Loading...
Note: File does not exist in v4.6.
1==========================
2Short users guide for SLUB
3==========================
4
5The basic philosophy of SLUB is very different from SLAB. SLAB
6requires rebuilding the kernel to activate debug options for all
7slab caches. SLUB always includes full debugging but it is off by default.
8SLUB can enable debugging only for selected slabs in order to avoid
9an impact on overall system performance which may make a bug more
10difficult to find.
11
12In order to switch debugging on one can add an option ``slab_debug``
13to the kernel command line. That will enable full debugging for
14all slabs.
15
16Typically one would then use the ``slabinfo`` command to get statistical
17data and perform operation on the slabs. By default ``slabinfo`` only lists
18slabs that have data in them. See "slabinfo -h" for more options when
19running the command. ``slabinfo`` can be compiled with
20::
21
22 gcc -o slabinfo tools/mm/slabinfo.c
23
24Some of the modes of operation of ``slabinfo`` require that slub debugging
25be enabled on the command line. F.e. no tracking information will be
26available without debugging on and validation can only partially
27be performed if debugging was not switched on.
28
29Some more sophisticated uses of slab_debug:
30-------------------------------------------
31
32Parameters may be given to ``slab_debug``. If none is specified then full
33debugging is enabled. Format:
34
35slab_debug=<Debug-Options>
36 Enable options for all slabs
37
38slab_debug=<Debug-Options>,<slab name1>,<slab name2>,...
39 Enable options only for select slabs (no spaces
40 after a comma)
41
42Multiple blocks of options for all slabs or selected slabs can be given, with
43blocks of options delimited by ';'. The last of "all slabs" blocks is applied
44to all slabs except those that match one of the "select slabs" block. Options
45of the first "select slabs" blocks that matches the slab's name are applied.
46
47Possible debug options are::
48
49 F Sanity checks on (enables SLAB_DEBUG_CONSISTENCY_CHECKS
50 Sorry SLAB legacy issues)
51 Z Red zoning
52 P Poisoning (object and padding)
53 U User tracking (free and alloc)
54 T Trace (please only use on single slabs)
55 A Enable failslab filter mark for the cache
56 O Switch debugging off for caches that would have
57 caused higher minimum slab orders
58 - Switch all debugging off (useful if the kernel is
59 configured with CONFIG_SLUB_DEBUG_ON)
60
61F.e. in order to boot just with sanity checks and red zoning one would specify::
62
63 slab_debug=FZ
64
65Trying to find an issue in the dentry cache? Try::
66
67 slab_debug=,dentry
68
69to only enable debugging on the dentry cache. You may use an asterisk at the
70end of the slab name, in order to cover all slabs with the same prefix. For
71example, here's how you can poison the dentry cache as well as all kmalloc
72slabs::
73
74 slab_debug=P,kmalloc-*,dentry
75
76Red zoning and tracking may realign the slab. We can just apply sanity checks
77to the dentry cache with::
78
79 slab_debug=F,dentry
80
81Debugging options may require the minimum possible slab order to increase as
82a result of storing the metadata (for example, caches with PAGE_SIZE object
83sizes). This has a higher likelihood of resulting in slab allocation errors
84in low memory situations or if there's high fragmentation of memory. To
85switch off debugging for such caches by default, use::
86
87 slab_debug=O
88
89You can apply different options to different list of slab names, using blocks
90of options. This will enable red zoning for dentry and user tracking for
91kmalloc. All other slabs will not get any debugging enabled::
92
93 slab_debug=Z,dentry;U,kmalloc-*
94
95You can also enable options (e.g. sanity checks and poisoning) for all caches
96except some that are deemed too performance critical and don't need to be
97debugged by specifying global debug options followed by a list of slab names
98with "-" as options::
99
100 slab_debug=FZ;-,zs_handle,zspage
101
102The state of each debug option for a slab can be found in the respective files
103under::
104
105 /sys/kernel/slab/<slab name>/
106
107If the file contains 1, the option is enabled, 0 means disabled. The debug
108options from the ``slab_debug`` parameter translate to the following files::
109
110 F sanity_checks
111 Z red_zone
112 P poison
113 U store_user
114 T trace
115 A failslab
116
117failslab file is writable, so writing 1 or 0 will enable or disable
118the option at runtime. Write returns -EINVAL if cache is an alias.
