xref: /linux/Documentation/admin-guide/pstore-blk.rst (revision 0ea5c948cb64bab5bc7a5516774eb8536f05aa0d)
1.. SPDX-License-Identifier: GPL-2.0
2
3pstore block oops/panic logger
4==============================
5
6Introduction
7------------
8
9pstore block (pstore/blk) is an oops/panic logger that writes its logs to a
10block device and non-block device before the system crashes. You can get
11these log files by mounting pstore filesystem like::
12
13    mount -t pstore pstore /sys/fs/pstore
14
15
16pstore block concepts
17---------------------
18
19pstore/blk provides efficient configuration method for pstore/blk, which
20divides all configurations into two parts, configurations for user and
21configurations for driver.
22
23Configurations for user determine how pstore/blk works, such as pmsg_size,
24kmsg_size and so on. All of them support both Kconfig and module parameters,
25but module parameters have priority over Kconfig.
26
27Configurations for driver are all about block device and non-block device,
28such as total_size of block device and read/write operations.
29
30Configurations for user
31-----------------------
32
33All of these configurations support both Kconfig and module parameters, but
34module parameters have priority over Kconfig.
35
36Here is an example for module parameters::
37
38        pstore_blk.blkdev=/dev/mmcblk0p7 pstore_blk.kmsg_size=64 best_effort=y
39
40The detail of each configurations may be of interest to you.
41
42blkdev
43~~~~~~
44
45The block device to use. Most of the time, it is a partition of block device.
46It's required for pstore/blk. It is also used for MTD device.
47
48When pstore/blk is built as a module, "blkdev" accepts the following variants:
49
501. /dev/<disk_name> represents the device number of disk
51#. /dev/<disk_name><decimal> represents the device number of partition - device
52   number of disk plus the partition number
53#. /dev/<disk_name>p<decimal> - same as the above; this form is used when disk
54   name of partitioned disk ends with a digit.
55
56When pstore/blk is built into the kernel, "blkdev" accepts the following variants:
57
58#. <hex_major><hex_minor> device number in hexadecimal representation,
59   with no leading 0x, for example b302.
60#. PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF represents the unique id of
61   a partition if the partition table provides it. The UUID may be either an
62   EFI/GPT UUID, or refer to an MSDOS partition using the format SSSSSSSS-PP,
63   where SSSSSSSS is a zero-filled hex representation of the 32-bit
64   "NT disk signature", and PP is a zero-filled hex representation of the
65   1-based partition number.
66#. PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to a
67   partition with a known unique id.
68#. <major>:<minor> major and minor number of the device separated by a colon.
69
70It accepts the following variants for MTD device:
71
721. <device name> MTD device name. "pstore" is recommended.
73#. <device number> MTD device number.
74
75kmsg_size
76~~~~~~~~~
77
78The chunk size in KB for oops/panic front-end. It **MUST** be a multiple of 4.
79It's optional if you do not care about the oops/panic log.
80
81There are multiple chunks for oops/panic front-end depending on the remaining
82space except other pstore front-ends.
83
84pstore/blk will log to oops/panic chunks one by one, and always overwrite the
85oldest chunk if there is no more free chunk.
86
87pmsg_size
88~~~~~~~~~
89
90The chunk size in KB for pmsg front-end. It **MUST** be a multiple of 4.
91It's optional if you do not care about the pmsg log.
92
93Unlike oops/panic front-end, there is only one chunk for pmsg front-end.
94
95Pmsg is a user space accessible pstore object. Writes to */dev/pmsg0* are
96appended to the chunk. On reboot the contents are available in
97*/sys/fs/pstore/pmsg-pstore-blk-0*.
98
99console_size
100~~~~~~~~~~~~
101
102The chunk size in KB for console front-end.  It **MUST** be a multiple of 4.
103It's optional if you do not care about the console log.
104
105Similar to pmsg front-end, there is only one chunk for console front-end.
106
107All log of console will be appended to the chunk. On reboot the contents are
108available in */sys/fs/pstore/console-pstore-blk-0*.
109
110ftrace_size
111~~~~~~~~~~~
112
113The chunk size in KB for ftrace front-end. It **MUST** be a multiple of 4.
114It's optional if you do not care about the ftrace log.
115
116Similar to oops front-end, there are multiple chunks for ftrace front-end
117depending on the count of cpu processors. Each chunk size is equal to
118ftrace_size / processors_count.
119
120All log of ftrace will be appended to the chunk. On reboot the contents are
121combined and available in */sys/fs/pstore/ftrace-pstore-blk-0*.
122
123Persistent function tracing might be useful for debugging software or hardware
124related hangs. Here is an example of usage::
125
126 # mount -t pstore pstore /sys/fs/pstore
127 # mount -t debugfs debugfs /sys/kernel/debug/
128 # echo 1 > /sys/kernel/debug/pstore/record_ftrace
129 # reboot -f
130 [...]
