Skip to contentSkip to conversation

perf: add cond_resched() to task_function_call()

PATCH Applied Barret Rhoden 4 participants 7 messages Apr 14, 2020 – May 23, 2020 perf events

lore
Share & tools
lore
#0 lore raw

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

---
 kernel/events/core.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
kernel/events/core.c +8 -4
index 55e44417f66d..65c2c05e24c2 100644
@@ -99,7 +99,7 @@ static void remote_function(void *data)
  *
  * returns: @func return value, or
  *	    -ESRCH  - when the process isn't running
- *	    -EAGAIN - when the process moved away
+ *	    -ENXIO  - when the cpu the process was on has gone offline
  */
 static int
 task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -112,11 +112,15 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
 	};
 	int ret;
 
-	do {
-		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);
+	while (1) {
+		ret = smp_call_function_single(task_cpu(p), remote_function,
+					       &data, 1);
 		if (!ret)
 			ret = data.ret;
-	} while (ret == -EAGAIN);
+		if (ret != -EAGAIN)
+			break;
+		cond_resched();
+	}
 
 	return ret;
 }
Signature · 1 line
2.26.0.110.g2183baf09c-goog
maintainer
#1 lore raw
Quoted 8 lines · Barret Rhoden

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

Sigh,.. virt again :/

Quoted 11 lines
---
 kernel/events/core.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
kernel/events/core.c +0 -0
index 55e44417f66d..65c2c05e24c2 100644

kernel/events/core.c static void remote_function(void *data) · line 103

Quoted 6 lines
kernel/events/core.c +1 -1
@@ -99,7 +99,7 @@ static void remote_function(void *data)
  *
  * returns: @func return value, or
  *	    -ESRCH  - when the process isn't running
- *	    -EAGAIN - when the process moved away
+ *	    -ENXIO  - when the cpu the process was on has gone offline
  */

Hurm.. I don't think that was actually intended behaviour. As long as the task lives we ought to retry. Luckily I don't think the current code cares much, it'll loop again on the caller side.

With the exception of perf_cgroup_attach() that is, that might actually be broken because of this.

Quoted 8 lines
kernel/events/core.c +0 -2
 static int
 task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -112,11 +112,15 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
 	};
 	int ret;
 
-	do {
-		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);

kernel/events/core.c task_function_call(struct task_struct *p, remote_function_f func, void *info) · line 123

Quoted 10 lines
kernel/events/core.c +7 -1
+	while (1) {
+		ret = smp_call_function_single(task_cpu(p), remote_function,
+					       &data, 1);
 		if (!ret)
 			ret = data.ret;
-	} while (ret == -EAGAIN);
+		if (ret != -EAGAIN)
+			break;
+		cond_resched();
+	}

So how about we make that:

	for (;;) {
		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);
		ret = !ret ? data.ret : -EAGAIN;

		if (ret != -EAGAIN)
			break;

		cond_resched();
	}

Or something like that, hmmm?

#2 lore raw
Quoted 18 lines · Peter Zijlstra
Quoted 8 lines · Barret Rhoden

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

Sigh,.. virt again :/

Quoted 6 lines
---
  kernel/events/core.c | 12 ++++++++----
  1 file changed, 8 insertions(+), 4 deletions(-)
Quoted 13 lines

kernel/events/core.c static void remote_function(void *data) · line 103

Quoted 11 lines
kernel/events/core.c +1 -1
index 55e44417f66d..65c2c05e24c2 100644
@@ -99,7 +99,7 @@ static void remote_function(void *data)
   *
   * returns: @func return value, or
   *	    -ESRCH  - when the process isn't running
- *	    -EAGAIN - when the process moved away
+ *	    -ENXIO  - when the cpu the process was on has gone offline
   */

Hurm.. I don't think that was actually intended behaviour.

Yeah, I noticed we never return EAGAIN, so I figured I'd look for any other potential return values.

Quoted 38 lines

As long as
the task lives we ought to retry. Luckily I don't think the current code cares much, it'll loop again on the caller side.

