sched/cputime: Fix getrusage(RUSAGE_THREAD) with nohz_full

PATCH NAK Sasha Levin 2 participants 7 messages Dec 13, 2021

Subsystems (3)

Stable branch KVM Kselftest

lore

Share & tools

lore

Sasha Levin

author maintainer

2021-12-13 14:20 UTC #0 lore raw

PATCH 1/4 sched/cputime: Fix getrusage(RUSAGE_THREAD) with nohz_full

From: Frederic Weisbecker <[email protected]>

[ Upstream commit e7f2be115f0746b969c0df14c0d182f65f005ca5 ]

getrusage(RUSAGE_THREAD) with nohz_full may return shorter utime/stime than the actual time.

task_cputime_adjusted() snapshots utime and stime and then adjust their sum to match the scheduler maintained cputime.sum_exec_runtime. Unfortunately in nohz_full, sum_exec_runtime is only updated once per second in the worst case, causing a discrepancy against utime and stime that can be updated anytime by the reader using vtime.

To fix this situation, perform an update of cputime.sum_exec_runtime when the cputime snapshot reports the task as actually running while the tick is disabled. The related overhead is then contained within the relevant situations.

Reported-by Hasegawa Hitomi <[email protected]>

Signed-off-by Frederic Weisbecker <[email protected]>

Signed-off-by Hasegawa Hitomi <[email protected]>

Signed-off-by Thomas Gleixner <[email protected]>

Tested-by Masayoshi Mizuma <[email protected]>

Acked-by Phil Auld <[email protected]>

Link https://lore.kernel.org/r/[email protected] open here

Signed-off-by Sasha Levin <[email protected]>

---
 include/linux/sched/cputime.h |  5 +++--
 kernel/sched/cputime.c        | 12 +++++++++---
 2 files changed, 12 insertions(+), 5 deletions(-)

include/linux/sched/cputime.h +3 -2

index 6c9f19a33865a..ce3c58286062c 100644
@@ -18,15 +18,16 @@
 #endif /* CONFIG_VIRT_CPU_ACCOUNTING_NATIVE */
 
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
-extern void task_cputime(struct task_struct *t,
+extern bool task_cputime(struct task_struct *t,
 			 u64 *utime, u64 *stime);
 extern u64 task_gtime(struct task_struct *t);
 #else
-static inline void task_cputime(struct task_struct *t,
+static inline bool task_cputime(struct task_struct *t,
 				u64 *utime, u64 *stime)
 {
 	*utime = t->utime;
 	*stime = t->stime;
+	return false;
 }
 
 static inline u64 task_gtime(struct task_struct *t)

kernel/sched/cputime.c +9 -3

index 5a55d23004524..9cd34730cf623 100644
@@ -617,7 +617,8 @@ void task_cputime_adjusted(struct task_struct *p, u64 *ut, u64 *st)
 		.sum_exec_runtime = p->se.sum_exec_runtime,
 	};
 
-	task_cputime(p, &cputime.utime, &cputime.stime);
+	if (task_cputime(p, &cputime.utime, &cputime.stime))
+		cputime.sum_exec_runtime = task_sched_runtime(p);
 	cputime_adjust(&cputime, &p->prev_cputime, ut, st);
 }
 EXPORT_SYMBOL_GPL(task_cputime_adjusted);

@@ -830,19 +831,21 @@ u64 task_gtime(struct task_struct *t)
  * add up the pending nohz execution time since the last
  * cputime snapshot.
  */
-void task_cputime(struct task_struct *t, u64 *utime, u64 *stime)
+bool task_cputime(struct task_struct *t, u64 *utime, u64 *stime)
 {
 	struct vtime *vtime = &t->vtime;
 	unsigned int seq;
 	u64 delta;
+	int ret;
 
 	if (!vtime_accounting_enabled()) {
 		*utime = t->utime;
 		*stime = t->stime;
-		return;
+		return false;
 	}
 
 	do {
+		ret = false;
 		seq = read_seqcount_begin(&vtime->seqcount);
 
