Based on kernel version 2.6.26. Page generated on 2008-07-16 21:13 EST.
1 Short users guide for SLUB 2 -------------------------- 3 4 The basic philosophy of SLUB is very different from SLAB. SLAB 5 requires rebuilding the kernel to activate debug options for all 6 slab caches. SLUB always includes full debugging but it is off by default. 7 SLUB can enable debugging only for selected slabs in order to avoid 8 an impact on overall system performance which may make a bug more 9 difficult to find. 10 11 In order to switch debugging on one can add a option "slub_debug" 12 to the kernel command line. That will enable full debugging for 13 all slabs. 14 15 Typically one would then use the "slabinfo" command to get statistical 16 data and perform operation on the slabs. By default slabinfo only lists 17 slabs that have data in them. See "slabinfo -h" for more options when 18 running the command. slabinfo can be compiled with 19 20 gcc -o slabinfo Documentation/vm/slabinfo.c 21 22 Some of the modes of operation of slabinfo require that slub debugging 23 be enabled on the command line. F.e. no tracking information will be 24 available without debugging on and validation can only partially 25 be performed if debugging was not switched on. 26 27 Some more sophisticated uses of slub_debug: 28 ------------------------------------------- 29 30 Parameters may be given to slub_debug. If none is specified then full 31 debugging is enabled. Format: 32 33 slub_debug=<Debug-Options> Enable options for all slabs 34 slub_debug=<Debug-Options>,<slab name> 35 Enable options only for select slabs 36 37 Possible debug options are 38 F Sanity checks on (enables SLAB_DEBUG_FREE. Sorry 39 SLAB legacy issues) 40 Z Red zoning 41 P Poisoning (object and padding) 42 U User tracking (free and alloc) 43 T Trace (please only use on single slabs) 44 - Switch all debugging off (useful if the kernel is 45 configured with CONFIG_SLUB_DEBUG_ON) 46 47 F.e. in order to boot just with sanity checks and red zoning one would specify: 48 49 slub_debug=FZ 50 51 Trying to find an issue in the dentry cache? Try 52 53 slub_debug=,dentry 54 55 to only enable debugging on the dentry cache. 56 57 Red zoning and tracking may realign the slab. We can just apply sanity checks 58 to the dentry cache with 59 60 slub_debug=F,dentry 61 62 In case you forgot to enable debugging on the kernel command line: It is 63 possible to enable debugging manually when the kernel is up. Look at the 64 contents of: 65 66 /sys/kernel/slab/<slab name>/ 67 68 Look at the writable files. Writing 1 to them will enable the 69 corresponding debug option. All options can be set on a slab that does 70 not contain objects. If the slab already contains objects then sanity checks 71 and tracing may only be enabled. The other options may cause the realignment 72 of objects. 73 74 Careful with tracing: It may spew out lots of information and never stop if 75 used on the wrong slab. 76 77 Slab merging 78 ------------ 79 80 If no debug options are specified then SLUB may merge similar slabs together 81 in order to reduce overhead and increase cache hotness of objects. 82 slabinfo -a displays which slabs were merged together. 83 84 Slab validation 85 --------------- 86 87 SLUB can validate all object if the kernel was booted with slub_debug. In 88 order to do so you must have the slabinfo tool. Then you can do 89 90 slabinfo -v 91 92 which will test all objects. Output will be generated to the syslog. 93 94 This also works in a more limited way if boot was without slab debug. 95 In that case slabinfo -v simply tests all reachable objects. Usually 96 these are in the cpu slabs and the partial slabs. Full slabs are not 97 tracked by SLUB in a non debug situation. 98 99 Getting more performance 100 ------------------------ 101 102 To some degree SLUB's performance is limited by the need to take the 103 list_lock once in a while to deal with partial slabs. That overhead is 104 governed by the order of the allocation for each slab. The allocations 105 can be influenced by kernel parameters: 106 107 slub_min_objects=x (default 4) 108 slub_min_order=x (default 0) 109 slub_max_order=x (default 1) 110 111 slub_min_objects allows to specify how many objects must at least fit 112 into one slab in order for the allocation order to be acceptable. 113 In general slub will be able to perform this number of allocations 114 on a slab without consulting centralized resources (list_lock) where 115 contention may occur. 116 117 slub_min_order specifies a minim order of slabs. A similar effect like 118 slub_min_objects. 119 120 slub_max_order specified the order at which slub_min_objects should no 121 longer be checked. This is useful to avoid SLUB trying to generate 122 super large order pages to fit slub_min_objects of a slab cache with 123 large object sizes into one high order page. 124 125 SLUB Debug output 126 ----------------- 127 128 Here is a sample of slub debug output: 129 130 ==================================================================== 131 BUG kmalloc-8: Redzone overwritten 132 -------------------------------------------------------------------- 133 134 INFO: 0xc90f6d28-0xc90f6d2b. First byte 0x00 instead of 0xcc 135 INFO: Slab 0xc528c530 flags=0x400000c3 inuse=61 fp=0xc90f6d58 136 INFO: Object 0xc90f6d20 @offset=3360 fp=0xc90f6d58 137 INFO: Allocated in get_modalias+0x61/0xf5 age=53 cpu=1 pid=554 138 139 Bytes b4 0xc90f6d10: 00 00 00 00 00 00 00 00 5a 5a 5a 5a 5a 5a 5a 5a ........ZZZZZZZZ 140 Object 0xc90f6d20: 31 30 31 39 2e 30 30 35 1019.005 141 Redzone 0xc90f6d28: 00 cc cc cc . 