With the exception of perf_cgroup_attach() that is, that might actually be broken because of this.

Quoted 8 lines
kernel/events/core.c +0 -2
  static int
  task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -112,11 +112,15 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
  	};
  	int ret;
 
-	do {
-		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);

kernel/events/core.c task_function_call(struct task_struct *p, remote_function_f func, void *info) · line 123

Quoted 10 lines
kernel/events/core.c +7 -1
+	while (1) {
+		ret = smp_call_function_single(task_cpu(p), remote_function,
+					       &data, 1);
  		if (!ret)
  			ret = data.ret;
-	} while (ret == -EAGAIN);
+		if (ret != -EAGAIN)
+			break;
+		cond_resched();
+	}

So how about we make that:

	for (;;) {
		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);
		ret = !ret ? data.ret : -EAGAIN;

		if (ret != -EAGAIN)
			break;

		cond_resched();
	}
Quoted 1 line

Or something like that, hmmm?

Can do. I can put a comment in too. /* Retry due to any failure in smp_call_function */ or something.

Thanks,

Barret

#3 lore raw
In reply to Barret Rhoden

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

---
 kernel/events/core.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)
kernel/events/core.c +14 -9
index 55e44417f66d..bcafd7b992b3 100644
@@ -95,11 +95,11 @@ static void remote_function(void *data)
  * @info:	the function call argument
  *
  * Calls the function @func when the task is currently running. This might
- * be on the current CPU, which just calls the function directly
+ * be on the current CPU, which just calls the function directly.  This will
+ * retry due to any failures in smp_call_function_single(), such as if the
+ * task_cpu() goes offline concurrently.
  *
- * returns: @func return value, or
- *	    -ESRCH  - when the process isn't running
- *	    -EAGAIN - when the process moved away
+ * returns @func return value or -ESRCH when the process isn't running
  */
 static int
 task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -112,11 +112,16 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
 	};
 	int ret;
 
-	do {
-		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);
-		if (!ret)
-			ret = data.ret;
-	} while (ret == -EAGAIN);
+	for (;;) {
+		ret = smp_call_function_single(task_cpu(p), remote_function,
+					       &data, 1);
+		ret = !ret ? data.ret : -EAGAIN;
+
+		if (ret != -EAGAIN)
+			break;
+
+		cond_resched();
+	}
 
 	return ret;
 }
Signature · 1 line
2.26.0.110.g2183baf09c-goog
merge bot Applied
#4 lore raw

[tip: perf/core] perf: Add cond_resched() to task_function_call()

The following commit has been merged into the perf/core branch of tip:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     2ed6edd33a214bca02bd2b45e3fc3038a059436b
Gitweb:        https://git.kernel.org/tip/2ed6edd33a214bca02bd2b45e3fc3038a059436b
Author:        Barret Rhoden <[email protected]>
AuthorDate:    Tue, 14 Apr 2020 18:29:20 -04:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 30 Apr 2020 20:14:36 +02:00

perf: Add cond_resched() to task_function_call()

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

---
 kernel/events/core.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)
kernel/events/core.c +14 -9
index 52951e9..80cf996 100644
@@ -95,11 +95,11 @@ static void remote_function(void *data)
  * @info:	the function call argument
  *
  * Calls the function @func when the task is currently running. This might
- * be on the current CPU, which just calls the function directly
+ * be on the current CPU, which just calls the function directly.  This will
+ * retry due to any failures in smp_call_function_single(), such as if the
+ * task_cpu() goes offline concurrently.
  *
- * returns: @func return value, or
- *	    -ESRCH  - when the process isn't running
- *	    -EAGAIN - when the process moved away
+ * returns @func return value or -ESRCH when the process isn't running
  */
 static int
 task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -112,11 +112,16 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
 	};
 	int ret;
 
