drivers/net/ethernet/microchip/lan743x_main.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
The driver allocates ring elements using GFP_ATOMIC and GFP_DMA
flags. The allocation is not done in atomic context and there is
no dependency from LAN743x hardware on memory allocation should be
in DMA_ZONE. Hence modifying the flags to use only GFP_KERNEL.
Signed-off-by: Thangaraj Samynathan <thangaraj.s@microchip.com>
---
drivers/net/ethernet/microchip/lan743x_main.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index 23760b613d3e..c10b0131d5fb 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -2495,8 +2495,7 @@ static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
/* save existing skb, allocate new skb and map to dma */
skb = buffer_info->skb;
- if (lan743x_rx_init_ring_element(rx, rx->last_head,
- GFP_ATOMIC | GFP_DMA)) {
+ if (lan743x_rx_init_ring_element(rx, rx->last_head, GFP_KERNEL)) {
/* failed to allocate next skb.
* Memory is very low.
* Drop this packet and reuse buffer.
--
2.25.1
On 14.03.25 08:02, Thangaraj Samynathan wrote:
> The driver allocates ring elements using GFP_ATOMIC and GFP_DMA
> flags. The allocation is not done in atomic context and there is
> no dependency from LAN743x hardware on memory allocation should be
> in DMA_ZONE. Hence modifying the flags to use only GFP_KERNEL.
>
> Signed-off-by: Thangaraj Samynathan <thangaraj.s@microchip.com>
> ---
> drivers/net/ethernet/microchip/lan743x_main.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
> index 23760b613d3e..c10b0131d5fb 100644
> --- a/drivers/net/ethernet/microchip/lan743x_main.c
> +++ b/drivers/net/ethernet/microchip/lan743x_main.c
> @@ -2495,8 +2495,7 @@ static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
>
> /* save existing skb, allocate new skb and map to dma */
> skb = buffer_info->skb;
> - if (lan743x_rx_init_ring_element(rx, rx->last_head,
> - GFP_ATOMIC | GFP_DMA)) {
> + if (lan743x_rx_init_ring_element(rx, rx->last_head, GFP_KERNEL)) {
I agree with removing GFP_DMA. If it would be needed, then everywhere
and not only here in NAPI context as it is intended for hardware
limitations.
I'm not sure if GFP_ATOMIC can be removed. Isn't NAPI an atomic context?
For example napi_alloc_skb() and page_pool_dev_alloc_pages() use
GFP_ATOMIC.
Gerhard
Hi Gerhard,
Thanks for reviewing the patch. Please find my response inline.
On Fri, 2025-03-14 at 21:37 +0100, Gerhard Engleder wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you
> know the content is safe
>
> On 14.03.25 08:02, Thangaraj Samynathan wrote:
> > The driver allocates ring elements using GFP_ATOMIC and GFP_DMA
> > flags. The allocation is not done in atomic context and there is
> > no dependency from LAN743x hardware on memory allocation should be
> > in DMA_ZONE. Hence modifying the flags to use only GFP_KERNEL.
> >
> > Signed-off-by: Thangaraj Samynathan <thangaraj.s@microchip.com>
> > ---
> > drivers/net/ethernet/microchip/lan743x_main.c | 3 +--
> > 1 file changed, 1 insertion(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/microchip/lan743x_main.c
> > b/drivers/net/ethernet/microchip/lan743x_main.c
> > index 23760b613d3e..c10b0131d5fb 100644
> > --- a/drivers/net/ethernet/microchip/lan743x_main.c
> > +++ b/drivers/net/ethernet/microchip/lan743x_main.c
> > @@ -2495,8 +2495,7 @@ static int lan743x_rx_process_buffer(struct
> > lan743x_rx *rx)
> >
> > /* save existing skb, allocate new skb and map to dma */
> > skb = buffer_info->skb;
> > - if (lan743x_rx_init_ring_element(rx, rx->last_head,
> > - GFP_ATOMIC | GFP_DMA)) {
> > + if (lan743x_rx_init_ring_element(rx, rx->last_head,
> > GFP_KERNEL)) {
>
> I agree with removing GFP_DMA. If it would be needed, then everywhere
> and not only here in NAPI context as it is intended for hardware
> limitations.
>
> I'm not sure if GFP_ATOMIC can be removed. Isn't NAPI an atomic
> context?
> For example napi_alloc_skb() and page_pool_dev_alloc_pages() use
> GFP_ATOMIC.
>
Yes, you are right. GFP ATOMIC cannot be removed. Will address this in
the next revision of the patch
> Gerhard
>
© 2016 - 2025 Red Hat, Inc.