Merge 4.4.78 into android-4.4

Changes in 4.4.78
	net_sched: fix error recovery at qdisc creation
	net: sched: Fix one possible panic when no destroy callback
	net/phy: micrel: configure intterupts after autoneg workaround
	ipv6: avoid unregistering inet6_dev for loopback
	net: dp83640: Avoid NULL pointer dereference.
	tcp: reset sk_rx_dst in tcp_disconnect()
	net: prevent sign extension in dev_get_stats()
	bpf: prevent leaking pointer via xadd on unpriviledged
	net: handle NAPI_GRO_FREE_STOLEN_HEAD case also in napi_frags_finish()
	ipv6: dad: don't remove dynamic addresses if link is down
	net: ipv6: Compare lwstate in detecting duplicate nexthops
	vrf: fix bug_on triggered by rx when destroying a vrf
	rds: tcp: use sock_create_lite() to create the accept socket
	brcmfmac: fix possible buffer overflow in brcmf_cfg80211_mgmt_tx()
	cfg80211: Define nla_policy for NL80211_ATTR_LOCAL_MESH_POWER_MODE
	cfg80211: Validate frequencies nested in NL80211_ATTR_SCAN_FREQUENCIES
	cfg80211: Check if PMKID attribute is of expected size
	irqchip/gic-v3: Fix out-of-bound access in gic_set_affinity
	parisc: Report SIGSEGV instead of SIGBUS when running out of stack
	parisc: use compat_sys_keyctl()
	parisc: DMA API: return error instead of BUG_ON for dma ops on non dma devs
	parisc/mm: Ensure IRQs are off in switch_mm()
	tools/lib/lockdep: Reduce MAX_LOCK_DEPTH to avoid overflowing lock_chain/: Depth
	kernel/extable.c: mark core_kernel_text notrace
	mm/list_lru.c: fix list_lru_count_node() to be race free
	fs/dcache.c: fix spin lockup issue on nlru->lock
	checkpatch: silence perl 5.26.0 unescaped left brace warnings
	binfmt_elf: use ELF_ET_DYN_BASE only for PIE
	arm: move ELF_ET_DYN_BASE to 4MB
	arm64: move ELF_ET_DYN_BASE to 4GB / 4MB
	powerpc: move ELF_ET_DYN_BASE to 4GB / 4MB
	s390: reduce ELF_ET_DYN_BASE
	exec: Limit arg stack to at most 75% of _STK_LIM
	vt: fix unchecked __put_user() in tioclinux ioctls
	mnt: In umount propagation reparent in a separate pass
	mnt: In propgate_umount handle visiting mounts in any order
	mnt: Make propagate_umount less slow for overlapping mount propagation trees
	selftests/capabilities: Fix the test_execve test
	tpm: Get rid of chip->pdev
	tpm: Provide strong locking for device removal
	Add "shutdown" to "struct class".
	tpm: Issue a TPM2_Shutdown for TPM2 devices.
	mm: fix overflow check in expand_upwards()
	crypto: talitos - Extend max key length for SHA384/512-HMAC and AEAD
	crypto: atmel - only treat EBUSY as transient if backlog
	crypto: sha1-ssse3 - Disable avx2
	crypto: caam - fix signals handling
	sched/topology: Fix overlapping sched_group_mask
	sched/topology: Optimize build_group_mask()
	PM / wakeirq: Convert to SRCU
	PM / QoS: return -EINVAL for bogus strings
	tracing: Use SOFTIRQ_OFFSET for softirq dectection for more accurate results
	KVM: x86: disable MPX if host did not enable MPX XSAVE features
	kvm: vmx: Do not disable intercepts for BNDCFGS
	kvm: x86: Guest BNDCFGS requires guest MPX support
	kvm: vmx: Check value written to IA32_BNDCFGS
	kvm: vmx: allow host to access guest MSR_IA32_BNDCFGS
	Linux 4.4.78

Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>

1 files changed, 19 insertions, 3 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index e353de860bfd..a4e23e7e6ede 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6386,6 +6386,9 @@ enum s_alloc {
  * Build an iteration mask that can exclude certain CPUs from the upwards
  * domain traversal.
  *
+ * Only CPUs that can arrive at this group should be considered to continue
+ * balancing.
+ *
  * Asymmetric node setups can result in situations where the domain tree is of
  * unequal depth, make sure to skip domains that already cover the entire
  * range.
@@ -6397,18 +6400,31 @@ enum s_alloc {
  */
 static void build_group_mask(struct sched_domain *sd, struct sched_group *sg)
 {
-	const struct cpumask *span = sched_domain_span(sd);
+	const struct cpumask *sg_span = sched_group_cpus(sg);
 	struct sd_data *sdd = sd->private;
 	struct sched_domain *sibling;
 	int i;
 
-	for_each_cpu(i, span) {
+	for_each_cpu(i, sg_span) {
 		sibling = *per_cpu_ptr(sdd->sd, i);
-		if (!cpumask_test_cpu(i, sched_domain_span(sibling)))
+
+		/*
+		 * Can happen in the asymmetric case, where these siblings are
+		 * unused. The mask will not be empty because those CPUs that
+		 * do have the top domain _should_ span the domain.
+		 */
+		if (!sibling->child)
+			continue;
+
+		/* If we would not end up here, we can't continue from here */
+		if (!cpumask_equal(sg_span, sched_domain_span(sibling->child)))
 			continue;
 
 		cpumask_set_cpu(i, sched_group_mask(sg));
 	}
+
+	/* We must not have empty masks here */
+	WARN_ON_ONCE(cpumask_empty(sched_group_mask(sg)));
 }
 
 /*