1 // Copyright 2020, The Android Open Source Project 2 // 3 // Licensed under the Apache License, Version 2.0 (the "License"); 4 // you may not use this file except in compliance with the License. 5 // You may obtain a copy of the License at 6 // 7 // http://www.apache.org/licenses/LICENSE-2.0 8 // 9 // Unless required by applicable law or agreed to in writing, software 10 // distributed under the License is distributed on an "AS IS" BASIS, 11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 // See the License for the specific language governing permissions and 13 // limitations under the License. 14 15 //! This crate implements the `IKeystoreOperation` AIDL interface, which represents 16 //! an ongoing key operation, as well as the operation database, which is mainly 17 //! required for tracking operations for the purpose of pruning. 18 //! This crate also implements an operation pruning strategy. 19 //! 20 //! Operations implement the API calls update, finish, and abort. 21 //! Additionally, an operation can be dropped and pruned. The former 22 //! happens if the client deletes a binder to the operation object. 23 //! An existing operation may get pruned when running out of operation 24 //! slots and a new operation takes precedence. 25 //! 26 //! ## Operation Lifecycle 27 //! An operation gets created when the client calls `IKeystoreSecurityLevel::create`. 28 //! It may receive zero or more update request. The lifecycle ends when: 29 //! * `update` yields an error. 30 //! * `finish` is called. 31 //! * `abort` is called. 32 //! * The operation gets dropped. 33 //! * The operation gets pruned. 34 //! `Operation` has an `Outcome` member. While the outcome is `Outcome::Unknown`, 35 //! the operation is active and in a good state. Any of the above conditions may 36 //! change the outcome to one of the defined outcomes Success, Abort, Dropped, 37 //! Pruned, or ErrorCode. The latter is chosen in the case of an unexpected error, during 38 //! `update` or `finish`. `Success` is chosen iff `finish` completes without error. 39 //! Note that all operations get dropped eventually in the sense that they lose 40 //! their last reference and get destroyed. At that point, the fate of the operation 41 //! gets logged. However, an operation will transition to `Outcome::Dropped` iff 42 //! the operation was still active (`Outcome::Unknown`) at that time. 43 //! 44 //! ## Operation Dropping 45 //! To observe the dropping of an operation, we have to make sure that there 46 //! are no strong references to the IBinder representing this operation. 47 //! This would be simple enough if the operation object would need to be accessed 48 //! only by transactions. But to perform pruning, we have to retain a reference to the 49 //! original operation object. 50 //! 51 //! ## Operation Pruning 52 //! Pruning an operation happens during the creation of a new operation. 53 //! We have to iterate through the operation database to find a suitable 54 //! candidate. Then we abort and finalize this operation setting its outcome to 55 //! `Outcome::Pruned`. The corresponding KeyMint operation slot will have been freed 56 //! up at this point, but the `Operation` object lingers. When the client 57 //! attempts to use the operation again they will receive 58 //! ErrorCode::INVALID_OPERATION_HANDLE indicating that the operation no longer 59 //! exits. This should be the cue for the client to destroy its binder. 60 //! At that point the operation gets dropped. 61 //! 62 //! ## Architecture 63 //! The `IKeystoreOperation` trait is implemented by `KeystoreOperation`. 64 //! This acts as a proxy object holding a strong reference to actual operation 65 //! implementation `Operation`. 66 //! 67 //! ``` 68 //! struct KeystoreOperation { 69 //! operation: Mutex<Option<Arc<Operation>>>, 70 //! } 71 //! ``` 72 //! 73 //! The `Mutex` serves two purposes. It provides interior mutability allowing 74 //! us to set the Option to None. We do this when the life cycle ends during 75 //! a call to `update`, `finish`, or `abort`. As a result most of the Operation 76 //! related resources are freed. The `KeystoreOperation` proxy object still 77 //! lingers until dropped by the client. 78 //! The second purpose is to protect operations against concurrent usage. 79 //! Failing to lock this mutex yields `ResponseCode::OPERATION_BUSY` and indicates 80 //! a programming error in the client. 