-	do {
-		ret = smp_call_function_single(task_cpu(p), remote_function, &data, 1);
-		if (!ret)
-			ret = data.ret;
-	} while (ret == -EAGAIN);
+	for (;;) {
+		ret = smp_call_function_single(task_cpu(p), remote_function,
+					       &data, 1);
+		ret = !ret ? data.ret : -EAGAIN;
+
+		if (ret != -EAGAIN)
+			break;
+
+		cond_resched();
+	}
 
 	return ret;
 }
#5 lore raw

Re: [tip: perf/core] perf: Add cond_resched() to task_function_call()

Quoted 20 lines · On Fri, May 01, 2020 at 06:22:05PM -0000, tip-bot2 for Barret Rhoden wrote:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     2ed6edd33a214bca02bd2b45e3fc3038a059436b
Gitweb:        https://git.kernel.org/tip/2ed6edd33a214bca02bd2b45e3fc3038a059436b
Author:        Barret Rhoden <[email protected]>
AuthorDate:    Tue, 14 Apr 2020 18:29:20 -04:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 30 Apr 2020 20:14:36 +02:00

perf: Add cond_resched() to task_function_call()

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

Quoted 4 lines

hi,
this gives me warning below and the machine gets stuck, if I revert it seems to be ok, running this:

  # while :; do ./perf stat -e cycles ./perf bench sched messaging; done

jirka

---

[ 1782.995609] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.004301] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10d
if_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_supp
ort ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm 
drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.049855] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G          I       5.7.0-rc6+ #57
[ 1783.057763] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.065250] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.070991] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb e
a 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.089736] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046
[ 1783.094962] RAX: 0000000000000000 RBX: ffff888c10eaab80 RCX: 0000000000000007
[ 1783.102095] RDX: ffffffff81108ab0 RSI: 0000000000000000 RDI: ffff888c10eaab40
[ 1783.109225] RBP: ffff888c10eaab80 R08: 0000019f228b1c77 R09: 0000000000000000
[ 1783.116359] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
[ 1783.123491] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 1783.130624] FS:  0000000000000000(0000) GS:ffff888c10e80000(0000) knlGS:0000000000000000
[ 1783.138711] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1783.144457] CR2: 00007fc8501b29f0 CR3: 0000000be6ac8001 CR4: 00000000007606e0
[ 1783.151589] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1783.158721] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 1783.165854] PKRU: 55555554
[ 1783.168566] Call Trace:
[ 1783.171019]  <IRQ>
[ 1783.173041]  ? irqtime_account_irq+0x32/0xa0
[ 1783.177317]  smp_call_function_single_interrupt+0x3e/0xf0
[ 1783.182712]  call_function_single_interrupt+0xf/0x20
[ 1783.187678] RIP: 0010:_nohz_idle_balance+0x57/0x210
[ 1783.192556] Code: 60 ea 00 00 89 f0 83 e0 03 89 74 24 28 89 54 24 2c 83 f8 01 0f 84 9a 01 00 00 c7 05 4b 5f e9 01 00 00 00 00 f0 83 44 24 fc 00 <65> 48 8b 0
4 25 c0 8b 01 00 48 89 04 24 41 bf ff ff ff ff 45 31 ed
[ 1783.211303] RSP: 0018:ffffc900068b0ed0 EFLAGS: 00000286 ORIG_RAX: ffffffffffffff04
