Re: [Qemu-devel] [PATCH v2 2/2] memory: fix possible NULL pointer dereference

[Peter Maydell <peter.maydell@linaro.org>]

On 11 July 2018 at 14:47, Philippe Mathieu-Daudé <f4bug@amsat.org> wrote:
> Hi Dima,
>
> On 07/11/2018 05:34 AM, Dima Stepanov wrote:
>> Gentle ping. CCing Paolo Bonzini.
>>
>> Regards, Dima.
>>
>> On Tue, Jun 19, 2018 at 05:12:16PM +0300, Dima Stepanov wrote:
>>> Ping.
>>>
>>> Regards, Dima.
>>>
>>> On Wed, Jun 13, 2018 at 11:19:55AM +0300, Dima Stepanov wrote:
>>>> In the memory_region_do_invalidate_mmio_ptr() routine the section
>>>> variable is intialized by the memory_region_find() call. The section.mr
>>>> field can be set to NULL.
>>>>
>>>> Add the check for NULL before trying to drop a section.
>>>>
>>>> Signed-off-by: Dima Stepanov <dimastep@yandex-team.ru>
>>>> ---
>>>>  memory.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/memory.c b/memory.c
>>>> index 3212acc..bb45248 100644
>>>> --- a/memory.c
>>>> +++ b/memory.c
>>>> @@ -2712,7 +2712,7 @@ static void memory_region_do_invalidate_mmio_ptr(CPUState *cpu,
>>>>      /* Reset dirty so this doesn't happen later. */
>>>>      cpu_physical_memory_test_and_clear_dirty(offset, size, 1);
>>>>
>>>> -    if (section.mr != mr) {
>>>> +    if (section.mr && (section.mr != mr)) {
>
> section.mr can't be NULL here.
>
> You can give the static analyzer a hint using "assert(section.mr);"

Not in my view much point in messing with this code, though:
(a) it's broken and unusable as it stands (race conditions)
(b) it's obsoleted by the execute-from-mmio patchset
http://patchwork.ozlabs.org/cover/942090/ and if/when that
goes in it will all just get deleted.

thanks
-- PMM