131 # mount -t pstore pstore /sys/fs/pstore
132 # tail /sys/fs/pstore/ftrace-pstore-blk-0
133 CPU:0 ts:5914676 c0063828  c0063b94  call_cpuidle <- cpu_startup_entry+0x1b8/0x1e0
134 CPU:0 ts:5914678 c039ecdc  c006385c  cpuidle_enter_state <- call_cpuidle+0x44/0x48
135 CPU:0 ts:5914680 c039e9a0  c039ecf0  cpuidle_enter_freeze <- cpuidle_enter_state+0x304/0x314
136 CPU:0 ts:5914681 c0063870  c039ea30  sched_idle_set_state <- cpuidle_enter_state+0x44/0x314
137 CPU:1 ts:5916720 c0160f59  c015ee04  kernfs_unmap_bin_file <- __kernfs_remove+0x140/0x204
138 CPU:1 ts:5916721 c05ca625  c015ee0c  __mutex_lock_slowpath <- __kernfs_remove+0x148/0x204
139 CPU:1 ts:5916723 c05c813d  c05ca630  yield_to <- __mutex_lock_slowpath+0x314/0x358
140 CPU:1 ts:5916724 c05ca2d1  c05ca638  __ww_mutex_lock <- __mutex_lock_slowpath+0x31c/0x358
141
142max_reason
143~~~~~~~~~~
144
145Limiting which kinds of kmsg dumps are stored can be controlled via
146the ``max_reason`` value, as defined in include/linux/kmsg_dump.h's
147``enum kmsg_dump_reason``. For example, to store both Oopses and Panics,
148``max_reason`` should be set to 2 (KMSG_DUMP_OOPS), to store only Panics
149``max_reason`` should be set to 1 (KMSG_DUMP_PANIC). Setting this to 0
150(KMSG_DUMP_UNDEF), means the reason filtering will be controlled by the
151``printk.always_kmsg_dump`` boot param: if unset, it'll be KMSG_DUMP_OOPS,
152otherwise KMSG_DUMP_MAX.
153
154Configurations for driver
155-------------------------
156
157A device driver uses ``register_pstore_device`` with
158``struct pstore_device_info`` to register to pstore/blk.
159
160.. kernel-doc:: fs/pstore/blk.c
161   :export:
162
163Compression and header
164----------------------
165
166Block device is large enough for uncompressed oops data. Actually we do not
167recommend data compression because pstore/blk will insert some information into
168the first line of oops/panic data. For example::
169
170        Panic: Total 16 times
171
172It means that it's OOPS|Panic for the 16th time since the first booting.
173Sometimes the number of occurrences of oops|panic since the first booting is
174important to judge whether the system is stable.
175
176The following line is inserted by pstore filesystem. For example::
177
178        Oops#2 Part1
179
180It means that it's OOPS for the 2nd time on the last boot.
181
182Reading the data
183----------------
184
185The dump data can be read from the pstore filesystem. The format for these
186files is ``dmesg-pstore-blk-[N]`` for oops/panic front-end,
187``pmsg-pstore-blk-0`` for pmsg front-end and so on.  The timestamp of the
188dump file records the trigger time. To delete a stored record from block
189device, simply unlink the respective pstore file.
190
191Attentions in panic read/write APIs
192-----------------------------------
193
194If on panic, the kernel is not going to run for much longer, the tasks will not
195be scheduled and most kernel resources will be out of service. It
196looks like a single-threaded program running on a single-core computer.
197
198The following points require special attention for panic read/write APIs:
199
2001. Can **NOT** allocate any memory.
201   If you need memory, just allocate while the block driver is initializing
202   rather than waiting until the panic.
203#. Must be polled, **NOT** interrupt driven.
204   No task schedule any more. The block driver should delay to ensure the write
205   succeeds, but NOT sleep.
206#. Can **NOT** take any lock.
207   There is no other task, nor any shared resource; you are safe to break all
208   locks.
209#. Just use CPU to transfer.
210   Do not use DMA to transfer unless you are sure that DMA will not keep lock.
211#. Control registers directly.
212   Please control registers directly rather than use Linux kernel resources.
213   Do I/O map while initializing rather than wait until a panic occurs.
214#. Reset your block device and controller if necessary.
215   If you are not sure of the state of your block device and controller when
216   a panic occurs, you are safe to stop and reset them.
217
218pstore/blk supports psblk_blkdev_info(), which is defined in
219*linux/pstore_blk.h*, to get information of using block device, such as the
220device number, sector count and start sector of the whole disk.
221
222pstore block internals
223----------------------
224
225For developer reference, here are all the important structures and APIs:
226
227.. kernel-doc:: fs/pstore/zone.c
228   :internal:
229
230.. kernel-doc:: include/linux/pstore_zone.h
231   :internal:
232
233.. kernel-doc:: include/linux/pstore_blk.h
234   :internal:
235