81 //! 82 //! Note that the Mutex only protects the operation against concurrent client calls. 83 //! We still retain weak references to the operation in the operation database: 84 //! 85 //! ``` 86 //! struct OperationDb { 87 //! operations: Mutex<Vec<Weak<Operation>>> 88 //! } 89 //! ``` 90 //! 91 //! This allows us to access the operations for the purpose of pruning. 92 //! We do this in three phases. 93 //! 1. We gather the pruning information. Besides non mutable information, 94 //! we access `last_usage` which is protected by a mutex. 95 //! We only lock this mutex for single statements at a time. During 96 //! this phase we hold the operation db lock. 97 //! 2. We choose a pruning candidate by computing the pruning resistance 98 //! of each operation. We do this entirely with information we now 99 //! have on the stack without holding any locks. 100 //! (See `OperationDb::prune` for more details on the pruning strategy.) 101 //! 3. During pruning we briefly lock the operation database again to get the 102 //! the pruning candidate by index. We then attempt to abort the candidate. 103 //! If the candidate was touched in the meantime or is currently fulfilling 104 //! a request (i.e., the client calls update, finish, or abort), 105 //! we go back to 1 and try again. 106 //! 107 //! So the outer Mutex in `KeystoreOperation::operation` only protects 108 //! operations against concurrent client calls but not against concurrent 109 //! pruning attempts. This is what the `Operation::outcome` mutex is used for. 110 //! 111 //! ``` 112 //! struct Operation { 113 //! ... 114 //! outcome: Mutex<Outcome>, 115 //! ... 116 //! } 117 //! ``` 118 //! 119 //! Any request that can change the outcome, i.e., `update`, `finish`, `abort`, 120 //! `drop`, and `prune` has to take the outcome lock and check if the outcome 121 //! is still `Outcome::Unknown` before entering. `prune` is special in that 122 //! it will `try_lock`, because we don't want to be blocked on a potentially 123 //! long running request at another operation. If it fails to get the lock 124 //! the operation is either being touched, which changes its pruning resistance, 125 //! or it transitions to its end-of-life, which means we may get a free slot. 126 //! Either way, we have to revaluate the pruning scores. 127 128 use crate::enforcements::AuthInfo; 129 use crate::error::{map_err_with, map_km_error, map_or_log_err, Error, ErrorCode, ResponseCode}; 130 use crate::ks_err; 131 use crate::metrics_store::log_key_operation_event_stats; 132 use crate::utils::watchdog as wd; 133 use android_hardware_security_keymint::aidl::android::hardware::security::keymint::{ 134 IKeyMintOperation::IKeyMintOperation, KeyParameter::KeyParameter, KeyPurpose::KeyPurpose, 135 SecurityLevel::SecurityLevel, 136 }; 137 use android_hardware_security_keymint::binder::{BinderFeatures, Strong}; 138 use android_system_keystore2::aidl::android::system::keystore2::{ 139 IKeystoreOperation::BnKeystoreOperation, IKeystoreOperation::IKeystoreOperation, 140 }; 141 use anyhow::{anyhow, Context, Result}; 142 use std::{ 143 collections::HashMap, 144 sync::{Arc, Mutex, MutexGuard, Weak}, 145 time::Duration, 146 time::Instant, 147 }; 148 149 /// Operations have `Outcome::Unknown` as long as they are active. They transition 150 /// to one of the other variants exactly once. The distinction in outcome is mainly 151 /// for the statistic. 152 #[derive(Debug, Copy, Clone, Eq, PartialEq, Ord, PartialOrd)] 153 pub enum Outcome { 154 /// Operations have `Outcome::Unknown` as long as they are active. 155 Unknown, 156 /// Operation is successful. 157 Success, 158 /// Operation is aborted. 159 Abort, 160 /// Operation is dropped. 161 Dropped, 162 /// Operation is pruned. 163 Pruned, 164 /// Operation is failed with the error code. 165 ErrorCode(ErrorCode), 166 } 167 168 /// Operation bundles all of the operation related resources and tracks the operation's 169 /// outcome. 170 #[derive(Debug)] 171 pub struct Operation { 172 // The index of this operation in the OperationDb. 173 index: usize, 174 km_op: Strong<dyn IKeyMintOperation>, 175 last_usage: Mutex<Instant>, 176 outcome: Mutex<Outcome>, 177 owner: u32, // Uid of the operation's owner. 178 auth_info: Mutex<AuthInfo>, 179 forced: bool, 180 logging_info: LoggingInfo, 181 } 182 183 /// Keeps track of the information required for logging operations. 