1 # Debug the DragonFly kernel
5 This chapter should give you an introduction how to obtain a crash dump after a kernel panic and how to extract useful information for the developers out of the dump.
9 ***Contributed by Matthias Schmidt***
13 ## Configure your system
15 Normally a crash dump is saved in your swap partition after a crash. If you reboot your machine the next time the dump is extracted by [savecore(8)](http://leaf.dragonflybsd.org/cgi/web-man?command#savecore§ion8) from the partition and stored into `/var/crash`. Due to the fact that `/var` is a relatively small partition it could be possible that the dump isn't saved, because the dump size is larger than the remaining size.
19 To circumvent this problem you can change the default settings in `/etc/rc.conf`:
23 [[!table header=no data="""
24 > `dumpdev` | | Indicates the device (usually a swap partition) to which a crash dump should be written in the event of a system crash.
25 > `dumpdir` | | savecore(8) will save that crash dump and a copy of the kernel to the directory specified by the dumpdir variable. The default value is /var/crash. You can set this to another directory on another partition with more space available to safely obtain the dump.
28 If you are unsure about your swap partition device, use [swapinfo(8)](http://leaf.dragonflybsd.org/cgi/web-man?command#swapinfo§ion8) or look into `/etc/fstab` :
33 Device 1K-blocks Used Avail Capacity Type
34 /dev/ad0s1b 1048448 0 1048448 0% Interleaved
36 # cat /etc/fstab | grep swap
37 /dev/ad0s1b none swap sw 0 0
40 ### Enable debugging options in your custom kernel config
42 If you run a custom kernel you have to add the following lines to compile your kernel with debugging symbols:
46 makeoptions DEBUG=-g #Build kernel with gdb(1) debug symbols
49 If you want additional support for the interactive kernel debugger [ddb(4)](http://leaf.dragonflybsd.org/cgi/web-man?command#ddb§ion4) and invariant debugging, also add these lines:
53 # Debugging for Development
59 You don't have to do anything to get debugging enabled in the default GENERIC kernel as it's already there.
61 ## How does a crash look like?
65 Easy answer: Your system stopped working. Complicated one: Your system occurred a [panic(9)](http://leaf.dragonflybsd.org/cgi/web-man?command#panic§ion9) and drops into [ddb(4)](http://leaf.dragonflybsd.org/cgi/web-man?command=ddb§ion=4), the interactive kernel debugger.
67 The output while seeing a crash might look this:
71 Fatal trap 12: page fault while in kernel mode
72 fault virtual address = 0xd0686f55
73 fault code = supervisor read, page not present
74 instruction pointer = 0x8:0xc02ddb9a
75 stack pointer = 0x10:0xcec0fb18
76 frame pointer = 0x10:0xcec0fb18
77 code segment = base 0x0, limit 0xfffff, type 0x1b
78 = DPL 0, pres 1, def32 1, gran 1
79 processor eflags # interrupt enabled, resume, IOPL 0
80 current process = 50725 (sysctl)
81 current thread = pri 6
86 Before your machine reboots a crash dump is saved into your swap partition (if you have one and don't disabled crash dumps). Writing the dump to disk takes some time depending on your machine and the amount of RAM installed. This might look like this:
90 dumping to dev #ad/0x20001, blockno 1049088
91 dump 511 510 509 508 507 506 505
93 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 succeeded
95 Now your machine reboots, checks its file system and finally extracts the crash dump from the swap partition to your `dumpdir` (see `rc.conf` setting above). If your `/var` partition is to small, you'll see an error similar to the following:
98 savecore: reboot after panic: from debugger
99 savecore: no dump, not enough free space on device (231420 available, need 541840)
102 If this happens, you have to extract the crash dump yourselves. See next Section how to do this.
104 ### Extract a crash dump manually
106 You can use [savecore(8)](http://leaf.dragonflybsd.org/cgi/web-man?command#savecore§ion8) to copy your currently running kernel and the associated crash dump to a particular directory you have to specify (we use `/usr/crash` here in the example):
110 # mkdir -p /usr/crash
111 # chmod 700 /usr/crash
112 # savecore /usr/crash/
117 This will take some time dependent on the speed of your machine. See the man page of `savecore(8)` for more available options.