119Careful with tracing: It may spew out lots of information and never stop if
120used on the wrong slab.
121
122Slab merging
123============
124
125If no debug options are specified then SLUB may merge similar slabs together
126in order to reduce overhead and increase cache hotness of objects.
127``slabinfo -a`` displays which slabs were merged together.
128
129Slab validation
130===============
131
132SLUB can validate all object if the kernel was booted with slab_debug. In
133order to do so you must have the ``slabinfo`` tool. Then you can do
134::
135
136 slabinfo -v
137
138which will test all objects. Output will be generated to the syslog.
139
140This also works in a more limited way if boot was without slab debug.
141In that case ``slabinfo -v`` simply tests all reachable objects. Usually
142these are in the cpu slabs and the partial slabs. Full slabs are not
143tracked by SLUB in a non debug situation.
144
145Getting more performance
146========================
147
148To some degree SLUB's performance is limited by the need to take the
149list_lock once in a while to deal with partial slabs. That overhead is
150governed by the order of the allocation for each slab. The allocations
151can be influenced by kernel parameters:
152
153.. slab_min_objects=x (default: automatically scaled by number of cpus)
154.. slab_min_order=x (default 0)
155.. slab_max_order=x (default 3 (PAGE_ALLOC_COSTLY_ORDER))
156
157``slab_min_objects``
158 allows to specify how many objects must at least fit into one
159 slab in order for the allocation order to be acceptable. In
160 general slub will be able to perform this number of
161 allocations on a slab without consulting centralized resources
162 (list_lock) where contention may occur.
163
164``slab_min_order``
165 specifies a minimum order of slabs. A similar effect like
166 ``slab_min_objects``.
167
168``slab_max_order``
169 specified the order at which ``slab_min_objects`` should no
170 longer be checked. This is useful to avoid SLUB trying to
171 generate super large order pages to fit ``slab_min_objects``
172 of a slab cache with large object sizes into one high order
173 page. Setting command line parameter
174 ``debug_guardpage_minorder=N`` (N > 0), forces setting
175 ``slab_max_order`` to 0, what cause minimum possible order of
176 slabs allocation.
177
178``slab_strict_numa``
179 Enables the application of memory policies on each
180 allocation. This results in more accurate placement of
181 objects which may result in the reduction of accesses
182 to remote nodes. The default is to only apply memory
183 policies at the folio level when a new folio is acquired
184 or a folio is retrieved from the lists. Enabling this
185 option reduces the fastpath performance of the slab allocator.
186
187SLUB Debug output
188=================
189
190Here is a sample of slub debug output::
