1*d5af79c0SRandy Dunlap.. highlight:: none
2*d5af79c0SRandy Dunlap
3*d5af79c0SRandy DunlapDebugging kernel and modules via gdb
4*d5af79c0SRandy Dunlap====================================
5*d5af79c0SRandy Dunlap
6*d5af79c0SRandy DunlapThe kernel debugger kgdb, hypervisors like QEMU or JTAG-based hardware
7*d5af79c0SRandy Dunlapinterfaces allow to debug the Linux kernel and its modules during runtime
8*d5af79c0SRandy Dunlapusing gdb. Gdb comes with a powerful scripting interface for python. The
9*d5af79c0SRandy Dunlapkernel provides a collection of helper scripts that can simplify typical
10*d5af79c0SRandy Dunlapkernel debugging steps. This is a short tutorial about how to enable and use
11*d5af79c0SRandy Dunlapthem. It focuses on QEMU/KVM virtual machines as target, but the examples can
12*d5af79c0SRandy Dunlapbe transferred to the other gdb stubs as well.
13*d5af79c0SRandy Dunlap
14*d5af79c0SRandy Dunlap
15*d5af79c0SRandy DunlapRequirements
16*d5af79c0SRandy Dunlap------------
17*d5af79c0SRandy Dunlap
18*d5af79c0SRandy Dunlap- gdb 7.2+ (recommended: 7.4+) with python support enabled (typically true
19*d5af79c0SRandy Dunlap  for distributions)
20*d5af79c0SRandy Dunlap
21*d5af79c0SRandy Dunlap
22*d5af79c0SRandy DunlapSetup
23*d5af79c0SRandy Dunlap-----
24*d5af79c0SRandy Dunlap
25*d5af79c0SRandy Dunlap- Create a virtual Linux machine for QEMU/KVM (see www.linux-kvm.org and
26*d5af79c0SRandy Dunlap  www.qemu.org for more details). For cross-development,
27*d5af79c0SRandy Dunlap  https://landley.net/aboriginal/bin keeps a pool of machine images and
28*d5af79c0SRandy Dunlap  toolchains that can be helpful to start from.
29*d5af79c0SRandy Dunlap
30*d5af79c0SRandy Dunlap- Build the kernel with CONFIG_GDB_SCRIPTS enabled, but leave
31*d5af79c0SRandy Dunlap  CONFIG_DEBUG_INFO_REDUCED off. If your architecture supports
32*d5af79c0SRandy Dunlap  CONFIG_FRAME_POINTER, keep it enabled.
33*d5af79c0SRandy Dunlap
34*d5af79c0SRandy Dunlap- Install that kernel on the guest, turn off KASLR if necessary by adding
35*d5af79c0SRandy Dunlap  "nokaslr" to the kernel command line.
36*d5af79c0SRandy Dunlap  Alternatively, QEMU allows to boot the kernel directly using -kernel,
37*d5af79c0SRandy Dunlap  -append, -initrd command line switches. This is generally only useful if
38*d5af79c0SRandy Dunlap  you do not depend on modules. See QEMU documentation for more details on
39*d5af79c0SRandy Dunlap  this mode. In this case, you should build the kernel with
40*d5af79c0SRandy Dunlap  CONFIG_RANDOMIZE_BASE disabled if the architecture supports KASLR.
41*d5af79c0SRandy Dunlap
42*d5af79c0SRandy Dunlap- Build the gdb scripts (required on kernels v5.1 and above)::
43*d5af79c0SRandy Dunlap
44*d5af79c0SRandy Dunlap    make scripts_gdb
45*d5af79c0SRandy Dunlap
46*d5af79c0SRandy Dunlap- Enable the gdb stub of QEMU/KVM, either
47*d5af79c0SRandy Dunlap
48*d5af79c0SRandy Dunlap    - at VM startup time by appending "-s" to the QEMU command line
49*d5af79c0SRandy Dunlap
50*d5af79c0SRandy Dunlap  or
51*d5af79c0SRandy Dunlap
52*d5af79c0SRandy Dunlap    - during runtime by issuing "gdbserver" from the QEMU monitor
53*d5af79c0SRandy Dunlap      console
54*d5af79c0SRandy Dunlap
55*d5af79c0SRandy Dunlap- cd /path/to/linux-build
56*d5af79c0SRandy Dunlap
57*d5af79c0SRandy Dunlap- Start gdb: gdb vmlinux
58*d5af79c0SRandy Dunlap
59*d5af79c0SRandy Dunlap  Note: Some distros may restrict auto-loading of gdb scripts to known safe
60*d5af79c0SRandy Dunlap  directories. In case gdb reports to refuse loading vmlinux-gdb.py, add::
61*d5af79c0SRandy Dunlap
62*d5af79c0SRandy Dunlap    add-auto-load-safe-path /path/to/linux-build
63*d5af79c0SRandy Dunlap
64*d5af79c0SRandy Dunlap  to ~/.gdbinit. See gdb help for more details.
