From 93de2a6a4b91b72607136dd656edf03fb399d27f Mon Sep 17 00:00:00 2001 From: Sean Christopherson Date: Wed, 23 Sep 2026 09:37:21 -0700 Subject: KVM: SEV: Do cache maintenance on the source VM during intra-host migration Manually perform cache maintenance on the source VM during intra-host migration to ensure no stale data is left in CPU caches after the VM is destroyed. Because the source VM is "converted" to a non-SEV VM, KVM's memory reclaim flows won't trigger cache maintenance, e.g. when all guest memory is reclaimed in response to detaching from the mmu_notifier. Note, relying on the destination VM to do cache maintenance isn't an option as KVM doesn't require identical guest memory configurations, i.e. the source VM may have access to memory that the destination VM does not. Enforcing equivalent memory configurations is infeasible, as it would require a *deep* comparison of memslots, e.g. to verify that not only are the memslot identical, but what the memslots point at is also identical. Fixes: b56639318bb2 ("KVM: SEV: Add support for SEV intra host migration") Cc: stable@vger.kernel.org Reported-by: Stefan Teodorescu Signed-off-by: Sean Christopherson Message-ID: <20260923163721.1584779-3-seanjc@google.com> Signed-off-by: Paolo Bonzini --- lib/syscall.c | 88 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 88 insertions(+) create mode 100644 lib/syscall.c (limited to 'lib/syscall.c') diff --git a/lib/syscall.c b/lib/syscall.c new file mode 100644 index 000000000..006e256d2 --- /dev/null +++ b/lib/syscall.c @@ -0,0 +1,88 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include +#include +#include + +static int collect_syscall(struct task_struct *target, struct syscall_info *info) +{ + unsigned long args[6] = { }; + struct pt_regs *regs; + + if (!try_get_task_stack(target)) { + /* Task has no stack, so the task isn't in a syscall. */ + memset(info, 0, sizeof(*info)); + info->data.nr = -1; + return 0; + } + + regs = task_pt_regs(target); + if (unlikely(!regs)) { + put_task_stack(target); + return -EAGAIN; + } + + info->sp = user_stack_pointer(regs); + info->data.instruction_pointer = instruction_pointer(regs); + + info->data.nr = syscall_get_nr(target, regs); + if (info->data.nr != -1L) + syscall_get_arguments(target, regs, args); + + info->data.args[0] = args[0]; + info->data.args[1] = args[1]; + info->data.args[2] = args[2]; + info->data.args[3] = args[3]; + info->data.args[4] = args[4]; + info->data.args[5] = args[5]; + + put_task_stack(target); + return 0; +} + +/** + * task_current_syscall - Discover what a blocked task is doing. + * @target: thread to examine + * @info: structure with the following fields: + * .sp - filled with user stack pointer + * .data.nr - filled with system call number or -1 + * .data.args - filled with @maxargs system call arguments + * .data.instruction_pointer - filled with user PC + * + * If @target is blocked in a system call, returns zero with @info.data.nr + * set to the call's number and @info.data.args filled in with its + * arguments. Registers not used for system call arguments may not be available + * and it is not kosher to use &struct user_regset calls while the system + * call is still in progress. Note we may get this result if @target + * has finished its system call but not yet returned to user mode, such + * as when it's stopped for signal handling or syscall exit tracing. + * + * If @target is blocked in the kernel during a fault or exception, + * returns zero with *@info.data.nr set to -1 and does not fill in + * @info.data.args. If so, it's now safe to examine @target using + * &struct user_regset get() calls as long as we're sure @target won't return + * to user mode. + * + * Returns -%EAGAIN if @target does not remain blocked. + */ +int task_current_syscall(struct task_struct *target, struct syscall_info *info) +{ + unsigned long ncsw; + unsigned int state; + + if (target == current) + return collect_syscall(target, info); + + state = READ_ONCE(target->__state); + if (unlikely(!state)) + return -EAGAIN; + + ncsw = wait_task_inactive(target, state); + if (unlikely(!ncsw) || + unlikely(collect_syscall(target, info)) || + unlikely(wait_task_inactive(target, state) != ncsw)) + return -EAGAIN; + + return 0; +} -- cgit v1.3.1