191
192 ====================================================================
193 BUG kmalloc-8: Right Redzone overwritten
194 --------------------------------------------------------------------
195
196 INFO: 0xc90f6d28-0xc90f6d2b. First byte 0x00 instead of 0xcc
197 INFO: Slab 0xc528c530 flags=0x400000c3 inuse=61 fp=0xc90f6d58
198 INFO: Object 0xc90f6d20 @offset=3360 fp=0xc90f6d58
199 INFO: Allocated in get_modalias+0x61/0xf5 age=53 cpu=1 pid=554
200
201 Bytes b4 (0xc90f6d10): 00 00 00 00 00 00 00 00 5a 5a 5a 5a 5a 5a 5a 5a ........ZZZZZZZZ
202 Object (0xc90f6d20): 31 30 31 39 2e 30 30 35 1019.005
203 Redzone (0xc90f6d28): 00 cc cc cc .
204 Padding (0xc90f6d50): 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZ
205
206 [<c010523d>] dump_trace+0x63/0x1eb
207 [<c01053df>] show_trace_log_lvl+0x1a/0x2f
208 [<c010601d>] show_trace+0x12/0x14
209 [<c0106035>] dump_stack+0x16/0x18
210 [<c017e0fa>] object_err+0x143/0x14b
211 [<c017e2cc>] check_object+0x66/0x234
212 [<c017eb43>] __slab_free+0x239/0x384
213 [<c017f446>] kfree+0xa6/0xc6
214 [<c02e2335>] get_modalias+0xb9/0xf5
215 [<c02e23b7>] dmi_dev_uevent+0x27/0x3c
216 [<c027866a>] dev_uevent+0x1ad/0x1da
217 [<c0205024>] kobject_uevent_env+0x20a/0x45b
218 [<c020527f>] kobject_uevent+0xa/0xf
219 [<c02779f1>] store_uevent+0x4f/0x58
220 [<c027758e>] dev_attr_store+0x29/0x2f
221 [<c01bec4f>] sysfs_write_file+0x16e/0x19c
222 [<c0183ba7>] vfs_write+0xd1/0x15a
223 [<c01841d7>] sys_write+0x3d/0x72
224 [<c0104112>] sysenter_past_esp+0x5f/0x99
225 [<b7f7b410>] 0xb7f7b410
226 =======================
227
228 FIX kmalloc-8: Restoring Redzone 0xc90f6d28-0xc90f6d2b=0xcc
229
230If SLUB encounters a corrupted object (full detection requires the kernel
231to be booted with slab_debug) then the following output will be dumped
232into the syslog:
233
2341. Description of the problem encountered
235
236 This will be a message in the system log starting with::
237
238 ===============================================
239 BUG <slab cache affected>: <What went wrong>
240 -----------------------------------------------
241
242 INFO: <corruption start>-<corruption_end> <more info>
243 INFO: Slab <address> <slab information>
244 INFO: Object <address> <object information>
245 INFO: Allocated in <kernel function> age=<jiffies since alloc> cpu=<allocated by
246 cpu> pid=<pid of the process>
247 INFO: Freed in <kernel function> age=<jiffies since free> cpu=<freed by cpu>
248 pid=<pid of the process>
249
250 (Object allocation / free information is only available if SLAB_STORE_USER is
251 set for the slab. slab_debug sets that option)
252
2532. The object contents if an object was involved.
254
255 Various types of lines can follow the BUG SLUB line:
256
257 Bytes b4 <address> : <bytes>
258 Shows a few bytes before the object where the problem was detected.
259 Can be useful if the corruption does not stop with the start of the
260 object.
261
262 Object <address> : <bytes>
263 The bytes of the object. If the object is inactive then the bytes
264 typically contain poison values. Any non-poison value shows a
265 corruption by a write after free.
266
267 Redzone <address> : <bytes>
268 The Redzone following the object. The Redzone is used to detect
269 writes after the object. All bytes should always have the same
270 value. If there is any deviation then it is due to a write after
271 the object boundary.
272
273 (Redzone information is only available if SLAB_RED_ZONE is set.
274 slab_debug sets that option)
275
276 Padding <address> : <bytes>
277 Unused data to fill up the space in order to get the next object
278 properly aligned. In the debug case we make sure that there are
279 at least 4 bytes of padding. This allows the detection of writes
280 before the object.
281
2823. A stackdump
283
284 The stackdump describes the location where the error was detected. The cause
285 of the corruption is may be more likely found by looking at the function that
286 allocated or freed the object.
287
2884. Report on how the problem was dealt with in order to ensure the continued
289 operation of the system.
290
291 These are messages in the system log beginning with::
292
293 FIX <slab cache affected>: <corrective action taken>
294
295 In the above sample SLUB found that the Redzone of an active object has
296 been overwritten. Here a string of 8 characters was written into a slab that
297 has the length of 8 characters. However, a 8 character string needs a
298 terminating 0. That zero has overwritten the first byte of the Redzone field.