65*d5af79c0SRandy Dunlap
66*d5af79c0SRandy Dunlap- Attach to the booted guest::
67*d5af79c0SRandy Dunlap
68*d5af79c0SRandy Dunlap    (gdb) target remote :1234
69*d5af79c0SRandy Dunlap
70*d5af79c0SRandy Dunlap
71*d5af79c0SRandy DunlapExamples of using the Linux-provided gdb helpers
72*d5af79c0SRandy Dunlap------------------------------------------------
73*d5af79c0SRandy Dunlap
74*d5af79c0SRandy Dunlap- Load module (and main kernel) symbols::
75*d5af79c0SRandy Dunlap
76*d5af79c0SRandy Dunlap    (gdb) lx-symbols
77*d5af79c0SRandy Dunlap    loading vmlinux
78*d5af79c0SRandy Dunlap    scanning for modules in /home/user/linux/build
79*d5af79c0SRandy Dunlap    loading @0xffffffffa0020000: /home/user/linux/build/net/netfilter/xt_tcpudp.ko
80*d5af79c0SRandy Dunlap    loading @0xffffffffa0016000: /home/user/linux/build/net/netfilter/xt_pkttype.ko
81*d5af79c0SRandy Dunlap    loading @0xffffffffa0002000: /home/user/linux/build/net/netfilter/xt_limit.ko
82*d5af79c0SRandy Dunlap    loading @0xffffffffa00ca000: /home/user/linux/build/net/packet/af_packet.ko
83*d5af79c0SRandy Dunlap    loading @0xffffffffa003c000: /home/user/linux/build/fs/fuse/fuse.ko
84*d5af79c0SRandy Dunlap    ...
85*d5af79c0SRandy Dunlap    loading @0xffffffffa0000000: /home/user/linux/build/drivers/ata/ata_generic.ko
86*d5af79c0SRandy Dunlap
87*d5af79c0SRandy Dunlap- Set a breakpoint on some not yet loaded module function, e.g.::
88*d5af79c0SRandy Dunlap
89*d5af79c0SRandy Dunlap    (gdb) b btrfs_init_sysfs
90*d5af79c0SRandy Dunlap    Function "btrfs_init_sysfs" not defined.
91*d5af79c0SRandy Dunlap    Make breakpoint pending on future shared library load? (y or [n]) y
92*d5af79c0SRandy Dunlap    Breakpoint 1 (btrfs_init_sysfs) pending.
93*d5af79c0SRandy Dunlap
94*d5af79c0SRandy Dunlap- Continue the target::
95*d5af79c0SRandy Dunlap
96*d5af79c0SRandy Dunlap    (gdb) c
97*d5af79c0SRandy Dunlap
98*d5af79c0SRandy Dunlap- Load the module on the target and watch the symbols being loaded as well as
99*d5af79c0SRandy Dunlap  the breakpoint hit::
100*d5af79c0SRandy Dunlap
101*d5af79c0SRandy Dunlap    loading @0xffffffffa0034000: /home/user/linux/build/lib/libcrc32c.ko
102*d5af79c0SRandy Dunlap    loading @0xffffffffa0050000: /home/user/linux/build/lib/lzo/lzo_compress.ko
103*d5af79c0SRandy Dunlap    loading @0xffffffffa006e000: /home/user/linux/build/lib/zlib_deflate/zlib_deflate.ko
104*d5af79c0SRandy Dunlap    loading @0xffffffffa01b1000: /home/user/linux/build/fs/btrfs/btrfs.ko
105*d5af79c0SRandy Dunlap
106*d5af79c0SRandy Dunlap    Breakpoint 1, btrfs_init_sysfs () at /home/user/linux/fs/btrfs/sysfs.c:36
107*d5af79c0SRandy Dunlap    36              btrfs_kset = kset_create_and_add("btrfs", NULL, fs_kobj);
108*d5af79c0SRandy Dunlap
109*d5af79c0SRandy Dunlap- Dump the log buffer of the target kernel::
110*d5af79c0SRandy Dunlap
111*d5af79c0SRandy Dunlap    (gdb) lx-dmesg
112*d5af79c0SRandy Dunlap    [     0.000000] Initializing cgroup subsys cpuset
113*d5af79c0SRandy Dunlap    [     0.000000] Initializing cgroup subsys cpu
114*d5af79c0SRandy Dunlap    [     0.000000] Linux version 3.8.0-rc4-dbg+ (...
115*d5af79c0SRandy Dunlap    [     0.000000] Command line: root=/dev/sda2 resume=/dev/sda1 vga=0x314
116*d5af79c0SRandy Dunlap    [     0.000000] e820: BIOS-provided physical RAM map:
117*d5af79c0SRandy Dunlap    [     0.000000] BIOS-e820: [mem 0x0000000000000000-0x000000000009fbff] usable
118*d5af79c0SRandy Dunlap    [     0.000000] BIOS-e820: [mem 0x000000000009fc00-0x000000000009ffff] reserved
119*d5af79c0SRandy Dunlap    ....
