Skip to contentSkip to conversation

x86/hyperv: Properly suspend/resume reenlightenment notifications

PATCH Applied Vitaly Kuznetsov 4 participants 4 messages May 12, 2020 – May 13, 2020

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/hyperv: Properly suspend/resume reenlightenment notifications

Errors during hibernation with reenlightenment notifications enabled were reported:

 [   51.730435] PM: hibernation entry
 [   51.737435] PM: Syncing filesystems ...
 ...
 [   54.102216] Disabling non-boot CPUs ...
 [   54.106633] smpboot: CPU 1 is now offline
 [   54.110006] unchecked MSR access error: WRMSR to 0x40000106 (tried to
     write 0x47c72780000100ee) at rIP: 0xffffffff90062f24
     native_write_msr+0x4/0x20)
 [   54.110006] Call Trace:
 [   54.110006]  hv_cpu_die+0xd9/0xf0
 ...

Normally, hv_cpu_die() just reassigns reenlightenment notifications to some other CPU when the CPU receiving them goes offline. Upon hibernation, there is no other CPU which is still online so cpumask_any_but(cpu_online_mask) returns >= nr_cpu_ids and using it as hv_vp_index index is incorrect. Disable the feature when cpumask_any_but() fails.

Also, as we now disable reenlightenment notifications upon hibernation we need to restore them on resume. Check if hv_reenlightenment_cb was previously set and restore from hv_resume().

---
 arch/x86/hyperv/hv_init.c | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)
arch/x86/hyperv/hv_init.c +17 -2
index fd51bac11b46..acf76b466db6 100644
@@ -226,10 +226,18 @@ static int hv_cpu_die(unsigned int cpu)
 
 	rdmsrl(HV_X64_MSR_REENLIGHTENMENT_CONTROL, *((u64 *)&re_ctrl));
 	if (re_ctrl.target_vp == hv_vp_index[cpu]) {
-		/* Reassign to some other online CPU */
+		/*
+		 * Reassign reenlightenment notifications to some other online
+		 * CPU or just disable the feature if there are no online CPUs
+		 * left (happens on hibernation).
+		 */
 		new_cpu = cpumask_any_but(cpu_online_mask, cpu);
 
-		re_ctrl.target_vp = hv_vp_index[new_cpu];
+		if (new_cpu < nr_cpu_ids)
+			re_ctrl.target_vp = hv_vp_index[new_cpu];
+		else
+			re_ctrl.enabled = 0;
+
 		wrmsrl(HV_X64_MSR_REENLIGHTENMENT_CONTROL, *((u64 *)&re_ctrl));
 	}
 
@@ -293,6 +301,13 @@ static void hv_resume(void)
 
 	hv_hypercall_pg = hv_hypercall_pg_saved;
 	hv_hypercall_pg_saved = NULL;
+
+	/*
+	 * Reenlightenment notifications are disabled by hv_cpu_die(0),
+	 * reenable them here if hv_reenlightenment_cb was previously set.
+	 */
+	if (hv_reenlightenment_cb)
+		set_hv_tscchange_cb(hv_reenlightenment_cb);
 }
 
 /* Note: when the ops are called, only CPU0 is online and IRQs are disabled. */
Signature · 1 line
2.25.4
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 32 lines

From: Vitaly Kuznetsov <[email protected]>
Sent: Tuesday, May 12, 2020 9:02 AM
To: [email protected]
Cc: Wei Liu <[email protected]>; [email protected]; [email protected]; [email protected]; Michael Kelley <[email protected]>; Dexuan Cui <[email protected]>; Tianyu Lan <[email protected]>
Subject: [PATCH] x86/hyperv: Properly suspend/resume reenlightenment notifications

Errors during hibernation with reenlightenment notifications enabled were reported:

 [   51.730435] PM: hibernation entry
 [   51.737435] PM: Syncing filesystems ...
 ...
 [   54.102216] Disabling non-boot CPUs ...
 [   54.106633] smpboot: CPU 1 is now offline
 [   54.110006] unchecked MSR access error: WRMSR to 0x40000106 (tried
to
     write 0x47c72780000100ee) at rIP: 0xffffffff90062f24
     native_write_msr+0x4/0x20)
 [   54.110006] Call Trace:
 [   54.110006]  hv_cpu_die+0xd9/0xf0
 ...

Normally, hv_cpu_die() just reassigns reenlightenment notifications to some other CPU when the CPU receiving them goes offline. Upon hibernation, there is no other CPU which is still online so cpumask_any_but(cpu_online_mask) returns >= nr_cpu_ids and using it as hv_vp_index index is incorrect. Disable the feature when cpumask_any_but() fails.

Quoted 6 lines

Also, as we now disable reenlightenment notifications upon hibernation we need to restore them on resume. Check if hv_reenlightenment_cb was previously set and restore from hv_resume().

