boehm-gc/doc/README.linux

   1 See README.alpha for Linux on DEC AXP info.
   2
   3 This file applies mostly to Linux/Intel IA32.  Ports to Linux on an M68K, IA64,
   4 SPARC, MIPS, Alpha and PowerPC are also integrated.  They should behave
   5 similarly, except that the PowerPC port lacks incremental GC support, and
   6 it is unknown to what extent the Linux threads code is functional.
   7 See below for M68K specific notes.
   8
   9 Incremental GC is generally supported.
  10
  11 Dynamic libraries are supported on an ELF system.  A static executable
  12 should be linked with the gcc option "-Wl,-defsym,_DYNAMIC=0".
  13
  14 The collector appears to work reliably with Linux threads, but beware
  15 of older versions of glibc and gdb.
  16
  17 The garbage collector uses SIGPWR and SIGXCPU if it is used with
  18 Linux threads.  These should not be touched by the client program.
  19
  20 To use threads, you need to abide by the following requirements:
  21
  22 1) You need to use LinuxThreads or NPTL (which are included in libc6).
  23
  24    The collector relies on some implementation details of the LinuxThreads
  25    package.  This code may not work on other
  26    pthread implementations (in particular it will *not* work with
  27    MIT pthreads).
  28
  29 2) You must compile the collector with -DGC_LINUX_THREADS and -D_REENTRANT
  30    specified in the Makefile.
  31
  32 3a) Every file that makes thread calls should define GC_LINUX_THREADS and
  33    _REENTRANT and then include gc.h.  Gc.h redefines some of the
  34    pthread primitives as macros which also provide the collector with
  35    information it requires.
  36
  37 3b) A new alternative to (3a) is to build the collector and compile GC clients
  38    with -DGC_USE_LD_WRAP, and to link the final program with
  39
  40    (for ld) --wrap read --wrap dlopen --wrap pthread_create \
  41             --wrap pthread_join --wrap pthread_detach \
  42             --wrap pthread_sigmask --wrap sleep
  43
  44    (for gcc) -Wl,--wrap -Wl,read -Wl,--wrap -Wl,dlopen -Wl,--wrap \
  45              -Wl,pthread_create -Wl,--wrap -Wl,pthread_join -Wl,--wrap \
  46              -Wl,pthread_detach -Wl,--wrap -Wl,pthread_sigmask \
  47              -Wl,--wrap -Wl,sleep
  48
  49    In any case, _REENTRANT should be defined during compilation.
  50
  51 4) Dlopen() disables collection during its execution.  (It can't run
  52    concurrently with the collector, since the collector looks at its
  53    data structures.  It can't acquire the allocator lock, since arbitrary
  54    user startup code may run as part of dlopen().)  Under unusual
  55    conditions, this may cause unexpected heap growth.
  56
  57 5) The combination of GC_LINUX_THREADS, REDIRECT_MALLOC, and incremental
  58    collection fails in seemingly random places.  This hasn't been tracked
  59    down yet, but is perhaps not completely astonishing.  The thread package
  60    uses malloc, and thus can presumably get SIGSEGVs while inside the
  61    package.  There is no real guarantee that signals are handled properly
  62    at that point.
  63
  64 6) Thread local storage may not be viewed as part of the root set by the
  65    collector.  This probably depends on the linuxthreads version.  For the
  66    time being, any collectable memory referenced by thread local storage should
  67    also be referenced from elsewhere, or be allocated as uncollectable.
  68    (This is really a bug that should be fixed somehow.)
  69
  70
  71 M68K LINUX:
  72 (From Richard Zidlicky)
  73 The bad news is that it can crash every linux-m68k kernel on a 68040,
  74 so an additional test is needed somewhere on startup. I have meanwhile
  75 patches to correct the problem in 68040 buserror handler but it is not
  76 yet in any standard kernel.
  77
  78 Here is a simple test program to detect whether the kernel has the
  79 problem. It could be run as a separate check in configure or tested
  80 upon startup. If it fails (return !0) than mprotect can't be used
  81 on that system.
  82
  83 /*
  84  * test for bug that may crash 68040 based Linux
  85  */
  86
  87 #include <sys/mman.h>
  88 #include <signal.h>
  89 #include <unistd.h>
  90 #include <stdio.h>
  91 #include <stdlib.h>
  92
  93
  94 char *membase;
  95 int pagesize=4096;
  96 int pageshift=12;
  97 int x_taken=0;
  98
  99 int sighandler(int sig)
 100 {
 101    mprotect(membase,pagesize,PROT_READ|PROT_WRITE);
 102    x_taken=1;
 103 }
 104
 105 main()
 106 {
 107   long l;
 108
 109    signal(SIGSEGV,sighandler);
 110    l=(long)mmap(NULL,pagesize,PROT_READ,MAP_PRIVATE | MAP_ANON,-1,0);
 111   if (l==-1)
 112      {
 113        perror("mmap/malloc");
 114        abort();
 115      }
 116   membase=(char*)l;
 117     *(long*)(membase+sizeof(long))=123456789;
 118   if (*(long*)(membase+sizeof(long)) != 123456789 )
 119     {
 120       fprintf(stderr,"writeback failed !\n");
 121       exit(1);
 122     }
 123   if (!x_taken)
 124     {
 125       fprintf(stderr,"exception not taken !\n");
 126       exit(1);
 127     }
 128   fprintf(stderr,"vmtest Ok\n");
 129   exit(0);
 130 }
 131
 132