184 #[derive(Debug)] 185 pub struct LoggingInfo { 186 sec_level: SecurityLevel, 187 purpose: KeyPurpose, 188 op_params: Vec<KeyParameter>, 189 key_upgraded: bool, 190 } 191 192 impl LoggingInfo { 193 /// Constructor new( sec_level: SecurityLevel, purpose: KeyPurpose, op_params: Vec<KeyParameter>, key_upgraded: bool, ) -> LoggingInfo194 pub fn new( 195 sec_level: SecurityLevel, 196 purpose: KeyPurpose, 197 op_params: Vec<KeyParameter>, 198 key_upgraded: bool, 199 ) -> LoggingInfo { 200 Self { sec_level, purpose, op_params, key_upgraded } 201 } 202 } 203 204 struct PruningInfo { 205 last_usage: Instant, 206 owner: u32, 207 index: usize, 208 forced: bool, 209 } 210 211 // We don't except more than 32KiB of data in `update`, `updateAad`, and `finish`. 212 const MAX_RECEIVE_DATA: usize = 0x8000; 213 214 impl Operation { 215 /// Constructor new( index: usize, km_op: binder::Strong<dyn IKeyMintOperation>, owner: u32, auth_info: AuthInfo, forced: bool, logging_info: LoggingInfo, ) -> Self216 pub fn new( 217 index: usize, 218 km_op: binder::Strong<dyn IKeyMintOperation>, 219 owner: u32, 220 auth_info: AuthInfo, 221 forced: bool, 222 logging_info: LoggingInfo, 223 ) -> Self { 224 Self { 225 index, 226 km_op, 227 last_usage: Mutex::new(Instant::now()), 228 outcome: Mutex::new(Outcome::Unknown), 229 owner, 230 auth_info: Mutex::new(auth_info), 231 forced, 232 logging_info, 233 } 234 } 235 get_pruning_info(&self) -> Option<PruningInfo>236 fn get_pruning_info(&self) -> Option<PruningInfo> { 237 // An operation may be finalized. 238 if let Ok(guard) = self.outcome.try_lock() { 239 match *guard { 240 Outcome::Unknown => {} 241 // If the outcome is any other than unknown, it has been finalized, 242 // and we can no longer consider it for pruning. 243 _ => return None, 244 } 245 } 246 // Else: If we could not grab the lock, this means that the operation is currently 247 // being used and it may be transitioning to finalized or it was simply updated. 248 // In any case it is fair game to consider it for pruning. If the operation 249 // transitioned to a final state, we will notice when we attempt to prune, and 250 // a subsequent attempt to create a new operation will succeed. 251 Some(PruningInfo { 252 // Expect safety: 253 // `last_usage` is locked only for primitive single line statements. 254 // There is no chance to panic and poison the mutex. 255 last_usage: *self.last_usage.lock().expect("In get_pruning_info."), 256 owner: self.owner, 257 index: self.index, 258 forced: self.forced, 259 }) 260 } 261 prune(&self, last_usage: Instant) -> Result<(), Error>262 fn prune(&self, last_usage: Instant) -> Result<(), Error> { 263 let mut locked_outcome = match self.outcome.try_lock() { 264 Ok(guard) => match *guard { 265 Outcome::Unknown => guard, 266 _ => return Err(Error::Km(ErrorCode::INVALID_OPERATION_HANDLE)), 267 }, 268 Err(_) => return Err(Error::Rc(ResponseCode::OPERATION_BUSY)), 269 }; 270 271 // In `OperationDb::prune`, which is our caller, we first gather the pruning 272 // information including the last usage. When we select a candidate 273 // we call `prune` on that candidate passing the last_usage 274 // that we gathered earlier. If the actual last usage 275 // has changed since than, it means the operation was busy in the 276 // meantime, which means that we have to reevaluate the pruning score. 277 // 278 // Expect safety: 279 // `last_usage` is locked only for primitive single line statements. 280 // There is no chance to panic and poison the mutex. 281 if *self.last_usage.lock().expect("In Operation::prune()") != last_usage { 282 return Err(Error::Rc(ResponseCode::OPERATION_BUSY)); 283 } 284 *locked_outcome = Outcome::Pruned; 285 286 let _wp = wd::watch_millis("In Operation::prune: calling abort()", 500); 287 288 // We abort the operation. If there was an error we log it but ignore it. 289 if let Err(e) = map_km_error(self.km_op.abort()) { 290 log::error!("In prune: KeyMint::abort failed with {:?}.", e); 291 } 292 293 Ok(()) 294 } 295 296 // This function takes a Result from a KeyMint call and inspects it for errors. 297 // If an error was found it updates the given `locked_outcome` accordingly. 298 // It forwards the Result unmodified. 299 // The precondition to this call must be *locked_outcome == Outcome::Unknown. 