Annotation of www/ddb.html, Revision 1.16
1.12 tb 1: <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
1.1 beck 2: <html>
3: <head>
1.10 tj 4: <title>OpenBSD: Crash Reports</title>
1.1 beck 5: <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
6: <meta name="description" content="How to report an OpenBSD kernel crash">
1.9 tb 7: <meta name="copyright" content="This document copyright 1998-2016 by OpenBSD.">
8: <meta name="viewport" content="width=device-width, initial-scale=1">
9: <link rel="stylesheet" type="text/css" href="openbsd.css">
1.13 tb 10: <link rel="canonical" href="https://www.openbsd.org/report.html">
1.14 tb 11: <style type="text/css">
12: h3, h4 { color: #0000e0; }
13: </style>
1.1 beck 14: </head>
15:
16: <body bgcolor="#ffffff" text="#000000" link="#23238e">
1.9 tb 17:
18: <h2>
19: <a href="index.html">
20: <font color="#0000ff"><i>Open</i></font><font color="#000084">BSD</font></a>
1.10 tj 21: <font color="#e00000">Crash Reports</font>
1.9 tb 22: </h2>
23: <hr>
1.1 beck 24: <p>
25:
1.14 tb 26: <h3>Minimum information for kernel problems</h3>
1.1 beck 27:
1.6 tb 28: Familiarize yourself with
29: <a href="report.html">the general bug reporting procedures</a>
30: first.
31: All of that will apply.
1.1 beck 32: When reporting a kernel panic or crash, please remember:
33:
34: <ul>
1.14 tb 35: <li><i>We need the console output on the screen</i>.
36: Capture it and save it.
37: Serial consoles are best, but if you are on a VGA console you can
38: <a href="faq/faq7.html">scroll the console back</a>
39: and take readable pictures with a phone or camera.<br>
40:
41: <li><i>If the kernel panicked we need the traceback.</i>
42: It may be displayed on the screen.
43: If you are at a
44: <tt><a href="http://man.openbsd.org/ddb.4">ddb</a>></tt>
45: prompt, type <tt>trace</tt>.
46: If you are running SMP, use the <tt>mach ddbcpu N</tt> command for each
47: of the <tt>N</tt> processors you have and repeat the <tt>trace</tt>
48: command for each processor.<br>
1.1 beck 49:
1.14 tb 50: <li><i>We need the process list.</i>
51: Use the command <tt>ps</tt> to get that.
1.1 beck 52: </ul>
53:
1.14 tb 54: <i>
55: Reports without the above information are useless.
56: This is the minimum we need to be able to track down the issue.
57: </i>
58:
59: <h3>Additional information you can send</h3>
1.1 beck 60:
1.6 tb 61: In some situations more information is desirable.
62: Below are outlined some additional steps you can take in certain situations:
1.14 tb 63:
1.1 beck 64: <ul>
1.14 tb 65: <li><i>If your crash appears to involve filesystems.</i>
66: The following additional things would be helpful
67: <ul>
68: <li>The output of the
69: <tt><a href="http://man.openbsd.org/ddb.4">ddb</a>></tt> command
70: <tt>show uvm</tt>
71: <li>The output of the
72: <tt><a href="http://man.openbsd.org/ddb.4">ddb</a>></tt>
73: command <tt>show bcstats</tt>
74: <li>The output of the <tt>mount</tt> command from your running machine, so
75: we know what filesystems are mounted and how.
76: </ul>
77: <li> ... XXX boot crash? XXX
78: <li> ... XXX show regs? XXX
1.1 beck 79: </ul>
1.14 tb 80:
81: <h3>Lost the panic message?</h3>
82:
83: Under some circumstances, you may lose the very first message of a panic,
84: stating the reason for the panic.
85:
86: <blockquote><pre>
87: ddb> <b>show panic</b>
88: 0: kernel: page fault trap, code=0
89: ddb>
90: </pre></blockquote>
91:
92: <h3>Note for SMP systems</h3>
93:
94: You should get a trace from each processor as part of your report:
95:
96: <blockquote><pre>
97: ddb{0}> <b>trace</b>
98: pool_get(d05e7c20,0,dab19ef8,d0169414,80) at pool_get+0x226
99: fxp_add_rfabuf(d0a62000,d3c12b00,dab19f10,dab19f10) at fxp_add_rfabuf+0xa5
100: fxp_intr(d0a62000) at fxp_intr+0x1e7
101: Xintr_ioapic0() at Xintr_ioapic0+0x6d
102: --- interrupt ---
103: idle_loop+0x21:
104: ddb{0}> <b>machine ddbcpu 1</b>
105: Stopped at Debugger+0x4: leave
106: ddb{1}> <b>trace</b>
107: Debugger(d0319e28,d05ff5a0,dab1bee8,d031cc6e,d0a61800) at Debugger+0x4
108: i386_ipi_db(d0a61800,d05ff5a0,dab1bef8,d01eb997) at i386_ipi_db+0xb
109: i386_ipi_handler(b0,d05f0058,dab10010,d01d0010,dab10010) at i386_ipi_handler+0x
110: 4a
111: Xintripi() at Xintripi+0x47
112: --- interrupt ---
113: i386_softintlock(0,58,dab10010,dab10010,d01e0010) at i386_softintlock+0x37
114: Xintrltimer() at Xintrltimer+0x47
115: --- interrupt ---
116: idle_loop+0x21:
117: ddb{1}>
118: </pre></blockquote>
119:
120: Repeat the <tt>machine ddbcpu x</tt> followed by <tt>trace</tt> for each
121: processor in your machine.
