Okay, so I'm trying to figure out whether this bug report is user-mode related or system-mode related. The issue involves a program crashing the kernel when run under qemu-system-aarch64. Let's break down what I know. The user provided a C program that uses SVE and Streaming SVE (SME) modes. They mention that switching between these states causes the kernel to crash during task switching in QEMU, but not on an Arm Fast Model. The kernel log shows a NULL pointer dereference when trying to save the SVE state, which suggests that the kernel isn't handling the context switch correctly for this process. The program uses prctl(PR_SVE_SET_VL) to set the vector length and uses asm volatile instructions to enter Streaming SVE mode. It then enters an infinite loop, waiting to be preempted. The crash happens when the task is switched, likely during a context save operation. Looking at the backtrace from the kernel log, the crash occurs in sve_save_state, which is part of the fpsimd_save function. This indicates that the kernel is attempting to save the SVE state but doesn't have the necessary memory allocated for it. The user mentioned that other SME programs run correctly, so it's not a general issue with SME support. The problem seems to occur when switching back from Streaming SVE mode after a syscall. The kernel expects to save the state but doesn't allocate space for it, leading to a NULL pointer dereference. Now, considering whether this is a user-mode or system-mode issue: User-mode issues are typically related to how applications behave in userspace, while system-mode involves higher-privilege operations, peripherals, devices, and the kernel itself. Since the crash happens within the kernel during task switching and context saving, it's more likely a system-mode problem. Therefore, this bug report is about an issue in the kernel's handling of SVE state during context switches under QEMU. system