300 // Ideally the `locked_outcome` came from a successful call to `check_active` 301 // see below. update_outcome<T>( &self, locked_outcome: &mut Outcome, err: Result<T, Error>, ) -> Result<T, Error>302 fn update_outcome<T>( 303 &self, 304 locked_outcome: &mut Outcome, 305 err: Result<T, Error>, 306 ) -> Result<T, Error> { 307 match &err { 308 Err(Error::Km(e)) => *locked_outcome = Outcome::ErrorCode(*e), 309 Err(_) => *locked_outcome = Outcome::ErrorCode(ErrorCode::UNKNOWN_ERROR), 310 Ok(_) => (), 311 } 312 err 313 } 314 315 // This function grabs the outcome lock and checks the current outcome state. 316 // If the outcome is still `Outcome::Unknown`, this function returns 317 // the locked outcome for further updates. In any other case it returns 318 // ErrorCode::INVALID_OPERATION_HANDLE indicating that this operation has 319 // been finalized and is no longer active. check_active(&self) -> Result<MutexGuard<Outcome>>320 fn check_active(&self) -> Result<MutexGuard<Outcome>> { 321 let guard = self.outcome.lock().expect("In check_active."); 322 match *guard { 323 Outcome::Unknown => Ok(guard), 324 _ => Err(Error::Km(ErrorCode::INVALID_OPERATION_HANDLE)) 325 .context(ks_err!("Call on finalized operation with outcome: {:?}.", *guard)), 326 } 327 } 328 329 // This function checks the amount of input data sent to us. We reject any buffer 330 // exceeding MAX_RECEIVE_DATA bytes as input to `update`, `update_aad`, and `finish` 331 // in order to force clients into using reasonable limits. check_input_length(data: &[u8]) -> Result<()>332 fn check_input_length(data: &[u8]) -> Result<()> { 333 if data.len() > MAX_RECEIVE_DATA { 334 // This error code is unique, no context required here. 335 return Err(anyhow!(Error::Rc(ResponseCode::TOO_MUCH_DATA))); 336 } 337 Ok(()) 338 } 339 340 // Update the last usage to now. touch(&self)341 fn touch(&self) { 342 // Expect safety: 343 // `last_usage` is locked only for primitive single line statements. 344 // There is no chance to panic and poison the mutex. 345 *self.last_usage.lock().expect("In touch.") = Instant::now(); 346 } 347 348 /// Implementation of `IKeystoreOperation::updateAad`. 349 /// Refer to the AIDL spec at system/hardware/interfaces/keystore2 for details. update_aad(&self, aad_input: &[u8]) -> Result<()>350 fn update_aad(&self, aad_input: &[u8]) -> Result<()> { 351 let mut outcome = self.check_active().context("In update_aad")?; 352 Self::check_input_length(aad_input).context("In update_aad")?; 353 self.touch(); 354 355 let (hat, tst) = self 356 .auth_info 357 .lock() 358 .unwrap() 359 .before_update() 360 .context(ks_err!("Trying to get auth tokens."))?; 361 362 self.update_outcome(&mut outcome, { 363 let _wp = wd::watch_millis("Operation::update_aad: calling updateAad", 500); 364 map_km_error(self.km_op.updateAad(aad_input, hat.as_ref(), tst.as_ref())) 365 }) 366 .context(ks_err!("Update failed."))?; 367 368 Ok(()) 369 } 370 371 /// Implementation of `IKeystoreOperation::update`. 372 /// Refer to the AIDL spec at system/hardware/interfaces/keystore2 for details. update(&self, input: &[u8]) -> Result<Option<Vec<u8>>>373 fn update(&self, input: &[u8]) -> Result<Option<Vec<u8>>> { 374 let mut outcome = self.check_active().context("In update")?; 375 Self::check_input_length(input).context("In update")?; 376 self.touch(); 377 378 let (hat, tst) = self 379 .auth_info 380 .lock() 381 .unwrap() 382 .before_update() 383 .context(ks_err!("Trying to get auth tokens."))?; 384 385 let output = self 386 .update_outcome(&mut outcome, { 387 let _wp = wd::watch_millis("Operation::update: calling update", 500); 388 map_km_error(self.km_op.update(input, hat.as_ref(), tst.as_ref())) 389 }) 390 .context(ks_err!("Update failed."))?; 391 392 if output.is_empty() { 393 Ok(None) 394 } else { 395 Ok(Some(output)) 396 } 397 } 398 399 /// Implementation of `IKeystoreOperation::finish`. 400 /// Refer to the AIDL spec at system/hardware/interfaces/keystore2 for details. finish(&self, input: Option<&[u8]>, signature: Option<&[u8]>) -> Result<Option<Vec<u8>>>401 fn finish(&self, input: Option<&[u8]>, signature: Option<&[u8]>) -> Result<Option<Vec<u8>>> { 402 let mut outcome = self.check_active().context("In finish")?; 403 if let Some(input) = input { 404 Self::check_input_length(input).context("In finish")?; 405 } 406 self.touch(); 407 408 let (hat, tst, confirmation_token) = self 409 .auth_info 410 .lock() 411 .unwrap() 412 .before_finish() 413 .context(ks_err!("Trying to get auth tokens."))