299 After reporting the details of the issue encountered the FIX SLUB message
300 tells us that SLUB has restored the Redzone to its proper value and then
301 system operations continue.
302
303Emergency operations
304====================
305
306Minimal debugging (sanity checks alone) can be enabled by booting with::
307
308 slab_debug=F
309
310This will be generally be enough to enable the resiliency features of slub
311which will keep the system running even if a bad kernel component will
312keep corrupting objects. This may be important for production systems.
313Performance will be impacted by the sanity checks and there will be a
314continual stream of error messages to the syslog but no additional memory
315will be used (unlike full debugging).
316
317No guarantees. The kernel component still needs to be fixed. Performance
318may be optimized further by locating the slab that experiences corruption
319and enabling debugging only for that cache
320
321I.e.::
322
323 slab_debug=F,dentry
324
325If the corruption occurs by writing after the end of the object then it
326may be advisable to enable a Redzone to avoid corrupting the beginning
327of other objects::
328
329 slab_debug=FZ,dentry
330
331Extended slabinfo mode and plotting
332===================================
333
334The ``slabinfo`` tool has a special 'extended' ('-X') mode that includes:
335 - Slabcache Totals
336 - Slabs sorted by size (up to -N <num> slabs, default 1)
337 - Slabs sorted by loss (up to -N <num> slabs, default 1)
338
339Additionally, in this mode ``slabinfo`` does not dynamically scale
340sizes (G/M/K) and reports everything in bytes (this functionality is
341also available to other slabinfo modes via '-B' option) which makes
342reporting more precise and accurate. Moreover, in some sense the `-X'
343mode also simplifies the analysis of slabs' behaviour, because its
344output can be plotted using the ``slabinfo-gnuplot.sh`` script. So it
345pushes the analysis from looking through the numbers (tons of numbers)
346to something easier -- visual analysis.
347
348To generate plots:
349
350a) collect slabinfo extended records, for example::
351
352 while [ 1 ]; do slabinfo -X >> FOO_STATS; sleep 1; done
353
354b) pass stats file(-s) to ``slabinfo-gnuplot.sh`` script::
355
356 slabinfo-gnuplot.sh FOO_STATS [FOO_STATS2 .. FOO_STATSN]
357
358 The ``slabinfo-gnuplot.sh`` script will pre-processes the collected records
359 and generates 3 png files (and 3 pre-processing cache files) per STATS
360 file:
361 - Slabcache Totals: FOO_STATS-totals.png
362 - Slabs sorted by size: FOO_STATS-slabs-by-size.png
363 - Slabs sorted by loss: FOO_STATS-slabs-by-loss.png
364
365Another use case, when ``slabinfo-gnuplot.sh`` can be useful, is when you
366need to compare slabs' behaviour "prior to" and "after" some code
367modification. To help you out there, ``slabinfo-gnuplot.sh`` script
368can 'merge' the `Slabcache Totals` sections from different
369measurements. To visually compare N plots:
370
371a) Collect as many STATS1, STATS2, .. STATSN files as you need::
372
373 while [ 1 ]; do slabinfo -X >> STATS<X>; sleep 1; done
374
375b) Pre-process those STATS files::
376
377 slabinfo-gnuplot.sh STATS1 STATS2 .. STATSN
378
379c) Execute ``slabinfo-gnuplot.sh`` in '-t' mode, passing all of the
380 generated pre-processed \*-totals::
381
382 slabinfo-gnuplot.sh -t STATS1-totals STATS2-totals .. STATSN-totals
383
384 This will produce a single plot (png file).
385
386 Plots, expectedly, can be large so some fluctuations or small spikes
387 can go unnoticed. To deal with that, ``slabinfo-gnuplot.sh`` has two
388 options to 'zoom-in'/'zoom-out':
389
390 a) ``-s %d,%d`` -- overwrites the default image width and height
391 b) ``-r %d,%d`` -- specifies a range of samples to use (for example,
392 in ``slabinfo -X >> FOO_STATS; sleep 1;`` case, using a ``-r
393 40,60`` range will plot only samples collected between 40th and
394 60th seconds).