---

Looks good to me. Thanks!

Reviewed-by on v1 1/1
#2 lore raw
In reply to Vitaly Kuznetsov
Quoted 60 lines · Vitaly Kuznetsov

Errors during hibernation with reenlightenment notifications enabled were reported:

  [   51.730435] PM: hibernation entry
  [   51.737435] PM: Syncing filesystems ...
  ...
  [   54.102216] Disabling non-boot CPUs ...
  [   54.106633] smpboot: CPU 1 is now offline
  [   54.110006] unchecked MSR access error: WRMSR to 0x40000106 (tried to
      write 0x47c72780000100ee) at rIP: 0xffffffff90062f24
      native_write_msr+0x4/0x20)
  [   54.110006] Call Trace:
  [   54.110006]  hv_cpu_die+0xd9/0xf0
  ...

Normally, hv_cpu_die() just reassigns reenlightenment notifications to some other CPU when the CPU receiving them goes offline. Upon hibernation, there is no other CPU which is still online so cpumask_any_but(cpu_online_mask) returns >= nr_cpu_ids and using it as hv_vp_index index is incorrect. Disable the feature when cpumask_any_but() fails.

Also, as we now disable reenlightenment notifications upon hibernation we need to restore them on resume. Check if hv_reenlightenment_cb was previously set and restore from hv_resume().

---
  arch/x86/hyperv/hv_init.c | 19 +++++++++++++++++--
  1 file changed, 17 insertions(+), 2 deletions(-)
arch/x86/hyperv/hv_init.c +10 -2
index fd51bac11b46..acf76b466db6 100644
@@ -226,10 +226,18 @@ static int hv_cpu_die(unsigned int cpu)
 
  	rdmsrl(HV_X64_MSR_REENLIGHTENMENT_CONTROL, *((u64 *)&re_ctrl));
  	if (re_ctrl.target_vp == hv_vp_index[cpu]) {
-		/* Reassign to some other online CPU */
+		/*
+		 * Reassign reenlightenment notifications to some other online
+		 * CPU or just disable the feature if there are no online CPUs
+		 * left (happens on hibernation).
+		 */
  		new_cpu = cpumask_any_but(cpu_online_mask, cpu);
 
-		re_ctrl.target_vp = hv_vp_index[new_cpu];
+		if (new_cpu < nr_cpu_ids)
+			re_ctrl.target_vp = hv_vp_index[new_cpu];
+		else
+			re_ctrl.enabled = 0;
+
  		wrmsrl(HV_X64_MSR_REENLIGHTENMENT_CONTROL, *((u64 *)&re_ctrl));
  	}
 
@@ -293,6 +301,13 @@ static void hv_resume(void)
 
  	hv_hypercall_pg = hv_hypercall_pg_saved;
  	hv_hypercall_pg_saved = NULL;

arch/x86/hyperv/hv_init.c static void hv_resume(void) · line 313

Quoted 10 lines
arch/x86/hyperv/hv_init.c +7 -0
+
+	/*
+	 * Reenlightenment notifications are disabled by hv_cpu_die(0),
+	 * reenable them here if hv_reenlightenment_cb was previously set.
+	 */
+	if (hv_reenlightenment_cb)
+		set_hv_tscchange_cb(hv_reenlightenment_cb);
  }
 
  /* Note: when the ops are called, only CPU0 is online and IRQs are disabled. */
maintainer Applied
#3 lore raw
In reply to Vitaly Kuznetsov
Quoted 21 lines · Vitaly Kuznetsov

Errors during hibernation with reenlightenment notifications enabled were reported:

 [   51.730435] PM: hibernation entry
 [   51.737435] PM: Syncing filesystems ...
 ...
 [   54.102216] Disabling non-boot CPUs ...
 [   54.106633] smpboot: CPU 1 is now offline
 [   54.110006] unchecked MSR access error: WRMSR to 0x40000106 (tried to
     write 0x47c72780000100ee) at rIP: 0xffffffff90062f24
     native_write_msr+0x4/0x20)
 [   54.110006] Call Trace:
 [   54.110006]  hv_cpu_die+0xd9/0xf0
 ...

Normally, hv_cpu_die() just reassigns reenlightenment notifications to some other CPU when the CPU receiving them goes offline. Upon hibernation, there is no other CPU which is still online so cpumask_any_but(cpu_online_mask) returns >= nr_cpu_ids and using it as hv_vp_index index is incorrect. Disable the feature when cpumask_any_but() fails.

Quoted 5 lines

Also, as we now disable reenlightenment notifications upon hibernation we need to restore them on resume. Check if hv_reenlightenment_cb was previously set and restore from hv_resume().

Applied to hyperv-fixes.

Thank you all.

Wei.