?; 414 415 let output = self 416 .update_outcome(&mut outcome, { 417 let _wp = wd::watch_millis("Operation::finish: calling finish", 500); 418 map_km_error(self.km_op.finish( 419 input, 420 signature, 421 hat.as_ref(), 422 tst.as_ref(), 423 confirmation_token.as_deref(), 424 )) 425 }) 426 .context(ks_err!("Finish failed."))?; 427 428 self.auth_info.lock().unwrap().after_finish().context("In finish.")?; 429 430 // At this point the operation concluded successfully. 431 *outcome = Outcome::Success; 432 433 if output.is_empty() { 434 Ok(None) 435 } else { 436 Ok(Some(output)) 437 } 438 } 439 440 /// Aborts the operation if it is active. IFF the operation is aborted the outcome is 441 /// set to `outcome`. `outcome` must reflect the reason for the abort. Since the operation 442 /// gets aborted `outcome` must not be `Operation::Success` or `Operation::Unknown`. abort(&self, outcome: Outcome) -> Result<()>443 fn abort(&self, outcome: Outcome) -> Result<()> { 444 let mut locked_outcome = self.check_active().context("In abort")?; 445 *locked_outcome = outcome; 446 447 { 448 let _wp = wd::watch_millis("Operation::abort: calling abort", 500); 449 map_km_error(self.km_op.abort()).context(ks_err!("KeyMint::abort failed.")) 450 } 451 } 452 } 453 454 impl Drop for Operation { drop(&mut self)455 fn drop(&mut self) { 456 let guard = self.outcome.lock().expect("In drop."); 457 log_key_operation_event_stats( 458 self.logging_info.sec_level, 459 self.logging_info.purpose, 460 &(self.logging_info.op_params), 461 &guard, 462 self.logging_info.key_upgraded, 463 ); 464 if let Outcome::Unknown = *guard { 465 drop(guard); 466 // If the operation was still active we call abort, setting 467 // the outcome to `Outcome::Dropped` 468 if let Err(e) = self.abort(Outcome::Dropped) { 469 log::error!("While dropping Operation: abort failed:\n {:?}", e); 470 } 471 } 472 } 473 } 474 475 /// The OperationDb holds weak references to all ongoing operations. 476 /// Its main purpose is to facilitate operation pruning. 477 #[derive(Debug, Default)] 478 pub struct OperationDb { 479 // TODO replace Vec with WeakTable when the weak_table crate becomes 480 // available. 481 operations: Mutex<Vec<Weak<Operation>>>, 482 } 483 484 impl OperationDb { 485 /// Creates a new OperationDb. new() -> Self486 pub fn new() -> Self { 487 Self { operations: Mutex::new(Vec::new()) } 488 } 489 490 /// Creates a new operation. 491 /// This function takes a KeyMint operation and an associated 492 /// owner uid and returns a new Operation wrapped in a `std::sync::Arc`. create_operation( &self, km_op: binder::Strong<dyn IKeyMintOperation>, owner: u32, auth_info: AuthInfo, forced: bool, logging_info: LoggingInfo, ) -> Arc<Operation>493 pub fn create_operation( 494 &self, 495 km_op: binder::Strong<dyn IKeyMintOperation>, 496 owner: u32, 497 auth_info: AuthInfo, 498 forced: bool, 499 logging_info: LoggingInfo, 500 ) -> Arc<Operation> { 501 // We use unwrap because we don't allow code that can panic while locked. 502 let mut operations = self.operations.lock().expect("In create_operation."); 503 504 let mut index: usize = 0; 505 // First we iterate through the operation slots to try and find an unused 506 // slot. If we don't find one, we append the new entry instead. 507 match (*operations).iter_mut().find(|s| { 508 index += 1; 509 s.upgrade().is_none() 510 }) { 511 Some(free_slot) => { 512 let new_op = Arc::new(Operation::new( 513 index - 1, 514 km_op, 515 owner, 516 auth_info, 517 forced, 518 logging_info, 519 )); 520 *free_slot = Arc::downgrade(&new_op); 521 new_op 522 } 523 None => { 524 let new_op = Arc::new(Operation::new( 525 operations.len(), 526 km_op, 527 owner, 528 auth_info, 529 forced, 530 logging_info, 531 )); 532 operations.push(Arc::downgrade(&new_op)); 533 new_op 534 } 535 } 536 } 537 get(&self, index: usize) -> Option<Arc<Operation>>538 fn get(&self, index: usize) -> Option<Arc<Operation>> { 539 self.operations.lock().expect("In OperationDb::get.").get(index).and_then(|op| op.upgrade()) 540 } 541 542 /// Attempts to prune an operation. 543 /// 544 /// This function is used during operation creation, i.e., by 545 /// `KeystoreSecurityLevel::create_operation`, to try and free up an operation slot 546 /// if it got `ErrorCode::TOO_MANY_OPERATIONS` from the KeyMint backend. It is not 547 /// guaranteed that an operation slot is available after this call successfully 548 /// returned for various reasons. E.g., another thread may have snatched up the newly 549 /// available slot. Callers may have to call prune multiple times before they get a 550 /// free operation slot. Prune may also return `Err(Error::Rc(ResponseCode::BACKEND_BUSY))` 551 /// which indicates that no prunable operation was found. 