module/zfs/zfs_debug.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 http://www.opensolaris.org/os/licensing.
  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) 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2013 by Delphix. All rights reserved.
  24  */
  25
  26 #include <sys/zfs_context.h>
  27
  28 #if !defined(_KERNEL) || !defined(__linux__)
  29 list_t zfs_dbgmsgs;
  30 int zfs_dbgmsg_size;
  31 kmutex_t zfs_dbgmsgs_lock;
  32 int zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
  33 #endif
  34
  35 /*
  36  * Enable various debugging features.
  37  */
  38 int zfs_flags = 0;
  39
  40 /*
  41  * zfs_recover can be set to nonzero to attempt to recover from
  42  * otherwise-fatal errors, typically caused by on-disk corruption.  When
  43  * set, calls to zfs_panic_recover() will turn into warning messages.
  44  * This should only be used as a last resort, as it typically results
  45  * in leaked space, or worse.
  46  */
  47 int zfs_recover = B_FALSE;
  48
  49 /*
  50  * If destroy encounters an EIO while reading metadata (e.g. indirect
  51  * blocks), space referenced by the missing metadata can not be freed.
  52  * Normally this causes the background destroy to become "stalled", as
  53  * it is unable to make forward progress.  While in this stalled state,
  54  * all remaining space to free from the error-encountering filesystem is
  55  * "temporarily leaked".  Set this flag to cause it to ignore the EIO,
  56  * permanently leak the space from indirect blocks that can not be read,
  57  * and continue to free everything else that it can.
  58  *
  59  * The default, "stalling" behavior is useful if the storage partially
  60  * fails (i.e. some but not all i/os fail), and then later recovers.  In
  61  * this case, we will be able to continue pool operations while it is
  62  * partially failed, and when it recovers, we can continue to free the
  63  * space, with no leaks.  However, note that this case is actually
  64  * fairly rare.
  65  *
  66  * Typically pools either (a) fail completely (but perhaps temporarily,
  67  * e.g. a top-level vdev going offline), or (b) have localized,
  68  * permanent errors (e.g. disk returns the wrong data due to bit flip or
  69  * firmware bug).  In case (a), this setting does not matter because the
  70  * pool will be suspended and the sync thread will not be able to make
  71  * forward progress regardless.  In case (b), because the error is
  72  * permanent, the best we can do is leak the minimum amount of space,
  73  * which is what setting this flag will do.  Therefore, it is reasonable
  74  * for this flag to normally be set, but we chose the more conservative
  75  * approach of not setting it, so that there is no possibility of
  76  * leaking space in the "partial temporary" failure case.
  77  */
  78 int zfs_free_leak_on_eio = B_FALSE;
  79
  80
  81 void
  82 zfs_panic_recover(const char *fmt, ...)
  83 {
  84         va_list adx;
  85
  86         va_start(adx, fmt);
  87         vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
  88         va_end(adx);
  89 }
  90
  91 /*
  92  * Debug logging is enabled by default for production kernel builds.
  93  * The overhead for this is negligible and the logs can be valuable when
  94  * debugging.  For non-production user space builds all debugging except
  95  * logging is enabled since performance is no longer a concern.
