On 2026/7/1 下午2:54, Song Gao wrote:
Validate guest-controlled cpu_num before using it to index the cpu[] array
or pass to async_run_on_cpu(). Without this check, a malicious guest can
trigger a NULL pointer dereference in async_run_on_cpu() and an
out-of-bounds array access in qemu_set_irq(), causing host crash.
Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3616
Fixes: 0d148eaf5a3e ("hw/loongarch: Implement dintc set irq")
Reported-by: Thomas Huth <[email protected]>
Signed-off-by: Song Gao <[email protected]>
---
hw/intc/loongarch_dintc.c | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/hw/intc/loongarch_dintc.c b/hw/intc/loongarch_dintc.c
index c877a8003b..e40292887c 100644
--- a/hw/intc/loongarch_dintc.c
+++ b/hw/intc/loongarch_dintc.c
@@ -19,6 +19,7 @@
#include "target/loongarch/cpu.h"
#include "qemu/error-report.h"
#include "system/hw_accel.h"
+#include "qemu/log.h"
/* msg addr field */
FIELD(MSG_ADDR, IRQ_NUM, 4, 8)
@@ -52,7 +53,19 @@ static void loongarch_dintc_mem_write(void *opaque, hwaddr
addr,
CPUState *cs;
cpu_num = FIELD_EX64(msg_addr, MSG_ADDR, CPU_NUM);
+
+ /* Validate cpu_num against the configured number of CPUs */
+ if (cpu_num >= s->num_cpu) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "loongarch-dintc: invalid cpu number%d\n", cpu_num);
+ return;
+ }
Here cpu_num decoded from MSG_ADDR is physical CPU ID, I think that it
is not meaningful to compare it with s->num_cpu.
cs = cpu_by_arch_id(cpu_num);
This is ok to use cpu_by_arch_id now, in later when CPU hotplug is
support with dintc, there should be self CPU search logic like ipi irqchip.
Regards
Bibo Mao
+ if (!cs) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "loongarch-dintc: no CPU for arch_id %d\n", cpu_num);
+ return;
+ }
irq_num = FIELD_EX64(msg_addr, MSG_ADDR, IRQ_NUM);
async_run_on_cpu(cs, do_set_vcpu_dintc_irq,