552 /// 553 /// To find a suitable candidate we compute the malus for the caller and each existing 554 /// operation. The malus is the inverse of the pruning power (caller) or pruning 555 /// resistance (existing operation). 556 /// 557 /// The malus is based on the number of sibling operations and age. Sibling 558 /// operations are operations that have the same owner (UID). 559 /// 560 /// Every operation, existing or new, starts with a malus of 1. Every sibling 561 /// increases the malus by one. The age is the time since an operation was last touched. 562 /// It increases the malus by log6(<age in seconds> + 1) rounded down to the next 563 /// integer. So the malus increases stepwise after 5s, 35s, 215s, ... 564 /// Of two operations with the same malus the least recently used one is considered 565 /// weaker. 566 /// 567 /// For the caller to be able to prune an operation it must find an operation 568 /// with a malus higher than its own. 569 /// 570 /// The malus can be expressed as 571 /// ``` 572 /// malus = 1 + no_of_siblings + floor(log6(age_in_seconds + 1)) 573 /// ``` 574 /// where the constant `1` accounts for the operation under consideration. 575 /// In reality we compute it as 576 /// ``` 577 /// caller_malus = 1 + running_siblings 578 /// ``` 579 /// because the new operation has no age and is not included in the `running_siblings`, 580 /// and 581 /// ``` 582 /// running_malus = running_siblings + floor(log6(age_in_seconds + 1)) 583 /// ``` 584 /// because a running operation is included in the `running_siblings` and it has 585 /// an age. 586 /// 587 /// ## Example 588 /// A caller with no running operations has a malus of 1. Young (age < 5s) operations 589 /// also with no siblings have a malus of one and cannot be pruned by the caller. 590 /// We have to find an operation that has at least one sibling or is older than 5s. 591 /// 592 /// A caller with one running operation has a malus of 2. Now even young siblings 593 /// or single child aging (5s <= age < 35s) operations are off limit. An aging 594 /// sibling of two, however, would have a malus of 3 and would be fair game. 595 /// 596 /// ## Rationale 597 /// Due to the limitation of KeyMint operation slots, we cannot get around pruning or 598 /// a single app could easily DoS KeyMint. 599 /// Keystore 1.0 used to always prune the least recently used operation. This at least 600 /// guaranteed that new operations can always be started. With the increased usage 601 /// of Keystore we saw increased pruning activity which can lead to a livelock 602 /// situation in the worst case. 603 /// 604 /// With the new pruning strategy we want to provide well behaved clients with 605 /// progress assurances while punishing DoS attempts. As a result of this 606 /// strategy we can be in the situation where no operation can be pruned and the 607 /// creation of a new operation fails. This allows single child operations which 608 /// are frequently updated to complete, thereby breaking up livelock situations 609 /// and facilitating system wide progress. 610 /// 611 /// ## Update 612 /// We also allow callers to cannibalize their own sibling operations if no other 613 /// slot can be found. In this case the least recently used sibling is pruned. prune(&self, caller: u32, forced: bool) -> Result<(), Error>614 pub fn prune(&self, caller: u32, forced: bool) -> Result<(), Error> { 615 loop { 616 // Maps the uid of the owner to the number of operations that owner has 617 // (running_siblings). More operations per owner lowers the pruning 618 // resistance of the operations of that owner. Whereas the number of 619 // ongoing operations of the caller lowers the pruning power of the caller. 620 let mut owners: HashMap<u32, u64> = HashMap::new(); 621 let mut pruning_info: Vec<PruningInfo> = Vec::new(); 622 623 let now = Instant::now(); 624 self.operations 625 .lock() 626 .expect("In OperationDb::prune: Trying to lock self.operations.") 627 .iter() 628 .for_each(|op| { 629 if let Some(op) = op.upgrade() { 630 if let Some(p_info) = op.get_pruning_info() { 631 let owner = p_info.owner; 632 pruning_info.push(p_info); 633 // Count operations per owner. 