| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <6dda193d-53e3-44e3-8d3c-734ab8b50869@kadam.mountain>
Date: Tue, 7 Nov 2023 07:12:55 +0300
From: Dan Carpenter <dan.carpenter@...aro.org>
To: oe-kbuild@...ts.linux.dev, Kent Overstreet <kmo@...erainc.com>
Cc: lkp@...el.com, oe-kbuild-all@...ts.linux.dev,
linux-kernel@...r.kernel.org
Subject: fs/bcachefs/six.c:283 __do_six_trylock() error: uninitialized symbol
'old'.
tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
head: d2f51b3516dade79269ff45eae2a7668ae711b25
commit: b60c8e9e7b082abac290ebdb9166b806e7d83fb7 six locks: lock->state.seq no longer used for write lock held
config: x86_64-randconfig-161-20231102 (https://download.01.org/0day-ci/archive/20231106/202311062227.0GUulHgD-lkp@intel.com/config)
compiler: gcc-7 (Ubuntu 7.5.0-6ubuntu2) 7.5.0
reproduce: (https://download.01.org/0day-ci/archive/20231106/202311062227.0GUulHgD-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@...el.com>
| Reported-by: Dan Carpenter <dan.carpenter@...aro.org>
| Closes: https://lore.kernel.org/r/202311062227.0GUulHgD-lkp@intel.com/
New smatch warnings:
fs/bcachefs/six.c:283 __do_six_trylock() error: uninitialized symbol 'old'.
vim +/old +283 fs/bcachefs/six.c
91d16f16d0fd4b Kent Overstreet 2023-05-21 199 static int __do_six_trylock(struct six_lock *lock, enum six_lock_type type,
91d16f16d0fd4b Kent Overstreet 2023-05-21 200 struct task_struct *task, bool try)
1c6fdbd8f2465d Kent Overstreet 2017-03-16 201 {
1c6fdbd8f2465d Kent Overstreet 2017-03-16 202 const struct six_lock_vals l[] = LOCK_VALS;
84a37cbf62e044 Kent Overstreet 2022-08-26 203 int ret;
1fb4fe63178881 Kent Overstreet 2023-05-20 204 u64 old, new, v;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 205
84a37cbf62e044 Kent Overstreet 2022-08-26 206 EBUG_ON(type == SIX_LOCK_write && lock->owner != task);
1fb4fe63178881 Kent Overstreet 2023-05-20 207 EBUG_ON(type == SIX_LOCK_write &&
b60c8e9e7b082a Kent Overstreet 2023-06-16 208 (try != !(atomic64_read(&lock->state) & SIX_LOCK_HELD_write)));
1fb4fe63178881 Kent Overstreet 2023-05-20 209 EBUG_ON(type == SIX_LOCK_write &&
b60c8e9e7b082a Kent Overstreet 2023-06-16 210 (try != !(atomic64_read(&lock->state) & SIX_STATE_WRITE_LOCK)));
1c6fdbd8f2465d Kent Overstreet 2017-03-16 211
1c6fdbd8f2465d Kent Overstreet 2017-03-16 212 /*
1c6fdbd8f2465d Kent Overstreet 2017-03-16 213 * Percpu reader mode:
1c6fdbd8f2465d Kent Overstreet 2017-03-16 214 *
1c6fdbd8f2465d Kent Overstreet 2017-03-16 215 * The basic idea behind this algorithm is that you can implement a lock
1c6fdbd8f2465d Kent Overstreet 2017-03-16 216 * between two threads without any atomics, just memory barriers:
1c6fdbd8f2465d Kent Overstreet 2017-03-16 217 *
1c6fdbd8f2465d Kent Overstreet 2017-03-16 218 * For two threads you'll need two variables, one variable for "thread a
1c6fdbd8f2465d Kent Overstreet 2017-03-16 219 * has the lock" and another for "thread b has the lock".
1c6fdbd8f2465d Kent Overstreet 2017-03-16 220 *
1c6fdbd8f2465d Kent Overstreet 2017-03-16 221 * To take the lock, a thread sets its variable indicating that it holds
1c6fdbd8f2465d Kent Overstreet 2017-03-16 222 * the lock, then issues a full memory barrier, then reads from the
1c6fdbd8f2465d Kent Overstreet 2017-03-16 223 * other thread's variable to check if the other thread thinks it has
1c6fdbd8f2465d Kent Overstreet 2017-03-16 224 * the lock. If we raced, we backoff and retry/sleep.
37f612bea5bd92 Kent Overstreet 2023-05-21 225 *
37f612bea5bd92 Kent Overstreet 2023-05-21 226 * Failure to take the lock may cause a spurious trylock failure in
37f612bea5bd92 Kent Overstreet 2023-05-21 227 * another thread, because we temporarily set the lock to indicate that
37f612bea5bd92 Kent Overstreet 2023-05-21 228 * we held it. This would be a problem for a thread in six_lock(), when
37f612bea5bd92 Kent Overstreet 2023-05-21 229 * they are calling trylock after adding themself to the waitlist and
37f612bea5bd92 Kent Overstreet 2023-05-21 230 * prior to sleeping.
37f612bea5bd92 Kent Overstreet 2023-05-21 231 *
37f612bea5bd92 Kent Overstreet 2023-05-21 232 * Therefore, if we fail to get the lock, and there were waiters of the
37f612bea5bd92 Kent Overstreet 2023-05-21 233 * type we conflict with, we will have to issue a wakeup.
