The Linux Kernel Mailing List
 help / color / mirror / Atom feed
* [PATCH 0/3] powerpc: Misc fixes for debug kernel splats
@ 2026-11-17  4:12 Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 1/3] powerpc/64: Leave lockdep hardirq context around replayed interrupts Ritesh Harjani (IBM)
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Ritesh Harjani (IBM) @ 2026-11-17  4:12 UTC (permalink / raw)
  To: linuxppc-dev
  Cc: Madhavan Srinivasan, Venkat Rao Bagalkote, linux-kernel, chleroy,
	Michael Ellerman, Nicholas Piggin, Shrikanth Hegde,
	Mukesh Kumar Chaurasiya, Ritesh Harjani (IBM)

Enabling debug options like PROVE_LOCKING, DEBUG_PREEMPT and
DEBUG_ATOMIC_SLEEP on powerpc64 showed these warnings.

Main motivation for me to look into this was - many folks internally reported
that after generic entry/exit feature, they aren't able to use lockdep because
of the warning splat it generates during boot, after which lockdep gets
disabled.

Therefore this series is an attempt to fix the warnings which are generated
- when we enable these various debug config options.

Ritesh Harjani (IBM) (3):
  powerpc/64: Leave lockdep hardirq context around replayed interrupts
  powerpc/kvm-hv-pmu: Don't allocate gsm/gsb GFP_KERNEL allocations under spinlock
  powerpc/watchdog: Disable preemption in arch_touch_nmi_watchdog()

 arch/powerpc/kernel/irq_64.c   | 11 +++---
 arch/powerpc/kernel/watchdog.c |  8 +++--
 arch/powerpc/perf/kvm-hv-pmu.c | 65 ++++++++++++++++------------------
 3 files changed, 41 insertions(+), 43 deletions(-)

--
2.39.5


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH 1/3] powerpc/64: Leave lockdep hardirq context around replayed interrupts
  2026-11-17  4:12 [PATCH 0/3] powerpc: Misc fixes for debug kernel splats Ritesh Harjani (IBM)
@ 2026-11-17  4:12 ` Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 2/3] powerpc/kvm-hv-pmu: Don't allocate gsm/gsb GFP_KERNEL allocations under spinlock Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 3/3] powerpc/watchdog: Disable preemption in arch_touch_nmi_watchdog() Ritesh Harjani (IBM)
  2 siblings, 0 replies; 4+ messages in thread
From: Ritesh Harjani (IBM) @ 2026-11-17  4:12 UTC (permalink / raw)
  To: linuxppc-dev
  Cc: Madhavan Srinivasan, Venkat Rao Bagalkote, linux-kernel, chleroy,
	Michael Ellerman, Nicholas Piggin, Shrikanth Hegde,
	Mukesh Kumar Chaurasiya, Ritesh Harjani (IBM), stable

With PROVE_LOCKING, boot warns once and lockdep turns itself off:
  DEBUG_LOCKS_WARN_ON(lockdep_hardirq_context())
  WARNING: kernel/locking/lockdep.c:4448 at lockdep_hardirqs_on_prepare+0x22c/0x2d4, CPU#0: swapper/0/0
  CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 7.4.0-rc2-00027-g074831f06e18-dirty #123 PREEMPT
  NIP:  c0000000002b2f2c LR: c0000000002b2f28 CTR: c000000002426f90
  REGS: c000000004f075c0 TRAP: 0700   Not tainted  (7.4.0-rc2-00027-g074831f06e18-dirty)
  MSR:  8000000002021033 <SF,VEC,ME,IR,DR,RI,LE>  CR: 48022221  XER: 20040000
  NIP [c0000000002b2f2c] lockdep_hardirqs_on_prepare+0x22c/0x2d4
  LR [c0000000002b2f28] lockdep_hardirqs_on_prepare+0x228/0x2d4
  Call Trace:
    lockdep_hardirqs_on_prepare+0x228/0x2d4 (unreliable)
    trace_hardirqs_on+0x98/0x11c
    irqentry_exit+0x64c/0x1064
    timer_interrupt+0x274/0x6bc
    __replay_soft_interrupts+0x158/0x380
    arch_local_irq_restore+0x1cc/0x244
    console_flush_one_record+0x438/0x654
    console_unlock+0xb8/0x1e0
    vprintk_emit+0x300/0x56c
    vprintk+0x20/0x34
    _printk+0x40/0x54
    lockdep_init+0xdc/0x104
    start_kernel+0xa24/0xc68
    start_here_common+0x1c/0x20

