CSD-lock diagnostic updates for v6.13

This commit switches from sched_clock() to ktime_get_mono_fast_ns(), which
 on x86 switches from the rdtsc instruction to the rdtscp instruction,
 thus avoiding instruction reorderings that cause false-positive reports
 of CSD-lock stalls of almost 2^46 nanoseconds.  These false positives
 are rare, but really are seen in the wild.
 -----BEGIN PGP SIGNATURE-----
 
 iQJHBAABCgAxFiEEbK7UrM+RBIrCoViJnr8S83LZ+4wFAmc5XvETHHBhdWxtY2tA
 a2VybmVsLm9yZwAKCRCevxLzctn7jNUCD/9NqeuxsVcumybbjlHs/IbJt47qTPVk
 1O+mpLiKfscw/ndfvqJe1RU+IOUJUPBPzBPUWvZQZ2SzeU03oOI4/szFttDdXSi3
 0uI9qOJn3auk2+cdU7CxXOLSiWYEWlMjWvN6d34QeLh7smLkendxH2wo2fkL9kf0
 DzvosOrlyNWGZPUQrb1TRW7RKGE7vap8x7tK/p1qMO2xmaPeIX7dfiY38CJC5fjj
 +n8i1aZIxLFc65I0/Z+nGTMFrktzbYjJik6k++QZzHx+GiXaCkgfidZFspj3uPXW
 CPa6KxheCrdmFV4A/TVnKYJyutoGeheMjwlVfz0YOSe8J5/N3F9RfDFBYedt2fL+
 11gRpOg5hz61AsyxZ1+iViW0guXoVzn2uwQ5rkou9184fBXPuwH1MAwBcsKYwQig
 Frd0ZzyrqGHCHwDWtBfAb+qC17b5krsa+fKkjiPFDRDRB5N2hh67tcquOE3wzvrG
 oAHEZgeFwxZQYGIZ7uITebyThe9NvkBRyrJLvxUEpvF2MoI0yJaqoAwkHieSl1vD
 KJJ+o+HxVa3D/WCWxTNCjDyxvCMJpFHFWB3h8+hi+X2UleRGrDiIDmIdddMM1/gr
 meYjZ/c1/t7Y14zwzp/SxUHFJ4U8U2jI23/K5ldJVH34k6XwccU06NYWjHYokXgT
 ZRMCJ8sAcTtlhw==
 =8GQJ
 -----END PGP SIGNATURE-----

Merge tag 'csd-lock.2024.11.16a' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu

Pull CSD-lock update from Paul McKenney:
 "This switches from sched_clock() to ktime_get_mono_fast_ns(), which on
  x86 switches from the rdtsc instruction to the rdtscp instruction,
  thus avoiding instruction reorderings that cause false-positive
  reports of CSD-lock stalls of almost 2^46 nanoseconds. These false
  positives are rare, but really are seen in the wild"

* tag 'csd-lock.2024.11.16a' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu:
  locking/csd-lock: Switch from sched_clock() to ktime_get_mono_fast_ns()
This commit is contained in:
Linus Torvalds 2024-11-19 10:18:45 -08:00
commit 311e062ad5

View File

@ -246,7 +246,7 @@ static bool csd_lock_wait_toolong(call_single_data_t *csd, u64 ts0, u64 *ts1, in
return true;
}
ts2 = sched_clock();
ts2 = ktime_get_mono_fast_ns();
/* How long since we last checked for a stuck CSD lock.*/
ts_delta = ts2 - *ts1;
if (likely(ts_delta <= csd_lock_timeout_ns * (*nmessages + 1) *
@ -321,7 +321,7 @@ static void __csd_lock_wait(call_single_data_t *csd)
int bug_id = 0;
u64 ts0, ts1;
ts1 = ts0 = sched_clock();
ts1 = ts0 = ktime_get_mono_fast_ns();
for (;;) {
if (csd_lock_wait_toolong(csd, ts0, &ts1, &bug_id, &nmessages))
break;