[ 1783.218869] RAX: 0000000000000003 RBX: 00000001001780ff RCX: 0000000000000003
[ 1783.226000] RDX: 0000000000000000 RSI: 0000000000000003 RDI: ffff888c10eaab40
[ 1783.233131] RBP: 0000000000000014 R08: 0000000000000003 R09: 0000000000000100
[ 1783.240258] R10: 0000019ea6c791d9 R11: 0000000000000394 R12: ffffffff82805138
[ 1783.247390] R13: 0000000000000007 R14: 0000000000000007 R15: ffffffff82805110
[ 1783.254523]  ? call_function_single_interrupt+0xa/0x20
[ 1783.259667]  __do_softirq+0xee/0x313
[ 1783.263243]  ? clockevents_program_event+0x8d/0xf0
[ 1783.268036]  ? sched_clock+0x5/0x10
[ 1783.271531]  irq_exit+0xe6/0xf0
[ 1783.274675]  smp_apic_timer_interrupt+0x7a/0x150
[ 1783.279292]  apic_timer_interrupt+0xf/0x20
[ 1783.283391]  </IRQ>
[ 1783.285500] RIP: 0010:finish_task_switch+0x7b/0x290
[ 1783.290377] Code: 44 00 00 65 48 8b 1c 25 c0 8b 01 00 e9 89 00 00 00 0f 1f 44 00 00 41 c7 45 38 00 00 00 00 41 c6 04 24 00 fb 66 0f 1f 44 00 00 <65> 48 8b 0
4 25 c0 8b 01 00 0f 1f 44 00 00 4d 85 f6 74 21 65 48 8b
[ 1783.309122] RSP: 0018:ffffc9000649be28 EFLAGS: 00000282 ORIG_RAX: ffffffffffffff13
[ 1783.316687] RAX: 0000000000000000 RBX: ffff888c0c602580 RCX: 0000000000000000
[ 1783.323820] RDX: 0000000000000000 RSI: ffff888c0c602580 RDI: ffff888be6c48000
[ 1783.330953] RBP: ffffc9000649be50 R08: ffff888c10ea7c00 R09: ffff888be8a1e800
[ 1783.338085] R10: 0000000000000000 R11: 000000000000000c R12: ffff888c10eaab40
[ 1783.345219] R13: ffff888be6c48000 R14: 0000000000000000 R15: 0000000000000001
[ 1783.352354]  ? finish_task_switch+0xf6/0x290
[ 1783.356624]  __schedule+0x2c8/0x750
[ 1783.360118]  ? recalibrate_cpu_khz+0x10/0x10
[ 1783.364389]  schedule_idle+0x28/0x40
[ 1783.367970]  do_idle+0x169/0x260
[ 1783.371201]  cpu_startup_entry+0x19/0x20
[ 1783.375128]  start_secondary+0x169/0x1c0
[ 1783.379058]  secondary_startup_64+0xa4/0xb0
[ 1783.383240] ---[ end trace 5ec3d41605e082c2 ]---
[ 1783.387861] ------------[ cut here ]------------
[ 1783.392481] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.401169] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_support ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.446705] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G        W I       5.7.0-rc6+ #57
[ 1783.454618] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.462098] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.467842] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb ea 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.486587] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046
[ 1783.491814] RAX: 0000000000000000 RBX: ffff888c10eaab80 RCX: ffff888c10e80000
[ 1783.498948] RDX: ffffffff81108ab0 RSI: 0000000000000001 RDI: ffff888c10eaab40
[ 1783.506079] RBP: ffff888c10eaab80 R08: 0000019f228b1c77 R09: 0000000000000000
[ 1783.513212] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
[ 1783.520346] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 1783.527477] FS:  0000000000000000(0000) GS:ffff888c10e80000(0000) knlGS:0000000000000000
[ 1783.535564] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1783.541308] CR2: 00007fc8501b29f0 CR3: 0000000be6ac8001 CR4: 00000000007606e0