 		*utime = t->utime;

@@ -852,6 +855,7 @@ void task_cputime(struct task_struct *t, u64 *utime, u64 *stime)
 		if (vtime->state < VTIME_SYS)
 			continue;
 
+		ret = true;
 		delta = vtime_delta(vtime);
 
 		/*

@@ -863,6 +867,8 @@ void task_cputime(struct task_struct *t, u64 *utime, u64 *stime)
 		else
 			*utime += vtime->utime + delta;
 	} while (read_seqcount_retry(&vtime->seqcount, seq));
+
+	return ret;
 }
 
 static int vtime_state_fetch(struct vtime *vtime, int cpu)

Signature · 1 line

2.33.0

Sasha Levin

author maintainer

2021-12-13 14:20 UTC #1 lore raw

PATCH 2/4 KVM: VMX: clear vmx_x86_ops.sync_pir_to_irr if APICv is disabled

From: Paolo Bonzini <[email protected]>

[ Upstream commit e90e51d5f01d2baae5dcce280866bbb96816e978 ]

There is nothing to synchronize if APICv is disabled, since neither other vCPUs nor assigned devices can set PIR.ON.

Signed-off-by Paolo Bonzini <[email protected]>

Signed-off-by Sasha Levin <[email protected]>

---
 arch/x86/kvm/vmx/vmx.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

arch/x86/kvm/vmx/vmx.c +2 -2

index 5b7664d51dc2b..dff8ab5a53280 100644
@@ -7814,10 +7814,10 @@ static __init int hardware_setup(void)
 		ple_window_shrink = 0;
 	}
 
-	if (!cpu_has_vmx_apicv()) {
+	if (!cpu_has_vmx_apicv())
 		enable_apicv = 0;
+	if (!enable_apicv)
 		vmx_x86_ops.sync_pir_to_irr = NULL;
-	}
 
 	if (cpu_has_vmx_tsc_scaling()) {
 		kvm_has_tsc_control = true;

Signature · 1 line

2.33.0

Paolo Bonzini

maintainer NAK

2021-12-13 14:29 UTC #2 lore raw

Re: [PATCH MANUALSEL 5.10 2/4] KVM: VMX: clear vmx_x86_ops.sync_pir_to_irr if APICv is disabled

Quoted 21 lines · Sasha Levin

From: Paolo Bonzini <[email protected]>
[ Upstream commit e90e51d5f01d2baae5dcce280866bbb96816e978 ]
There is nothing to synchronize if APICv is disabled, since neither other vCPUs nor assigned devices can set PIR.ON.
Signed-off-by Paolo Bonzini <[email protected]>
Signed-off-by Sasha Levin <[email protected]>
---
  arch/x86/kvm/vmx/vmx.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/kvm/vmx/vmx.c +0 -0
index 5b7664d51dc2b..dff8ab5a53280 100644
@@ -7814,10 +7814,10 @@ static __init int hardware_setup(void)
  		ple_window_shrink = 0;
  	}

arch/x86/kvm/vmx/vmx.c static __init int hardware_setup(void) · line 7823

Quoted 10 lines

arch/x86/kvm/vmx/vmx.c +2 -2

 
-	if (!cpu_has_vmx_apicv()) {
+	if (!cpu_has_vmx_apicv())
  		enable_apicv = 0;
+	if (!enable_apicv)
  		vmx_x86_ops.sync_pir_to_irr = NULL;
-	}
 
  	if (cpu_has_vmx_tsc_scaling()) {
  		kvm_has_tsc_control = true;

NACK - the patch is only okay to backport for 5.15

Paolo

Sasha Levin

author maintainer

2021-12-13 14:20 UTC #3 lore raw

PATCH 3/4 KVM: selftests: Make sure kvm_create_max_vcpus test won't hit RLIMIT_NOFILE

