4.9-stable review patch. If anyone has any objections, please let me know.
------------------
From: Andre Przywara <andre.przyw...@arm.com>
commit bf308242ab98b5d1648c3663e753556bef9bec01 upstream.
kvm_read_guest() will eventually look up in kvm_memslots(), which requires
either to hold the kvm->slots_lock or to be inside a kvm->srcu critical
section.
In contrast to x86 and s390 we don't take the SRCU lock on every guest
exit, so we have to do it individually for each kvm_read_guest() call.
Provide a wrapper which does that and use that everywhere.
Note that ending the SRCU critical section before returning from the
kvm_read_guest() wrapper is safe, because the data has been *copied*, so
we don't need to rely on valid references to the memslot anymore.
Cc: Stable <sta...@vger.kernel.org> # 4.8+
Reported-by: Jan Glauber <jan.glau...@caviumnetworks.com>
Signed-off-by: Andre Przywara <andre.przyw...@arm.com>
Acked-by: Christoffer Dall <christoffer.d...@arm.com>
Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gre...@linuxfoundation.org>
---
arch/arm/include/asm/kvm_mmu.h | 16 ++++++++++++++++
arch/arm64/include/asm/kvm_mmu.h | 16 ++++++++++++++++
virt/kvm/arm/vgic/vgic-its.c | 15 ++++++++-------
3 files changed, 40 insertions(+), 7 deletions(-)
--- a/arch/arm/include/asm/kvm_mmu.h
+++ b/arch/arm/include/asm/kvm_mmu.h
@@ -223,6 +223,22 @@ static inline unsigned int kvm_get_vmid_
return 8;
}
+/*
+ * We are not in the kvm->srcu critical section most of the time, so we take
+ * the SRCU read lock here. Since we copy the data from the user page, we
+ * can immediately drop the lock again.
+ */
+static inline int kvm_read_guest_lock(struct kvm *kvm,
+ gpa_t gpa, void *data, unsigned long len)
+{
+ int srcu_idx = srcu_read_lock(&kvm->srcu);
+ int ret = kvm_read_guest(kvm, gpa, data, len);
+
+ srcu_read_unlock(&kvm->srcu, srcu_idx);
+
+ return ret;
+}
+
static inline void *kvm_get_hyp_vector(void)
{
return kvm_ksym_ref(__kvm_hyp_vector);
--- a/arch/arm64/include/asm/kvm_mmu.h
+++ b/arch/arm64/include/asm/kvm_mmu.h
@@ -313,6 +313,22 @@ static inline unsigned int kvm_get_vmid_
return (cpuid_feature_extract_unsigned_field(reg,
ID_AA64MMFR1_VMIDBITS_SHIFT) == 2) ? 16 : 8;
}
+/*
+ * We are not in the kvm->srcu critical section most of the time, so we take
+ * the SRCU read lock here. Since we copy the data from the user page, we
+ * can immediately drop the lock again.
+ */
+static inline int kvm_read_guest_lock(struct kvm *kvm,
+ gpa_t gpa, void *data, unsigned long len)
+{
+ int srcu_idx = srcu_read_lock(&kvm->srcu);
+ int ret = kvm_read_guest(kvm, gpa, data, len);
+
+ srcu_read_unlock(&kvm->srcu, srcu_idx);
+
+ return ret;
+}
+
#ifdef CONFIG_HARDEN_BRANCH_PREDICTOR
#include <asm/mmu.h>
--- a/virt/kvm/arm/vgic/vgic-its.c
+++ b/virt/kvm/arm/vgic/vgic-its.c
@@ -208,8 +208,8 @@ static int update_lpi_config(struct kvm
u8 prop;
int ret;
- ret = kvm_read_guest(kvm, propbase + irq->intid - GIC_LPI_OFFSET,
- &prop, 1);
+ ret = kvm_read_guest_lock(kvm, propbase + irq->intid - GIC_LPI_OFFSET,
+ &prop, 1);
if (ret)
return ret;
@@ -339,8 +339,9 @@ static int its_sync_lpi_pending_table(st
* this very same byte in the last iteration. Reuse that.
*/
if (byte_offset != last_byte_offset) {
- ret = kvm_read_guest(vcpu->kvm, pendbase + byte_offset,
- &pendmask, 1);
+ ret = kvm_read_guest_lock(vcpu->kvm,
+ pendbase + byte_offset,
+ &pendmask, 1);
if (ret) {
kfree(intids);
return ret;
@@ -628,7 +629,7 @@ static bool vgic_its_check_id(struct vgi
return false;
/* Each 1st level entry is represented by a 64-bit value. */
- if (kvm_read_guest(its->dev->kvm,
+ if (kvm_read_guest_lock(its->dev->kvm,
BASER_ADDRESS(baser) + index * sizeof(indirect_ptr),
&indirect_ptr, sizeof(indirect_ptr)))
return false;
@@ -1152,8 +1153,8 @@ static void vgic_its_process_commands(st
cbaser = CBASER_ADDRESS(its->cbaser);
while (its->cwriter != its->creadr) {
- int ret = kvm_read_guest(kvm, cbaser + its->creadr,
- cmd_buf, ITS_CMD_SIZE);
+ int ret = kvm_read_guest_lock(kvm, cbaser + its->creadr,
+ cmd_buf, ITS_CMD_SIZE);
/*
* If kvm_read_guest() fails, this could be due to the guest
* programming a bogus value in CBASER or something else going