37f612bea5bd92 Kent Overstreet 2023-05-21 234 *
37f612bea5bd92 Kent Overstreet 2023-05-21 235 * Since we may be called under wait_lock (and by the wakeup code
37f612bea5bd92 Kent Overstreet 2023-05-21 236 * itself), we return that the wakeup has to be done instead of doing it
37f612bea5bd92 Kent Overstreet 2023-05-21 237 * here.
1c6fdbd8f2465d Kent Overstreet 2017-03-16 238 */
1c6fdbd8f2465d Kent Overstreet 2017-03-16 239 if (type == SIX_LOCK_read && lock->readers) {
1c6fdbd8f2465d Kent Overstreet 2017-03-16 240 preempt_disable();
1c6fdbd8f2465d Kent Overstreet 2017-03-16 241 this_cpu_inc(*lock->readers); /* signal that we own lock */
1c6fdbd8f2465d Kent Overstreet 2017-03-16 242
1c6fdbd8f2465d Kent Overstreet 2017-03-16 243 smp_mb();
1c6fdbd8f2465d Kent Overstreet 2017-03-16 244
1fb4fe63178881 Kent Overstreet 2023-05-20 245 old = atomic64_read(&lock->state);
1fb4fe63178881 Kent Overstreet 2023-05-20 246 ret = !(old & l[type].lock_fail);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 247
1c6fdbd8f2465d Kent Overstreet 2017-03-16 248 this_cpu_sub(*lock->readers, !ret);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 249 preempt_enable();
1c6fdbd8f2465d Kent Overstreet 2017-03-16 250
37f612bea5bd92 Kent Overstreet 2023-05-21 251 if (!ret && (old & SIX_STATE_WAITING_WRITE))
84a37cbf62e044 Kent Overstreet 2022-08-26 252 ret = -1 - SIX_LOCK_write;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 253 } else if (type == SIX_LOCK_write && lock->readers) {
1c6fdbd8f2465d Kent Overstreet 2017-03-16 254 if (try) {
b60c8e9e7b082a Kent Overstreet 2023-06-16 255 atomic64_add(SIX_STATE_WRITE_LOCK, &lock->state);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 256 smp_mb__after_atomic();
1c6fdbd8f2465d Kent Overstreet 2017-03-16 257 }
1c6fdbd8f2465d Kent Overstreet 2017-03-16 258
1c6fdbd8f2465d Kent Overstreet 2017-03-16 259 ret = !pcpu_read_count(lock);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 260
b60c8e9e7b082a Kent Overstreet 2023-06-16 261 if (try && !ret) {
b60c8e9e7b082a Kent Overstreet 2023-06-16 262 old = atomic64_sub_return(SIX_STATE_WRITE_LOCK, &lock->state);
old only set when ret is zero
b60c8e9e7b082a Kent Overstreet 2023-06-16 263 if (old & SIX_STATE_WAITING_READ)
84a37cbf62e044 Kent Overstreet 2022-08-26 264 ret = -1 - SIX_LOCK_read;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 265 }
1c6fdbd8f2465d Kent Overstreet 2017-03-16 266 } else {
1fb4fe63178881 Kent Overstreet 2023-05-20 267 v = atomic64_read(&lock->state);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 268 do {
1fb4fe63178881 Kent Overstreet 2023-05-20 269 new = old = v;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 270
b60c8e9e7b082a Kent Overstreet 2023-06-16 271 ret = !(old & l[type].lock_fail);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 272
b60c8e9e7b082a Kent Overstreet 2023-06-16 273 if (!ret || (type == SIX_LOCK_write && !try))
d2c86b77de5894 Kent Overstreet 2023-05-20 274 break;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 275
b60c8e9e7b082a Kent Overstreet 2023-06-16 276 new += l[type].lock_val;
b60c8e9e7b082a Kent Overstreet 2023-06-16 277 } while ((v = atomic64_cmpxchg_acquire(&lock->state, old, new)) != old);
1c6fdbd8f2465d Kent Overstreet 2017-03-16 278
1fb4fe63178881 Kent Overstreet 2023-05-20 279 EBUG_ON(ret && !(atomic64_read(&lock->state) & l[type].held_mask));
1c6fdbd8f2465d Kent Overstreet 2017-03-16 280 }
1c6fdbd8f2465d Kent Overstreet 2017-03-16 281
84a37cbf62e044 Kent Overstreet 2022-08-26 282 if (ret > 0)
84a37cbf62e044 Kent Overstreet 2022-08-26 @283 six_set_owner(lock, type, old, task);
^^^
We're allowed to pass uninitialized data here because it's not used and
it's an inline function. But the compiler is going to zero it anyway.
Just zero it and make the checker happy.
1c6fdbd8f2465d Kent Overstreet 2017-03-16 284
b60c8e9e7b082a Kent Overstreet 2023-06-16 285 EBUG_ON(type == SIX_LOCK_write && try && ret <= 0 &&
b60c8e9e7b082a Kent Overstreet 2023-06-16 286 (atomic64_read(&lock->state) & SIX_STATE_WRITE_LOCK));
1c6fdbd8f2465d Kent Overstreet 2017-03-16 287
1c6fdbd8f2465d Kent Overstreet 2017-03-16 288 return ret;
1c6fdbd8f2465d Kent Overstreet 2017-03-16 289 }
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
Powered by blists - more mailing lists