st); here.
}

void os::print_os_info(outputStream* st) {
  st->print_cr("OS:");

  os::Posix::print_uname_info(st);

  uint32_t ver = os::Aix::os_version();
  st->print_cr("AIX kernel version %u.%u.%u.%u",
               (ver >> 24) & 0xFF, (ver >> 16) & 0xFF, (ver >> 8) & 0xFF, ver & 0xFF);

  os::Posix::print_uptime_info(st);

  os::Posix::print_rlimit_info(st);

  os::Posix::print_load_average(st);

  // _SC_THREAD_THREADS_MAX is the maximum number of threads within a process.
  long tmax = sysconf(_SC_THREAD_THREADS_MAX);
  st->print_cr("maximum #threads within a process:%ld", tmax);

  // print wpar info
  libperfstat::wparinfo_t wi;
  if (libperfstat::get_wparinfo(&wi)) {
    st->print_cr("wpar info");
    st->print_cr("name: %s", wi.name);
    st->print_cr("id:   %d", wi.wpar_id);
    st->print_cr("type: %s", (wi.app_wpar ? "application" : "system"));
  }

  VM_Version::print_platform_virtualization_info(st);
}

void os::print_memory_info(outputStream* st) {

  st->print_cr("Memory:");

  st->print_cr("  Base page size (sysconf _SC_PAGESIZE):  %s",
    describe_pagesize(g_multipage_support.pagesize));
  st->print_cr("  Data page size (C-Heap, bss, etc):      %s",
    describe_pagesize(g_multipage_support.datapsize));
  st->print_cr("  Text page size:                         %s",
    describe_pagesize(g_multipage_support.textpsize));
  st->print_cr("  Thread stack page size (pthread):       %s",
    describe_pagesize(g_multipage_support.pthr_stack_pagesize));
  st->print_cr("  Can use 64K pages with mmap memory:     %s",
    (g_multipage_support.can_use_64K_mmap_pages ? "yes" :"no"));
  st->print_cr("  Default shared memory page size:        %s",
    describe_pagesize(g_multipage_support.shmpsize));
  st->print_cr("  Can use 64K pages dynamically with shared memory:  %s",
    (g_multipage_support.can_use_64K_pages ? "yes" :"no"));
  st->print_cr("  Can use 16M pages dynamically with shared memory: %s",
    (g_multipage_support.can_use_16M_pages ? "yes" :"no"));
  st->print_cr("  Multipage error: %d",
    g_multipage_support.error);
  st->cr();
  st->print_cr("  os::vm_page_size:       %s", describe_pagesize(os::vm_page_size()));

  // print out LDR_CNTRL because it affects the default page sizes
  const char* const ldr_cntrl = ::getenv("LDR_CNTRL");
  st->print_cr("  LDR_CNTRL=%s.", ldr_cntrl ? ldr_cntrl : "<unset>");

  // Print out AIXTHREAD_GUARDPAGES because it affects the size of pthread stacks.
  const char* const aixthread_guardpages = ::getenv("AIXTHREAD_GUARDPAGES");
  st->print_cr("  AIXTHREAD_GUARDPAGES=%s.",
      aixthread_guardpages ? aixthread_guardpages : "<unset>");
  st->cr();

  os::Aix::meminfo_t mi;
  if (os::Aix::get_meminfo(&mi)) {
    st->print_cr("physical total : %zu", mi.real_total);
    st->print_cr("physical free  : %zu", mi.real_free);
    st->print_cr("swap total     : %zu", mi.pgsp_total);
    st->print_cr("swap free      : %zu", mi.pgsp_free);
  }
  st->cr();

  // Print program break.
  st->print_cr("Program break at VM startup: " PTR_FORMAT ".", p2i(g_brk_at_startup));
  address brk_now = (address)::sbrk(0);
  if (brk_now != (address)-1) {
    st->print_cr("Program break now          : " PTR_FORMAT " (distance: %zuk).",
                 p2i(brk_now), (size_t)((brk_now - g_brk_at_startup) / K));
  }
  st->print_cr("MaxExpectedDataSegmentSize    : %zuk.", MaxExpectedDataSegmentSize / K);
  st->cr();

  // Print segments allocated with os::reserve_memory.
  st->print_cr("internal virtual memory regions used by vm:");
  vmembk_print_on(st);
}

// Get a string for the cpuinfo that is a summary of the cpu type
void os::get_summary_cpu_info(char* buf, size_t buflen) {
  // read _system_configuration.version
  switch (_system_configuration.version) {
  case PV_11:
    strncpy(buf, "Power PC 11", buflen);
    break;
  case PV_10:
    strncpy(buf, "Power PC 10", buflen);
    break;
  case PV_9:
    strncpy(buf, "Power PC 9", buflen);
    break;
  case PV_8:
    strncpy(buf, "Power PC 8", buflen);
    break;
  case PV_8_Compat:
    strncpy(buf, "PV_8_Compat", buflen);
    break;
  case PV_9_Compat:
    strncpy(buf, "PV_9_Compat", buflen);
    break;
  case PV_10_Compat:
    strncpy(buf, "PV_10_Compat", buflen);
    break;
  case PV_11_Compat:
    strncpy(buf, "PV_11_Compat", buflen);
    break;
  default:
    strncpy(buf, "unknown", buflen);
  }
}

void os::pd_print_cpu_info(outputStream* st, char* buf, size_t buflen) {
  // Nothing to do beyond of what os::print_cpu_info() does.
}

////////////////////////////////////////////////////////////////////////////////
// Virtual Memory

// We need to keep small simple bookkeeping for os::reserve_memory and friends.

#define VMEM_MAPPED  1
#define VMEM_SHMATED 2

struct vmembk_t {
  int type;         // 1 - mmap, 2 - shmat
  char* addr;
  size_t size;      // Real size, may be larger than usersize.
  size_t pagesize;  // page size of area
  vmembk_t* next;

  bool contains_addr(char* p) const {
    return p >= addr && p < (addr + size);
  }

  bool contains_range(char* p, size_t s) const {
    return contains_addr(p) && contains_addr(p + s - 1);
  }

  void print_on(outputStream* os) const {
    os->print("[" PTR_FORMAT " - " PTR_FORMAT "] (%zu"
      " bytes, %ld %s pages), %s",
      p2i(addr), p2i(addr) + size - 1, size, size / pagesize, describe_pagesize(pagesize),
      (type == VMEM_SHMATED ? "shmat" : "mmap")
    );
  }