  96  */
  97 void
  98 zfs_dbgmsg_init(void)
  99 {
 100 #if !defined(_KERNEL) || !defined(__linux__)
 101         list_create(&zfs_dbgmsgs, sizeof (zfs_dbgmsg_t),
 102             offsetof(zfs_dbgmsg_t, zdm_node));
 103         mutex_init(&zfs_dbgmsgs_lock, NULL, MUTEX_DEFAULT, NULL);
 104 #endif
 105
 106         if (zfs_flags == 0) {
 107 #if defined(_KERNEL)
 108                 zfs_flags = ZFS_DEBUG_DPRINTF;
 109                 spl_debug_set_mask(spl_debug_get_mask() | SD_DPRINTF);
 110                 spl_debug_set_subsys(spl_debug_get_subsys() | SS_USER1);
 111 #else
 112                 zfs_flags = ~ZFS_DEBUG_DPRINTF;
 113 #endif /* _KERNEL */
 114         }
 115 }
 116
 117 void
 118 zfs_dbgmsg_fini(void)
 119 {
 120 #if !defined(_KERNEL) || !defined(__linux__)
 121         zfs_dbgmsg_t *zdm;
 122
 123         while ((zdm = list_remove_head(&zfs_dbgmsgs)) != NULL) {
 124                 int size = sizeof (zfs_dbgmsg_t) + strlen(zdm->zdm_msg);
 125                 kmem_free(zdm, size);
 126                 zfs_dbgmsg_size -= size;
 127         }
 128         mutex_destroy(&zfs_dbgmsgs_lock);
 129         ASSERT0(zfs_dbgmsg_size);
 130 #endif
 131 }
 132
 133 #if !defined(_KERNEL) || !defined(__linux__)
 134 /*
 135  * Print these messages by running:
 136  * echo ::zfs_dbgmsg | mdb -k
 137  *
 138  * Monitor these messages by running:
 139  *      dtrace -q -n 'zfs-dbgmsg{printf("%s\n", stringof(arg0))}'
 140  */
 141 void
 142 zfs_dbgmsg(const char *fmt, ...)
 143 {
 144         int size;
 145         va_list adx;
 146         zfs_dbgmsg_t *zdm;
 147
 148         va_start(adx, fmt);
 149         size = vsnprintf(NULL, 0, fmt, adx);
 150         va_end(adx);
 151
 152         /*
 153          * There is one byte of string in sizeof (zfs_dbgmsg_t), used
 154          * for the terminating null.
 155          */
 156         zdm = kmem_alloc(sizeof (zfs_dbgmsg_t) + size, KM_SLEEP);
 157         zdm->zdm_timestamp = gethrestime_sec();
 158
 159         va_start(adx, fmt);
 160         (void) vsnprintf(zdm->zdm_msg, size + 1, fmt, adx);
 161         va_end(adx);
 162
 163         DTRACE_PROBE1(zfs__dbgmsg, char *, zdm->zdm_msg);
 164
 165         mutex_enter(&zfs_dbgmsgs_lock);
 166         list_insert_tail(&zfs_dbgmsgs, zdm);
 167         zfs_dbgmsg_size += sizeof (zfs_dbgmsg_t) + size;
 168         while (zfs_dbgmsg_size > zfs_dbgmsg_maxsize) {
 169                 zdm = list_remove_head(&zfs_dbgmsgs);
 170                 size = sizeof (zfs_dbgmsg_t) + strlen(zdm->zdm_msg);
 171                 kmem_free(zdm, size);
 172                 zfs_dbgmsg_size -= size;
 173         }
 174         mutex_exit(&zfs_dbgmsgs_lock);
 175 }
 176
 177 void
 178 zfs_dbgmsg_print(const char *tag)
 179 {
 180         zfs_dbgmsg_t *zdm;
 181
 182         (void) printf("ZFS_DBGMSG(%s):\n", tag);
 183         mutex_enter(&zfs_dbgmsgs_lock);
 184         for (zdm = list_head(&zfs_dbgmsgs); zdm;
 185             zdm = list_next(&zfs_dbgmsgs, zdm))
 186                 (void) printf("%s\n", zdm->zdm_msg);
 187         mutex_exit(&zfs_dbgmsgs_lock);
 188 }
 189 #endif
 190
 191 #if defined(_KERNEL)
 192 module_param(zfs_flags, int, 0644);
 193 MODULE_PARM_DESC(zfs_flags, "Set additional debugging flags");
 194
 195 module_param(zfs_recover, int, 0644);
 196 MODULE_PARM_DESC(zfs_recover, "Set to attempt to recover from fatal errors");
 197
 198 module_param(zfs_free_leak_on_eio, int, 0644);
 199 MODULE_PARM_DESC(zfs_free_leak_on_eio,
 200         "Set to ignore IO errors during free and permanently leak the space");
 201 #endif /* _KERNEL */