[ 1783.548435] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1783.555566] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 1783.562699] PKRU: 55555554
[ 1783.565411] Call Trace:
[ 1783.567863]  <IRQ>
[ 1783.569883]  ? irqtime_account_irq+0x32/0xa0
[ 1783.574157]  smp_call_function_single_interrupt+0x3e/0xf0
[ 1783.579556]  call_function_single_interrupt+0xf/0x20
[ 1783.584522] RIP: 0010:_nohz_idle_balance+0x57/0x210
[ 1783.589401] Code: 60 ea 00 00 89 f0 83 e0 03 89 74 24 28 89 54 24 2c 83 f8 01 0f 84 9a 01 00 00 c7 05 4b 5f e9 01 00 00 00 00 f0 83 44 24 fc 00 <65> 48 8b 04 25 c0 8b 01 00 48 89 04 24 41 bf ff ff ff ff 45 31 ed
[ 1783.608146] RSP: 0018:ffffc900068b0ed0 EFLAGS: 00000286 ORIG_RAX: ffffffffffffff04
[ 1783.615713] RAX: 0000000000000003 RBX: 00000001001780ff RCX: 0000000000000003
[ 1783.622844] RDX: 0000000000000000 RSI: 0000000000000003 RDI: ffff888c10eaab40
[ 1783.629978] RBP: 0000000000000014 R08: 0000000000000003 R09: 0000000000000100
[ 1783.637111] R10: 0000019ea6c791d9 R11: 0000000000000394 R12: ffffffff82805138
[ 1783.644243] R13: 0000000000000007 R14: 0000000000000007 R15: ffffffff82805110
[ 1783.651378]  ? call_function_single_interrupt+0xa/0x20
[ 1783.656519]  __do_softirq+0xee/0x313
[ 1783.660094]  ? clockevents_program_event+0x8d/0xf0
[ 1783.664887]  ? sched_clock+0x5/0x10
[ 1783.668382]  irq_exit+0xe6/0xf0
[ 1783.671528]  smp_apic_timer_interrupt+0x7a/0x150
[ 1783.676146]  apic_timer_interrupt+0xf/0x20
[ 1783.680243]  </IRQ>
[ 1783.682351] RIP: 0010:finish_task_switch+0x7b/0x290
[ 1783.687231] Code: 44 00 00 65 48 8b 1c 25 c0 8b 01 00 e9 89 00 00 00 0f 1f 44 00 00 41 c7 45 38 00 00 00 00 41 c6 04 24 00 fb 66 0f 1f 44 00 00 <65> 48 8b 04 25 c0 8b 01 00 0f 1f 44 00 00 4d 85 f6 74 21 65 48 8b
[ 1783.705974] RSP: 0018:ffffc9000649be28 EFLAGS: 00000282 ORIG_RAX: ffffffffffffff13
[ 1783.713542] RAX: 0000000000000000 RBX: ffff888c0c602580 RCX: 0000000000000000
[ 1783.720673] RDX: 0000000000000000 RSI: ffff888c0c602580 RDI: ffff888be6c48000
[ 1783.727807] RBP: ffffc9000649be50 R08: ffff888c10ea7c00 R09: ffff888be8a1e800
[ 1783.734938] R10: 0000000000000000 R11: 000000000000000c R12: ffff888c10eaab40
[ 1783.742070] R13: ffff888be6c48000 R14: 0000000000000000 R15: 0000000000000001
[ 1783.749199]  ? finish_task_switch+0xf6/0x290
[ 1783.753469]  __schedule+0x2c8/0x750
[ 1783.756961]  ? recalibrate_cpu_khz+0x10/0x10
[ 1783.761234]  schedule_idle+0x28/0x40
[ 1783.764814]  do_idle+0x169/0x260
[ 1783.768047]  cpu_startup_entry+0x19/0x20
[ 1783.771973]  start_secondary+0x169/0x1c0
[ 1783.775897]  secondary_startup_64+0xa4/0xb0
[ 1783.780085] ---[ end trace 5ec3d41605e082c3 ]---
[ 1783.784704] ------------[ cut here ]------------
[ 1783.789323] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.798015] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_support ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.843543] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G        W I       5.7.0-rc6+ #57
[ 1783.851452] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.858933] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.864680] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb ea 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.883425] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046