  // Check that range is a sub range of memory block (or equal to memory block);
  // also check that range is fully page aligned to the page size if the block.
  void assert_is_valid_subrange(char* p, size_t s) const {
    if (!contains_range(p, s)) {
      fatal(RANGEFMT " is not a sub range of " RANGEFMT, RANGEFMTARGS(p, s),
            RANGEFMTARGS(addr, size));
    }
    if (!is_aligned_to(p, pagesize) || !is_aligned_to(p + s, pagesize)) {
      fatal("range " RANGEFMT " is not aligned to pagesize (%lu)",
            RANGEFMTARGS(p, s), (unsigned long)pagesize);
    }
  }
};

static struct {
  vmembk_t* first;
  MiscUtils::CritSect cs;
} vmem;

static void vmembk_add(char* addr, size_t size, size_t pagesize, int type) {
  vmembk_t* p = (vmembk_t*) permit_forbidden_function::malloc(sizeof(vmembk_t));
  assert0(p);
  if (p) {
    MiscUtils::AutoCritSect lck(&vmem.cs);
    p->addr = addr; p->size = size;
    p->pagesize = pagesize;
    p->type = type;
    p->next = vmem.first;
    vmem.first = p;
  }
}

static vmembk_t* vmembk_find(char* addr) {
  MiscUtils::AutoCritSect lck(&vmem.cs);
  for (vmembk_t* p = vmem.first; p; p = p->next) {
    if (p->addr <= addr && (p->addr + p->size) > addr) {
      return p;
    }
  }
  return nullptr;
}

static void vmembk_remove(vmembk_t* p0) {
  MiscUtils::AutoCritSect lck(&vmem.cs);
  assert0(p0);
  assert0(vmem.first); // List should not be empty.
  for (vmembk_t** pp = &(vmem.first); *pp; pp = &((*pp)->next)) {
    if (*pp == p0) {
      *pp = p0->next;
      permit_forbidden_function::free(p0);
      return;
    }
  }
  assert0(false); // Not found?
}

static void vmembk_print_on(outputStream* os) {
  MiscUtils::AutoCritSect lck(&vmem.cs);
  for (vmembk_t* vmi = vmem.first; vmi; vmi = vmi->next) {
    vmi->print_on(os);
    os->cr();
  }
}

// Reserve and attach a section of System V memory.
// If <requested_addr> is not null, function will attempt to attach the memory at the given
// address. Failing that, it will attach the memory anywhere.
// If <requested_addr> is null, function will attach the memory anywhere.
static char* reserve_shmated_memory (size_t bytes, char* requested_addr) {

  trcVerbose("reserve_shmated_memory %zu bytes, wishaddress "
    PTR_FORMAT "...", bytes, p2i(requested_addr));

  // We must prevent anyone from attaching too close to the
  // BRK because that may cause malloc OOM.
  if (requested_addr != nullptr && is_close_to_brk((address)requested_addr)) {
    log_info(os, map)("Wish address " PTR_FORMAT
                      " is too close to the BRK segment.",
                      p2i(requested_addr));
    // Since we treat an attach to the wrong address as an error later anyway,
    // we return null here
    return nullptr;
  }

  // Align size of shm up to 64K to avoid errors if we later try to change the page size.
  const size_t size = align_up(bytes, 64*K);

  // Reserve the shared segment.
  int shmid = shmget(IPC_PRIVATE, size, IPC_CREAT | S_IRUSR | S_IWUSR);
  if (shmid == -1) {
    ErrnoPreserver ep;
    log_trace(os, map)("shmget(.., %zu, ..) failed (errno=%s).",
                       size, os::strerror(ep.saved_errno()));
    return nullptr;
  }

  // Important note:
  // It is very important that we, upon leaving this function, do not leave a shm segment alive.
  // We must right after attaching it remove it from the system. System V shm segments are global and
  // survive the process.
  // So, from here on: Do not assert, do not return, until we have called shmctl(IPC_RMID) (A).

  struct shmid_ds shmbuf;
  memset(&shmbuf, 0, sizeof(shmbuf));
  shmbuf.shm_pagesize = 64*K;
  if (shmctl(shmid, SHM_PAGESIZE, &shmbuf) != 0) {
    assert(false,
           "Failed to set page size (need %zu"
           " 64K pages) - shmctl failed. (errno=%s).",
           size / (64 * K), os::strerror(os::get_last_error()));
  }

  // Now attach the shared segment.
  // Note that we deliberately *don't* pass SHM_RND. The contract of os::attempt_reserve_memory_at() -
  // which invokes this function with a request address != nullptr - is to map at the specified address
  // excactly, or to fail. If the caller passed us an address that is not usable (aka not a valid segment
  // boundary), shmat should not round down the address, or think up a completely new one.
  // (In places where this matters, e.g. when reserving the heap, we take care of passing segment-aligned
  // addresses on Aix. See, e.g., ReservedHeapSpace.
  char* const addr = (char*) shmat(shmid, requested_addr, 0);
  const int errno_shmat = errno;

  // (A) Right after shmat and before handing shmat errors delete the shm segment.
  if (::shmctl(shmid, IPC_RMID, nullptr) == -1) {
    ErrnoPreserver ep;
    log_trace(os, map)("shmctl(%u, IPC_RMID) failed (errno=%s)\n",
                       shmid,
                       os::strerror(ep.saved_errno()));
    assert(false, "failed to remove shared memory segment!");
  }

  // Handle shmat error. If we failed to attach, just return.
  if (addr == (char*)-1) {
    ErrnoPreserver ep;
    log_trace(os, map)("Failed to attach segment at " PTR_FORMAT " (errno=%s).",
                       p2i(requested_addr),
                       os::strerror(ep.saved_errno()));
    return nullptr;
  }

  // Just for info: query the real page size. In case setting the page size did not
  // work (see above), the system may have given us something other then 4K (LDR_CNTRL).
  const size_t real_pagesize = os::Aix::query_pagesize(addr);
  if (real_pagesize != (size_t)shmbuf.shm_pagesize) {
    log_trace(os, map)("pagesize is, surprisingly, %zu",
                       real_pagesize);
  }

  if (addr) {
    log_trace(os, map)("shm-allocated succeeded: " RANGEFMT
                       " (%zu %s pages)",
                       RANGEFMTARGS(addr, size),
                       size / real_pagesize,
                       describe_pagesize(real_pagesize));
  } else {
    if (requested_addr != nullptr) {
      log_trace(os, map)("shm-allocate failed: " RANGEFMT,
                         RANGEFMTARGS(requested_addr, size));
    } else {
      log_trace(os, map)("failed to shm-allocate %zu"
                         " bytes at any address.",
                         size);
    }
  }

  // book-keeping
  vmembk_add(addr, size, real_pagesize, VMEM_SHMATED);
  assert0(is_aligned_to(addr, os::vm_page_size()));

  return addr;
}

static bool release_shmated_memory(char* addr, size_t size) {

  trcVerbose("release_shmated_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
    p2i(addr), p2i(addr + size - 1));

  bool rc = false;

  // TODO: is there a way to verify shm size without doing bookkeeping?
  if (::shmdt(addr) != 0) {
    ErrnoPreserver ep;
    log_trace(os, map)("shmdt failed: " RANGEFMT " errno=(%s)",
                       RANGEFMTARGS(addr, size),
                       os::strerror(ep.saved_errno()));
  } else {
    log_trace(os, map)("shmdt succeded: " RANGEFMT,
                       RANGEFMTARGS(addr, size));
    rc = true;
  }
  return rc;
}

static bool uncommit_shmated_memory(char* addr, size_t size) {
  trcVerbose("uncommit_shmated_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
    p2i(addr), p2i(addr + size - 1));

  const int rc = disclaim64(addr, size, DISCLAIM_ZEROMEM);

  if (rc != 0) {
    ErrnoPreserver ep;
    log_warning(os)("disclaim64(" PTR_FORMAT ", %zu) failed, %s\n", p2i(addr), size, os::strerror(ep.saved_errno()));
    return false;
  }
  return true;
}

