lib/libspl/backtrace.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or https://opensource.org/licenses/CDDL-1.0.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
  23  * Copyright (c) 2024, Klara Inc.
  24  */
  25
  26 #include <sys/backtrace.h>
  27 #include <sys/types.h>
  28 #include <sys/debug.h>
  29 #include <unistd.h>
  30
  31 /*
  32  * Output helpers. libspl_backtrace() must not block, must be thread-safe and
  33  * must be safe to call from a signal handler. At least, that means not having
  34  * printf, so we end up having to call write() directly on the fd. That's
  35  * awkward, as we always have to pass through a length, and some systems will
  36  * complain if we don't consume the return. So we have some macros to make
  37  * things a little more palatable.
  38  */
  39 #define spl_bt_write_n(fd, s, n) \
  40         do { ssize_t r __maybe_unused = write(fd, s, n); } while (0)
  41 #define spl_bt_write(fd, s)             spl_bt_write_n(fd, s, sizeof (s))
  42
  43 #if defined(HAVE_LIBUNWIND)
  44 #define UNW_LOCAL_ONLY
  45 #include <libunwind.h>
  46
  47 /*
  48  * Convert `v` to ASCII hex characters. The bottom `n` nybbles (4-bits ie one
  49  * hex digit) will be written, up to `buflen`. The buffer will not be
  50  * null-terminated. Returns the number of digits written.
  51  */
  52 static size_t
  53 spl_bt_u64_to_hex_str(uint64_t v, size_t n, char *buf, size_t buflen)
  54 {
  55         static const char hexdigits[] = {
  56             '0', '1', '2', '3', '4', '5', '6', '7',
  57             '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'
  58         };
  59
  60         size_t pos = 0;
  61         boolean_t want = (n == 0);
  62         for (int i = 15; i >= 0; i--) {
  63                 const uint64_t d = v >> (i * 4) & 0xf;
  64                 if (!want && (d != 0 || n > i))
  65                         want = B_TRUE;
  66                 if (want) {
  67                         buf[pos++] = hexdigits[d];
  68                         if (pos == buflen)
  69                                 break;
  70                 }
  71         }
  72         return (pos);
  73 }
  74
  75 void
  76 libspl_backtrace(int fd)
  77 {
  78         unw_context_t uc;
  79         unw_cursor_t cp;
  80         unw_word_t v;
  81         char buf[128];
  82         size_t n;
  83         int err;
  84
  85         /* Snapshot the current frame and state. */
  86         unw_getcontext(&uc);
  87
  88         /*
  89          * TODO: walk back to the frame that tripped the assertion / the place
  90          *       where the signal was recieved.
  91          */
  92
  93         /*
  94          * Register dump. We're going to loop over all the registers in the
  95          * top frame, and show them, with names, in a nice three-column
  96          * layout, which keeps us within 80 columns.
  97          */
  98         spl_bt_write(fd, "Registers:\n");
  99
 100         /* Initialise a frame cursor, starting at the current frame */
 101         unw_init_local(&cp, &uc);
 102
 103         /*
 104          * libunwind's list of possible registers for this architecture is an
 105          * enum, unw_regnum_t. UNW_TDEP_LAST_REG is the highest-numbered
 106          * register in that list, however, not all register numbers in this
 107          * range are defined by the architecture, and not all defined registers
 108          * will be present on every implementation of that architecture.
 109          * Moreover, libunwind provides nice names for most, but not all
 110          * registers, but these are hardcoded; a name being available does not
 111          * mean that register is available.
 112          *
 113          * So, we have to pull this all together here. We try to get the value
 114          * of every possible register. If we get a value for it, then the
 115          * register must exist, and so we get its name. If libunwind has no
 116          * name for it, we synthesize something. These cases should be rare,
 117          * and they're usually for uninteresting or niche registers, so it
 118          * shouldn't really matter. We can see the value, and that's the main
 119          * thing.
 120          */
 121         uint_t cols = 0;
 122         for (uint_t regnum = 0; regnum <= UNW_TDEP_LAST_REG; regnum++) {
 123                 /*
 124                  * Get the value. Any error probably means the register
 125                  * doesn't exist, and we skip it.
 126                  */
 127                 if (unw_get_reg(&cp, regnum, &v) < 0)
 128                         continue;
 129
 130                 /*
 131                  * Register name. If libunwind doesn't have a name for it,
 132                  * it will return "???". As a shortcut, we just treat '?'
 133                  * is an alternate end-of-string character.
 134                  */
 135                 const char *name = unw_regname(regnum);
 136                 for (n = 0; name[n] != '\0' && name[n] != '?'; n++) {}
 137                 if (n == 0) {
 138                         /*
 139                          * No valid name, so make one of the form "?xx", where
 140                          * "xx" is the two-char hex of libunwind's register
 141                          * number.
 142                          */
 143                         buf[0] = '?';
 144                         n = spl_bt_u64_to_hex_str(regnum, 2,
 145                             &buf[1], sizeof (buf)-1) + 1;
 146                         name = buf;
 147                 }
 148
 149                 /*
 150                  * Two spaces of padding before each column, plus extra
 151                  * spaces to align register names shorter than three chars.