In reply to Sasha Levin

From: Vitaly Kuznetsov <[email protected]>

[ Upstream commit 908fa88e420f30dde6d80f092795a18ec72ca6d3 ]

With the elevated 'KVM_CAP_MAX_VCPUS' value kvm_create_max_vcpus test may hit RLIMIT_NOFILE limits:

 # ./kvm_create_max_vcpus
 KVM_CAP_MAX_VCPU_ID: 4096
 KVM_CAP_MAX_VCPUS: 1024
 Testing creating 1024 vCPUs, with IDs 0...1023.
 /dev/kvm not available (errno: 24), skipping test

Adjust RLIMIT_NOFILE limits to make sure KVM_CAP_MAX_VCPUS fds can be opened. Note, raising hard limit ('rlim_max') requires CAP_SYS_RESOURCE capability which is generally not needed to run kvm selftests (but without raising the limit the test is doomed to fail anyway).

Signed-off-by Vitaly Kuznetsov <[email protected]>

Message-Id: <[email protected]>

[Skip the test if the hard limit can be raised. - Paolo]

Reviewed-by Sean Christopherson <[email protected]>

Tested-by Sean Christopherson <[email protected]>

Signed-off-by Paolo Bonzini <[email protected]>

Signed-off-by Sasha Levin <[email protected]>

---
 .../selftests/kvm/kvm_create_max_vcpus.c      | 30 +++++++++++++++++++
 1 file changed, 30 insertions(+)

tools/testing/selftests/kvm/kvm_create_max_vcpus.c +30 -0

index 0299cd81b8ba2..aa3795cd7bd3d 100644
@@ -12,6 +12,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <sys/resource.h>
 
 #include "test_util.h"

@@ -40,10 +41,39 @@ int main(int argc, char *argv[])
 {
 	int kvm_max_vcpu_id = kvm_check_cap(KVM_CAP_MAX_VCPU_ID);
 	int kvm_max_vcpus = kvm_check_cap(KVM_CAP_MAX_VCPUS);
+	/*
+	 * Number of file descriptors reqired, KVM_CAP_MAX_VCPUS for vCPU fds +
+	 * an arbitrary number for everything else.
+	 */
+	int nr_fds_wanted = kvm_max_vcpus + 100;
+	struct rlimit rl;
 
 	pr_info("KVM_CAP_MAX_VCPU_ID: %d\n", kvm_max_vcpu_id);
 	pr_info("KVM_CAP_MAX_VCPUS: %d\n", kvm_max_vcpus);
 
+	/*
+	 * Check that we're allowed to open nr_fds_wanted file descriptors and
+	 * try raising the limits if needed.
+	 */
+	TEST_ASSERT(!getrlimit(RLIMIT_NOFILE, &rl), "getrlimit() failed!");
+
+	if (rl.rlim_cur < nr_fds_wanted) {
+		rl.rlim_cur = nr_fds_wanted;
+		if (rl.rlim_max < nr_fds_wanted) {
+			int old_rlim_max = rl.rlim_max;
+			rl.rlim_max = nr_fds_wanted;
+
+			int r = setrlimit(RLIMIT_NOFILE, &rl);
+			if (r < 0) {
+				printf("RLIMIT_NOFILE hard limit is too low (%d, wanted %d)\n",
+				       old_rlim_max, nr_fds_wanted);
+				exit(KSFT_SKIP);
+			}
+		} else {
+			TEST_ASSERT(!setrlimit(RLIMIT_NOFILE, &rl), "setrlimit() failed!");
+		}
+	}
+
 	/*
 	 * Upstream KVM prior to 4.8 does not support KVM_CAP_MAX_VCPU_ID.
 	 * Userspace is supposed to use KVM_CAP_MAX_VCPUS as the maximum ID

Signature · 1 line

2.33.0

Paolo Bonzini

maintainer Acked-by on v1 3/4

2021-12-13 14:22 UTC #4 lore raw

Re: [PATCH MANUALSEL 5.10 3/4] KVM: selftests: Make sure kvm_create_max_vcpus test won't hit RLIMIT_NOFILE

