1========= 2Livepatch 3========= 4 5This document outlines basic information about kernel livepatching. 6 7.. Table of Contents: 8 9 1. Motivation 10 2. Kprobes, Ftrace, Livepatching 11 3. Consistency model 12 4. Livepatch module 13 4.1. New functions 14 4.2. Metadata 15 5. Livepatch life-cycle 16 5.1. Loading 17 5.2. Enabling 18 5.3. Replacing 19 5.4. Disabling 20 5.5. Removing 21 6. Sysfs 22 7. Limitations 23 24 251. Motivation 26============= 27 28There are many situations where users are reluctant to reboot a system. It may 29be because their system is performing complex scientific computations or under 30heavy load during peak usage. In addition to keeping systems up and running, 31users want to also have a stable and secure system. Livepatching gives users 32both by allowing for function calls to be redirected; thus, fixing critical 33functions without a system reboot. 34 35 362. Kprobes, Ftrace, Livepatching 37================================ 38 39There are multiple mechanisms in the Linux kernel that are directly related 40to redirection of code execution; namely: kernel probes, function tracing, 41and livepatching: 42 43 - The kernel probes are the most generic. The code can be redirected by 44 putting a breakpoint instruction instead of any instruction. 45 46 - The function tracer calls the code from a predefined location that is 47 close to the function entry point. This location is generated by the 48 compiler using the '-pg' gcc option. 49 50 - Livepatching typically needs to redirect the code at the very beginning 51 of the function entry before the function parameters or the stack 52 are in any way modified. 53 54All three approaches need to modify the existing code at runtime. Therefore 55they need to be aware of each other and not step over each other's toes. 56Most of these problems are solved by using the dynamic ftrace framework as 57a base. A Kprobe is registered as a ftrace handler when the function entry 58is probed, see CONFIG_KPROBES_ON_FTRACE. Also an alternative function from 59a live patch is called with the help of a custom ftrace handler. But there are 60some limitations, see below. 61 62 633. Consistency model 64==================== 65 66Functions are there for a reason. They take some input parameters, get or 67release locks, read, process, and even write some data in a defined way, 68have return values. In other words, each function has a defined semantic. 69 70Many fixes do not change the semantic of the modified functions. For 71example, they add a NULL pointer or a boundary check, fix a race by adding 72a missing memory barrier, or add some locking around a critical section. 73Most of these changes are self contained and the function presents itself 74the same way to the rest of the system. In this case, the functions might 75be updated independently one by one. 76 77But there are more complex fixes. For example, a patch might change 78ordering of locking in multiple functions at the same time. Or a patch 79might exchange meaning of some temporary structures and update 80all the relevant functions. In this case, the affected unit 81(thread, whole kernel) need to start using all new versions of 82the functions at the same time. Also the switch must happen only 83when it is safe to do so, e.g. when the affected locks are released 84or no data are stored in the modified structures at the moment. 85 86The theory about how to apply functions a safe way is rather complex. 87The aim is to define a so-called consistency model. It attempts to define 88conditions when the new implementation could be used so that the system 89stays consistent. 90 91Livepatch has a consistency model which is a hybrid of kGraft and 92kpatch: it uses kGraft's per-task consistency and syscall barrier 93switching combined with kpatch's stack trace switching. There are also 94a number of fallback options which make it quite flexible. 95 96Patches are applied on a per-task basis, when the task is deemed safe to 97switch over. When a patch is enabled, livepatch enters into a 98transition state where tasks are converging to the patched state. 99Usually this transition state can complete in a few seconds. The same 100sequence occurs when a patch is disabled, except the tasks converge from 101the patched state to the unpatched state. 102 103An interrupt handler inherits the patched state of the task it 104interrupts. The same is true for forked tasks: the child inherits the 105patched state of the parent. 106 107Livepatch uses several complementary approaches to determine when it's 108safe to patch tasks: 109 1101. The first and most effective approach is stack checking of sleeping 111 tasks. If no affected functions are on the stack of a given task, 112 the task is patched. In most cases this will patch most or all of 113 the tasks on the first try. Otherwise it'll keep trying 114 periodically. This option is only available if the architecture has 115 reliable stacks (HAVE_RELIABLE_STACKTRACE). 