linux-kernel - RE: [PATCH net] net: enetc: correct the value of ENETC_RXB

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID:
 <PAXPR04MB85106C49873A85DFB61DFD7A88EBA@PAXPR04MB8510.eurprd04.prod.outlook.com>
Date: Tue, 14 Oct 2025 10:00:14 +0000
From: Wei Fang <wei.fang@....com>
To: Claudiu Manoil <claudiu.manoil@....com>, Vladimir Oltean
	<vladimir.oltean@....com>
CC: Clark Wang <xiaoning.wang@....com>, "andrew+netdev@...n.ch"
	<andrew+netdev@...n.ch>, "davem@...emloft.net" <davem@...emloft.net>,
	"edumazet@...gle.com" <edumazet@...gle.com>, "kuba@...nel.org"
	<kuba@...nel.org>, "pabeni@...hat.com" <pabeni@...hat.com>, Frank Li
	<frank.li@....com>, "imx@...ts.linux.dev" <imx@...ts.linux.dev>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH net] net: enetc: correct the value of ENETC_RXB_TRUESIZE

> > > > On Fri, Oct 10, 2025 at 05:26:08PM +0800, Wei Fang wrote:
> > > > > ENETC_RXB_TRUESIZE indicates the size of half a page, but the page
> > > > > size is adjustable, for ARM64 platform, the PAGE_SIZE can be 4K,
> > > > > 16K and 64K, so a fixed value '2048' is not correct when the
> > > > > PAGE_SIZE is 16K or 64K.
> > > > >
> > > > > Fixes: d4fd0404c1c9 ("enetc: Introduce basic PF and VF ENETC
> > > > > ethernet
> > > > > drivers")
> > > > > Signed-off-by: Wei Fang <wei.fang@....com>
> > > > > ---
> > > > >  drivers/net/ethernet/freescale/enetc/enetc.h | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/drivers/net/ethernet/freescale/enetc/enetc.h
> > > > > b/drivers/net/ethernet/freescale/enetc/enetc.h
> > > > > index 0ec010a7d640..f279fa597991 100644
> > > > > --- a/drivers/net/ethernet/freescale/enetc/enetc.h
> > > > > +++ b/drivers/net/ethernet/freescale/enetc/enetc.h
> > > > > @@ -76,7 +76,7 @@ struct enetc_lso_t {
> > > > >  #define ENETC_LSO_MAX_DATA_LEN		SZ_256K
> > > > >
> > > > >  #define ENETC_RX_MAXFRM_SIZE	ENETC_MAC_MAXFRM_SIZE
> > > > > -#define ENETC_RXB_TRUESIZE	2048 /* PAGE_SIZE >> 1 */
> > > > > +#define ENETC_RXB_TRUESIZE	(PAGE_SIZE >> 1)
> > > > >  #define ENETC_RXB_PAD		NET_SKB_PAD /* add extra space if
> > > > needed */
> > > > >  #define ENETC_RXB_DMA_SIZE	\
> > > > >  	(SKB_WITH_OVERHEAD(ENETC_RXB_TRUESIZE) - ENETC_RXB_PAD)
> > > > > --
> > > > > 2.34.1
> > > > >
> > > >
> > > > I wonder why 2048 was preferred, even though PAGE_SIZE >> 1 was in a
> > > > comment.
> > > > Claudiu, do you remember?
> > >
> > > Initial driver implementation for enetcv1 was bound to 4k pages, I
> > > need to recheck why and get back to you.
> > >
> >
> > Any updates?
> 
> Reviewed RXB_TRUESIZE usage, and confirmed that it's not only used for
> building
> skbs and in the page halves flipping mechanism, but this setting is also
> accordingly
> propagated to the hardware level Rx buffer size configuration (e.g. for Rx S/G).
> The allowed h/w Rx buffer size can be up to 64KB, so we would be safe even
> with
> 128K pages.
> The 'PAGE_SIZE >> 1' comment is actually a TODO item, and I crudely limited
> TRUESIZE to 2K to enforce 4K pages in the initial driver (i.e. for simplification).
> 
> So, pls go ahead with the configurable RX_TRUESIZE, and note that this opens
> up
> the driver to new usage / performance scenarios, and user configurable Rx
> buffer
> sizes.
> 

Thanks for the update, since PAGE_SIZE is configurable, so the old definition
of RXB_TRUESIZE is not consistent with the page halves flipping mechanism
if the PAGE_SIZE is not 4K, I would like to fix it in stable kernels, due to users
may change PAGE_SIZE for some use cases. Then I will add a separate patch
to support changing the RX buffer size at runtime.