 152                  */
 153                 spl_bt_write_n(fd, "      ", 5-MIN(n, 3));
 154
 155                 /* Register name and column punctuation */
 156                 spl_bt_write_n(fd, name, n);
 157                 spl_bt_write(fd, ": 0x");
 158
 159                 /*
 160                  * Convert register value (from unw_get_reg()) to hex. We're
 161                  * assuming that all registers are 64-bits wide, which is
 162                  * probably fine for any general-purpose registers on any
 163                  * machine currently in use. A more generic way would be to
 164                  * look at the width of unw_word_t, but that would also
 165                  * complicate the column code a bit. This is fine.
 166                  */
 167                 n = spl_bt_u64_to_hex_str(v, 16, buf, sizeof (buf));
 168                 spl_bt_write_n(fd, buf, n);
 169
 170                 /* Every third column, emit a newline */
 171                 if (!(++cols % 3))
 172                         spl_bt_write(fd, "\n");
 173         }
 174
 175         /* If we finished before the third column, emit a newline. */
 176         if (cols % 3)
 177                 spl_bt_write(fd, "\n");
 178
 179         /* Now the main event, the backtrace. */
 180         spl_bt_write(fd, "Call trace:\n");
 181
 182         /* Reset the cursor to the top again. */
 183         unw_init_local(&cp, &uc);
 184
 185         do {
 186                 /*
 187                  * Getting the IP should never fail; libunwind handles it
 188                  * specially, because its used a lot internally. Still, no
 189                  * point being silly about it, as the last thing we want is
 190                  * our crash handler to crash. So if it ever does fail, we'll
 191                  * show an error line, but keep going to the next frame.
 192                  */
 193                 if (unw_get_reg(&cp, UNW_REG_IP, &v) < 0) {
 194                         spl_bt_write(fd, "  [couldn't get IP register; "
 195                             "corrupt frame?]");
 196                         continue;
 197                 }
 198
 199                 /* IP & punctuation */
 200                 n = spl_bt_u64_to_hex_str(v, 16, buf, sizeof (buf));
 201                 spl_bt_write(fd, "  [0x");
 202                 spl_bt_write_n(fd, buf, n);
 203                 spl_bt_write(fd, "] ");
 204
 205                 /*
 206                  * Function ("procedure") name for the current frame. `v`
 207                  * receives the offset from the named function to the IP, which
 208                  * we show as a "+offset" suffix.
 209                  *
 210                  * If libunwind can't determine the name, we just show "???"
 211                  * instead. We've already displayed the IP above; that will
 212                  * have to do.
 213                  *
 214                  * unw_get_proc_name() will return ENOMEM if the buffer is too
 215                  * small, instead truncating the name. So we treat that as a
 216                  * success and use whatever is in the buffer.
 217                  */
 218                 err = unw_get_proc_name(&cp, buf, sizeof (buf), &v);
 219                 if (err == 0 || err == -UNW_ENOMEM) {
 220                         for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
 221                         spl_bt_write_n(fd, buf, n);
 222
 223                         /* Offset from proc name */
 224                         spl_bt_write(fd, "+0x");
 225                         n = spl_bt_u64_to_hex_str(v, 2, buf, sizeof (buf));
 226                         spl_bt_write_n(fd, buf, n);
 227                 } else
 228                         spl_bt_write(fd, "???");
 229
 230 #ifdef HAVE_LIBUNWIND_ELF
 231                 /*
 232                  * Newer libunwind has unw_get_elf_filename(), which gets
 233                  * the name of the ELF object that the frame was executing in.
 234                  * Like `unw_get_proc_name()`, `v` recieves the offset within
 235                  * the file, and UNW_ENOMEM indicates that a truncate filename
 236                  * was left in the buffer.
 237                  */
 238                 err = unw_get_elf_filename(&cp, buf, sizeof (buf), &v);
 239                 if (err == 0 || err == -UNW_ENOMEM) {
 240                         for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
 241                         spl_bt_write(fd, " (in ");
 242                         spl_bt_write_n(fd, buf, n);
 243
 244                         /* Offset within file */
 245                         spl_bt_write(fd, " +0x");
 246                         n = spl_bt_u64_to_hex_str(v, 2, buf, sizeof (buf));
 247                         spl_bt_write_n(fd, buf, n);
 248                         spl_bt_write(fd, ")");
 249                 }
 250 #endif
 251                 spl_bt_write(fd, "\n");
 252         } while (unw_step(&cp) > 0);
 253 }
 254 #elif defined(HAVE_BACKTRACE)
 255 #include <execinfo.h>
 256
 257 void
 258 libspl_backtrace(int fd)
 259 {
 260         void *btptrs[64];
 261         size_t nptrs = backtrace(btptrs, 64);
 262         spl_bt_write(fd, "Call trace:\n");
 263         backtrace_symbols_fd(btptrs, nptrs, fd);
 264 }
 265 #else
 266 void
 267 libspl_backtrace(int fd __maybe_unused)
 268 {
 269 }
 270 #endif