120*d5af79c0SRandy Dunlap
121*d5af79c0SRandy Dunlap- Examine fields of the current task struct(supported by x86 and arm64 only)::
122*d5af79c0SRandy Dunlap
123*d5af79c0SRandy Dunlap    (gdb) p $lx_current().pid
124*d5af79c0SRandy Dunlap    $1 = 4998
125*d5af79c0SRandy Dunlap    (gdb) p $lx_current().comm
126*d5af79c0SRandy Dunlap    $2 = "modprobe\000\000\000\000\000\000\000"
127*d5af79c0SRandy Dunlap
128*d5af79c0SRandy Dunlap- Make use of the per-cpu function for the current or a specified CPU::
129*d5af79c0SRandy Dunlap
130*d5af79c0SRandy Dunlap    (gdb) p $lx_per_cpu("runqueues").nr_running
131*d5af79c0SRandy Dunlap    $3 = 1
132*d5af79c0SRandy Dunlap    (gdb) p $lx_per_cpu("runqueues", 2).nr_running
133*d5af79c0SRandy Dunlap    $4 = 0
134*d5af79c0SRandy Dunlap
135*d5af79c0SRandy Dunlap- Dig into hrtimers using the container_of helper::
136*d5af79c0SRandy Dunlap
137*d5af79c0SRandy Dunlap    (gdb) set $next = $lx_per_cpu("hrtimer_bases").clock_base[0].active.next
138*d5af79c0SRandy Dunlap    (gdb) p *$container_of($next, "struct hrtimer", "node")
139*d5af79c0SRandy Dunlap    $5 = {
140*d5af79c0SRandy Dunlap      node = {
141*d5af79c0SRandy Dunlap        node = {
142*d5af79c0SRandy Dunlap          __rb_parent_color = 18446612133355256072,
143*d5af79c0SRandy Dunlap          rb_right = 0x0 <irq_stack_union>,
144*d5af79c0SRandy Dunlap          rb_left = 0x0 <irq_stack_union>
145*d5af79c0SRandy Dunlap        },
146*d5af79c0SRandy Dunlap        expires = {
147*d5af79c0SRandy Dunlap          tv64 = 1835268000000
148*d5af79c0SRandy Dunlap        }
149*d5af79c0SRandy Dunlap      },
150*d5af79c0SRandy Dunlap      _softexpires = {
151*d5af79c0SRandy Dunlap        tv64 = 1835268000000
152*d5af79c0SRandy Dunlap      },
153*d5af79c0SRandy Dunlap      function = 0xffffffff81078232 <tick_sched_timer>,
154*d5af79c0SRandy Dunlap      base = 0xffff88003fd0d6f0,
155*d5af79c0SRandy Dunlap      state = 1,
156*d5af79c0SRandy Dunlap      start_pid = 0,
157*d5af79c0SRandy Dunlap      start_site = 0xffffffff81055c1f <hrtimer_start_range_ns+20>,
158*d5af79c0SRandy Dunlap      start_comm = "swapper/2\000\000\000\000\000\000"
159*d5af79c0SRandy Dunlap    }
160*d5af79c0SRandy Dunlap
161*d5af79c0SRandy Dunlap
162*d5af79c0SRandy DunlapList of commands and functions
163*d5af79c0SRandy Dunlap------------------------------
164*d5af79c0SRandy Dunlap
165*d5af79c0SRandy DunlapThe number of commands and convenience functions may evolve over the time,
166*d5af79c0SRandy Dunlapthis is just a snapshot of the initial version::
167*d5af79c0SRandy Dunlap
168*d5af79c0SRandy Dunlap (gdb) apropos lx
169*d5af79c0SRandy Dunlap function lx_current -- Return current task
170*d5af79c0SRandy Dunlap function lx_module -- Find module by name and return the module variable
171*d5af79c0SRandy Dunlap function lx_per_cpu -- Return per-cpu variable
172*d5af79c0SRandy Dunlap function lx_task_by_pid -- Find Linux task by PID and return the task_struct variable
173*d5af79c0SRandy Dunlap function lx_thread_info -- Calculate Linux thread_info from task variable
174*d5af79c0SRandy Dunlap lx-dmesg -- Print Linux kernel log buffer
175*d5af79c0SRandy Dunlap lx-lsmod -- List currently loaded modules
176*d5af79c0SRandy Dunlap lx-symbols -- (Re-)load symbols of Linux kernel and currently loaded modules
177*d5af79c0SRandy Dunlap
178*d5af79c0SRandy DunlapDetailed help can be obtained via "help <command-name>" for commands and "help
179*d5af79c0SRandy Dunlapfunction <function-name>" for convenience functions.
180