122:
123: <h3>How do I gather further information from a kernel crash?</h3><p>
124:
125: A typical kernel crash on OpenBSD might look like this:
126: (things to watch for are marked with bold font)
127:
128: <blockquote><pre>
129: kernel: page fault trap, code=0
130: Stopped at <b>_pf_route+0x263</b>: mov 0x40(%edi),%edx
131: ddb>
132: </pre></blockquote>
133:
134: The first command to run from the <tt>ddb></tt> prompt is <tt>trace</tt>
135: (see <a href="http://man.openbsd.org/ddb">ddb(4)</a> for details):
136:
137: <blockquote><pre>
138: ddb> <b>trace</b>
139: <b>_pf_route</b>(e28cb7e4,e28bc978,2,1fad,d0b8b120) at <b>_pf_route+0x263</b>
140: _pf_test(2,1f4ad,e28cb7e4,b4c1) at _pf_test+0x706
141: _pf_route(e28cbb00,e28bc978,2,d0a65440,d0b8b120) at _pf_route+0x207
142: _pf_test(2,d0a65440,e28cbb00,d023c282) at _pf_test+0x706
143: _ip_output(d0b6a200,0,0,0,0) at _ip_output+0xb67
144: _icmp_send(d0b6a200,0,1,a012) at _icmp_send+0x57
145: _icmp_reflect(d0b6a200,0,1,0,3) at _icmp_reflect+0x26b
146: _icmp_input(d0b6a200,14,0,0,d0b6a200) at _icmp_input+0x42c
147: _ipv4_input(d0b6a200,e289f140,d0a489e0,e289f140) at _ipv4_input+0x6eb
148: _ipintr(10,10,e289f140,e289f140,e28cbd38) at _ipintr+0x8d
149: Bad frame pointer: 0xe28cbcac
150: ddb>
151: </pre></blockquote>
152:
153: This tells us what function calls lead to the crash.
154:
155: <p>
156: To find out the particular line of C code that caused the crash, you can
157: do the following:
158:
159: <p>
160: Find the source file where the crashing function is defined in.
161: In this example, that would be <tt>pf_route()</tt> in <tt>sys/net/pf.c</tt>.
162: Recompile that source file with debug information:
163:
164: <blockquote><pre>
165: # <b>cd /usr/src/sys/arch/$(uname -m)/compile/GENERIC</b>
1.15 tb 166: # <b>rm obj/pf.o</b>
1.14 tb 167: # <b>DEBUG=-g make pf.o</b>
168: </pre></blockquote>
169:
170: Then use <a href="http://man.openbsd.org/objdump">objdump(1)</a> to get the
171: disassembly:
172:
173: <blockquote><pre>
1.16 ! tb 174: # <b>objdump --line --disassemble --reloc obj/pf.o >pf.dis</b>
1.14 tb 175: </pre></blockquote>
176:
177: In the output, grep for the function name:
178:
179: <blockquote><pre>
180: # <b>grep "<_pf_route>:" pf.dis</b>
181: 0000<b>7d88</b> <_pf_route>:
182: </pre></blockquote>
183:
184: Take this first hex number and add the offset from the <tt>Stopped at</tt> line:
185: <tt>0x7d88 + 0x263 == 0x7feb</tt>.
186:
187: <p>
188: Scroll down to that line (the assembler instruction should match the one
189: quoted in the <tt>Stopped at</tt> line), then up to the nearest C line number:
190:
191: <blockquote><pre>
192: # <b>more pf.dis</b>
193: /usr/src/sys/arch/i386/compile/GENERIC/../../../../net/pf.c:<b>3872</b>
194: 7fe7: 0f b7 43 02 movzwl 0x2(%ebx),%eax
195: <b>7feb</b>: 8b 57 40 mov 0x40(%edi),%edx
196: 7fee: 39 d0 cmp %edx,%eax
197: 7ff0: 0f 87 92 00 00 00 ja 8088 <_pf_route+0x300>
198: </pre></blockquote>
199:
200: So, it's precisely line <tt>3872</tt> of <tt>pf.c</tt> that crashes:
201:
202: <blockquote><pre>
203: # <b>cat -n pf.c | head -n 3872 | tail -n 1</b>
204: 3872 if ((u_int16_t)ip->ip_len <= ifp->if_mtu) {
205: </pre></blockquote>
206:
207: Note that the kernel that produced the crash output and the object file
208: for objdump must be compiled from the exact same source file, otherwise
209: the offsets won't match.
210:
211: <p>
212: If you provide both the ddb trace output and the relevant objdump section,
213: that's very helpful.
1.1 beck 214:
215: <p>
216: </body>
217: </html>