Re: [Qemu-devel] [Qemu-arm] [PATCH] intc: arm_gicv3: limit GICR ipriority index

["Philippe Mathieu-Daudé" <f4bug@amsat.org>]

On 09/05/2017 08:58 AM, Peter Maydell wrote:
> On 5 September 2017 at 12:21, P J P <ppandit@redhat.com> wrote:
>> From: Prasad J Pandit <pjp@fedoraproject.org>
>>
>> When reading or writing to GICR ipriority array, index 'irq'
>> could go beyond its bounds; Restrict it within array limits.
>>
>> Reported-by: Guoxiang Niu <niuguoxiang@huawei.com>
>> Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
>> ---
>>   hw/intc/arm_gicv3_redist.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/hw/intc/arm_gicv3_redist.c b/hw/intc/arm_gicv3_redist.c
>> index 77e5cfa327..7683c4cc7f 100644
>> --- a/hw/intc/arm_gicv3_redist.c
>> +++ b/hw/intc/arm_gicv3_redist.c
>> @@ -187,7 +187,7 @@ static MemTxResult gicr_readl(GICv3CPUState *cs, hwaddr offset,
>>       case GICR_ICACTIVER0:
>>           *data = gicr_read_bitmap_reg(cs, attrs, cs->gicr_iactiver0);
>>           return MEMTX_OK;
>> -    case GICR_IPRIORITYR ... GICR_IPRIORITYR + 0x1f:

0x1f are only the cpu (private) irqs, then the range is valid up-to:
(extract64(cs->gicr_typer, 0, 5) + 1) * 32 - 1 supported irqs

>> +    case GICR_IPRIORITYR ... GICR_IPRIORITYR + 0x1c:
>>       {
>>           int i, irq = offset - GICR_IPRIORITYR;
>>           uint32_t value = 0;
>> @@ -310,7 +310,7 @@ static MemTxResult gicr_writel(GICv3CPUState *cs, hwaddr offset,
>>       case GICR_ICACTIVER0:
>>           gicr_write_clear_bitmap_reg(cs, attrs, &cs->gicr_iactiver0, value);
>>           return MEMTX_OK;
>> -    case GICR_IPRIORITYR ... GICR_IPRIORITYR + 0x1f:
>> +    case GICR_IPRIORITYR ... GICR_IPRIORITYR + 0x1c:
>>       {
>>           int i, irq = offset - GICR_IPRIORITYR;
> 
> Why do you think the buffer can be overrun? These functions
> are the word (4 byte) access functions, and they cannot
> be called with a non-4-aligned offset (see the asserts in
> gicv3_redist_read() and gicv3_redist_write()).
> 
> thanks
> -- PMM
>