Various (possible) problems of aarch64

Here a list of possible problems of NetBSD/evbarm aarch64 that needs further investigation in order to write proper PR or better yet to fix them!

mpv SIGSEGVs (strnlen(s, (size_t)-1) always returns -1)

UPDATE: This was fixed by <ryo>, thanks!

Just by invoking mpv via:

% mpv

It SEGV as follows:

% mpv
/usr/pkg/lib/ffmpeg4/libavcodec.so.58: text relocations
mpv 0.29.0 Copyright © 2000-2018 mpv/MPlayer/mplayer2 projects
 built on Sat Jul 28 21:02:57 CEST 2018
ffmpeg library versions:
   libavutil       56.14.100
   libavcodec      58.18.100
   libavformat     58.12.100
   libswscale      5.1.100
   libavfilter     7.16.100
   libswresample   3.1.100
ffmpeg version: 4.0.2

Usage:   mpv [options] [url|path/]filename

Basic options:
 --start=<time>    seek to given (percent, seconds, or hh:mm:ss) position
 --no-audio        do not play sound
 --no-video        do not play video
 --fs              fullscreen playback
 --sub-file=<file> specify subtitle file to use
 --playlist=<file> specify playlist file

 --list-options    list all mpv options
 --h=<string>      print options which contain the given string in their name

[1]   Segmentation fault (core dumped) /usr/pkg/bin/mpv...
% gdb -core mpv.core /usr/pkg/bin/mpv
Reading symbols from /usr/pkg/bin/mpv...done.
[New process 5]
[New process 6]
[New process 4]
[New process 3]
[New process 2]
[New process 1]
Core was generated by `mpv'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x0000f96f727f40b0 in memcpy () from /usr/lib/libc.so.12
[Current thread is 1 (process 5)]
(gdb) bt
#0  0x0000f96f727f40b0 in memcpy () from /usr/lib/libc.so.12
#1  0x00000002001cba34 in __memcpy_ichk (len=18446744073709551615, src=0xf96f6e10fec0, dst=<optimized out>)
    at /usr/include/ssp/string.h:82
#2  strndup_append_at (str=str@entry=0xf96f6f4ff348, at=0, append=0xf96f6e10fec0 "speed", append_len=<optimized out>,
    append_len@entry=18446744073709551615) at ../ta/ta_utils.c:113
#3  0x00000002001cbea4 in ta_strdup_append_buffer (str=str@entry=0xf96f6f4ff348, a=<optimized out>) at ../ta/ta_utils.c:165
#4  0x00000002001cb7b8 in ta_talloc_strdup_append_buffer (s=<optimized out>, s@entry=0x0, a=<optimized out>) at ../ta/ta_talloc.c:31
#5  0x0000000200157eec in print_str_list (opt=<optimized out>, src=<optimized out>) at ../options/m_option.c:1477
[...]

This happens because strnlen(s, (size_t)-1), always returns -1, e.g.:

% cat strnlen_size_max.c
#include <stdio.h>
#include <string.h>


int
main(int argc, char *argv[])
{
    int i;
    for (i = 1; i < argc; i++) {
        printf("strnlen(\"%s\", (size_t)-1) -> %ld\n",
            argv[i], strnlen(argv[i], ~(size_t)0));
    }

    return 0;
}
% gcc -Wall strnlen_size_max.c
% ./a.out foo
strnlen("foo", (size_t)-1) -> -1

(This should returns 3, not -1!)

The following patch workaround the mpv problem by avoiding all strnlen(s, (size_t)-1) calls (and just using strlen() instead):

$NetBSD: aarch64_problems.mdwn,v 1.12 2019/01/04 22:26:31 leot Exp $

Avoid to directly call:

 strnlen(s, (size_t)-1)

because on aarch64 ATM it (incorrectly) always returns -1 (also
when s is a string with less than SIZE_MAX characters).

Add a kludge in order to use strlen() in these cases.

--- ta/ta_utils.c.orig  2018-07-31 21:45:47.492269366 +0000
+++ ta/ta_utils.c
@@ -98,7 +98,11 @@ static bool strndup_append_at(char **str
     if (!*str && !append)
         return true; // stays NULL, but not an OOM condition

-    size_t real_len = append ? strnlen(append, append_len) : 0;
+    size_t real_len = append ?
+   append_len == (size_t)-1 ?
+      strlen(append) :
+      strnlen(append, append_len) :
+   0;
     if (append_len > real_len)
         append_len = real_len;

Python import requests SIGILLs

UPDATE: A kludge was added since py-cffi-1.11.5nb1 to avoid that. Doing a (please note that also python36 is affected):

$ python2.7 -c 'import requests'
Illegal instruction (core dumped)
Exit 132

...where:

% gdb -core python2.7.core `which python2.7`
Reading symbols from /usr/pkg/bin/python2.7...(no debugging symbols found)...done.
[New process 1]
Core was generated by `python2.7'.
Program terminated with signal SIGILL, Illegal instruction.
#0  0x0000f42f5fd00000 in ?? ()
(gdb) bt 5
#0  0x0000f42f5fd00000 in ?? ()
#1  0x0000f42f5fd0a3d0 in restore_errno_only () at c/misc_thread_common.h:43
#2  0x0000f42f5fbf0f90 in _cffi_f_SSL_library_init (self=<optimized out>, noarg=<optimized out>)
    at build/temp.netbsd-8.99.22-evbarm-2.7/_openssl.c:51839
#3  0x0000f42f617606ac in PyEval_EvalFrameEx () from /usr/pkg/lib/libpython2.7.so.1.0
#4  0x0000f42f6175e5c0 in PyEval_EvalCodeEx () from /usr/pkg/lib/libpython2.7.so.1.0
(More stack frames follow...)
(gdb) f 2
#2  0x0000f42f5fbf0f90 in _cffi_f_SSL_library_init (self=<optimized out>, noarg=<optimized out>)
    at build/temp.netbsd-8.99.22-evbarm-2.7/_openssl.c:51839
warning: Source file is more recent than executable.
51839     _cffi_restore_errno();
(gdb) list
51834   _cffi_f_SSL_library_init(PyObject *self, PyObject *noarg)
51835   {
51836     int result;
51837
51838     Py_BEGIN_ALLOW_THREADS
51839     _cffi_restore_errno();
51840     { result = SSL_library_init(); }
51841     _cffi_save_errno();
51842     Py_END_ALLOW_THREADS
51843
(gdb) f 1
#1  0x0000f42f5fd0a3d0 in restore_errno_only () at c/misc_thread_common.h:43
43      static void restore_errno_only(void) { errno = cffi_saved_errno; }
(gdb) list
38         syntactically valid to use "__thread" with this C compiler. */
39      #ifdef USE__THREAD
40
41      static __thread int cffi_saved_errno = 0;
42      static void save_errno_only(void) { cffi_saved_errno = errno; }
43      static void restore_errno_only(void) { errno = cffi_saved_errno; }
44
45      #else
46
47      static void save_errno_only(void)
(gdb) disas
Dump of assembler code for function restore_errno_only:
   0x0000f42f5fd0a3b0 <+0>:     str     x30, [sp, #-16]!
   0x0000f42f5fd0a3b4 <+4>:     bl      0xf42f5fd075c0 <__errno@plt>
   0x0000f42f5fd0a3b8 <+8>:     mov     x2, x0
   0x0000f42f5fd0a3bc <+12>:    mrs     x1, tpidr_el0
   0x0000f42f5fd0a3c0 <+16>:    adrp    x0, 0xf42f5fd34000
   0x0000f42f5fd0a3c4 <+20>:    ldr     x3, [x0, #4088]
   0x0000f42f5fd0a3c8 <+24>:    add     x0, x0, #0xff8
   0x0000f42f5fd0a3cc <+28>:    blr     x3
=> 0x0000f42f5fd0a3d0 <+32>:    ldr     w0, [x1, x0]
   0x0000f42f5fd0a3d4 <+36>:    str     w0, [x2]
   0x0000f42f5fd0a3d8 <+40>:    ldr     x30, [sp], #16
   0x0000f42f5fd0a3dc <+44>:    ret
End of assembler dump.
(gdb) info reg
x0             0xf42f5fd34ff8   268484308324344
x1             0xf42f61875080   268484336898176
x2             0xf42f6186c848   268484336863304
x3             0xf42f5fd00000   268484308107264
x4             0xf42f5ff3c0b8   268484310450360
x5             0xf42f60000570   268484311254384
x6             0x65000000000000 28428972647776256
x7             0x65     101
x8             0xf42f6185e000   268484336803840
x9             0xf42f6185e668   268484336805480
x10            0xf42f6185e668   268484336805480
x11            0x1      1
x12            0x30     48
x13            0xffffff9264f0   281474969527536
x14            0x65     101
x15            0x3      3
x16            0xf42f615bd730   268484334049072
x17            0xf42f615ec2c8   268484334240456
x18            0xf42f615ca1a0   268484334100896
x19            0xf42f5fc9cea8   268484307701416
x20            0xf42f5fd9c3c0   268484308747200
x21            0xf42f61307090   268484331204752
x22            0xf42f5fcd38e8   268484307925224
x23            0xf42f5fcd10a0   268484307914912
x24            0xf42f5fb8c5a8   268484306585000
x25            0xf42f5fd9c3c0   268484308747200
x26            0xf42f5fbf0f70   268484306997104
x27            0xf42f5fdd467e   268484308977278
x28            0xf42f61307090   268484331204752
x29            0x0      0
x30            0xf42f5fd0a3d0   268484308149200
sp             0xffffff926760   0xffffff926760
pc             0xf42f5fd0a3d0   0xf42f5fd0a3d0 <restore_errno_only+32>
cpsr           0x60000000       [ EL=0 C Z ]
fpsr           0x3000000        50331648
fpcr           0x10     16
(gdb) quit

...and _cffi_restore_errno and _cffi_save_errno are defined in cffi/_cffi_include.h as follows:

#define _cffi_restore_errno                                              \
    ((void(*)(void))_cffi_exports[13])
#define _cffi_save_errno                                                 \
    ((void(*)(void))_cffi_exports[14])

Since py-cffi-1.11.5nb1 a kludge to workaround the issue is present (by disabling __thread).

polkitd crashes

pinebook# /usr/pkg/lib/polkit-1/polkitd
Successfully changed to user polkitd
Error loading /var/run/ConsoleKit/database: Error statting file /var/run/ConsoleKit/database: No such file or directory[1]   Segmentation fault      /usr/pkg/lib/polkit-1/polkitd

Please run $PREFIX/sbin/console-kit-daemon to get the database and pid files in /var/run :) I'm currently debugging the memory fault on the pinebook

Comment by youri late Sunday afternoon, July 29th, 2018