...

#6 lore raw

Re: [tip: perf/core] perf: Add cond_resched() to task_function_call()

Quoted 26 lines · Jiri Olsa
Quoted 20 lines · On Fri, May 01, 2020 at 06:22:05PM -0000, tip-bot2 for Barret Rhoden wrote:

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     2ed6edd33a214bca02bd2b45e3fc3038a059436b
Gitweb:        https://git.kernel.org/tip/2ed6edd33a214bca02bd2b45e3fc3038a059436b
Author:        Barret Rhoden <[email protected]>
AuthorDate:    Tue, 14 Apr 2020 18:29:20 -04:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 30 Apr 2020 20:14:36 +02:00

perf: Add cond_resched() to task_function_call()

Under rare circumstances, task_function_call() can repeatedly fail and cause a soft lockup.

There is a slight race where the process is no longer running on the cpu we targeted by the time remote_function() runs. The code will simply try again. If we are very unlucky, this will continue to fail, until a watchdog fires. This can happen in a heavily loaded, multi-core virtual machine.

Quoted 4 lines
Quoted 5 lines

hi,
this gives me warning below and the machine gets stuck, if I revert it seems to be ok, running this:

  # while :; do ./perf stat -e cycles ./perf bench sched messaging; done

ugh, I take it back.. it got stuck as well after longer time

jirka

Quoted 133 lines

jirka

---

[ 1782.995609] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.004301] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10d
if_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_supp
ort ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm 
drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.049855] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G          I       5.7.0-rc6+ #57
[ 1783.057763] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.065250] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.070991] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb e
a 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.089736] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046
[ 1783.094962] RAX: 0000000000000000 RBX: ffff888c10eaab80 RCX: 0000000000000007
[ 1783.102095] RDX: ffffffff81108ab0 RSI: 0000000000000000 RDI: ffff888c10eaab40
[ 1783.109225] RBP: ffff888c10eaab80 R08: 0000019f228b1c77 R09: 0000000000000000
[ 1783.116359] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
[ 1783.123491] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 1783.130624] FS:  0000000000000000(0000) GS:ffff888c10e80000(0000) knlGS:0000000000000000
[ 1783.138711] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1783.144457] CR2: 00007fc8501b29f0 CR3: 0000000be6ac8001 CR4: 00000000007606e0
[ 1783.151589] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1783.158721] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 1783.165854] PKRU: 55555554
[ 1783.168566] Call Trace:
[ 1783.171019]  <IRQ>
[ 1783.173041]  ? irqtime_account_irq+0x32/0xa0
[ 1783.177317]  smp_call_function_single_interrupt+0x3e/0xf0
[ 1783.182712]  call_function_single_interrupt+0xf/0x20
[ 1783.187678] RIP: 0010:_nohz_idle_balance+0x57/0x210
[ 1783.192556] Code: 60 ea 00 00 89 f0 83 e0 03 89 74 24 28 89 54 24 2c 83 f8 01 0f 84 9a 01 00 00 c7 05 4b 5f e9 01 00 00 00 00 f0 83 44 24 fc 00 <65> 48 8b 0
4 25 c0 8b 01 00 48 89 04 24 41 bf ff ff ff ff 45 31 ed
[ 1783.211303] RSP: 0018:ffffc900068b0ed0 EFLAGS: 00000286 ORIG_RAX: ffffffffffffff04