116 1172. The second approach, if needed, is kernel exit switching. A 118 task is switched when it returns to user space from a system call, a 119 user space IRQ, or a signal. It's useful in the following cases: 120 121 a) Patching I/O-bound user tasks which are sleeping on an affected 122 function. In this case you have to send SIGSTOP and SIGCONT to 123 force it to exit the kernel and be patched. 124 b) Patching CPU-bound user tasks. If the task is highly CPU-bound 125 then it will get patched the next time it gets interrupted by an 126 IRQ. 127 1283. For idle "swapper" tasks, since they don't ever exit the kernel, they 129 instead have a klp_update_patch_state() call in the idle loop which 130 allows them to be patched before the CPU enters the idle state. 131 132 (Note there's not yet such an approach for kthreads.) 133 134Architectures which don't have HAVE_RELIABLE_STACKTRACE solely rely on 135the second approach. It's highly likely that some tasks may still be 136running with an old version of the function, until that function 137returns. In this case you would have to signal the tasks. This 138especially applies to kthreads. They may not be woken up and would need 139to be forced. See below for more information. 140 141Unless we can come up with another way to patch kthreads, architectures 142without HAVE_RELIABLE_STACKTRACE are not considered fully supported by 143the kernel livepatching. 144 145The /sys/kernel/livepatch/<patch>/transition file shows whether a patch 146is in transition. Only a single patch can be in transition at a given 147time. A patch can remain in transition indefinitely, if any of the tasks 148are stuck in the initial patch state. 149 150A transition can be reversed and effectively canceled by writing the 151opposite value to the /sys/kernel/livepatch/<patch>/enabled file while 152the transition is in progress. Then all the tasks will attempt to 153converge back to the original patch state. 154 155There's also a /proc/<pid>/patch_state file which can be used to 156determine which tasks are blocking completion of a patching operation. 157If a patch is in transition, this file shows 0 to indicate the task is 158unpatched and 1 to indicate it's patched. Otherwise, if no patch is in 159transition, it shows -1. Any tasks which are blocking the transition 160can be signaled with SIGSTOP and SIGCONT to force them to change their 161patched state. This may be harmful to the system though. Sending a fake signal 162to all remaining blocking tasks is a better alternative. No proper signal is 163actually delivered (there is no data in signal pending structures). Tasks are 164interrupted or woken up, and forced to change their patched state. The fake 165signal is automatically sent every 15 seconds. 166 167Administrator can also affect a transition through 168/sys/kernel/livepatch/<patch>/force attribute. Writing 1 there clears 169TIF_PATCH_PENDING flag of all tasks and thus forces the tasks to the patched 170state. Important note! The force attribute is intended for cases when the 171transition gets stuck for a long time because of a blocking task. Administrator 172is expected to collect all necessary data (namely stack traces of such blocking 173tasks) and request a clearance from a patch distributor to force the transition. 174Unauthorized usage may cause harm to the system. It depends on the nature of the 175patch, which functions are (un)patched, and which functions the blocking tasks 176are sleeping in (/proc/<pid>/stack may help here). Removal (rmmod) of patch 177modules is permanently disabled when the force feature is used. It cannot be 178guaranteed there is no task sleeping in such module. It implies unbounded 179reference count if a patch module is disabled and enabled in a loop. 180 181Moreover, the usage of force may also affect future applications of live 182patches and cause even more harm to the system. Administrator should first 183consider to simply cancel a transition (see above). If force is used, reboot 184should be planned and no more live patches applied. 185 1863.1 Adding consistency model support to new architectures 187--------------------------------------------------------- 188 189For adding consistency model support to new architectures, there are a 190few options: 191 1921) Add CONFIG_HAVE_RELIABLE_STACKTRACE. This means porting objtool, and 193 for non-DWARF unwinders, also making sure there's a way for the stack 194 tracing code to detect interrupts on the stack. 195 1962) Alternatively, ensure that every kthread has a call to 197 klp_update_patch_state() in a safe location. Kthreads are typically 198 in an infinite loop which does some action repeatedly. The safe 199 location to switch the kthread's patch state would be at a designated 200 point in the loop where there are no locks taken and all data 201 structures are in a well-defined state. 202 203 The location is clear when using workqueues or the kthread worker 204 API. These kthreads process independent actions in a generic loop. 205 206 It's much more complicated with kthreads which have a custom loop. 207 There the safe location must be carefully selected on a case-by-case 208 basis. 