[PATCH] misc: mic: fix memory leak

[PATCH] misc: mic: fix memory leak

[Sudip Mukherjee]

In scif_node_connect() we were returning if the initialization of p2p_ji
fails. But at that time p2p_ij has already been initialized and
resources allocated for it. And since p2p_ij is not added to the list
till now so we will have a leak.
Lets deinitialize and release the resources connected to p2p_ij.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
---

One sudip sending patch to another sudeep... :)

 drivers/misc/mic/scif/scif_nodeqp.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/drivers/misc/mic/scif/scif_nodeqp.c b/drivers/misc/mic/scif/scif_nodeqp.c
index 7180d56..5e94fac 100644
--- a/drivers/misc/mic/scif/scif_nodeqp.c
+++ b/drivers/misc/mic/scif/scif_nodeqp.c
@@ -435,6 +435,18 @@ free_p2p:
 	return NULL;
 }
 
+/* Uninitialize and release resources from a p2p mapping */
+void scif_deinit_p2p_info(struct scif_dev *scifdev, struct scif_p2p_info *p2p)
+{
+	struct scif_hw_dev *sdev = scifdev->sdev;
+
+	dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_MMIO],
+		     p2p->sg_nentries[SCIF_PPI_MMIO], DMA_BIDIRECTIONAL);
+	scif_p2p_freesg(p2p->ppi_sg[SCIF_PPI_MMIO]);
+	scif_p2p_freesg(p2p->ppi_sg[SCIF_PPI_APER]);
+	kfree(p2p);
+}
+
 /**
  * scif_node_connect: Respond to SCIF_NODE_CONNECT interrupt message
  * @dst: Destination node
@@ -477,8 +489,10 @@ static void scif_node_connect(struct scif_dev *scifdev, int dst)
 	if (!p2p_ij)
 		return;
 	p2p_ji = scif_init_p2p_info(dev_j, dev_i);
-	if (!p2p_ji)
+	if (!p2p_ji) {
+		scif_deinit_p2p_info(dev_i, p2p_ij);
 		return;
+	}
 	list_add_tail(&p2p_ij->ppi_list, &dev_i->p2p);
 	list_add_tail(&p2p_ji->ppi_list, &dev_j->p2p);
 
-- 
1.9.1

Re: [PATCH] misc: mic: fix memory leak

[kbuild test robot]

Hi Sudip,

[auto build test WARNING on char-misc/char-misc-next -- if it's inappropriate base, please suggest rules for selecting the more suitable base]

url:    https://github.com/0day-ci/linux/commits/Sudip-Mukherjee/misc-mic-fix-memory-leak/20151013-171026
reproduce:
        # apt-get install sparse
        make ARCH=x86_64 allmodconfig
        make C=1 CF=-D__CHECK_ENDIAN__


sparse warnings: (new ones prefixed by >>)

>> drivers/misc/mic/scif/scif_nodeqp.c:439:6: sparse: symbol 'scif_deinit_p2p_info' was not declared. Should it be static?

Please review and possibly fold the followup patch.

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

Re: [PATCH] misc: mic: fix memory leak

[Sudip Mukherjee]

On Tue, Oct 13, 2015 at 05:50:12PM +0800, kbuild test robot wrote:
> Hi Sudip,
> 
> [auto build test WARNING on char-misc/char-misc-next -- if it's inappropriate base, please suggest rules for selecting the more suitable base]
> 
> url:    https://github.com/0day-ci/linux/commits/Sudip-Mukherjee/misc-mic-fix-memory-leak/20151013-171026
> reproduce:
>         # apt-get install sparse
>         make ARCH=x86_64 allmodconfig
>         make C=1 CF=-D__CHECK_ENDIAN__
> 
> 
> sparse warnings: (new ones prefixed by >>)
> 
> >> drivers/misc/mic/scif/scif_nodeqp.c:439:6: sparse: symbol 'scif_deinit_p2p_info' was not declared. Should it be static?
> 
> Please review and possibly fold the followup patch.

Thanks. I have sent a v2.

regards
sudip

[RFC PATCH] misc: mic: scif_deinit_p2p_info() can be static

[kbuild test robot]

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
---
 scif_nodeqp.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/misc/mic/scif/scif_nodeqp.c b/drivers/misc/mic/scif/scif_nodeqp.c
index 5e94fac..81f0b23 100644
--- a/drivers/misc/mic/scif/scif_nodeqp.c
+++ b/drivers/misc/mic/scif/scif_nodeqp.c
@@ -436,7 +436,7 @@ free_p2p:
 }
 
 /* Uninitialize and release resources from a p2p mapping */
-void scif_deinit_p2p_info(struct scif_dev *scifdev, struct scif_p2p_info *p2p)
+static void scif_deinit_p2p_info(struct scif_dev *scifdev, struct scif_p2p_info *p2p)
 {
 	struct scif_hw_dev *sdev = scifdev->sdev;
 

Re: [PATCH] misc: mic: fix memory leak

[Sudeep Dutt]

On Tue, 2015-10-13 at 14:38 +0530, Sudip Mukherjee wrote:
> In scif_node_connect() we were returning if the initialization of p2p_ji
> fails. But at that time p2p_ij has already been initialized and
> resources allocated for it. And since p2p_ij is not added to the list
> till now so we will have a leak.
> Lets deinitialize and release the resources connected to p2p_ij.
> 
> Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> ---
> 
> One sudip sending patch to another sudeep... :)
> 
>  drivers/misc/mic/scif/scif_nodeqp.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/misc/mic/scif/scif_nodeqp.c b/drivers/misc/mic/scif/scif_nodeqp.c
> index 7180d56..5e94fac 100644
> --- a/drivers/misc/mic/scif/scif_nodeqp.c
> +++ b/drivers/misc/mic/scif/scif_nodeqp.c
> @@ -435,6 +435,18 @@ free_p2p:
>  	return NULL;
>  }
>  
> +/* Uninitialize and release resources from a p2p mapping */
> +void scif_deinit_p2p_info(struct scif_dev *scifdev, struct scif_p2p_info *p2p)
> +{
> +	struct scif_hw_dev *sdev = scifdev->sdev;
> +
> +	dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_MMIO],
> +		     p2p->sg_nentries[SCIF_PPI_MMIO], DMA_BIDIRECTIONAL);

Hi Sudip,

You missed unmapping the aperture here:

dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_APER],
	     p2p->sg_nentries[SCIF_PPI_APER], DMA_BIDIRECTIONAL);

Can you please resubmit this patch with the change above?

Also you should mention [PATCH char-misc-next] in the next revision.

Thanks for the fix!

Sudeep Dutt

> +	scif_p2p_freesg(p2p->ppi_sg[SCIF_PPI_MMIO]);
> +	scif_p2p_freesg(p2p->ppi_sg[SCIF_PPI_APER]);
> +	kfree(p2p);
> +}
> +
>  /**
>   * scif_node_connect: Respond to SCIF_NODE_CONNECT interrupt message
>   * @dst: Destination node
> @@ -477,8 +489,10 @@ static void scif_node_connect(struct scif_dev *scifdev, int dst)
>  	if (!p2p_ij)
>  		return;
>  	p2p_ji = scif_init_p2p_info(dev_j, dev_i);
> -	if (!p2p_ji)
> +	if (!p2p_ji) {
> +		scif_deinit_p2p_info(dev_i, p2p_ij);
>  		return;
> +	}
>  	list_add_tail(&p2p_ij->ppi_list, &dev_i->p2p);
>  	list_add_tail(&p2p_ji->ppi_list, &dev_j->p2p);
>  

Re: [PATCH] misc: mic: fix memory leak

[Sudip Mukherjee]

On Tue, Oct 13, 2015 at 09:34:21AM -0700, Sudeep Dutt wrote:
> On Tue, 2015-10-13 at 14:38 +0530, Sudip Mukherjee wrote:
<snip>
> > +	struct scif_hw_dev *sdev = scifdev->sdev;
> > +
> > +	dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_MMIO],
> > +		     p2p->sg_nentries[SCIF_PPI_MMIO], DMA_BIDIRECTIONAL);
> 
> Hi Sudip,
> 
> You missed unmapping the aperture here:
> 
> dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_APER],
> 	     p2p->sg_nentries[SCIF_PPI_APER], DMA_BIDIRECTIONAL);
> 
> Can you please resubmit this patch with the change above?

Hi Sudeep,
v3 sent for your review.   
> 
> Also you should mention [PATCH char-misc-next] in the next revision.

Ohhh... yes, I keep forgetting that Fengguang will need this tag for
his build bot.

But I have one more doubt. If scif_nodeqp_send() fails then shouldn't we
remove these two nodes of the p2p connection from the lists and release
the resources?

regards
sudip

Re: [PATCH] misc: mic: fix memory leak

[Sudeep Dutt]

On Tue, 2015-10-13 at 22:46 +0530, Sudip Mukherjee wrote:
> On Tue, Oct 13, 2015 at 09:34:21AM -0700, Sudeep Dutt wrote:
> > On Tue, 2015-10-13 at 14:38 +0530, Sudip Mukherjee wrote:
> <snip>
> > > +	struct scif_hw_dev *sdev = scifdev->sdev;
> > > +
> > > +	dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_MMIO],
> > > +		     p2p->sg_nentries[SCIF_PPI_MMIO], DMA_BIDIRECTIONAL);
> > 
> > Hi Sudip,
> > 
> > You missed unmapping the aperture here:
> > 
> > dma_unmap_sg(&sdev->dev, p2p->ppi_sg[SCIF_PPI_APER],
> > 	     p2p->sg_nentries[SCIF_PPI_APER], DMA_BIDIRECTIONAL);
> > 
> > Can you please resubmit this patch with the change above?
> 
> Hi Sudeep,
> v3 sent for your review.   
> > 
> > Also you should mention [PATCH char-misc-next] in the next revision.
> 
> Ohhh... yes, I keep forgetting that Fengguang will need this tag for
> his build bot.
> 
> But I have one more doubt. If scif_nodeqp_send() fails then shouldn't we
> remove these two nodes of the p2p connection from the lists and release
> the resources?

The resources will get cleaned up via scif_destroy_p2p(..) once they are
added to the lists.

Sudeep Dutt