634 *owners.entry(owner).or_insert(0) += 1; 635 } 636 } 637 }); 638 639 // If the operation is forced, the caller has a malus of 0. 640 let caller_malus = if forced { 0 } else { 1u64 + *owners.entry(caller).or_default() }; 641 642 // We iterate through all operations computing the malus and finding 643 // the candidate with the highest malus which must also be higher 644 // than the caller_malus. 645 struct CandidateInfo { 646 index: usize, 647 malus: u64, 648 last_usage: Instant, 649 age: Duration, 650 } 651 let mut oldest_caller_op: Option<CandidateInfo> = None; 652 let candidate = pruning_info.iter().fold( 653 None, 654 |acc: Option<CandidateInfo>, &PruningInfo { last_usage, owner, index, forced }| { 655 // Compute the age of the current operation. 656 let age = now 657 .checked_duration_since(last_usage) 658 .unwrap_or_else(|| Duration::new(0, 0)); 659 660 // Find the least recently used sibling as an alternative pruning candidate. 661 if owner == caller { 662 if let Some(CandidateInfo { age: a, .. }) = oldest_caller_op { 663 if age > a { 664 oldest_caller_op = 665 Some(CandidateInfo { index, malus: 0, last_usage, age }); 666 } 667 } else { 668 oldest_caller_op = 669 Some(CandidateInfo { index, malus: 0, last_usage, age }); 670 } 671 } 672 673 // Compute the malus of the current operation. 674 let malus = if forced { 675 // Forced operations have a malus of 0. And cannot even be pruned 676 // by other forced operations. 677 0 678 } else { 679 // Expect safety: Every owner in pruning_info was counted in 680 // the owners map. So this unwrap cannot panic. 681 *owners.get(&owner).expect( 682 "This is odd. We should have counted every owner in pruning_info.", 683 ) + ((age.as_secs() + 1) as f64).log(6.0).floor() as u64 684 }; 685 686 // Now check if the current operation is a viable/better candidate 687 // the one currently stored in the accumulator. 688 match acc { 689 // First we have to find any operation that is prunable by the caller. 690 None => { 691 if caller_malus < malus { 692 Some(CandidateInfo { index, malus, last_usage, age }) 693 } else { 694 None 695 } 696 } 697 // If we have found one we look for the operation with the worst score. 698 // If there is a tie, the older operation is considered weaker. 699 Some(CandidateInfo { index: i, malus: m, last_usage: l, age: a }) => { 700 if malus > m || (malus == m && age > a) { 701 Some(CandidateInfo { index, malus, last_usage, age }) 702 } else { 703 Some(CandidateInfo { index: i, malus: m, last_usage: l, age: a }) 704 } 705 } 706 } 707 }, 708 ); 709 710 // If we did not find a suitable candidate we may cannibalize our oldest sibling. 711 let candidate = candidate.or(oldest_caller_op); 712 713 match candidate { 714 Some(CandidateInfo { index, malus: _, last_usage, age: _ }) => { 715 match self.get(index) { 716 Some(op) => { 717 match op.prune(last_usage) { 718 // We successfully freed up a slot. 719 Ok(()) => break Ok(()), 720 // This means the operation we tried to prune was on its way 721 // out. It also means that the slot it had occupied was freed up. 722 Err(Error::Km(ErrorCode::INVALID_OPERATION_HANDLE)) => break Ok(()), 723 // This means the operation we tried to prune was currently 724 // servicing a request. There are two options. 725 // * Assume that it was touched, which means that its 726 // pruning resistance increased. In that case we have 727 // to start over and find another candidate. 728 // * Assume that the operation is transitioning to end-of-life. 729 // which means that we got a free slot for free. 730 // If we assume the first but the second is true, we prune 731 // a good operation without need (aggressive approach). 732 // If we assume the second but the first is true, our 733 // caller will attempt to create a new KeyMint operation, 734 // fail with `ErrorCode::TOO_MANY_OPERATIONS`, and call 735 // us again (conservative approach). 736 Err(Error::Rc(ResponseCode::OPERATION_BUSY)) => { 737 // We choose the conservative approach, because 738 // every needlessly pruned operation can impact 739 // the user experience. 740 // To switch to the aggressive approach replace 741 // the following line with `continue`. 742 break Ok(()); 743 } 744 745 // The candidate may have been touched so the score 746 // has changed since our evaluation. 