209 210 In that case, arches without HAVE_RELIABLE_STACKTRACE would still be 211 able to use the non-stack-checking parts of the consistency model: 212 213 a) patching user tasks when they cross the kernel/user space 214 boundary; and 215 216 b) patching kthreads and idle tasks at their designated patch points. 217 218 This option isn't as good as option 1 because it requires signaling 219 user tasks and waking kthreads to patch them. But it could still be 220 a good backup option for those architectures which don't have 221 reliable stack traces yet. 222 223 2244. Livepatch module 225=================== 226 227Livepatches are distributed using kernel modules, see 228samples/livepatch/livepatch-sample.c. 229 230The module includes a new implementation of functions that we want 231to replace. In addition, it defines some structures describing the 232relation between the original and the new implementation. Then there 233is code that makes the kernel start using the new code when the livepatch 234module is loaded. Also there is code that cleans up before the 235livepatch module is removed. All this is explained in more details in 236the next sections. 237 238 2394.1. New functions 240------------------ 241 242New versions of functions are typically just copied from the original 243sources. A good practice is to add a prefix to the names so that they 244can be distinguished from the original ones, e.g. in a backtrace. Also 245they can be declared as static because they are not called directly 246and do not need the global visibility. 247 248The patch contains only functions that are really modified. But they 249might want to access functions or data from the original source file 250that may only be locally accessible. This can be solved by a special 251relocation section in the generated livepatch module, see 252Documentation/livepatch/module-elf-format.rst for more details. 253 254 2554.2. Metadata 256------------- 257 258The patch is described by several structures that split the information 259into three levels: 260 261 - struct klp_func is defined for each patched function. It describes 262 the relation between the original and the new implementation of a 263 particular function. 264 265 The structure includes the name, as a string, of the original function. 266 The function address is found via kallsyms at runtime. 267 268 Then it includes the address of the new function. It is defined 269 directly by assigning the function pointer. Note that the new 270 function is typically defined in the same source file. 271 272 As an optional parameter, the symbol position in the kallsyms database can 273 be used to disambiguate functions of the same name. This is not the 274 absolute position in the database, but rather the order it has been found 275 only for a particular object ( vmlinux or a kernel module ). Note that 276 kallsyms allows for searching symbols according to the object name. 277 278 - struct klp_object defines an array of patched functions (struct 279 klp_func) in the same object. Where the object is either vmlinux 280 (NULL) or a module name. 281 282 The structure helps to group and handle functions for each object 283 together. Note that patched modules might be loaded later than 284 the patch itself and the relevant functions might be patched 285 only when they are available. 286 287 288 - struct klp_patch defines an array of patched objects (struct 289 klp_object). 290 291 This structure handles all patched functions consistently and eventually, 292 synchronously. The whole patch is applied only when all patched 293 symbols are found. The only exception are symbols from objects 294 (kernel modules) that have not been loaded yet. 295 296 For more details on how the patch is applied on a per-task basis, 297 see the "Consistency model" section. 298 299 3005. Livepatch life-cycle 301======================= 302 303Livepatching can be described by five basic operations: 304loading, enabling, replacing, disabling, removing. 305 306Where the replacing and the disabling operations are mutually 307exclusive. They have the same result for the given patch but 308not for the system. 309 310 3115.1. Loading 312------------ 313 314The only reasonable way is to enable the patch when the livepatch kernel 315module is being loaded. For this, klp_enable_patch() has to be called 316in the module_init() callback. There are two main reasons: 317 318First, only the module has an easy access to the related struct klp_patch. 319 320Second, the error code might be used to refuse loading the module when 321the patch cannot get enabled. 322 323 3245.2. Enabling 325------------- 326 327The livepatch gets enabled by calling klp_enable_patch() from 328the module_init() callback. The system will start using the new 329implementation of the patched functions at this stage. 330 331First, the addresses of the patched functions are found according to their 332names. The special relocations, mentioned in the section "New functions", 333are applied. The relevant entries are created under 334/sys/kernel/livepatch/<name>. The patch is rejected when any above 335operation fails. 