Quoted 72 lines · Sasha Levin

From: Vitaly Kuznetsov <[email protected]>
[ Upstream commit 908fa88e420f30dde6d80f092795a18ec72ca6d3 ]
With the elevated 'KVM_CAP_MAX_VCPUS' value kvm_create_max_vcpus test may hit RLIMIT_NOFILE limits:
  # ./kvm_create_max_vcpus
  KVM_CAP_MAX_VCPU_ID: 4096
  KVM_CAP_MAX_VCPUS: 1024
  Testing creating 1024 vCPUs, with IDs 0...1023.
  /dev/kvm not available (errno: 24), skipping test
Adjust RLIMIT_NOFILE limits to make sure KVM_CAP_MAX_VCPUS fds can be opened. Note, raising hard limit ('rlim_max') requires CAP_SYS_RESOURCE capability which is generally not needed to run kvm selftests (but without raising the limit the test is doomed to fail anyway).
Signed-off-by Vitaly Kuznetsov <[email protected]>
Message-Id: <[email protected]>
[Skip the test if the hard limit can be raised. - Paolo]
Reviewed-by Sean Christopherson <[email protected]>
Tested-by Sean Christopherson <[email protected]>
Signed-off-by Paolo Bonzini <[email protected]>
Signed-off-by Sasha Levin <[email protected]>
---
  .../selftests/kvm/kvm_create_max_vcpus.c      | 30 +++++++++++++++++++
  1 file changed, 30 insertions(+)
tools/testing/selftests/kvm/kvm_create_max_vcpus.c +23 -0
index 0299cd81b8ba2..aa3795cd7bd3d 100644
@@ -12,6 +12,7 @@
  #include <stdio.h>
  #include <stdlib.h>
  #include <string.h>
+#include <sys/resource.h>
 
  #include "test_util.h"
 
@@ -40,10 +41,39 @@ int main(int argc, char *argv[])
  {
  	int kvm_max_vcpu_id = kvm_check_cap(KVM_CAP_MAX_VCPU_ID);
  	int kvm_max_vcpus = kvm_check_cap(KVM_CAP_MAX_VCPUS);
+	/*
+	 * Number of file descriptors reqired, KVM_CAP_MAX_VCPUS for vCPU fds +
+	 * an arbitrary number for everything else.
+	 */
+	int nr_fds_wanted = kvm_max_vcpus + 100;
+	struct rlimit rl;
 
  	pr_info("KVM_CAP_MAX_VCPU_ID: %d\n", kvm_max_vcpu_id);
  	pr_info("KVM_CAP_MAX_VCPUS: %d\n", kvm_max_vcpus);
 
+	/*
+	 * Check that we're allowed to open nr_fds_wanted file descriptors and
+	 * try raising the limits if needed.
+	 */
+	TEST_ASSERT(!getrlimit(RLIMIT_NOFILE, &rl), "getrlimit() failed!");
+
+	if (rl.rlim_cur < nr_fds_wanted) {
+		rl.rlim_cur = nr_fds_wanted;
+		if (rl.rlim_max < nr_fds_wanted) {
+			int old_rlim_max = rl.rlim_max;
+			rl.rlim_max = nr_fds_wanted;
+
+			int r = setrlimit(RLIMIT_NOFILE, &rl);
+			if (r < 0) {
+				printf("RLIMIT_NOFILE hard limit is too low (%d, wanted %d)\n",
+				       old_rlim_max, nr_fds_wanted);

tools/testing/selftests/kvm/kvm_create_max_vcpus.c int main(int argc, char *argv[]) · line 79