////////////////////////////////  mmap-based routines /////////////////////////////////

// Reserve memory via mmap.
// If <requested_addr> is given, an attempt is made to attach at the given address.
// Failing that, memory is allocated at any address.
static char* reserve_mmaped_memory(size_t bytes, char* requested_addr) {
  trcVerbose("reserve_mmaped_memory %zu bytes, wishaddress " PTR_FORMAT "...",
    bytes, p2i(requested_addr));

  if (requested_addr && !is_aligned_to(requested_addr, os::vm_page_size()) != 0) {
    log_trace(os, map)("Wish address " PTR_FORMAT
                       " not aligned to page boundary.",
                       p2i(requested_addr));
    return nullptr;
  }

  // We must prevent anyone from attaching too close to the
  // BRK because that may cause malloc OOM.
  if (requested_addr != nullptr && is_close_to_brk((address)requested_addr)) {
    log_trace(os, map)("Wish address " PTR_FORMAT
                       " is too close to the BRK segment.",
                       p2i(requested_addr));
    // Since we treat an attach to the wrong address as an error later anyway,
    // we return null here
    return nullptr;
  }

  // In 64K mode, we lie and claim the global page size (os::vm_page_size()) is 64K
  //  (complicated story). This mostly works just fine since 64K is a multiple of the
  //  actual 4K lowest page size. Only at a few seams light shines thru, e.g. when
  //  calling mmap. mmap will return memory aligned to the lowest pages size - 4K -
  //  so we must make sure - transparently - that the caller only ever sees 64K
  //  aligned mapping start addresses.
  const size_t alignment = os::vm_page_size();

  // Size shall always be a multiple of os::vm_page_size (esp. in 64K mode).
  const size_t size = align_up(bytes, os::vm_page_size());

  // alignment: Allocate memory large enough to include an aligned range of the right size and
  // cut off the leading and trailing waste pages.
  assert0(alignment != 0 && is_aligned_to(alignment, os::vm_page_size())); // see above
  const size_t extra_size = size + alignment;

  // Note: MAP_SHARED (instead of MAP_PRIVATE) needed to be able to
  // later use msync(MS_INVALIDATE) (see os::uncommit_memory).
  int flags = MAP_ANONYMOUS | MAP_SHARED;

  if (os::vm_page_size() == 64*K && g_multipage_support.can_use_64K_mmap_pages) {
    flags |= MAP_ANON_64K;
  }

  // MAP_FIXED is needed to enforce requested_addr - manpage is vague about what
  // it means if wishaddress is given but MAP_FIXED is not set.
  //
  // Important! Behaviour differs depending on whether SPEC1170 mode is active or not.
  // SPEC1170 mode active: behaviour like POSIX, MAP_FIXED will clobber existing mappings.
  // SPEC1170 mode not active: behaviour, unlike POSIX, is that no existing mappings will
  // get clobbered.
  if (requested_addr != nullptr) {
    if (!os::Aix::xpg_sus_mode()) {  // not SPEC1170 Behaviour
      flags |= MAP_FIXED;
    }
  }

  char* addr = (char*)::mmap(requested_addr, extra_size,
      PROT_READ|PROT_WRITE|PROT_EXEC, flags, -1, 0);

  if (addr == MAP_FAILED) {
    ErrnoPreserver ep;
    log_trace(os, map)("mmap failed: " RANGEFMT " errno=(%s)",
                       RANGEFMTARGS(requested_addr, size),
                       os::strerror(ep.saved_errno()));
    return nullptr;
  } else if (requested_addr != nullptr && addr != requested_addr) {
    log_trace(os, map)("mmap succeeded: " RANGEFMT
                       ", but at a different address than"
                       "requested (" PTR_FORMAT "), will unmap",
                       RANGEFMTARGS(requested_addr, size),
                       p2i(addr));
    ::munmap(addr, extra_size);
    return nullptr;
  }

  // Handle alignment.
  char* const addr_aligned = align_up(addr, alignment);
  const size_t waste_pre = addr_aligned - addr;
  char* const addr_aligned_end = addr_aligned + size;
  const size_t waste_post = extra_size - waste_pre - size;
  if (waste_pre > 0) {
    ::munmap(addr, waste_pre);
  }
  if (waste_post > 0) {
    ::munmap(addr_aligned_end, waste_post);
  }
  addr = addr_aligned;

  trcVerbose("mmap-allocated " PTR_FORMAT " .. " PTR_FORMAT " (%zu bytes)",
    p2i(addr), p2i(addr + bytes), bytes);

  // bookkeeping
  if (os::vm_page_size() == 64*K && g_multipage_support.can_use_64K_mmap_pages) {
    vmembk_add(addr, size, 64*K, VMEM_MAPPED);
  } else {
    vmembk_add(addr, size, 4*K, VMEM_MAPPED);
  }

  // Test alignment, see above.
  assert0(is_aligned_to(addr, os::vm_page_size()));

  return addr;
}

static bool release_mmaped_memory(char* addr, size_t size) {
  assert0(is_aligned_to(addr, os::vm_page_size()));
  assert0(is_aligned_to(size, os::vm_page_size()));

  trcVerbose("release_mmaped_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
    p2i(addr), p2i(addr + size - 1));
  bool rc = false;

  if (::munmap(addr, size) != 0) {
    ErrnoPreserver ep;
    log_trace(os, map)("munmap failed: " RANGEFMT " errno=(%s)",
                       RANGEFMTARGS(addr, size),
                       os::strerror(ep.saved_errno()));
    rc = false;
  } else {
    log_trace(os, map)("munmap succeeded: " RANGEFMT,
                       RANGEFMTARGS(addr, size));
    rc = true;
  }

  return rc;
}

static bool uncommit_mmaped_memory(char* addr, size_t size) {

  assert0(is_aligned_to(addr, os::vm_page_size()));
  assert0(is_aligned_to(size, os::vm_page_size()));

  trcVerbose("uncommit_mmaped_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
    p2i(addr), p2i(addr + size - 1));
  bool rc = false;