Since commit bee25f97ad24 ("powerpc: Enable GENERIC_ENTRY feature") a
replayed handler always returns via irqentry_exit(), which traces hardirqs
on while the replay is still inside irq_enter() (w/o fix):
  arch_local_irq_restore()
    irq_enter()                        hardirq_context = 1
    __replay_soft_interrupts()
      timer_interrupt(&regs)
        irq_enter_rcu()                hardirq_context = 2
        irq_exit_rcu()                 hardirq_context = 1
        irqentry_exit()
          trace_hardirqs_on()          hardirq_context != 0 -> WARN
      next_interrupt()
        lockdep_hardirq_exit()         too late

Leave the hardirq context before each replayed handler instead of
after it (with the fix):
  arch_local_irq_restore()
    irq_enter()                        hardirq_context = 1
    __replay_soft_interrupts()
      lockdep_hardirq_exit()           hardirq_context = 0 // fix
      timer_interrupt(&regs)
        irq_enter_rcu()                hardirq_context = 1
        irq_exit_rcu()                 hardirq_context = 0
        irqentry_exit()
          trace_hardirqs_on()          ok
      next_interrupt()
        trace_hardirqs_off()
        lockdep_hardirq_enter()        hardirq_context = 1
    irq_exit()                         hardirq_context = 0

Fixes: bee25f97ad24 ("powerpc: Enable GENERIC_ENTRY feature")
Cc: stable@vger.kernel.org
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
---
 arch/powerpc/kernel/irq_64.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/arch/powerpc/kernel/irq_64.c b/arch/powerpc/kernel/irq_64.c
index d5c48d1b0a31..b189e796eb34 100644
--- a/arch/powerpc/kernel/irq_64.c
+++ b/arch/powerpc/kernel/irq_64.c
@@ -75,12 +75,6 @@ static inline void next_interrupt(struct pt_regs *regs)
 		WARN_ON(irq_soft_mask_return() != IRQS_ALL_DISABLED);
 	}

-	/*
-	 * We are responding to the next interrupt, so interrupt-off
-	 * latencies should be reset here.
-	 */
-	lockdep_hardirq_exit();
-	trace_hardirqs_on();
 	trace_hardirqs_off();
 	lockdep_hardirq_enter();
 }
@@ -137,18 +131,21 @@ static __no_kcsan void __replay_soft_interrupts(void)
 	if (IS_ENABLED(CONFIG_PPC_BOOK3S) &&
 	    irq_happened_test_and_clear(PACA_IRQ_HMI)) {
 		regs.trap = INTERRUPT_HMI;
+		lockdep_hardirq_exit();
 		handle_hmi_exception(&regs);
 		next_interrupt(&regs);
 	}

 	if (irq_happened_test_and_clear(PACA_IRQ_DEC)) {
 		regs.trap = INTERRUPT_DECREMENTER;
+		lockdep_hardirq_exit();
 		timer_interrupt(&regs);
 		next_interrupt(&regs);
 	}

 	if (irq_happened_test_and_clear(PACA_IRQ_EE)) {
 		regs.trap = INTERRUPT_EXTERNAL;
+		lockdep_hardirq_exit();
 		do_IRQ(&regs);
 		next_interrupt(&regs);
 	}
@@ -156,6 +153,7 @@ static __no_kcsan void __replay_soft_interrupts(void)
 	if (IS_ENABLED(CONFIG_PPC_DOORBELL) &&
 	    irq_happened_test_and_clear(PACA_IRQ_DBELL)) {
 		regs.trap = INTERRUPT_DOORBELL;
+		lockdep_hardirq_exit();
 		doorbell_exception(&regs);
 		next_interrupt(&regs);
 	}
@@ -164,6 +162,7 @@ static __no_kcsan void __replay_soft_interrupts(void)
 	if (IS_ENABLED(CONFIG_PPC_BOOK3S) &&
 	    irq_happened_test_and_clear(PACA_IRQ_PMI)) {
 		regs.trap = INTERRUPT_PERFMON;
+		lockdep_hardirq_exit();
 		performance_monitor_exception(&regs);
 		next_interrupt(&regs);
 	}
--
2.39.5