[ 1783.218869] RAX: 0000000000000003 RBX: 00000001001780ff RCX: 0000000000000003
[ 1783.226000] RDX: 0000000000000000 RSI: 0000000000000003 RDI: ffff888c10eaab40
[ 1783.233131] RBP: 0000000000000014 R08: 0000000000000003 R09: 0000000000000100
[ 1783.240258] R10: 0000019ea6c791d9 R11: 0000000000000394 R12: ffffffff82805138
[ 1783.247390] R13: 0000000000000007 R14: 0000000000000007 R15: ffffffff82805110
[ 1783.254523]  ? call_function_single_interrupt+0xa/0x20
[ 1783.259667]  __do_softirq+0xee/0x313
[ 1783.263243]  ? clockevents_program_event+0x8d/0xf0
[ 1783.268036]  ? sched_clock+0x5/0x10
[ 1783.271531]  irq_exit+0xe6/0xf0
[ 1783.274675]  smp_apic_timer_interrupt+0x7a/0x150
[ 1783.279292]  apic_timer_interrupt+0xf/0x20
[ 1783.283391]  </IRQ>
[ 1783.285500] RIP: 0010:finish_task_switch+0x7b/0x290
[ 1783.290377] Code: 44 00 00 65 48 8b 1c 25 c0 8b 01 00 e9 89 00 00 00 0f 1f 44 00 00 41 c7 45 38 00 00 00 00 41 c6 04 24 00 fb 66 0f 1f 44 00 00 <65> 48 8b 0
4 25 c0 8b 01 00 0f 1f 44 00 00 4d 85 f6 74 21 65 48 8b
[ 1783.309122] RSP: 0018:ffffc9000649be28 EFLAGS: 00000282 ORIG_RAX: ffffffffffffff13
[ 1783.316687] RAX: 0000000000000000 RBX: ffff888c0c602580 RCX: 0000000000000000
[ 1783.323820] RDX: 0000000000000000 RSI: ffff888c0c602580 RDI: ffff888be6c48000
[ 1783.330953] RBP: ffffc9000649be50 R08: ffff888c10ea7c00 R09: ffff888be8a1e800
[ 1783.338085] R10: 0000000000000000 R11: 000000000000000c R12: ffff888c10eaab40
[ 1783.345219] R13: ffff888be6c48000 R14: 0000000000000000 R15: 0000000000000001
[ 1783.352354]  ? finish_task_switch+0xf6/0x290
[ 1783.356624]  __schedule+0x2c8/0x750
[ 1783.360118]  ? recalibrate_cpu_khz+0x10/0x10
[ 1783.364389]  schedule_idle+0x28/0x40
[ 1783.367970]  do_idle+0x169/0x260
[ 1783.371201]  cpu_startup_entry+0x19/0x20
[ 1783.375128]  start_secondary+0x169/0x1c0
[ 1783.379058]  secondary_startup_64+0xa4/0xb0
[ 1783.383240] ---[ end trace 5ec3d41605e082c2 ]---
[ 1783.387861] ------------[ cut here ]------------
[ 1783.392481] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.401169] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_support ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.446705] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G        W I       5.7.0-rc6+ #57
[ 1783.454618] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.462098] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.467842] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb ea 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.486587] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046
[ 1783.491814] RAX: 0000000000000000 RBX: ffff888c10eaab80 RCX: ffff888c10e80000
[ 1783.498948] RDX: ffffffff81108ab0 RSI: 0000000000000001 RDI: ffff888c10eaab40
[ 1783.506079] RBP: ffff888c10eaab80 R08: 0000019f228b1c77 R09: 0000000000000000
[ 1783.513212] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
[ 1783.520346] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[ 1783.527477] FS:  0000000000000000(0000) GS:ffff888c10e80000(0000) knlGS:0000000000000000
[ 1783.535564] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1783.541308] CR2: 00007fc8501b29f0 CR3: 0000000be6ac8001 CR4: 00000000007606e0
[ 1783.548435] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1783.555566] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 1783.562699] PKRU: 55555554
[ 1783.565411] Call Trace:
[ 1783.567863]  <IRQ>
[ 1783.569883]  ? irqtime_account_irq+0x32/0xa0
