RISC-V: KVM: Fix SRCU deadlock caused by kvm_riscv_check_vcpu_requests()
authorAnup Patel <apatel@ventanamicro.com>
Mon, 11 Jul 2022 04:06:32 +0000 (09:36 +0530)
committerAnup Patel <anup@brainfault.org>
Mon, 11 Jul 2022 04:06:32 +0000 (09:36 +0530)
commitbe82abe6a76ba8e76f25312566182b0f13c4fbf9
tree66d4c7e41c5055d0fa865316547b3085dce1ec64
parent88573389aaa34640b9ecde15622c68e8737d8f8c
RISC-V: KVM: Fix SRCU deadlock caused by kvm_riscv_check_vcpu_requests()

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>
arch/riscv/kvm/vcpu.c