  // Uncommit mmap memory with msync MS_INVALIDATE.
  if (::msync(addr, size, MS_INVALIDATE) != 0) {
    ErrnoPreserver ep;
    log_trace(os, map)("msync failed: " RANGEFMT " errno=(%s)",
                       RANGEFMTARGS(addr, size),
                       os::strerror(ep.saved_errno()));
    rc = false;
  } else {
    log_trace(os, map)("msync succeeded: " RANGEFMT,
                       RANGEFMTARGS(addr, size));
    rc = true;
  }

  return rc;
}

#ifdef PRODUCT
static void warn_fail_commit_memory(char* addr, size_t size, bool exec,
                                    int err) {
  warning("INFO: os::commit_memory(" PTR_FORMAT ", %zu, %d) failed; error='%s' (errno=%d)",
          p2i(addr), size, exec, os::errno_name(err), err);
}
#endif

void os::pd_commit_memory_or_exit(char* addr, size_t size, bool exec,
                                  const char* mesg) {
  assert(mesg != nullptr, "mesg must be specified");
  if (!pd_commit_memory(addr, size, exec)) {
    // Add extra info in product mode for vm_exit_out_of_memory():
    PRODUCT_ONLY(warn_fail_commit_memory(addr, size, exec, errno);)
    vm_exit_out_of_memory(size, OOM_MMAP_ERROR, "%s", mesg);
  }
}

bool os::pd_commit_memory(char* addr, size_t size, bool exec) {

  assert(is_aligned_to(addr, os::vm_page_size()),
    "addr " PTR_FORMAT " not aligned to vm_page_size (%zu)",
    p2i(addr), os::vm_page_size());
  assert(is_aligned_to(size, os::vm_page_size()),
    "size " PTR_FORMAT " not aligned to vm_page_size (%zu)",
    size, os::vm_page_size());

  vmembk_t* const vmi = vmembk_find(addr);
  guarantee0(vmi);
  vmi->assert_is_valid_subrange(addr, size);

  log_info(os)("commit_memory [" PTR_FORMAT " - " PTR_FORMAT "].", p2i(addr), p2i(addr + size - 1));

  if (UseExplicitCommit) {
    // AIX commits memory on touch. So, touch all pages to be committed.
    for (char* p = addr; p < (addr + size); p += 4*K) {
      *p = '\0';
    }
  }

  return true;
}

bool os::pd_commit_memory(char* addr, size_t size, size_t alignment_hint, bool exec) {
  return pd_commit_memory(addr, size, exec);
}

void os::pd_commit_memory_or_exit(char* addr, size_t size,
                                  size_t alignment_hint, bool exec,
                                  const char* mesg) {
  // Alignment_hint is ignored on this OS.
  pd_commit_memory_or_exit(addr, size, exec, mesg);
}

bool os::pd_uncommit_memory(char* addr, size_t size, bool exec) {
  assert(is_aligned_to(addr, os::vm_page_size()),
    "addr " PTR_FORMAT " not aligned to vm_page_size (%zu)",
    p2i(addr), os::vm_page_size());
  assert(is_aligned_to(size, os::vm_page_size()),
    "size " PTR_FORMAT " not aligned to vm_page_size (%zu)",
    size, os::vm_page_size());

  // Dynamically do different things for mmap/shmat.
  const vmembk_t* const vmi = vmembk_find(addr);
  guarantee0(vmi);
  vmi->assert_is_valid_subrange(addr, size);

  if (vmi->type == VMEM_SHMATED) {
    return uncommit_shmated_memory(addr, size);
  } else {
    return uncommit_mmaped_memory(addr, size);
  }
}

bool os::pd_create_stack_guard_pages(char* addr, size_t size) {
  // Do not call this; no need to commit stack pages on AIX.
  ShouldNotReachHere();
  return true;
}

void os::remove_stack_guard_pages(char* addr, size_t size) {
  // Do not call this; no need to commit stack pages on AIX.
  ShouldNotReachHere();
}

void os::pd_realign_memory(char *addr, size_t bytes, size_t alignment_hint) {
}

void os::pd_disclaim_memory(char *addr, size_t bytes) {
}

size_t os::pd_pretouch_memory(void* first, void* last, size_t page_size) {
  return page_size;
}

void os::numa_set_thread_affinity(Thread *thread, int node) {
}

void os::numa_make_global(char *addr, size_t bytes) {
}

void os::numa_make_local(char *addr, size_t bytes, int lgrp_hint) {
}

size_t os::numa_get_groups_num() {
  return 1;
}

int os::numa_get_group_id() {
  return 0;
}

size_t os::numa_get_leaf_groups(uint *ids, size_t size) {
  if (size > 0) {
    ids[0] = 0;
    return 1;
  }
  return 0;
}

int os::numa_get_group_id_for_address(const void* address) {
  return 0;
}

bool os::numa_get_group_ids_for_range(const void** addresses, int* lgrp_ids, size_t count) {
  return false;
}

// Reserves and attaches a shared memory segment.
char* os::pd_reserve_memory(size_t bytes, bool exec) {
  // Always round to os::vm_page_size(), which may be larger than 4K.
  bytes = align_up(bytes, os::vm_page_size());

  // In 4K mode always use mmap.
  // In 64K mode allocate with mmap if it supports 64K pages, otherwise use 64K shmatted.
  if (os::vm_page_size() == 4*K || g_multipage_support.can_use_64K_mmap_pages) {
    return reserve_mmaped_memory(bytes, nullptr /* requested_addr */);
  } else {
    return reserve_shmated_memory(bytes, nullptr /* requested_addr */);
  }
}

bool os::pd_release_memory(char* addr, size_t size) {

  // Dynamically do different things for mmap/shmat.
  vmembk_t* const vmi = vmembk_find(addr);
  guarantee0(vmi);
  vmi->assert_is_valid_subrange(addr, size);

  // Always round to os::vm_page_size(), which may be larger than 4K.
  size = align_up(size, os::vm_page_size());
  addr = align_up(addr, os::vm_page_size());

  bool rc = false;
  bool remove_bookkeeping = false;
  if (vmi->type == VMEM_SHMATED) {
    // For shmatted memory, we do:
    // - If user wants to release the whole range, release the memory (shmdt).
    // - If user only wants to release a partial range, uncommit (disclaim) that
    //   range. That way, at least, we do not use memory anymore (bust still page
    //   table space).
    if (addr == vmi->addr && size == vmi->size) {
      rc = release_shmated_memory(addr, size);
      remove_bookkeeping = true;
    } else {
      rc = uncommit_shmated_memory(addr, size);
    }
  } else {
    // In mmap-mode:
    //  - If the user wants to release the full range, we do that and remove the mapping.
    //  - If the user wants to release part of the range, we release that part, but need
    //    to adjust bookkeeping.
    assert(is_aligned(size, 4 * K), "Sanity");
    rc = release_mmaped_memory(addr, size);
    if (addr == vmi->addr && size == vmi->size) {
      remove_bookkeeping = true;
    } else {
      if (addr == vmi->addr && size < vmi->size) {
        // Chopped from head
        vmi->addr += size;
        vmi->size -= size;
      } else if (addr + size == vmi->addr + vmi->size) {
        // Chopped from tail
        vmi->size -= size;
      } else {
        // releasing a mapping in the middle of the original mapping:
        // For now we forbid this, since this is an invalid scenario
        // (the bookkeeping is easy enough to fix if needed but there
        //  is no use case for it; any occurrence is likely an error.
        ShouldNotReachHere();
      }
    }
  }

  // update bookkeeping
  if (rc && remove_bookkeeping) {
    vmembk_remove(vmi);
  }

  return rc;
}

static bool checked_mprotect(char* addr, size_t size, int prot) {

  // Little problem here: if SPEC1170 behaviour is off, mprotect() on AIX will
  // not tell me if protection failed when trying to protect an un-protectable range.
  //
  // This means if the memory was allocated using shmget/shmat, protection won't work
  // but mprotect will still return 0:
  //
  // See http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.basetechref/doc/basetrf1/mprotect.htm

  Events::log_memprotect(nullptr, "Protecting memory [" INTPTR_FORMAT "," INTPTR_FORMAT "] with protection modes %x", p2i(addr), p2i(addr+size), prot);
  bool rc = ::mprotect(addr, size, prot) == 0 ? true : false;

  if (!rc) {
    const char* const s_errno = os::errno_name(errno);
    warning("mprotect(" PTR_FORMAT "-" PTR_FORMAT ", 0x%X) failed (%s).", p2i(addr), p2i(addr) + size, prot, s_errno);
    return false;
  }

  // mprotect success check
  //
  // Mprotect said it changed the protection but can I believe it?
  //
  // To be sure I need to check the protection afterwards. Try to
  // read from protected memory and check whether that causes a segfault.
  //
  if (!os::Aix::xpg_sus_mode()) {

    const bool read_protected =
      (SafeFetch32((int*)addr, 0x12345678) == 0x12345678 &&
       SafeFetch32((int*)addr, 0x76543210) == 0x76543210) ? true : false;

    if (prot & PROT_READ) {
      rc = !read_protected;
    } else {
      rc = read_protected;
    }
  }

  if (!rc) {
    // Reporting success via mprotect but failing to mprotect is a symptom of calling mprotect
    // on SystemV-memory. It should not happen for mmap-mode.
    assert(!g_multipage_support.can_use_64K_mmap_pages, "Should only happen in old-style shmat mode");
  }

  return rc;
}

// Set protections specified
bool os::protect_memory(char* addr, size_t size, ProtType prot, bool is_committed) {
  unsigned int p = 0;
  switch (prot) {
  case MEM_PROT_NONE: p = PROT_NONE; break;
  case MEM_PROT_READ: p = PROT_READ; break;
  case MEM_PROT_RW:   p = PROT_READ|PROT_WRITE; break;
  case MEM_PROT_RWX:  p = PROT_READ|PROT_WRITE|PROT_EXEC; break;
  default:
    ShouldNotReachHere();
  }
  // is_committed is unused.
  return checked_mprotect(addr, size, p);
}

bool os::guard_memory(char* addr, size_t size) {
  return checked_mprotect(addr, size, PROT_NONE);
}

bool os::unguard_memory(char* addr, size_t size) {
  return checked_mprotect(addr, size, PROT_READ|PROT_WRITE|PROT_EXEC);
}

// Large page support

static size_t _large_page_size = 0;

// Enable large page support if OS allows that.
void os::large_page_init() {
  return; // Nothing to do. See query_multipage_support and friends.
}

char* os::pd_reserve_memory_special(size_t bytes, size_t alignment, size_t page_size, char* req_addr, bool exec) {
  fatal("os::reserve_memory_special should not be called on AIX.");
  return nullptr;
}

size_t os::large_page_size() {
  return _large_page_size;
}

bool os::can_commit_large_page_memory() {
  // Does not matter, we do not support huge pages.
  return false;
}

char* os::pd_attempt_map_memory_to_file_at(char* requested_addr, size_t bytes, int file_desc) {
  assert(file_desc >= 0, "file_desc is not valid");
  char* result = nullptr;

  // Always round to os::vm_page_size(), which may be larger than 4K.
  bytes = align_up(bytes, os::vm_page_size());
  result = reserve_mmaped_memory(bytes, requested_addr);

  if (result != nullptr) {
    if (replace_existing_mapping_with_file_mapping(result, bytes, file_desc) == nullptr) {
      vm_exit_during_initialization(err_msg("Error in mapping Java heap at the given filesystem directory"));
    }
  }
  return result;
}

// Reserve memory at an arbitrary address, only if that area is
// available (and not reserved for something else).
char* os::pd_attempt_reserve_memory_at(char* requested_addr, size_t bytes, bool exec) {
  char* addr = nullptr;

  // Always round to os::vm_page_size(), which may be larger than 4K.
  bytes = align_up(bytes, os::vm_page_size());

  // In 4K mode always use mmap.
  // In 64K mode allocate with mmap if it supports 64K pages, otherwise use 64K shmatted.
  if (os::vm_page_size() == 4*K || g_multipage_support.can_use_64K_mmap_pages) {
    return reserve_mmaped_memory(bytes, requested_addr);
  } else {
    return reserve_shmated_memory(bytes, requested_addr);
  }

  return addr;
}

size_t os::vm_min_address() {
  // On AIX, we need to make sure we don't block the sbrk. However, this is
  // done at actual reservation time, where we honor a "no-mmap" area following
  // the break. See MaxExpectedDataSegmentSize. So we can return a very low
  // address here.
  assert(is_aligned(_vm_min_address_default, os::vm_allocation_granularity()), "Sanity");
  return _vm_min_address_default;
}

////////////////////////////////////////////////////////////////////////////////
// thread priority support

// From AIX manpage to pthread_setschedparam
// (see: http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?
//    topic=/com.ibm.aix.basetechref/doc/basetrf1/pthread_setschedparam.htm):
//
// "If schedpolicy is SCHED_OTHER, then sched_priority must be in the
// range from 40 to 80, where 40 is the least favored priority and 80
// is the most favored."
//
// (Actually, I doubt this even has an impact on AIX, as we do kernel
// scheduling there; however, this still leaves iSeries.)
//
int os::java_to_os_priority[CriticalPriority + 1] = {
  54,             // 0 Entry should never be used

  55,             // 1 MinPriority
  55,             // 2
  56,             // 3

  56,             // 4
  57,             // 5 NormPriority
  57,             // 6

  58,             // 7
  58,             // 8
  59,             // 9 NearMaxPriority

  60,             // 10 MaxPriority

  60              // 11 CriticalPriority
};

static int prio_init() {
  if (ThreadPriorityPolicy == 1) {
    if (geteuid() != 0) {
      if (!FLAG_IS_DEFAULT(ThreadPriorityPolicy) && !FLAG_IS_JIMAGE_RESOURCE(ThreadPriorityPolicy)) {
        warning("-XX:ThreadPriorityPolicy=1 may require system level permission, " \
                "e.g., being the root user. If the necessary permission is not " \
                "possessed, changes to priority will be silently ignored.");
      }
    }
  }
  if (UseCriticalJavaThreadPriority) {
    os::java_to_os_priority[MaxPriority] = os::java_to_os_priority[CriticalPriority];
  }
  return 0;
}

OSReturn os::set_native_priority(Thread* thread, int newpri) {
  if (!UseThreadPriorities || ThreadPriorityPolicy == 0) return OS_OK;
  pthread_t thr = thread->osthread()->pthread_id();
  int policy = SCHED_OTHER;
  struct sched_param param;
  param.sched_priority = newpri;
  int ret = pthread_setschedparam(thr, policy, &param);

  if (ret != 0) {
    log_warning(os)("Could not change priority for thread %d to %d (error %d, %s)",
        (int)thr, newpri, ret, os::errno_name(ret));
  }
  return (ret == 0) ? OS_OK : OS_ERR;
}