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH 2/3] powerpc/kvm-hv-pmu: Don't allocate gsm/gsb GFP_KERNEL allocations under spinlock
  2026-11-17  4:12 [PATCH 0/3] powerpc: Misc fixes for debug kernel splats Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 1/3] powerpc/64: Leave lockdep hardirq context around replayed interrupts Ritesh Harjani (IBM)
@ 2026-11-17  4:12 ` Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 3/3] powerpc/watchdog: Disable preemption in arch_touch_nmi_watchdog() Ritesh Harjani (IBM)
  2 siblings, 0 replies; 4+ messages in thread
From: Ritesh Harjani (IBM) @ 2026-11-17  4:12 UTC (permalink / raw)
  To: linuxppc-dev
  Cc: Madhavan Srinivasan, Venkat Rao Bagalkote, linux-kernel, chleroy,
	Michael Ellerman, Nicholas Piggin, Shrikanth Hegde,
	Mukesh Kumar Chaurasiya, Ritesh Harjani (IBM), stable

kvmppc_init_hostwide() takes lock_l0_stats with spin_lock_irqsave() and
then allocates the guest state message and buffer with GFP_KERNEL. On
a nested API v2 host with DEBUG_ATOMIC_SLEEP this warns at boot:

  BUG: sleeping function called from invalid context at include/linux/sched/mm.h:322
  in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0
  preempt_count: 1, expected: 0
  RCU nest depth: 0, expected: 0
  locks held by swapper/0/1: 1, on CPU#0:
   #0: c000000004ad95b8 (lock_l0_stats){....}-{3:3}, at: kvmppc_register_pmu+0x50/0x270
  CPU: 0 UID: 0 PID: 1 Comm: swapper/0 Not tainted 7.4.0-rc2-00028-ge5a8bf3a3ee9-dirty #124 PREEMPT
  Call Trace:
   dump_stack_lvl+0xe8/0x150 (unreliable)
   __might_resched+0x190/0x30c
   __kmalloc_cache_noprof+0x4dc/0x760
   kvmppc_gsm_new+0x68/0xcc
   kvmppc_register_pmu+0xd4/0x270
   do_one_initcall+0xac/0x48c
   kernel_init_freeable+0x3cc/0x7f0
   kernel_init+0x3c/0x274
   ret_from_kernel_user_thread+0x14/0x1c

The lock is not needed for the allocation of gsm and gsb buffers.
Allocate and fill the message and buffer without the lock and only take
it to publish the two global pointers.

Fixes: f0c9c49c5067 ("powerpc/kvm-hv-pmu: Implement GSB message-ops for hostwide counters")
Cc: stable@vger.kernel.org
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
---
 arch/powerpc/perf/kvm-hv-pmu.c | 65 ++++++++++++++++------------------
 1 file changed, 30 insertions(+), 35 deletions(-)

diff --git a/arch/powerpc/perf/kvm-hv-pmu.c b/arch/powerpc/perf/kvm-hv-pmu.c
index aa72b96b5a8c..a58634f656e9 100644
--- a/arch/powerpc/perf/kvm-hv-pmu.c
+++ b/arch/powerpc/perf/kvm-hv-pmu.c
@@ -314,53 +314,48 @@ static struct kvmppc_gs_msg_ops gsb_ops_l0_stats = {

 static int kvmppc_init_hostwide(void)
 {
-	int rc = 0;
+	struct kvmppc_gs_msg *gsm;
+	struct kvmppc_gs_buff *gsb;
 	unsigned long flags;
+	int rc;