120 ### Upload the crash dump
122 If you don't have the ability or skills to debug the crash yourselves, please upload the complete content of your crash directory to somewhere private (e.g., cloud share with a password, your leaf account), and open a detailed bug report. Please tar and compress the crash dump to save disk space and bandwidth.
124 **WARNING**: The saved core file (in `/var/crash`) contains *sensitive* data, e.g., passwords, certificates, decrypted private keys. Therefore, do *not* upload it to somewhere that can be publicly accessed!
127 ## Debug the crash dump with kgdb
129 The [kgdb(1)](http://leaf.dragonflybsd.org/cgi/web-man?command#kgdb§ion1) utility is a debugger based on gdb(1) that allows debugging of kernel core files.
133 To get some handy helper command execute the following command before starting kgdb:
137 source /usr/src/test/debug/gdb.kernel
139 This gives you several new commands like kldstat (displays all loaded modules) or psx (displays all running processes).
141 Start kgdb as follows:
147 -rw-r--r-- 1 root wheel 2B Jan 7 17:07 bounds
148 -rw-r--r-- 1 root wheel 17M Jan 7 17:08 kernel.0
149 -rw------- 1 root wheel 512M Jan 7 17:08 vmcore.0
150 # kgdb kernel.0 vmcore.0
153 kgdb(1) will show you the panic message after start. The first thing to do is to obtain a ***backtrace*** with the ***bt*** command:
157 Unread portion of the kernel message buffer:
163 Fatal trap 12: page fault while in kernel mode
164 fault virtual address = 0xd0686f55
165 fault code = supervisor read, page not present
166 instruction pointer = 0x8:0xc02ddb9a
167 stack pointer = 0x10:0xcec0fb18
168 frame pointer = 0x10:0xcec0fb18
169 code segment = base 0x0, limit 0xfffff, type 0x1b
170 current process = 50725 (sysctl)
171 current thread = pri 6
179 Fatal trap 3: breakpoint instruction fault while in kernel mode
180 instruction pointer = 0x8:0xc03136a4
181 stack pointer = 0x10:0xcec0f92c
182 frame pointer = 0x10:0xcec0f934
183 code segment = base 0x0, limit 0xfffff, type 0x1b
184 = DPL 0, pres 1, def32 1, gran 1
185 processor eflags # interrupt enabled, IOPL 0
186 current process = 50725 (sysctl)
187 current thread = pri 6
195 dumping to dev #ad/0x20001, blockno 1049088
196 dump 511 510 509 508 507 506 505 504 503 502 501 500 499 498
198 40 39 38 37 36 35 34 33 32 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
203 Copyright 2004 Free Software Foundation, Inc.
204 GDB is free software, covered by the GNU General Public License, and you are
205 welcome to change it and/or distribute copies of it under certain conditions.
206 Type "show copying" to see the conditions.
207 There is absolutely no warranty for GDB. Type "show warranty" for details.
208 This GDB was configured as "i386-dragonfly".
210 #0 dumpsys () at thread.h:83
211 #1 0xc01c4e1b in boot (howto=260) at /usr/src/sys/kern/kern_shutdown.c:375
212 #2 0xc01c4f3c in panic (fmt=Variable "fmt" is not available.
213 ) at /usr/src/sys/kern/kern_shutdown.c:800
214 #3 0xc0149be5 in db_panic (addr=Could not find the frame base for "db_panic".
215 ) at /usr/src/sys/ddb/db_command.c:447
216 #4 0xc014a250 in db_command_loop () at /usr/src/sys/ddb/db_command.c:343
217 #5 0xc014c7bc in db_trap (type#12, code0) at /usr/src/sys/ddb/db_trap.c:71
218 #6 0xc03137f7 in kdb_trap (type#12, code0, regs=0xcec0fad0) at /usr/src/sys/platform/pc32/i386/db_interface.c:148
219 #7 0xc032384b in trap_fatal (frame#0xcec0fad0, evaVariable "eva" is not available.
220 ) at /usr/src/sys/platform/pc32/i386/trap.c:1091
221 #8 0xc03239b0 in trap_pfault (frame#0xcec0fad0, usermode0, eva=3496505173)
222 at /usr/src/sys/platform/pc32/i386/trap.c:997
223 #9 0xc03241a0 in trap (frame=0xcec0fad0) at /usr/src/sys/platform/pc32/i386/trap.c:680
224 #10 0xc0314506 in calltrap () at /usr/src/sys/platform/pc32/i386/exception.s:783
225 #11 0xc02ddb9a in strlen (str=0xd0686f55 <Address 0xd0686f55 out of bounds>) at /usr/src/sys/libkern/strlen.c:41
226 #12 0xc02c2153 in sysctl_vm_zone (oidp#0xc03b42a0, arg10x0, arg2=0, req=0xcec0fc08) at /usr/src/sys/vm/vm_zone.c:447
227 #13 0xc01cf935 in sysctl_root (oidp=Variable "oidp" is not available.
228 ) at /usr/src/sys/kern/kern_sysctl.c:1193
229 #14 0xc01cfa27 in userland_sysctl (name#0xcec0fc90, namelen2, old=0x0, oldlenp=0xbfbfe8f0, inkernel=0, new=0x0,
230 newlen#0, retval0xcec0fc8c) at /usr/src/sys/kern/kern_sysctl.c:1268
231 #15 0xc01cfc28 in sys___sysctl (uap=0xcec0fcf0) at /usr/src/sys/kern/kern_sysctl.c:1211
232 #16 0xc0323ccb in syscall2 (frame=0xcec0fd40) at /usr/src/sys/platform/pc32/i386/trap.c:1339
233 #17 0xc03145a5 in Xint0x80_syscall () at /usr/src/sys/platform/pc32/i386/exception.s:872
234 #18 0x08055d38 in ?? ()
235 #19 0xbfbfe86c in ?? ()
236 #20 0x0000002f in ?? ()
237 #21 0x00000000 in ?? ()
238 #22 0x00000000 in ?? ()
239 #23 0x00000000 in ?? ()
240 #24 0x00000000 in ?? ()
241 #25 0x13c4b000 in ?? ()
242 #26 0x00000001 in ?? ()
243 #27 0xc03c2bf8 in intr_info_ary ()
244 #28 0xcec0f8d4 in ?? ()
245 #29 0xcec0f8c4 in ?? ()
246 #30 0xc8076300 in ?? ()
247 #31 0xc01cac5a in lwkt_preempt (ntd#0x2, critpriCannot access memory at address 0xbfbfe8a4
248 ) at /usr/src/sys/kern/lwkt_thread.c:893
249 Previous frame inner to this frame (corrupt stack?)
253 kgdb(1) gives you the ability to look into specific frames, display variable content and obtain the source code (if your kernel was compiled with -g):
258 #13 0xc01cf935 in sysctl_root (oidp=Variable "oidp" is not available.
259 ) at /usr/src/sys/kern/kern_sysctl.c:1193
260 1193 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
263 1189 if ((oid->oid_kind & CTLTYPE) # CTLTYPE_NODE)
264 1190 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
267 1193 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
273 $1 # {oid_parent 0xc03cbda8, oid_link = {sle_next = 0x0}, oid_number = 283, oid_kind = -2147483645, oid_arg1 = 0x0,
274 oid_arg2 # 0, oid_name 0xc03616ad "zone", oid_handler = 0xc02c20fa <sysctl_vm_zone>, oid_fmt = 0xc036a56f "A",
275 oid_refcnt # 0, oid_descr 0xc036906a "Zone Info"}
279 ## Further Information
281 To get more information about how to use a debugger, look here:
284 * [Man page of kgdb(1)](http://leaf.dragonflybsd.org/cgi/web-man?command#kgdb§ion1)
287 * [Man page of gdb(1)](http://leaf.dragonflybsd.org/cgi/web-man?command#gdb§ion1)
290 * [How to retrieve symbols from kernel modules](http://leaf.dragonflybsd.org/mailarchive/kernel/2005-11/msg00065.html)
293 * [FreeBSD Developers Handbook](http://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/book.html#KERNELDEBUG)
296 * [GDB Manual](http://sourceware.org/gdb/documentation/)
299 * [Debug tutorial from Greg Lehey](http://www.lemis.com/grog/Papers/Debug-tutorial/tutorial.pdf)