On 09/24/2015 11:22 AM, Alex Williamson wrote:
> On Thu, 2015-09-24 at 10:54 +0800, Wen Congyang wrote:
>> On 09/24/2015 04:23 AM, Alex Williamson wrote:
>>> @@ -275,12 +276,15 @@ static void
>>> vfio_vga_probe_ati_3c3_quirk(VFIOPCIDevice *vdev)
>>> }
>>>
>>> quirk = g_malloc0(sizeof(*quirk));
>>> - quirk->vdev = vdev;
>>> + legacy = quirk->data = g_malloc0(sizeof(*legacy));
>>> + quirk->mem = legacy->mem = g_malloc0_n(sizeof(MemoryRegion), 1);
>>
>> Why do you use g_malloc0_n() here? It is introduced in glib 2.24, but we
>> only require glib 2.22.
>
> Because I saw it, I guess in scripts/coverity-model.c, and used it. In
> this particular instance it seems irrelevant, but VFIOQuirk.mem points
> to an array of MemoryRegions, other users here have more than one array
> entry and I chose to use the same allocator throughout for consistency.
> What's the preferred helper here, simply calloc()? Thanks,
Yes, there is g_malloc0_n() in scripts/coverity-model.c. But it still breaks
building. calloc()? I guess we don't use it directly in qemu.
g_malloc0_n ()
gpointer
g_malloc0_n (gsize n_blocks,
gsize n_block_bytes);
This function is similar to g_malloc0(), allocating (n_blocks * n_block_bytes )
bytes, but care is taken to detect possible overflow during multiplication.
So I think we can use g_malloc0(sizeof(MemoryRegion) * n).
Thanks
Wen Congyang
>
> Alex
>
> .
>
