Date: Wed, 29 May 2024 17:50:49 +0000
From: Sunil Kovvuri Goutham <sgoutham@...vell.com>
To: Bc-bocun Chen (陳柏村)
	<bc-bocun.chen@...iatek.com>,
        "daniel@...rotopia.org" <daniel@...rotopia.org>
CC: Mark-MC Lee (李明昌) <Mark-MC.Lee@...iatek.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-mediatek@...ts.infradead.org" <linux-mediatek@...ts.infradead.org>,
        SkyLake Huang (黃啟澤)
	<SkyLake.Huang@...iatek.com>,
        Sam Shih (史碩三)
	<Sam.Shih@...iatek.com>,
        "linux@...web.de" <linux@...web.de>, "nbd@....name"
	<nbd@....name>,
        "john@...ozen.org" <john@...ozen.org>,
        "lorenzo@...nel.org"
	<lorenzo@...nel.org>,
        "frank-w@...lic-files.de" <frank-w@...lic-files.de>,
        Sean Wang <Sean.Wang@...iatek.com>,
        "kuba@...nel.org" <kuba@...nel.org>,
        "edumazet@...gle.com" <edumazet@...gle.com>,
        "pabeni@...hat.com"
	<pabeni@...hat.com>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Steven Liu (劉人豪) <steven.liu@...iatek.com>,
        "matthias.bgg@...il.com" <matthias.bgg@...il.com>,
        "davem@...emloft.net"
	<davem@...emloft.net>,
        "linux-arm-kernel@...ts.infradead.org"
	<linux-arm-kernel@...ts.infradead.org>,
        "angelogioacchino.delregno@...labora.com"
	<angelogioacchino.delregno@...labora.com>
Subject: RE: [net v2] net: ethernet: mtk_eth_soc: handle dma buffer size soc
 specific

On Mon, 2024-05-27 at 17:13 +0100, Daniel Golle wrote:
> > On Mon, May 27, 2024 at 03:55:55PM GMT, Sunil Kovvuri Goutham
> wrote:
> > > >
> > > >
> > > > > > -----Original Message-----
> > > > > > From: Frank Wunderlich <linux@...web.de>
> > > > > > Sent: Monday, May 27, 2024 7:52 PM
> > > > > > To: Felix Fietkau <nbd@....name>; Sean Wang <
> > > > > > sean.wang@...iatek.com>;
> > > > > > Mark Lee <Mark-MC.Lee@...iatek.com>; Lorenzo Bianconi
> > > > > > <lorenzo@...nel.org>; David S. Miller <davem@...emloft.net>
> > > ; Eric
> > > > > > Dumazet
> > > > > > <edumazet@...gle.com>; Jakub Kicinski <kuba@...nel.org>;
> > > Paolo
> > > > > > Abeni
> > > > > > <pabeni@...hat.com>; Matthias Brugger <
> > > matthias.bgg@...il.com>;
> > > > > > AngeloGioacchino Del Regno <
> > > > > > angelogioacchino.delregno@...labora.com>
> > > > > > Cc: Frank Wunderlich <frank-w@...lic-files.de>; John
> > > Crispin
> > > > > > <john@...ozen.org>; netdev@...r.kernel.org; 
> > > > > > linux-kernel@...r.kernel.org;
> > > > > > linux-arm-kernel@...ts.infradead.org; 
> > > > > > linux-mediatek@...ts.infradead.org;
> > > > > > Daniel Golle <daniel@...rotopia.org>
> > > > > > Subject: [net v2] net: ethernet: mtk_eth_soc: handle dma
> > > buffer
> > > > > > size soc specific
> > > > > >
> > > > > > From: Frank Wunderlich <frank-w@...lic-files.de>
> > > > > >
> > > > > > The mainline MTK ethernet driver suffers long time from
> > > rarly but
> > > > > > annoying tx
> > > > > > queue timeouts. We think that this is caused by fixed dma
> > > sizes
> > > > > > hardcoded for
> > > > > > all SoCs.
> > > > > >
> > > > > > Use the dma-size implementation from SDK in a per SoC
> > > manner.
> > > > > >
> > > > > > Fixes: 656e705243fd ("net-next: mediatek: add support for
> > > MT7623
> > > > > > ethernet")
> > > > > > Suggested-by: Daniel Golle <daniel@...rotopia.org>
> > > > > > Signed-off-by: Frank Wunderlich <frank-w@...lic-files.de>
> > 
> > > >
> > > > ..............
> > > > > >
> > > > > > diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> > > > > > b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> > > > > > index cae46290a7ae..f1ff1be73926 100644
> > > > > > --- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> > > > > > +++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
> > 
> > > >
> > > > .............
> > > > > > @@ -1142,40 +1142,46 @@ static int mtk_init_fq_dma(struct
> > > mtk_eth
> > > > > > *eth)
> > > > > >                             cnt * soc-
> > > > > > >tx.desc_size,
> > > > > >                             &eth-
> > > > > > >phy_scratch_ring,
> > > > > >                             GFP_KERNEL);
> > 
> > > >
> > > > ..............
> > > > > > -  for (i = 0; i < cnt; i++) {
> > > > > > -      dma_addr_t addr = dma_addr + i * MTK_QDMA_PAGE_SIZE;
> > > > > > -      struct mtk_tx_dma_v2 *txd;
> > > > > > +      dma_addr = dma_map_single(eth->dma_dev,
> > > > > > +                  eth->scratch_head[j], len *
> > > > > > MTK_QDMA_PAGE_SIZE,
> > > > > > +                  DMA_FROM_DEVICE);
> > > > > >
> > 
> > > >
> > > > As per commit msg, the fix is for transmit queue timeouts.
> > > > But the DMA buffer changes seems for receive pkts.
> > > > Can you please elaborate the connection here.
> 
> >
> > *I guess* the memory window used for both, TX and RX DMA
> descriptors
> > needs to be wisely split to not risk TX queue overruns, depending
> on
> > the
> > SoC speed and without hurting RX performance...
> >
> > Maybe someone inside MediaTek (I've added to Cc now) and more
> > familiar
> > with the design can elaborate in more detail.
 
>We've encountered a transmit queue timeout issue on the MT79888 and
>have identified it as being related to the RSS feature.
>We suspect this problem arises from a low level of free TX DMADs, the
>TX Ring alomost full.
>Since RSS is enabled, there are 4 Rx Rings, with each containing 2048
>DMADs, totaling 8192 for Rx. In contrast, the Tx Ring has only 2048
>DMADs. Tx DMADs will be consumed rapidly during a 10G LAN to 10G WAN
>forwarding test, subsequently causing the transmit queue to stop.
>Therefore, we reduced the number of Rx DMADs for each ring to balance
>both Tx and Rx DMADs, which resolves this issue.

Okay, but it’s still not clear why it’s resulting in a transmit timeout.
When transmit queue is stopped and after some Tx pkts in the pipeline are flushed out, isn’t
Tx queue wakeup not happening ?
 
 
Thanks,
Sunil.

 

Powered by blists - more mailing lists