336 337Second, livepatch enters into a transition state where tasks are converging 338to the patched state. If an original function is patched for the first 339time, a function specific struct klp_ops is created and an universal 340ftrace handler is registered\ [#]_. This stage is indicated by a value of '1' 341in /sys/kernel/livepatch/<name>/transition. For more information about 342this process, see the "Consistency model" section. 343 344Finally, once all tasks have been patched, the 'transition' value changes 345to '0'. 346 347.. [#] 348 349 Note that functions might be patched multiple times. The ftrace handler 350 is registered only once for a given function. Further patches just add 351 an entry to the list (see field `func_stack`) of the struct klp_ops. 352 The right implementation is selected by the ftrace handler, see 353 the "Consistency model" section. 354 355 That said, it is highly recommended to use cumulative livepatches 356 because they help keeping the consistency of all changes. In this case, 357 functions might be patched two times only during the transition period. 358 359 3605.3. Replacing 361-------------- 362 363All enabled patches might get replaced by a cumulative patch that 364has the .replace flag set. 365 366Once the new patch is enabled and the 'transition' finishes then 367all the functions (struct klp_func) associated with the replaced 368patches are removed from the corresponding struct klp_ops. Also 369the ftrace handler is unregistered and the struct klp_ops is 370freed when the related function is not modified by the new patch 371and func_stack list becomes empty. 372 373See Documentation/livepatch/cumulative-patches.rst for more details. 374 375 3765.4. Disabling 377-------------- 378 379Enabled patches might get disabled by writing '0' to 380/sys/kernel/livepatch/<name>/enabled. 381 382First, livepatch enters into a transition state where tasks are converging 383to the unpatched state. The system starts using either the code from 384the previously enabled patch or even the original one. This stage is 385indicated by a value of '1' in /sys/kernel/livepatch/<name>/transition. 386For more information about this process, see the "Consistency model" 387section. 388 389Second, once all tasks have been unpatched, the 'transition' value changes 390to '0'. All the functions (struct klp_func) associated with the to-be-disabled 391patch are removed from the corresponding struct klp_ops. The ftrace handler 392is unregistered and the struct klp_ops is freed when the func_stack list 393becomes empty. 394 395Third, the sysfs interface is destroyed. 396 397 3985.5. Removing 399------------- 400 401Module removal is only safe when there are no users of functions provided 402by the module. This is the reason why the force feature permanently 403disables the removal. Only when the system is successfully transitioned 404to a new patch state (patched/unpatched) without being forced it is 405guaranteed that no task sleeps or runs in the old code. 406 407 4086. Sysfs 409======== 410 411Information about the registered patches can be found under 412/sys/kernel/livepatch. The patches could be enabled and disabled 413by writing there. 414 415/sys/kernel/livepatch/<patch>/force attributes allow administrator to affect a 416patching operation. 417 418See Documentation/ABI/testing/sysfs-kernel-livepatch for more details. 419 420 4217. Limitations 422============== 423 424The current Livepatch implementation has several limitations: 425 426 - Only functions that can be traced could be patched. 427 428 Livepatch is based on the dynamic ftrace. In particular, functions 429 implementing ftrace or the livepatch ftrace handler could not be 430 patched. Otherwise, the code would end up in an infinite loop. A 431 potential mistake is prevented by marking the problematic functions 432 by "notrace". 433 434 435 436 - Livepatch works reliably only when the dynamic ftrace is located at 437 the very beginning of the function. 438 439 The function need to be redirected before the stack or the function 440 parameters are modified in any way. For example, livepatch requires 441 using -fentry gcc compiler option on x86_64. 442 443 One exception is the PPC port. It uses relative addressing and TOC. 444 Each function has to handle TOC and save LR before it could call 445 the ftrace handler. This operation has to be reverted on return. 446 Fortunately, the generic ftrace code has the same problem and all 447 this is handled on the ftrace level. 448 449 450 - Kretprobes using the ftrace framework conflict with the patched 451 functions. 452 453 Both kretprobes and livepatches use a ftrace handler that modifies 454 the return address. The first user wins. Either the probe or the patch 455 is rejected when the handler is already in use by the other. 456 457 458 - Kprobes in the original function are ignored when the code is 459 redirected to the new implementation. 460 461 There is a work in progress to add warnings about this situation. 462