-	spin_lock_irqsave(&lock_l0_stats, flags);
-
-	/* already registered ? */
-	if (gsm_l0_stats) {
-		rc = 0;
-		goto out;
-	}
-
-	/* setup the Guest state message/buffer to talk to L0 */
-	gsm_l0_stats = kvmppc_gsm_new(&gsb_ops_l0_stats, &l0_stats,
-				      GSM_SEND, GFP_KERNEL);
-	if (!gsm_l0_stats) {
-		rc = -ENOMEM;
-		goto out;
-	}
+	/* allocate the Guest state message/buffer to talk to L0 */
+	gsm = kvmppc_gsm_new(&gsb_ops_l0_stats, &l0_stats, GSM_SEND,
+			     GFP_KERNEL);
+	if (!gsm)
+		return -ENOMEM;

 	/* Populate the Idents */
-	kvmppc_gsm_include(gsm_l0_stats, KVMPPC_GSID_L0_GUEST_HEAP);
-	kvmppc_gsm_include(gsm_l0_stats, KVMPPC_GSID_L0_GUEST_HEAP_MAX);
-	kvmppc_gsm_include(gsm_l0_stats, KVMPPC_GSID_L0_GUEST_PGTABLE_SIZE);
-	kvmppc_gsm_include(gsm_l0_stats, KVMPPC_GSID_L0_GUEST_PGTABLE_SIZE_MAX);
-	kvmppc_gsm_include(gsm_l0_stats, KVMPPC_GSID_L0_GUEST_PGTABLE_RECLAIM);
+	kvmppc_gsm_include(gsm, KVMPPC_GSID_L0_GUEST_HEAP);
+	kvmppc_gsm_include(gsm, KVMPPC_GSID_L0_GUEST_HEAP_MAX);
+	kvmppc_gsm_include(gsm, KVMPPC_GSID_L0_GUEST_PGTABLE_SIZE);
+	kvmppc_gsm_include(gsm, KVMPPC_GSID_L0_GUEST_PGTABLE_SIZE_MAX);
+	kvmppc_gsm_include(gsm, KVMPPC_GSID_L0_GUEST_PGTABLE_RECLAIM);

 	/* allocate GSB. Guest/Vcpu Id is ignored */
-	gsb_l0_stats = kvmppc_gsb_new(kvmppc_gsm_size(gsm_l0_stats), 0, 0,
-				      GFP_KERNEL);
-	if (!gsb_l0_stats) {
+	gsb = kvmppc_gsb_new(kvmppc_gsm_size(gsm), 0, 0, GFP_KERNEL);
+	if (!gsb) {
 		rc = -ENOMEM;
-		goto out;
+		goto free_gsm;
 	}

 	/* ask the ops to fill in the info */
-	rc = kvmppc_gsm_fill_info(gsm_l0_stats, gsb_l0_stats);
+	rc = kvmppc_gsm_fill_info(gsm, gsb);
+	if (rc)
+		goto free_gsb;

-out:
-	if (rc) {
-		if (gsm_l0_stats)
-			kvmppc_gsm_free(gsm_l0_stats);
-		if (gsb_l0_stats)
-			kvmppc_gsb_free(gsb_l0_stats);
-		gsm_l0_stats = NULL;
-		gsb_l0_stats = NULL;
-	}
+	/* setup the Guest state message/buffer to talk to L0 */
+	spin_lock_irqsave(&lock_l0_stats, flags);
+	gsm_l0_stats = gsm;
+	gsb_l0_stats = gsb;
 	spin_unlock_irqrestore(&lock_l0_stats, flags);
+
+	return 0;
+
+free_gsb:
+	kvmppc_gsb_free(gsb);
+free_gsm:
+	kvmppc_gsm_free(gsm);
 	return rc;
 }

