All of lore.kernel.org
 help / color / mirror / Atom feed
From: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
To: Marios Makassikis <mmakassikis@freebox.fr>
Cc: netdev@vger.kernel.org, lorenzo@kernel.org, mcroce@microsoft.com,
	marcin.s.wojtas@gmail.com, linux@armlinux.org.uk,
	andrew+netdev@lunn.ch, davem@davemloft.net, edumazet@google.com,
	kuba@kernel.org, pabeni@redhat.com, linux-kernel@vger.kernel.org
Subject: Re: [REPOST PATCH] drivers: net: mvpp2: attempt to refill rx before allocating skb
Date: Tue, 27 May 2025 09:52:39 +0200	[thread overview]
Message-ID: <aDVvR8XXPMEUzztC@lore-desk> (raw)
In-Reply-To: <20250521171021.2978273-2-mmakassikis@freebox.fr>

[-- Attachment #1: Type: text/plain, Size: 2743 bytes --]

> on mvpp2_rx_refill() failure, the freshly allocated skb is freed,
> the rx error counter is incremented and the descriptor currently
> being processed is rearmed through mvpp2_bm_pool_put().
> 
> the logic is that the system is low on memory so it's not possible
> to allocate both a rx descriptor and an skb, so we might as well
> drop the skb and return the descriptor to the rx pool to avoid
> draining it (and preventing any future packet reception).

Hi Marios,

Can we just run the mvpp2_rx_refill() when the skb has been already
successfully sent to the networking stack? Doing so, if the skb allocation
fails, we will refill the in-flight rx descriptor (the one consumed by the
skb) to the descriptor ring, reducing the pressure on it.
What do you think?

Regards,
Lorenzo

> 
> the skb freeing is unfortunate, as build_skb() takes ownership
> of the 'data' buffer:
>  - build_skb() calls  __finalize_skb_around() which sets skb->head
>  and skb->data to point to 'data'
>  - dev_free_skb_any() may call skb_free_frag() on skb->head
> 
> thus, the final mvpp2_bm_pool_put() rearms a descriptor that was
> just freed.
> 
> call mvpp2_rx_refill() first, so there's no skb to free.
> 
> incidentally, doing rx refill prior to skb allocation is what is
> done in marvell's mvneta driver for armada 370 (mvneta_rx_hwbm() in
> mvneta.c)
> 
> Fixes: d6526926de739 ("net: mvpp2: fix memory leak in mvpp2_rx")
> Signed-off-by: Marios Makassikis <mmakassikis@freebox.fr>
> ---
>  drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> index 416a926a8281..e13055ec4483 100644
> --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
> @@ -4003,6 +4003,12 @@ static int mvpp2_rx(struct mvpp2_port *port, struct napi_struct *napi,
>  			metasize = xdp.data - xdp.data_meta;
>  		}
>  
> +		err = mvpp2_rx_refill(port, bm_pool, pp, pool);
> +		if (err) {
> +			netdev_err(port->dev, "failed to refill BM pools\n");
> +			goto err_drop_frame;
> +		}
> +
>  		if (frag_size)
>  			skb = build_skb(data, frag_size);
>  		else
> @@ -4021,13 +4027,6 @@ static int mvpp2_rx(struct mvpp2_port *port, struct napi_struct *napi,
>  					 skb_hwtstamps(skb));
>  		}
>  
> -		err = mvpp2_rx_refill(port, bm_pool, pp, pool);
> -		if (err) {
> -			netdev_err(port->dev, "failed to refill BM pools\n");
> -			dev_kfree_skb_any(skb);
> -			goto err_drop_frame;
> -		}
> -
>  		if (pp)
>  			skb_mark_for_recycle(skb);
>  		else
> -- 
> 2.49.0
> 
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

      reply	other threads:[~2025-05-27  7:52 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-14 17:34 [PATCH] drivers: net: mvpp2: attempt to refill rx before allocating skb Marios Makassikis
2025-05-15 15:00 ` Jakub Kicinski
2025-05-21 17:10   ` [REPOST PATCH] " Marios Makassikis
2025-05-27  7:52     ` Lorenzo Bianconi [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=aDVvR8XXPMEUzztC@lore-desk \
    --to=lorenzo.bianconi@redhat.com \
    --cc=andrew+netdev@lunn.ch \
    --cc=davem@davemloft.net \
    --cc=edumazet@google.com \
    --cc=kuba@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux@armlinux.org.uk \
    --cc=lorenzo@kernel.org \
    --cc=marcin.s.wojtas@gmail.com \
    --cc=mcroce@microsoft.com \
    --cc=mmakassikis@freebox.fr \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.