[ 1783.574157]  smp_call_function_single_interrupt+0x3e/0xf0
[ 1783.579556]  call_function_single_interrupt+0xf/0x20
[ 1783.584522] RIP: 0010:_nohz_idle_balance+0x57/0x210
[ 1783.589401] Code: 60 ea 00 00 89 f0 83 e0 03 89 74 24 28 89 54 24 2c 83 f8 01 0f 84 9a 01 00 00 c7 05 4b 5f e9 01 00 00 00 00 f0 83 44 24 fc 00 <65> 48 8b 04 25 c0 8b 01 00 48 89 04 24 41 bf ff ff ff ff 45 31 ed
[ 1783.608146] RSP: 0018:ffffc900068b0ed0 EFLAGS: 00000286 ORIG_RAX: ffffffffffffff04
[ 1783.615713] RAX: 0000000000000003 RBX: 00000001001780ff RCX: 0000000000000003
[ 1783.622844] RDX: 0000000000000000 RSI: 0000000000000003 RDI: ffff888c10eaab40
[ 1783.629978] RBP: 0000000000000014 R08: 0000000000000003 R09: 0000000000000100
[ 1783.637111] R10: 0000019ea6c791d9 R11: 0000000000000394 R12: ffffffff82805138
[ 1783.644243] R13: 0000000000000007 R14: 0000000000000007 R15: ffffffff82805110
[ 1783.651378]  ? call_function_single_interrupt+0xa/0x20
[ 1783.656519]  __do_softirq+0xee/0x313
[ 1783.660094]  ? clockevents_program_event+0x8d/0xf0
[ 1783.664887]  ? sched_clock+0x5/0x10
[ 1783.668382]  irq_exit+0xe6/0xf0
[ 1783.671528]  smp_apic_timer_interrupt+0x7a/0x150
[ 1783.676146]  apic_timer_interrupt+0xf/0x20
[ 1783.680243]  </IRQ>
[ 1783.682351] RIP: 0010:finish_task_switch+0x7b/0x290
[ 1783.687231] Code: 44 00 00 65 48 8b 1c 25 c0 8b 01 00 e9 89 00 00 00 0f 1f 44 00 00 41 c7 45 38 00 00 00 00 41 c6 04 24 00 fb 66 0f 1f 44 00 00 <65> 48 8b 04 25 c0 8b 01 00 0f 1f 44 00 00 4d 85 f6 74 21 65 48 8b
[ 1783.705974] RSP: 0018:ffffc9000649be28 EFLAGS: 00000282 ORIG_RAX: ffffffffffffff13
[ 1783.713542] RAX: 0000000000000000 RBX: ffff888c0c602580 RCX: 0000000000000000
[ 1783.720673] RDX: 0000000000000000 RSI: ffff888c0c602580 RDI: ffff888be6c48000
[ 1783.727807] RBP: ffffc9000649be50 R08: ffff888c10ea7c00 R09: ffff888be8a1e800
[ 1783.734938] R10: 0000000000000000 R11: 000000000000000c R12: ffff888c10eaab40
[ 1783.742070] R13: ffff888be6c48000 R14: 0000000000000000 R15: 0000000000000001
[ 1783.749199]  ? finish_task_switch+0xf6/0x290
[ 1783.753469]  __schedule+0x2c8/0x750
[ 1783.756961]  ? recalibrate_cpu_khz+0x10/0x10
[ 1783.761234]  schedule_idle+0x28/0x40
[ 1783.764814]  do_idle+0x169/0x260
[ 1783.768047]  cpu_startup_entry+0x19/0x20
[ 1783.771973]  start_secondary+0x169/0x1c0
[ 1783.775897]  secondary_startup_64+0xa4/0xb0
[ 1783.780085] ---[ end trace 5ec3d41605e082c3 ]---
[ 1783.784704] ------------[ cut here ]------------
[ 1783.789323] WARNING: CPU: 20 PID: 0 at kernel/smp.c:127 flush_smp_call_function_queue+0x86/0xf0
[ 1783.798015] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul ipmi_ssif ghash_clmulni_intel intel_cstate dell_smbios intel_uncore dell_wmi_descriptor iTCO_wdt dcdbas mei_me wmi_bmof iTCO_vendor_support ipmi_si mei lpc_ich ipmi_devintf i2c_i801 wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper cec drm_vram_helper drm_ttm_helper ttm drm tg3 megaraid_sas crc32c_intel i2c_algo_bit
[ 1783.843543] CPU: 20 PID: 0 Comm: swapper/20 Tainted: G        W I       5.7.0-rc6+ #57
[ 1783.851452] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018
[ 1783.858933] RIP: 0010:flush_smp_call_function_queue+0x86/0xf0
[ 1783.864680] Code: 00 00 48 85 ed 75 d3 48 83 c4 08 5b 5d 41 5c e9 80 06 07 00 a8 01 74 12 c7 43 18 00 00 00 00 e8 d0 24 c8 00 eb af 0f 0b eb d1 <0f> 0b eb ea 65 8b 05 9f 3b e9 7e 89 c0 48 0f a3 05 c5 a8 7c 01 72
[ 1783.883425] RSP: 0018:ffffc900068b0dd0 EFLAGS: 00010046

...