OSReturn os::get_native_priority(const Thread* const thread, int *priority_ptr) {
  if (!UseThreadPriorities || ThreadPriorityPolicy == 0) {
    *priority_ptr = java_to_os_priority[NormPriority];
    return OS_OK;
  }
  pthread_t thr = thread->osthread()->pthread_id();
  int policy = SCHED_OTHER;
  struct sched_param param;
  int ret = pthread_getschedparam(thr, &policy, &param);
  *priority_ptr = param.sched_priority;

  return (ret == 0) ? OS_OK : OS_ERR;
}

// To install functions for atexit system call
extern "C" {
  static void perfMemory_exit_helper() {
    perfMemory_exit();
  }
}

static void set_page_size(size_t page_size) {
  OSInfo::set_vm_page_size(page_size);
  OSInfo::set_vm_allocation_granularity(page_size);
}

// This is called _before_ the most of global arguments have been parsed.
void os::init(void) {
  // This is basic, we want to know if that ever changes.
  // (Shared memory boundary is supposed to be a 256M aligned.)
  assert(SHMLBA == ((uint64_t)0x10000000ULL)/*256M*/, "unexpected");

  // Record process break at startup.
  g_brk_at_startup = (address) ::sbrk(0);
  assert(g_brk_at_startup != (address) -1, "sbrk failed");

  // First off, we need to know the OS level we run on.
  os::Aix::initialize_os_info();

  // Scan environment (SPEC1170 behaviour, etc).
  os::Aix::scan_environment();

  // Probe multipage support.
  query_multipage_support();

  // Act like we only have one page size by eliminating corner cases which
  // we did not support very well anyway.
  // We have two input conditions:
  // 1) Data segment page size. This is controlled by linker setting (datapsize) on the
  //    launcher, and/or by LDR_CNTRL environment variable. The latter overrules the linker
  //    setting.
  //    Data segment page size is important for us because it defines the thread stack page
  //    size, which is needed for guard page handling, stack banging etc.
  // 2) The ability to allocate 64k pages dynamically. If this is a given, java heap can
  //    and should be allocated with 64k pages.
  //
  // So, we do the following:
  // LDR_CNTRL    can_use_64K_pages_dynamically(mmap or shm)       what we do                      remarks
  // 4K           no                                               4K                              old systems (aix 5.2) or new systems with AME activated
  // 4k           yes                                              64k (treat 4k stacks as 64k)    different loader than java and standard settings
  // 64k          no              --- AIX 5.2 ? ---
  // 64k          yes                                              64k                             new systems and standard java loader (we set datapsize=64k when linking)

  // datapsize = 64k. Data segment, thread stacks are 64k paged.
  // This normally means that we can allocate 64k pages dynamically.
  assert0(g_multipage_support.can_use_64K_pages || g_multipage_support.can_use_64K_mmap_pages);
  set_page_size(64*K);

  // For now UseLargePages is just ignored.
  FLAG_SET_ERGO(UseLargePages, false);
  _page_sizes.add(os::vm_page_size());

  // debug trace
  trcVerbose("os::vm_page_size %s", describe_pagesize(os::vm_page_size()));

  // Next, we need to initialize libperfstat
  os::Aix::initialize_libperfstat();

  // Reset the perfstat information provided by ODM.
  libperfstat::perfstat_reset();

  // Now initialize basic system properties. Note that for some of the values we
  // need libperfstat etc.
  os::Aix::initialize_system_info();

  // _main_thread points to the thread that created/loaded the JVM.
  Aix::_main_thread = pthread_self();

  os::Posix::init();
}

// This is called _after_ the global arguments have been parsed.
jint os::init_2(void) {

  // This could be set after os::Posix::init() but all platforms
  // have to set it the same so we have to mirror Solaris.
  DEBUG_ONLY(os::set_mutex_init_done();)

  os::Posix::init_2();

  trcVerbose("processor count: %d", os::_processor_count);
  trcVerbose("physical memory: " PHYS_MEM_TYPE_FORMAT, Aix::_physical_memory);

  // Initially build up the loaded dll map.
  LoadedLibraries::reload();
  if (Verbose) {
    trcVerbose("Loaded Libraries: ");
    LoadedLibraries::print(tty);
  }

  if (PosixSignals::init() == JNI_ERR) {
    return JNI_ERR;
  }

  // Check and sets minimum stack sizes against command line options
  if (set_minimum_stack_sizes() == JNI_ERR) {
    return JNI_ERR;
  }

  // Not supported.
  FLAG_SET_ERGO(UseNUMA, false);
  FLAG_SET_ERGO(UseNUMAInterleaving, false);

  if (MaxFDLimit) {
    // Set the number of file descriptors to max. print out error
    // if getrlimit/setrlimit fails but continue regardless.
    struct rlimit nbr_files;
    int status = getrlimit(RLIMIT_NOFILE, &nbr_files);
    if (status != 0) {
      log_info(os)("os::init_2 getrlimit failed: %s", os::strerror(errno));
    } else {
      nbr_files.rlim_cur = nbr_files.rlim_max;
      status = setrlimit(RLIMIT_NOFILE, &nbr_files);
      if (status != 0) {
        log_info(os)("os::init_2 setrlimit failed: %s", os::strerror(errno));
      }
    }
  }

  if (PerfAllowAtExitRegistration) {
    // Only register atexit functions if PerfAllowAtExitRegistration is set.
    // At exit functions can be delayed until process exit time, which
    // can be problematic for embedded VM situations. Embedded VMs should
    // call DestroyJavaVM() to assure that VM resources are released.

    // Note: perfMemory_exit_helper atexit function may be removed in
    // the future if the appropriate cleanup code can be added to the
    // VM_Exit VMOperation's doit method.
    if (atexit(perfMemory_exit_helper) != 0) {
      warning("os::init_2 atexit(perfMemory_exit_helper) failed");
    }
  }

  // initialize thread priority policy
  prio_init();

  return JNI_OK;
}

int os::active_processor_count() {
  // User has overridden the number of active processors
  if (ActiveProcessorCount > 0) {
    log_trace(os)("active_processor_count: "
                  "active processor count set by user : %d",
                  ActiveProcessorCount);
    return ActiveProcessorCount;
  }

  return Machine::active_processor_count();
}

int os::Machine::active_processor_count() {
  int online_cpus = ::sysconf(_SC_NPROCESSORS_ONLN);
  assert(online_cpus > 0 && online_cpus <= processor_count(), "sanity check");
  return online_cpus;
}

void os::set_native_thread_name(const char *name) {
  // Not yet implemented.
  return;
}

////////////////////////////////////////////////////////////////////////////////
// debug support

bool os::find(address addr, outputStream* st) {

  st->print(PTR_FORMAT ": ", p2i(addr));

  loaded_module_t lm;
  if (LoadedLibraries::find_for_text_address(addr, &lm) ||
      LoadedLibraries::find_for_data_address(addr, &lm)) {
    st->print_cr("%s", lm.path);
    return true;
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
// misc

// This does not do anything on Aix. This is basically a hook for being
// able to use structured exception handling (thread-local exception filters)
// on, e.g., Win32.
void
os::os_exception_wrapper(java_call_t f, JavaValue* value, const methodHandle& method,
                         JavaCallArguments* args, JavaThread* thread) {
  f(value, method, args, thread);
}

