Re: [PATCH 4.19] dmaengine: at_hdmac: Fix memory leak

From: Greg KH
Date: Wed Jan 06 2021 - 11:52:39 EST


On Tue, Jan 05, 2021 at 08:03:20AM +0000, Tudor.Ambarus@xxxxxxxxxxxxx wrote:
> On 1/4/21 1:02 PM, Greg KH wrote:
> > EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> >
> > On Wed, Sep 23, 2020 at 08:19:18AM +0000, Tudor.Ambarus@xxxxxxxxxxxxx wrote:
> >> On 9/23/20 11:13 AM, Tudor.Ambarus@xxxxxxxxxxxxx wrote:
> >>> Hi, Pavel,
> >>>
> >>> On 9/20/20 11:28 AM, Pavel Machek wrote:
> >>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> >>>>
> >>>> This fixes memory leak in at_hdmac. Mainline does not have the same
> >>>> problem.
> >>>>
> >>>> Signed-off-by: Pavel Machek (CIP) <pavel@xxxxxxx>
> >>>>
> >>>> diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
> >>>> index 86427f6ba78c..0847b2055857 100644
> >>>> --- a/drivers/dma/at_hdmac.c
> >>>> +++ b/drivers/dma/at_hdmac.c
> >>>> @@ -1714,8 +1714,10 @@ static struct dma_chan *at_dma_xlate(struct of_phandle_args *dma_spec,
> >>>> atslave->dma_dev = &dmac_pdev->dev;
> >>>>
> >>>> chan = dma_request_channel(mask, at_dma_filter, atslave);
> >>>> - if (!chan)
> >>>> + if (!chan) {
> >>>> + kfree(atslave);
> >>>> return NULL;
> >>>> + }
> >>>
> >>> Thanks for submitting this to stable. While the fix is good, you can instead
> >>> cherry-pick the commit that hit upstream. In order to do that cleanly on top
> >>> of v4.19.145, you have to pick two other fixes:
> >>>
> >>> commit a6e7f19c9100 ("dmaengine: at_hdmac: Substitute kzalloc with kmalloc")
> >>> commit 3832b78b3ec2 ("dmaengine: at_hdmac: add missing put_device() call in at_dma_xlate()")>>> commit a6e7f19c9100 ("dmaengine: at_hdmac: Substitute kzalloc with kmalloc")
> >>
> >> this last commit should have been
> >> commit e097eb7473d9 ("dmaengine: at_hdmac: add missing kfree() call in at_dma_xlate()")
> >>
> >> bad copy and paste :)
> >
> > So are all 3 of those needed on both 5.4.y and 4.19.y to resolve this
> > issue?
> >
>
> Yes. I've just cherry-picked all three commits on both v5.4.86 and v4.19.164,
> everything looks ok. I also compiled using sama5_defconfig, all good.
>
> The order in which they should be taken is 1/ a6e7f19c9100, then 2/ 3832b78b3ec2,
> and 3/ e097eb7473d9.

Thanks, will go queue those up now.

greg k-h