Am 01.03.22 um 15:33 schrieb Claudio Imbrenda:
When handling the SCK instruction, the kvm lock is taken, even though
the vcpu lock is already being held. The normal locking order is kvm
lock first and then vcpu lock. This is can (and in some circumstances
does) lead to deadlocks.
The function kvm_s390_set_tod_clock is called both by the SCK handler
and by some IOCTLs to set the clock. The IOCTLs will not hold the vcpu
lock, so they can safely take the kvm lock. The SCK handler holds the
vcpu lock, but will also somehow need to acquire the kvm lock without
relinquishing the vcpu lock.
The solution is to factor out the code to set the clock, and provide
two wrappers. One is called like the original function and does the
locking, the other is called kvm_s390_try_set_tod_clock and uses
trylock to try to acquire the kvm lock. This new wrapper is then used
in the SCK handler. If locking fails, -EAGAIN is returned, which is
eventually propagated to userspace, thus also freeing the vcpu lock and
allowing for forward progress.
This is not the most efficient or elegant way to solve this issue, but
the SCK instruction is deprecated and its performance is not critical.
The goal of this patch is just to provide a simple but correct way to
fix the bug.
Fixes: 6a3f95a6b04c ("KVM: s390: Intercept SCK instruction")
Signed-off-by: Claudio Imbrenda <imbrenda@xxxxxxxxxxxxx>
Reviewed-by: Christian Borntraeger <borntraeger@xxxxxxxxxx>
---
arch/s390/kvm/kvm-s390.c | 19 ++++++++++++++++---
arch/s390/kvm/kvm-s390.h | 4 ++--
arch/s390/kvm/priv.c | 14 +++++++++++++-
3 files changed, 31 insertions(+), 6 deletions(-)
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index 2296b1ff1e02..4e3db4004bfd 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -3869,14 +3869,12 @@ static int kvm_s390_handle_requests(struct kvm_vcpu *vcpu)
return 0;
}
-void kvm_s390_set_tod_clock(struct kvm *kvm,
- const struct kvm_s390_vm_tod_clock *gtod)
+static void __kvm_s390_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod)
{
struct kvm_vcpu *vcpu;
union tod_clock clk;
unsigned long i;
- mutex_lock(&kvm->lock);
preempt_disable();
store_tod_clock_ext(&clk);
@@ -3897,7 +3895,22 @@ void kvm_s390_set_tod_clock(struct kvm *kvm,
kvm_s390_vcpu_unblock_all(kvm);
preempt_enable();
+}
+
+void kvm_s390_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod)
+{
+ mutex_lock(&kvm->lock);
+ __kvm_s390_set_tod_clock(kvm, gtod);
+ mutex_unlock(&kvm->lock);
+}
+
+int kvm_s390_try_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod)
+{
+ if (!mutex_trylock(&kvm->lock))
+ return 0;
+ __kvm_s390_set_tod_clock(kvm, gtod);
mutex_unlock(&kvm->lock);
+ return 1;
}
/**
diff --git a/arch/s390/kvm/kvm-s390.h b/arch/s390/kvm/kvm-s390.h
index 098831e815e6..f2c910763d7f 100644
--- a/arch/s390/kvm/kvm-s390.h
+++ b/arch/s390/kvm/kvm-s390.h
@@ -349,8 +349,8 @@ int kvm_s390_handle_sigp(struct kvm_vcpu *vcpu);
int kvm_s390_handle_sigp_pei(struct kvm_vcpu *vcpu);
/* implemented in kvm-s390.c */
-void kvm_s390_set_tod_clock(struct kvm *kvm,
- const struct kvm_s390_vm_tod_clock *gtod);
+void kvm_s390_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod);
+int kvm_s390_try_set_tod_clock(struct kvm *kvm, const struct kvm_s390_vm_tod_clock *gtod);
long kvm_arch_fault_in_page(struct kvm_vcpu *vcpu, gpa_t gpa, int writable);
int kvm_s390_store_status_unloaded(struct kvm_vcpu *vcpu, unsigned long addr);
int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr);
diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c
index 417154b314a6..7f3e7990ef82 100644
--- a/arch/s390/kvm/priv.c
+++ b/arch/s390/kvm/priv.c
@@ -102,7 +102,19 @@ static int handle_set_clock(struct kvm_vcpu *vcpu)
return kvm_s390_inject_prog_cond(vcpu, rc);
VCPU_EVENT(vcpu, 3, "SCK: setting guest TOD to 0x%llx", gtod.tod);
- kvm_s390_set_tod_clock(vcpu->kvm, >od);
+ /*
+ * To set the TOD clock we need to take the kvm lock, but we are
+ * already holding the vcpu lock, and the usual lock order is the
+ * opposite. Therefore we use trylock instead of lock, and if the
+ * kvm lock cannot be taken, we retry the instruction and return
+ * -EAGAIN to userspace, thus freeing the vcpu lock.
+ * The SCK instruction is considered legacy and at this point it's
+ * not worth the effort to find a nicer solution.
+ */
+ if (!kvm_s390_try_set_tod_clock(vcpu->kvm, >od)) {
+ kvm_s390_retry_instr(vcpu);
+ return -EAGAIN;
+ }
kvm_s390_set_psw_cc(vcpu, 0);
return 0;
