diff mbox series

[SRU,J/allwinner-5.17,J/starfive-5.17,v2,1/1] RISC-V: KVM: Fix SRCU deadlock caused by kvm_riscv_check_vcpu_requests()

Message ID 20220729091632.777999-2-emil.renner.berthing@canonical.com
State New
Headers show
Series Fix RISC-V KVM with u-boot | expand

Commit Message

Emil Renner Berthing July 29, 2022, 9:16 a.m. UTC 
From: Anup Patel <apatel@ventanamicro.com>

BugLink: https://bugs.launchpad.net/bugs/1980594

The kvm_riscv_check_vcpu_requests() is called with SRCU read lock held
and for KVM_REQ_SLEEP request it will block the VCPU without releasing
SRCU read lock. This causes KVM ioctls (such as KVM_IOEVENTFD) from
other VCPUs of the same Guest/VM to hang/deadlock if there is any
synchronize_srcu() or synchronize_srcu_expedited() in the path.

To fix the above in kvm_riscv_check_vcpu_requests(), we should do SRCU
read unlock before blocking the VCPU and do SRCU read lock after VCPU
wakeup.

Fixes: cce69aff689e ("RISC-V: KVM: Implement VCPU interrupts and requests handling")
Reported-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
(backported from commit be82abe6a76ba8e76f25312566182b0f13c4fbf9)
[esmil: open code the srcu lock/unlock that now has static inline wrappers]
Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
---
 arch/riscv/kvm/vcpu.c | 2 ++
 1 file changed, 2 insertions(+)

Comments

Stefan Bader July 29, 2022, 9:43 a.m. UTC | #1 
On 29.07.22 11:16, Emil Renner Berthing wrote:
> From: Anup Patel <apatel@ventanamicro.com>
> 
> BugLink: https://bugs.launchpad.net/bugs/1980594
> 
> The kvm_riscv_check_vcpu_requests() is called with SRCU read lock held
> and for KVM_REQ_SLEEP request it will block the VCPU without releasing
> SRCU read lock. This causes KVM ioctls (such as KVM_IOEVENTFD) from
> other VCPUs of the same Guest/VM to hang/deadlock if there is any
> synchronize_srcu() or synchronize_srcu_expedited() in the path.
> 
> To fix the above in kvm_riscv_check_vcpu_requests(), we should do SRCU
> read unlock before blocking the VCPU and do SRCU read lock after VCPU
> wakeup.
> 
> Fixes: cce69aff689e ("RISC-V: KVM: Implement VCPU interrupts and requests handling")
> Reported-by: Bin Meng <bmeng.cn@gmail.com>
> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> Reviewed-by: Atish Patra <atishp@rivosinc.com>
> Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> Tested-by: Bin Meng <bmeng.cn@gmail.com>
> Signed-off-by: Anup Patel <anup@brainfault.org>
> (backported from commit be82abe6a76ba8e76f25312566182b0f13c4fbf9)
> [esmil: open code the srcu lock/unlock that now has static inline wrappers]
> Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>
Acked-by: Stefan Bader <stefan.bader@canonical.com>
> ---
>   arch/riscv/kvm/vcpu.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> index aad430668bb4..5c00a6526dca 100644
> --- a/arch/riscv/kvm/vcpu.c
> +++ b/arch/riscv/kvm/vcpu.c
> @@ -673,9 +673,11 @@ static void kvm_riscv_check_vcpu_requests(struct kvm_vcpu *vcpu)
>   
>   	if (kvm_request_pending(vcpu)) {
>   		if (kvm_check_request(KVM_REQ_SLEEP, vcpu)) {
> +			srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
>   			rcuwait_wait_event(wait,
>   				(!vcpu->arch.power_off) && (!vcpu->arch.pause),
>   				TASK_INTERRUPTIBLE);
> +			vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
>   
>   			if (vcpu->arch.power_off || vcpu->arch.pause) {
>   				/*
Kleber Souza July 29, 2022, 10:12 a.m. UTC | #2 
On 29.07.22 11:16, Emil Renner Berthing wrote:
> From: Anup Patel <apatel@ventanamicro.com>
> 
> BugLink: https://bugs.launchpad.net/bugs/1980594
> 
> The kvm_riscv_check_vcpu_requests() is called with SRCU read lock held
> and for KVM_REQ_SLEEP request it will block the VCPU without releasing
> SRCU read lock. This causes KVM ioctls (such as KVM_IOEVENTFD) from
> other VCPUs of the same Guest/VM to hang/deadlock if there is any
> synchronize_srcu() or synchronize_srcu_expedited() in the path.
> 
> To fix the above in kvm_riscv_check_vcpu_requests(), we should do SRCU
> read unlock before blocking the VCPU and do SRCU read lock after VCPU
> wakeup.
> 
> Fixes: cce69aff689e ("RISC-V: KVM: Implement VCPU interrupts and requests handling")
> Reported-by: Bin Meng <bmeng.cn@gmail.com>
> Signed-off-by: Anup Patel <apatel@ventanamicro.com>
> Reviewed-by: Atish Patra <atishp@rivosinc.com>
> Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> Tested-by: Bin Meng <bmeng.cn@gmail.com>
> Signed-off-by: Anup Patel <anup@brainfault.org>
> (backported from commit be82abe6a76ba8e76f25312566182b0f13c4fbf9)
> [esmil: open code the srcu lock/unlock that now has static inline wrappers]
> Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com>

Acked-by: Kleber Sacilotto de Souza <kleber.souza@canonical.com>

Thanks

> ---
>   arch/riscv/kvm/vcpu.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> index aad430668bb4..5c00a6526dca 100644
> --- a/arch/riscv/kvm/vcpu.c
> +++ b/arch/riscv/kvm/vcpu.c
> @@ -673,9 +673,11 @@ static void kvm_riscv_check_vcpu_requests(struct kvm_vcpu *vcpu)
>   
>   	if (kvm_request_pending(vcpu)) {
>   		if (kvm_check_request(KVM_REQ_SLEEP, vcpu)) {
> +			srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
>   			rcuwait_wait_event(wait,
>   				(!vcpu->arch.power_off) && (!vcpu->arch.pause),
>   				TASK_INTERRUPTIBLE);
> +			vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
>   
>   			if (vcpu->arch.power_off || vcpu->arch.pause) {
>   				/*
diff mbox series

Patch

diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
index aad430668bb4..5c00a6526dca 100644
--- a/arch/riscv/kvm/vcpu.c
+++ b/arch/riscv/kvm/vcpu.c
@@ -673,9 +673,11 @@  static void kvm_riscv_check_vcpu_requests(struct kvm_vcpu *vcpu)
 
 	if (kvm_request_pending(vcpu)) {
 		if (kvm_check_request(KVM_REQ_SLEEP, vcpu)) {
+			srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 			rcuwait_wait_event(wait,
 				(!vcpu->arch.power_off) && (!vcpu->arch.pause),
 				TASK_INTERRUPTIBLE);
+			vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 			if (vcpu->arch.power_off || vcpu->arch.pause) {
 				/*