Quoted 10 lines

tools/testing/selftests/kvm/kvm_create_max_vcpus.c +7 -0

+				exit(KSFT_SKIP);
+			}
+		} else {
+			TEST_ASSERT(!setrlimit(RLIMIT_NOFILE, &rl), "setrlimit() failed!");
+		}
+	}
+
  	/*
  	 * Upstream KVM prior to 4.8 does not support KVM_CAP_MAX_VCPU_ID.
  	 * Userspace is supposed to use KVM_CAP_MAX_VCPUS as the maximum ID

Acked-by Paolo Bonzini <[email protected]>

Sasha Levin

author maintainer

2021-12-13 14:20 UTC #5 lore raw

PATCH 4/4 KVM: downgrade two BUG_ONs to WARN_ON_ONCE

In reply to Sasha Levin

From: Paolo Bonzini <[email protected]>

[ Upstream commit 5f25e71e311478f9bb0a8ef49e7d8b95316491d7 ]

This is not an unrecoverable situation. Users of kvm_read_guest_offset_cached and kvm_write_guest_offset_cached must expect the read/write to fail, and therefore it is possible to just return early with an error value.

Signed-off-by Paolo Bonzini <[email protected]>

Signed-off-by Sasha Levin <[email protected]>

---
 virt/kvm/kvm_main.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

virt/kvm/kvm_main.c +4 -2

index 97ac3c6fd4441..4a7d377b3a500 100644
@@ -2590,7 +2590,8 @@ int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 	int r;
 	gpa_t gpa = ghc->gpa + offset;
 
-	BUG_ON(len + offset > ghc->len);
+	if (WARN_ON_ONCE(len + offset > ghc->len))
+		return -EINVAL;
 
 	if (slots->generation != ghc->generation) {
 		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))

@@ -2627,7 +2628,8 @@ int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
 	int r;
 	gpa_t gpa = ghc->gpa + offset;
 
-	BUG_ON(len + offset > ghc->len);
+	if (WARN_ON_ONCE(len + offset > ghc->len))
+		return -EINVAL;
 
 	if (slots->generation != ghc->generation) {
 		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))

Signature · 1 line

2.33.0

Paolo Bonzini

maintainer Acked-by on v1 4/4

2021-12-13 14:22 UTC #6 lore raw

Re: [PATCH MANUALSEL 5.10 4/4] KVM: downgrade two BUG_ONs to WARN_ON_ONCE

Quoted 30 lines · Sasha Levin

From: Paolo Bonzini <[email protected]>
[ Upstream commit 5f25e71e311478f9bb0a8ef49e7d8b95316491d7 ]
This is not an unrecoverable situation. Users of kvm_read_guest_offset_cached and kvm_write_guest_offset_cached must expect the read/write to fail, and therefore it is possible to just return early with an error value.
Signed-off-by Paolo Bonzini <[email protected]>
Signed-off-by Sasha Levin <[email protected]>
---
  virt/kvm/kvm_main.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
virt/kvm/kvm_main.c +2 -1
index 97ac3c6fd4441..4a7d377b3a500 100644
@@ -2590,7 +2590,8 @@ int kvm_write_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
  	int r;
  	gpa_t gpa = ghc->gpa + offset;
 
-	BUG_ON(len + offset > ghc->len);
+	if (WARN_ON_ONCE(len + offset > ghc->len))
+		return -EINVAL;
 
  	if (slots->generation != ghc->generation) {
  		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))

virt/kvm/kvm_main.c int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, · line 2635

Quoted 9 lines

virt/kvm/kvm_main.c +2 -1

@@ -2627,7 +2628,8 @@ int kvm_read_guest_offset_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
  	int r;
  	gpa_t gpa = ghc->gpa + offset;
 
-	BUG_ON(len + offset > ghc->len);
+	if (WARN_ON_ONCE(len + offset > ghc->len))
+		return -EINVAL;
 
  	if (slots->generation != ghc->generation) {
  		if (__kvm_gfn_to_hva_cache_init(slots, ghc, ghc->gpa, ghc->len))

Acked-by Paolo Bonzini <[email protected]>