// This code originates from JDK's sysOpen and open64_w
// from src/solaris/hpi/src/system_md.c

int os::open(const char *path, int oflag, int mode) {

  if (strlen(path) > MAX_PATH - 1) {
    errno = ENAMETOOLONG;
    return -1;
  }
  // AIX 7.X now supports O_CLOEXEC too, like modern Linux; but we have to be careful, see
  // IV90804: OPENING A FILE IN AFS WITH O_CLOEXEC FAILS WITH AN EINVAL ERROR APPLIES TO AIX 7100-04 17/04/14 PTF PECHANGE
  int oflag_with_o_cloexec = oflag | O_CLOEXEC;

  int fd = ::open(path, oflag_with_o_cloexec, mode);
  if (fd == -1) {
    // we might fail in the open call when O_CLOEXEC is set, so try again without (see IV90804)
    fd = ::open(path, oflag, mode);
    if (fd == -1) {
      return -1;
    }
  }

  // If the open succeeded, the file might still be a directory.
  {
    struct stat buf64;
    int ret = ::fstat(fd, &buf64);
    int st_mode = buf64.st_mode;

    if (ret != -1) {
      if ((st_mode & S_IFMT) == S_IFDIR) {
        ::close(fd);
        errno = EISDIR;
        return -1;
      }
    } else {
      ::close(fd);
      return -1;
    }
  }

  // All file descriptors that are opened in the JVM and not
  // specifically destined for a subprocess should have the
  // close-on-exec flag set. If we don't set it, then careless 3rd
  // party native code might fork and exec without closing all
  // appropriate file descriptors, and this in turn might:
  //
  // - cause end-of-file to fail to be detected on some file
  //   descriptors, resulting in mysterious hangs, or
  //
  // - might cause an fopen in the subprocess to fail on a system
  //   suffering from bug 1085341.

  // Validate that the use of the O_CLOEXEC flag on open above worked.
  static sig_atomic_t O_CLOEXEC_is_known_to_work = 0;
  if (O_CLOEXEC_is_known_to_work == 0) {
    int flags = ::fcntl(fd, F_GETFD);
    if (flags != -1) {
      if ((flags & FD_CLOEXEC) != 0) {
        O_CLOEXEC_is_known_to_work = 1;
      } else { // it does not work
        ::fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
        O_CLOEXEC_is_known_to_work = -1;
      }
    }
  } else if (O_CLOEXEC_is_known_to_work == -1) {
    int flags = ::fcntl(fd, F_GETFD);
    if (flags != -1) {
      ::fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
    }
  }

  return fd;
}

// current_thread_cpu_time(bool) and thread_cpu_time(Thread*, bool)
// are used by JVM M&M and JVMTI to get user+sys or user CPU time
// of a thread.
//
// current_thread_cpu_time() and thread_cpu_time(Thread*) returns
// the fast estimate available on the platform.

jlong os::current_thread_cpu_time() {
  // return user + sys since the cost is the same
  const jlong n = os::thread_cpu_time(Thread::current(), true /* user + sys */);
  assert(n >= 0, "negative CPU time");
  return n;
}

jlong os::thread_cpu_time(Thread* thread) {
  // consistent with what current_thread_cpu_time() returns
  const jlong n = os::thread_cpu_time(thread, true /* user + sys */);
  assert(n >= 0, "negative CPU time");
  return n;
}

jlong os::current_thread_cpu_time(bool user_sys_cpu_time) {
  const jlong n = os::thread_cpu_time(Thread::current(), user_sys_cpu_time);
  assert(n >= 0, "negative CPU time");
  return n;
}

static bool thread_cpu_time_unchecked(Thread* thread, jlong* p_sys_time, jlong* p_user_time) {
  bool error = false;

  jlong sys_time = 0;
  jlong user_time = 0;

  // Reimplemented using getthrds64().
  //
  // Works like this:
  // For the thread in question, get the kernel thread id. Then get the
  // kernel thread statistics using that id.
  //
  // This only works of course when no pthread scheduling is used,
  // i.e. there is a 1:1 relationship to kernel threads.
  // On AIX, see AIXTHREAD_SCOPE variable.

  pthread_t pthtid = thread->osthread()->pthread_id();

  // retrieve kernel thread id for the pthread:
  tid64_t tid = 0;
  struct __pthrdsinfo pinfo;
  // I just love those otherworldly IBM APIs which force me to hand down
  // dummy buffers for stuff I dont care for...
  char dummy[1];
  int dummy_size = sizeof(dummy);
  if (pthread_getthrds_np(&pthtid, PTHRDSINFO_QUERY_TID, &pinfo, sizeof(pinfo),
                          dummy, &dummy_size) == 0) {
    tid = pinfo.__pi_tid;
  } else {
    tty->print_cr("pthread_getthrds_np failed, errno: %d.", errno);
    error = true;
  }

  // retrieve kernel timing info for that kernel thread
  if (!error) {
    struct thrdentry64 thrdentry;
    if (getthrds64(getpid(), &thrdentry, sizeof(thrdentry), &tid, 1) == 1) {
      sys_time = thrdentry.ti_ru.ru_stime.tv_sec * 1000000000LL + thrdentry.ti_ru.ru_stime.tv_usec * 1000LL;
      user_time = thrdentry.ti_ru.ru_utime.tv_sec * 1000000000LL + thrdentry.ti_ru.ru_utime.tv_usec * 1000LL;
    } else {
      tty->print_cr("getthrds64 failed, errno: %d.", errno);
      error = true;
    }
  }

  if (p_sys_time) {
    *p_sys_time = sys_time;
  }

  if (p_user_time) {
    *p_user_time = user_time;
  }

  if (error) {
    return false;
  }

  return true;
}

jlong os::thread_cpu_time(Thread *thread, bool user_sys_cpu_time) {
  jlong sys_time;
  jlong user_time;

  if (!thread_cpu_time_unchecked(thread, &sys_time, &user_time)) {
    return -1;
  }

  return user_sys_cpu_time ? sys_time + user_time : user_time;
}

void os::current_thread_cpu_time_info(jvmtiTimerInfo *info_ptr) {
  info_ptr->max_value = all_bits_jlong;    // will not wrap in less than 64 bits
  info_ptr->may_skip_backward = false;     // elapsed time not wall time
  info_ptr->may_skip_forward = false;      // elapsed time not wall time
  info_ptr->kind = JVMTI_TIMER_TOTAL_CPU;  // user+system time is returned
}

void os::thread_cpu_time_info(jvmtiTimerInfo *info_ptr) {
  info_ptr->max_value = all_bits_jlong;    // will not wrap in less than 64 bits
  info_ptr->may_skip_backward = false;     // elapsed time not wall time
  info_ptr->may_skip_forward = false;      // elapsed time not wall time
  info_ptr->kind = JVMTI_TIMER_TOTAL_CPU;  // user+system time is returned
}

bool os::is_thread_cpu_time_supported() {
  return true;
}

// System loadavg support. Returns -1 if load average cannot be obtained.
// For now just return the system wide load average (no processor sets).
int os::loadavg(double values[], int nelem) {

  guarantee(nelem >= 0 && nelem <= 3, "argument error");
  guarantee(values, "argument error");

  // AIX: use libperfstat
  libperfstat::cpuinfo_t ci;
  if (libperfstat::get_cpuinfo(&ci)) {
    for (int i = 0; i < nelem; i++) {
      values[i] = ci.loadavg[i];
    }
  } else {
    return -1;
  }
  return nelem;
}

bool os::is_primordial_thread(void) {
  if (pthread_self() == (pthread_t)1) {
    return true;
  } else {
    return false;
  }
}