142 Padding 0xc90f6d50: 5a 5a 5a 5a 5a 5a 5a 5a ZZZZZZZZ 143 144 [<c010523d>] dump_trace+0x63/0x1eb 145 [<c01053df>] show_trace_log_lvl+0x1a/0x2f 146 [<c010601d>] show_trace+0x12/0x14 147 [<c0106035>] dump_stack+0x16/0x18 148 [<c017e0fa>] object_err+0x143/0x14b 149 [<c017e2cc>] check_object+0x66/0x234 150 [<c017eb43>] __slab_free+0x239/0x384 151 [<c017f446>] kfree+0xa6/0xc6 152 [<c02e2335>] get_modalias+0xb9/0xf5 153 [<c02e23b7>] dmi_dev_uevent+0x27/0x3c 154 [<c027866a>] dev_uevent+0x1ad/0x1da 155 [<c0205024>] kobject_uevent_env+0x20a/0x45b 156 [<c020527f>] kobject_uevent+0xa/0xf 157 [<c02779f1>] store_uevent+0x4f/0x58 158 [<c027758e>] dev_attr_store+0x29/0x2f 159 [<c01bec4f>] sysfs_write_file+0x16e/0x19c 160 [<c0183ba7>] vfs_write+0xd1/0x15a 161 [<c01841d7>] sys_write+0x3d/0x72 162 [<c0104112>] sysenter_past_esp+0x5f/0x99 163 [<b7f7b410>] 0xb7f7b410 164 ======================= 165 166 FIX kmalloc-8: Restoring Redzone 0xc90f6d28-0xc90f6d2b=0xcc 167 168 If SLUB encounters a corrupted object (full detection requires the kernel 169 to be booted with slub_debug) then the following output will be dumped 170 into the syslog: 171 172 1. Description of the problem encountered 173 174 This will be a message in the system log starting with 175 176 =============================================== 177 BUG <slab cache affected>: <What went wrong> 178 ----------------------------------------------- 179 180 INFO: <corruption start>-<corruption_end> <more info> 181 INFO: Slab <address> <slab information> 182 INFO: Object <address> <object information> 183 INFO: Allocated in <kernel function> age=<jiffies since alloc> cpu=<allocated by 184 cpu> pid=<pid of the process> 185 INFO: Freed in <kernel function> age=<jiffies since free> cpu=<freed by cpu> 186 pid=<pid of the process> 187 188 (Object allocation / free information is only available if SLAB_STORE_USER is 189 set for the slab. slub_debug sets that option) 190 191 2. The object contents if an object was involved. 192 193 Various types of lines can follow the BUG SLUB line: 194 195 Bytes b4 <address> : <bytes> 196 Shows a few bytes before the object where the problem was detected. 197 Can be useful if the corruption does not stop with the start of the 198 object. 199 200 Object <address> : <bytes> 201 The bytes of the object. If the object is inactive then the bytes 202 typically contain poison values. Any non-poison value shows a 203 corruption by a write after free. 204 205 Redzone <address> : <bytes> 206 The Redzone following the object. The Redzone is used to detect 207 writes after the object. All bytes should always have the same 208 value. If there is any deviation then it is due to a write after 209 the object boundary. 210 211 (Redzone information is only available if SLAB_RED_ZONE is set. 212 slub_debug sets that option) 213 214 Padding <address> : <bytes> 215 Unused data to fill up the space in order to get the next object 216 properly aligned. In the debug case we make sure that there are 217 at least 4 bytes of padding. This allows the detection of writes 218 before the object. 219 220 3. A stackdump 221 222 The stackdump describes the location where the error was detected. The cause 223 of the corruption is may be more likely found by looking at the function that 224 allocated or freed the object. 225 226 4. Report on how the problem was dealt with in order to ensure the continued 227 operation of the system. 228 229 These are messages in the system log beginning with 230 231 FIX <slab cache affected>: <corrective action taken> 232 233 In the above sample SLUB found that the Redzone of an active object has 234 been overwritten. Here a string of 8 characters was written into a slab that 235 has the length of 8 characters. However, a 8 character string needs a 236 terminating 0. That zero has overwritten the first byte of the Redzone field. 237 After reporting the details of the issue encountered the FIX SLUB message 238 tell us that SLUB has restored the Redzone to its proper value and then 239 system operations continue. 240 241 Emergency operations: 242 --------------------- 243 244 Minimal debugging (sanity checks alone) can be enabled by booting with 245 246 slub_debug=F 247 248 This will be generally be enough to enable the resiliency features of slub 249 which will keep the system running even if a bad kernel component will 250 keep corrupting objects. This may be important for production systems. 251 Performance will be impacted by the sanity checks and there will be a 252 continual stream of error messages to the syslog but no additional memory 253 will be used (unlike full debugging). 254 255 No guarantees. The kernel component still needs to be fixed. Performance 256 may be optimized further by locating the slab that experiences corruption 257 and enabling debugging only for that cache 258 259 I.e. 260 261 slub_debug=F,dentry 262 263 If the corruption occurs by writing after the end of the object then it 264 may be advisable to enable a Redzone to avoid corrupting the beginning 265 of other objects. 266 267 slub_debug=FZ,dentry 268 269 Christoph Lameter, <clameter[AT]sgi[DOT]com>, May 30, 2007