747 _ => continue, 748 } 749 } 750 // This index does not exist any more. The operation 751 // in this slot was dropped. Good news, a slot 752 // has freed up. 753 None => break Ok(()), 754 } 755 } 756 // We did not get a pruning candidate. 757 None => break Err(Error::Rc(ResponseCode::BACKEND_BUSY)), 758 } 759 } 760 } 761 } 762 763 /// Implementation of IKeystoreOperation. 764 pub struct KeystoreOperation { 765 operation: Mutex<Option<Arc<Operation>>>, 766 } 767 768 impl KeystoreOperation { 769 /// Creates a new operation instance wrapped in a 770 /// BnKeystoreOperation proxy object. It also enables 771 /// `BinderFeatures::set_requesting_sid` on the new interface, because 772 /// we need it for checking Keystore permissions. new_native_binder(operation: Arc<Operation>) -> binder::Strong<dyn IKeystoreOperation>773 pub fn new_native_binder(operation: Arc<Operation>) -> binder::Strong<dyn IKeystoreOperation> { 774 BnKeystoreOperation::new_binder( 775 Self { operation: Mutex::new(Some(operation)) }, 776 BinderFeatures { set_requesting_sid: true, ..BinderFeatures::default() }, 777 ) 778 } 779 780 /// Grabs the outer operation mutex and calls `f` on the locked operation. 781 /// The function also deletes the operation if it returns with an error or if 782 /// `delete_op` is true. with_locked_operation<T, F>(&self, f: F, delete_op: bool) -> Result<T> where for<'a> F: FnOnce(&'a Operation) -> Result<T>,783 fn with_locked_operation<T, F>(&self, f: F, delete_op: bool) -> Result<T> 784 where 785 for<'a> F: FnOnce(&'a Operation) -> Result<T>, 786 { 787 let mut delete_op: bool = delete_op; 788 match self.operation.try_lock() { 789 Ok(mut mutex_guard) => { 790 let result = match &*mutex_guard { 791 Some(op) => { 792 let result = f(op); 793 // Any error here means we can discard the operation. 794 if result.is_err() { 795 delete_op = true; 796 } 797 result 798 } 799 None => Err(Error::Km(ErrorCode::INVALID_OPERATION_HANDLE)) 800 .context(ks_err!("KeystoreOperation::with_locked_operation")), 801 }; 802 803 if delete_op { 804 // We give up our reference to the Operation, thereby freeing up our 805 // internal resources and ending the wrapped KeyMint operation. 806 // This KeystoreOperation object will still be owned by an SpIBinder 807 // until the client drops its remote reference. 808 *mutex_guard = None; 809 } 810 result 811 } 812 Err(_) => Err(Error::Rc(ResponseCode::OPERATION_BUSY)) 813 .context(ks_err!("KeystoreOperation::with_locked_operation")), 814 } 815 } 816 } 817 818 impl binder::Interface for KeystoreOperation {} 819 820 impl IKeystoreOperation for KeystoreOperation { updateAad(&self, aad_input: &[u8]) -> binder::Result<()>821 fn updateAad(&self, aad_input: &[u8]) -> binder::Result<()> { 822 let _wp = wd::watch_millis("IKeystoreOperation::updateAad", 500); 823 map_or_log_err( 824 self.with_locked_operation( 825 |op| op.update_aad(aad_input).context(ks_err!("KeystoreOperation::updateAad")), 826 false, 827 ), 828 Ok, 829 ) 830 } 831 update(&self, input: &[u8]) -> binder::Result<Option<Vec<u8>>>832 fn update(&self, input: &[u8]) -> binder::Result<Option<Vec<u8>>> { 833 let _wp = wd::watch_millis("IKeystoreOperation::update", 500); 834 map_or_log_err( 835 self.with_locked_operation( 836 |op| op.update(input).context(ks_err!("KeystoreOperation::update")), 837 false, 838 ), 839 Ok, 840 ) 841 } finish( &self, input: Option<&[u8]>, signature: Option<&[u8]>, ) -> binder::Result<Option<Vec<u8>>>842 fn finish( 843 &self, 844 input: Option<&[u8]>, 845 signature: Option<&[u8]>, 846 ) -> binder::Result<Option<Vec<u8>>> { 847 let _wp = wd::watch_millis("IKeystoreOperation::finish", 500); 848 map_or_log_err( 849 self.with_locked_operation( 850 |op| op.finish(input, signature).context(ks_err!("KeystoreOperation::finish")), 851 true, 852 ), 853 Ok, 854 ) 855 } 856 abort(&self) -> binder::Result<()>857 fn abort(&self) -> binder::Result<()> { 858 let _wp = wd::watch_millis("IKeystoreOperation::abort", 500); 859 map_err_with( 860 self.with_locked_operation( 861 |op| op.abort(Outcome::Abort).context(ks_err!("KeystoreOperation::abort")), 862 true, 863 ), 864 |e| { 865 match e.root_cause().downcast_ref::<Error>() { 866 // Calling abort on expired operations is something very common. 867 // There is no reason to clutter the log with it. It is never the cause 868 // for a true problem. 869 Some(Error::Km(ErrorCode::INVALID_OPERATION_HANDLE)) => {} 870 _ => log::error!("{:?}", e), 871 }; 872 e 873 }, 874 Ok, 875 ) 876 } 877 } 878