// OS recognitions (OS level) call this before calling Aix::os_version()
void os::Aix::initialize_os_info() {

  assert(_os_version == 0, "already called.");

  struct utsname uts;
  memset(&uts, 0, sizeof(uts));
  strcpy(uts.sysname, "?");
  if (::uname(&uts) == -1) {
    log_warning(os)("uname failed (%d)", errno);
    guarantee(0, "Could not determine uname information");
  } else {
    log_info(os)("uname says: sysname \"%s\" version \"%s\" release \"%s\" "
               "node \"%s\" machine \"%s\"\n",
               uts.sysname, uts.version, uts.release, uts.nodename, uts.machine);
    const int major = atoi(uts.version);
    assert(major > 0, "invalid OS version");
    const int minor = atoi(uts.release);
    assert(minor > 0, "invalid OS release");
    _os_version = (major << 24) | (minor << 16);
    char ver_str[20] = {0};

    // We do not support versions older than AIX 7.2 TL 5.
    // Determine detailed AIX version: Version, Release, Modification, Fix Level.
    odmWrapper::determine_os_kernel_version(&_os_version);
    if (_os_version < 0x07020500) {
      log_warning(os)("AIX releases older than AIX 7.2 TL 5 are not supported.");
      assert(false, "AIX release too old.");
    }

    jio_snprintf(ver_str, sizeof(ver_str), "%u.%u.%u.%u",
                 major, minor, (_os_version >> 8) & 0xFF, _os_version & 0xFF);
    log_info(os)("We run on AIX %s", ver_str);
  }

  guarantee(_os_version, "Could not determine AIX release");
} // end: os::Aix::initialize_os_info()

// Scan environment for important settings which might effect the VM.
// Trace out settings. Warn about invalid settings and/or correct them.
//
// Must run after os::Aix::initialue_os_info().
void os::Aix::scan_environment() {

  char* p;

  // Reject EXTSHM=ON. That switch changes how System V shared memory behaves
  // and prevents allocation of 64k pages for the heap.
  p = ::getenv("EXTSHM");
  trcVerbose("EXTSHM=%s.", p ? p : "<unset>");
  if (p && strcasecmp(p, "ON") == 0) {
    vm_exit_during_initialization("EXTSHM is ON. Please remove EXTSHM from your environment.");
  }

  // SPEC1170 behaviour: will change the behaviour of a number of POSIX APIs.
  // Not tested, not supported.
  //
  // Note that it might be worth the trouble to test and to require it, if only to
  // get useful return codes for mprotect.
  //
  // Note: Setting XPG_SUS_ENV in the process is too late. Must be set earlier (before
  // exec() ? before loading the libjvm ? ....)
  p = ::getenv("XPG_SUS_ENV");
  trcVerbose("XPG_SUS_ENV=%s.", p ? p : "<unset>");
  if (p && strcmp(p, "ON") == 0) {
    _xpg_sus_mode = 1;
    log_warning(os)("Unsupported setting: XPG_SUS_ENV=ON");
    // This is not supported. Worst of all, it changes behaviour of mmap MAP_FIXED to
    // clobber address ranges. If we ever want to support that, we have to do some
    // testing first.
    guarantee(false, "XPG_SUS_ENV=ON not supported");
  } else {
    _xpg_sus_mode = 0;
  }

  p = ::getenv("LDR_CNTRL");
  trcVerbose("LDR_CNTRL=%s.", p ? p : "<unset>");

  p = ::getenv("AIXTHREAD_GUARDPAGES");
  trcVerbose("AIXTHREAD_GUARDPAGES=%s.", p ? p : "<unset>");

} // end: os::Aix::scan_environment()

void os::Aix::initialize_libperfstat() {
  if (!libperfstat::init()) {
    log_warning(os)("libperfstat initialization failed.");
    assert(false, "libperfstat initialization failed");
  } else {
    trcVerbose("libperfstat initialized.");
  }
}

bool os::Aix::supports_64K_mmap_pages() {
  return g_multipage_support.can_use_64K_mmap_pages;
}

/////////////////////////////////////////////////////////////////////////////
// thread stack

// Get the current stack base and size from the OS (actually, the pthread library).
// Note: base usually not page aligned.
// Returned size is such that (base - size) is always aligned to page size.
void os::current_stack_base_and_size(address* stack_base, size_t* stack_size) {
  AixMisc::stackbounds_t bounds;
  bool rc = AixMisc::query_stack_bounds_for_current_thread(&bounds);
  guarantee(rc, "Unable to retrieve stack bounds.");
  *stack_base = bounds.base;

  // Align the reported stack size such that the stack low address
  // is aligned to page size (Note: base is usually not and we do not care).
  // We need to do this because caller code will assume stack low address is
  // page aligned and will place guard pages without checking.
  address low = bounds.base - bounds.size;
  address low_aligned = (address)align_up(low, os::vm_page_size());
  *stack_size = bounds.base - low_aligned;
}

// Get the default path to the core file
// Returns the length of the string
int os::get_core_path(char* buffer, size_t bufferSize) {
  const char* p = get_current_directory(buffer, bufferSize);

  if (p == nullptr) {
    assert(p != nullptr, "failed to get current directory");
    return 0;
  }

  jio_snprintf(buffer, bufferSize, "%s/core or core.%d",
                                               p, current_process_id());

  return checked_cast<int>(strlen(buffer));
}

bool os::start_debugging(char *buf, int buflen) {
  int len = (int)strlen(buf);
  char *p = &buf[len];

  jio_snprintf(p, buflen -len,
                 "\n\n"
                 "Do you want to debug the problem?\n\n"
                 "To debug, run 'dbx -a %d'; then switch to thread tid %zd, k-tid %zd\n"
                 "Enter 'yes' to launch dbx automatically (PATH must include dbx)\n"
                 "Otherwise, press RETURN to abort...",
                 os::current_process_id(),
                 os::current_thread_id(), thread_self());

  bool yes = os::message_box("Unexpected Error", buf);

  if (yes) {
    // yes, user asked VM to launch debugger
    jio_snprintf(buf, buflen, "dbx -a %d", os::current_process_id());

    os::fork_and_exec(buf);
    yes = false;
  }
  return yes;
}

static inline time_t get_mtime(const char* filename) {
  struct stat st;
  int ret = os::stat(filename, &st);
  assert(ret == 0, "failed to stat() file '%s': %s", filename, os::strerror(errno));
  return st.st_mtime;
}

int os::compare_file_modified_times(const char* file1, const char* file2) {
  time_t t1 = get_mtime(file1);
  time_t t2 = get_mtime(file2);
  return primitive_compare(t1, t2);
}

bool os::supports_map_sync() {
  return false;
}

void os::print_memory_mappings(char* addr, size_t bytes, outputStream* st) {}

#if INCLUDE_JFR

void os::jfr_report_memory_info() {}

#endif // INCLUDE_JFR

void os::print_open_file_descriptors(outputStream* st) {
  // File descriptor counting not implemented on AIX
}