Re: [PATCH 2/3] hw/intc: Add NULL pointer check on LoongArch ipi device

[Song Gao]

Hi,  Philippe

在 2023/5/12 上午3:03, Philippe Mathieu-Daudé 写道:
> On 6/4/23 12:00, Song Gao wrote:
> > When ipi mailbox is used, cpu index is decoded from iocsr register.
> > cpu maybe does not exist. This patch adss NULL pointer check on
> > ipi device.
>
> How can that happens from a guest vcpu context?
cpuid(cs->cpu_index)  is decoded from iocsr register.

    cpuid = (val >> 16) & 0x3ff;   // ipi_sned [25:16]

The value maybe invalid.  qemu only support 4 vcpu.

you can find more about ipi_send registers at:
https://github.com/loongson/LoongArch-Documentation/releases/download/2023.04.20/Loongson-3A5000-usermanual-v1.03-EN.pdf
Table 63. Processor core inter-processor communication registers

> > Signed-off-by: Song Gao <gaosong@loongson.cn>
> > ---
> >   hw/intc/loongarch_ipi.c | 31 +++++++++++++++++++------------
> >   1 file changed, 19 insertions(+), 12 deletions(-)
> >
> > diff --git a/hw/intc/loongarch_ipi.c b/hw/intc/loongarch_ipi.c
> > index 0563d83a35..39e899df46 100644
> > --- a/hw/intc/loongarch_ipi.c
> > +++ b/hw/intc/loongarch_ipi.c
> > @@ -86,11 +86,12 @@ static void ipi_send(uint64_t val)
> >       /* IPI status vector */
> >       data = 1 << (val & 0x1f);
> >       cs = qemu_get_cpu(cpuid);
> > -    cpu = LOONGARCH_CPU(cs);
> > -    env = &cpu->env;
> > -    address_space_stl(&env->address_space_iocsr, 0x1008,
> > -                      data, MEMTXATTRS_UNSPECIFIED, NULL);
> > -
> > +    if (cs) {
> > +        cpu = LOONGARCH_CPU(cs);
> > +        env = &cpu->env;
> > +        address_space_stl(&env->address_space_iocsr, 0x1008,
> > +                          data, MEMTXATTRS_UNSPECIFIED, NULL);
> > +    }
>
> Is that the hardware behavior?
Yes.
> Could logging the invalid cpuid request be useful?
Sure.

Thanks.
Song Gao