--
2.39.5


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH 3/3] powerpc/watchdog: Disable preemption in arch_touch_nmi_watchdog()
  2026-11-17  4:12 [PATCH 0/3] powerpc: Misc fixes for debug kernel splats Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 1/3] powerpc/64: Leave lockdep hardirq context around replayed interrupts Ritesh Harjani (IBM)
  2026-11-17  4:12 ` [PATCH 2/3] powerpc/kvm-hv-pmu: Don't allocate gsm/gsb GFP_KERNEL allocations under spinlock Ritesh Harjani (IBM)
@ 2026-11-17  4:12 ` Ritesh Harjani (IBM)
  2 siblings, 0 replies; 4+ messages in thread
From: Ritesh Harjani (IBM) @ 2026-11-17  4:12 UTC (permalink / raw)
  To: linuxppc-dev
  Cc: Madhavan Srinivasan, Venkat Rao Bagalkote, linux-kernel, chleroy,
	Michael Ellerman, Nicholas Piggin, Shrikanth Hegde,
	Mukesh Kumar Chaurasiya, Ritesh Harjani (IBM), stable

touch_nmi_watchdog() can be called from preemptible context, e.g.
sysrq-w via /proc/sysrq-trigger. The powerpc implementation uses
smp_processor_id(), so with DEBUG_PREEMPT we get:

  root@buildroot:/# echo w > /proc/sysrq-trigger
  sysrq: Show Blocked State
  BUG: using smp_processor_id() in preemptible [00000000] code: bash/238
  caller is arch_touch_nmi_watchdog+0x3c/0xd4
  CPU: 0 UID: 0 PID: 238 Comm: bash Not tainted 7.4.0-rc2-00029-g7664547149ae-dirty #126 PREEMPT
  Call Trace:
   dump_stack_lvl+0x138/0x150 (unreliable)
   check_preemption_disabled+0x120/0x158
   arch_touch_nmi_watchdog+0x3c/0xd4
   show_state_filter+0xf0/0x338
   sysrq_handle_showstate_blocked+0x24/0x38
   __handle_sysrq+0x134/0x3a8
   write_sysrq_trigger+0x178/0x22c
   proc_reg_write+0xf4/0x150
   vfs_write+0xfc/0x71c
   ksys_write+0x8c/0x150
   system_call_exception+0x310/0x590
   system_call_vectored_common+0x15c/0x2ec
  ---- interrupt: 3000 at 0x7ffd01755b98

Unlike the generic version, which only sets a flag, we update the CPU's
timebase stamp and its SMP watchdog pending bit, so a migration in
between touches the state of a CPU we are no longer on.

Disable preemption across the update. While at it, let's also mark the
function as notrace (since we might be within a tracer when this
function gets called).

Fixes: 2104180a5369 ("powerpc/64s: implement arch-specific hardlockup watchdog")
Cc: stable@vger.kernel.org
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
---
 arch/powerpc/kernel/watchdog.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/kernel/watchdog.c b/arch/powerpc/kernel/watchdog.c
index 44f178e07531..61b67dde04f2 100644
--- a/arch/powerpc/kernel/watchdog.c
+++ b/arch/powerpc/kernel/watchdog.c
@@ -456,12 +456,16 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
 	return HRTIMER_RESTART;
 }

-void arch_touch_nmi_watchdog(void)
+notrace void arch_touch_nmi_watchdog(void)
 {
 	unsigned long ticks = tb_ticks_per_usec * wd_timer_period_ms * 1000;
-	int cpu = smp_processor_id();
+	int cpu;
 	u64 tb;

+	/* Callers may be preemptible. Stay on this CPU for the update. */
+	guard(preempt_notrace)();
+	cpu = smp_processor_id();
+
 	if (!cpumask_test_cpu(cpu, &watchdog_cpumask))
 		return;

--
2.39.5


^ permalink raw reply related	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-11-17  4:13 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-17  4:12 [PATCH 0/3] powerpc: Misc fixes for debug kernel splats Ritesh Harjani (IBM)
2026-11-17  4:12 ` [PATCH 1/3] powerpc/64: Leave lockdep hardirq context around replayed interrupts Ritesh Harjani (IBM)
2026-11-17  4:12 ` [PATCH 2/3] powerpc/kvm-hv-pmu: Don't allocate gsm/gsb GFP_KERNEL allocations under spinlock Ritesh Harjani (IBM)
2026-11-17  4:12 ` [PATCH 3/3] powerpc/watchdog: Disable preemption in arch_touch_nmi_watchdog() Ritesh Harjani (IBM)

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox