diff mbox series

net: korina: free array used for rx/tx descriptors

Message ID 20201011113955.19511-1-vvidic@valentin-vidic.from.hr (mailing list archive)
State Not Applicable
Headers show
Series net: korina: free array used for rx/tx descriptors | expand

Commit Message

Valentin Vidić Oct. 11, 2020, 11:39 a.m. UTC
Memory was not freed when driver is unloaded from the kernel.

Signed-off-by: Valentin Vidic <vvidic@valentin-vidic.from.hr>
---
 drivers/net/ethernet/korina.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Willem de Bruijn Oct. 11, 2020, 6:37 p.m. UTC | #1
On Sun, Oct 11, 2020 at 7:46 AM Valentin Vidic
<vvidic@valentin-vidic.from.hr> wrote:
>
> Memory was not freed when driver is unloaded from the kernel.
>
> Signed-off-by: Valentin Vidic <vvidic@valentin-vidic.from.hr>

Makes sense.

Fixes: ef11291bcd5f ("Add support the Korina (IDT RC32434) Ethernet MAC")

Slightly off-topic, but I don't fully fathom what goes on with this
pointer straight after the initial kmalloc.

        lp->td_ring = (struct dma_desc *)KSEG1ADDR(lp->td_ring);

> ---
>  drivers/net/ethernet/korina.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/net/ethernet/korina.c b/drivers/net/ethernet/korina.c
> index 03e034918d14..99146145f020 100644
> --- a/drivers/net/ethernet/korina.c
> +++ b/drivers/net/ethernet/korina.c
> @@ -1133,6 +1133,7 @@ static int korina_remove(struct platform_device *pdev)
>         iounmap(lp->eth_regs);
>         iounmap(lp->rx_dma_regs);
>         iounmap(lp->tx_dma_regs);
> +       kfree(lp->td_ring);
>
>         unregister_netdev(bif->dev);
>         free_netdev(bif->dev);

In general it is nice to release in reverse of acquire. But the driver
already does not follow this practice.
Valentin Vidić Oct. 11, 2020, 9:21 p.m. UTC | #2
On Sun, Oct 11, 2020 at 02:37:33PM -0400, Willem de Bruijn wrote:
> Slightly off-topic, but I don't fully fathom what goes on with this
> pointer straight after the initial kmalloc.
> 
>         lp->td_ring = (struct dma_desc *)KSEG1ADDR(lp->td_ring);

KSEG1ADDR should rewrite the memory address into the uncached region
for memory mapped I/O. Not sure if this would case problems for kfree
since there is another kfree on the fail path:

probe_err_register:
        kfree(lp->td_ring);
diff mbox series

Patch

diff --git a/drivers/net/ethernet/korina.c b/drivers/net/ethernet/korina.c
index 03e034918d14..99146145f020 100644
--- a/drivers/net/ethernet/korina.c
+++ b/drivers/net/ethernet/korina.c
@@ -1133,6 +1133,7 @@  static int korina_remove(struct platform_device *pdev)
 	iounmap(lp->eth_regs);
 	iounmap(lp->rx_dma_regs);
 	iounmap(lp->tx_dma_regs);
+	kfree(lp->td_ring);
 
 	unregister_netdev(bif->dev);
 	free_netdev(bif->dev);