[PATCH] timekeeping: Register default clocksource before taking tk_core.lock

[Mikhail Gavrilov]

Commit f24df84cbe05 ("time/jiffies: Register jiffies clocksource before
usage") moved the jiffies clocksource registration into
clocksource_default_clock(), so that it is registered lazily on the first
call. __clocksource_register() acquires clocksource_mutex, but the first
caller is timekeeping_init(), which invokes clocksource_default_clock()
while holding tk_core.lock, a raw spinlock.

Acquiring a sleeping mutex while holding a raw spinlock is invalid, and
lockdep reports it on every boot when CONFIG_PROVE_RAW_LOCK_NESTING is
enabled:

  =============================
  [ BUG: Invalid wait context ]
  -----------------------------
  swapper/0/0 is trying to lock:
  ... (clocksource_mutex){....}-{4:4}, at: __clocksource_register_scale
  other info that might help us debug this:
  context-{5:5}
  1 lock held by swapper/0/0:
   #0: ... (&tkd->lock){....}-{2:2}, at: timekeeping_init+0x1d5/0x320
  stack backtrace:
   dump_stack_lvl+0x84/0xd0
   __lock_acquire.cold+0x39/0x63
   lock_acquire.part.0+0xc7/0x280
   __mutex_lock+0x1b2/0x2440
   __clocksource_register_scale.cold+0x289/0x609
   clocksource_default_clock+0x31/0x50
   timekeeping_init+0x1e3/0x320
   start_kernel+0x243/0x3f0
   x86_64_start_reservations+0x24/0x30
   x86_64_start_kernel+0x134/0x140
   common_startup_64+0x13e/0x151

On PREEMPT_RT this is a genuine sleeping-while-atomic bug. On !RT
clocksource_mutex is uncontended at this point, so the boot survives, but
the locking rule is violated either way.

The default clocksource only has to be registered before
tk_setup_internals() consumes its mult/shift/maxadj. Neither
clocksource_default_clock(), the ->enable() callback, nor the registration
itself need tk_core.lock, so fetch and enable the clock before acquiring
the lock. This preserves the "register before usage" ordering while
keeping clocksource_mutex out of the raw spinlock section.

clocksource_default_clock() has a second caller,
clocksource_done_booting(), which invokes it with clocksource_mutex
already held. That path avoids a recursive lock only because
timekeeping_init() has already run and set cs_jiffies_registered, so the
registration is skipped there. This change does not alter that; it only
fixes the invalid wait context in timekeeping_init().

Fixes: f24df84cbe05 ("time/jiffies: Register jiffies clocksource before usage")
Signed-off-by: Mikhail Gavrilov <mikhail.v.gavrilov@xxxxxxxxx>
---
 kernel/time/timekeeping.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index 0d5b67f609bb..b1b5ec43c0f2 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -2061,13 +2061,14 @@ void __init timekeeping_init(void)
 	 */
 	wall_to_mono = timespec64_sub(boot_offset, wall_time);
 
+	clock = clocksource_default_clock();
+	if (clock->enable)
+		clock->enable(clock);
+
 	guard(raw_spinlock_irqsave)(&tk_core.lock);
 
 	ntp_init();
 
-	clock = clocksource_default_clock();
-	if (clock->enable)
-		clock->enable(clock);
 	tk_setup_internals(tks, clock);
 
 	tk_set_xtime(tks, &wall_time);
-- 
2.54.0