395
396
397DebugFS files for SLUB
398======================
399
400For more information about current state of SLUB caches with the user tracking
401debug option enabled, debugfs files are available, typically under
402/sys/kernel/debug/slab/<cache>/ (created only for caches with enabled user
403tracking). There are 2 types of these files with the following debug
404information:
405
4061. alloc_traces::
407
408 Prints information about unique allocation traces of the currently
409 allocated objects. The output is sorted by frequency of each trace.
410
411 Information in the output:
412 Number of objects, allocating function, possible memory wastage of
413 kmalloc objects(total/per-object), minimal/average/maximal jiffies
414 since alloc, pid range of the allocating processes, cpu mask of
415 allocating cpus, numa node mask of origins of memory, and stack trace.
416
417 Example:::
418
419 338 pci_alloc_dev+0x2c/0xa0 waste=521872/1544 age=290837/291891/293509 pid=1 cpus=106 nodes=0-1
420 __kmem_cache_alloc_node+0x11f/0x4e0
421 kmalloc_trace+0x26/0xa0
422 pci_alloc_dev+0x2c/0xa0
423 pci_scan_single_device+0xd2/0x150
424 pci_scan_slot+0xf7/0x2d0
425 pci_scan_child_bus_extend+0x4e/0x360
426 acpi_pci_root_create+0x32e/0x3b0
427 pci_acpi_scan_root+0x2b9/0x2d0
428 acpi_pci_root_add.cold.11+0x110/0xb0a
429 acpi_bus_attach+0x262/0x3f0
430 device_for_each_child+0xb7/0x110
431 acpi_dev_for_each_child+0x77/0xa0
432 acpi_bus_attach+0x108/0x3f0
433 device_for_each_child+0xb7/0x110
434 acpi_dev_for_each_child+0x77/0xa0
435 acpi_bus_attach+0x108/0x3f0
436
4372. free_traces::
438
439 Prints information about unique freeing traces of the currently allocated
440 objects. The freeing traces thus come from the previous life-cycle of the
441 objects and are reported as not available for objects allocated for the first
442 time. The output is sorted by frequency of each trace.
443
444 Information in the output:
445 Number of objects, freeing function, minimal/average/maximal jiffies since free,
446 pid range of the freeing processes, cpu mask of freeing cpus, and stack trace.
447
448 Example:::
449
450 1980 <not-available> age=4294912290 pid=0 cpus=0
451 51 acpi_ut_update_ref_count+0x6a6/0x782 age=236886/237027/237772 pid=1 cpus=1
452 kfree+0x2db/0x420
453 acpi_ut_update_ref_count+0x6a6/0x782
454 acpi_ut_update_object_reference+0x1ad/0x234
455 acpi_ut_remove_reference+0x7d/0x84
456 acpi_rs_get_prt_method_data+0x97/0xd6
457 acpi_get_irq_routing_table+0x82/0xc4
458 acpi_pci_irq_find_prt_entry+0x8e/0x2e0
459 acpi_pci_irq_lookup+0x3a/0x1e0
460 acpi_pci_irq_enable+0x77/0x240
461 pcibios_enable_device+0x39/0x40
462 do_pci_enable_device.part.0+0x5d/0xe0
463 pci_enable_device_flags+0xfc/0x120
464 pci_enable_device+0x13/0x20
465 virtio_pci_probe+0x9e/0x170
466 local_pci_probe+0x48/0x80
467 pci_device_probe+0x105/0x1c0
468
469Christoph Lameter, May 30, 2007
470Sergey Senozhatsky, October 23, 2015