> Hi,
>
> On 2021/1/2 17:50, Dinghao Liu wrote:
> > When irq_domain_get_irq_data() or irqd_cfg() fails
> > meanwhile i == 0, data allocated by kzalloc() has not
> > been freed before returning, which leads to memleak.
> >
> > Fixes: b106ee63abccb ("irq_remapping/vt-d: Enhance Intel IR driver to
> > support hierarchical irqdomains")
> > Signed-off-by: Dinghao Liu <dinghao....@zju.edu.cn>
> > ---
> > drivers/iommu/intel/irq_remapping.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/drivers/iommu/intel/irq_remapping.c
> > b/drivers/iommu/intel/irq_remapping.c
> > index aeffda92b10b..cdaeed36750f 100644
> > --- a/drivers/iommu/intel/irq_remapping.c
> > +++ b/drivers/iommu/intel/irq_remapping.c
> > @@ -1354,6 +1354,8 @@ static int intel_irq_remapping_alloc(struct
> > irq_domain *domain,
> > irq_cfg = irqd_cfg(irq_data);
> > if (!irq_data || !irq_cfg) {
> > ret = -EINVAL;
> > + kfree(data);
> > + data = NULL;
>
> Do you need to check (i == 0) here? @data will not be used anymore as it
> goes to out branch, why setting it to NULL here?
>
data will be passed to ire_data->chip_data when i == 0 and
intel_free_irq_resources() will free it on failure. Thus I
set it to NULL to prevent double-free. However, if we add
a check (i == 0) here, we will not need to set it to NULL.
If this is better, I will resend a new patch soon.
Regards,
Dinghao
