[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANr-f5yiJG_+K63NpSuiAefqM4Cb=UWEdBGKv6mkS5fMZ96new@mail.gmail.com>
Date: Tue, 16 Nov 2021 22:47:22 +0100
From: Gerhard Engleder <gerhard@...leder-embedded.com>
To: Heiner Kallweit <hkallweit1@...il.com>
Cc: David Miller <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>, Andrew Lunn <andrew@...n.ch>,
netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next v5 3/3] tsnep: Add TSN endpoint Ethernet MAC driver
On Mon, Nov 15, 2021 at 11:21 PM Heiner Kallweit <hkallweit1@...il.com> wrote:
> On 15.11.2021 21:50, Gerhard Engleder wrote:
> > + /* entry->properties shall be valid before write pointer is
> > + * incrememted
> > + */
> > + wmb();
>
> A lighter variant like smp_wmb() is not sufficient here?
This is some leftover from lockless descriptor ring processing, which has
been replaced by a spinlock. I will remove it.
> > + dma_rmb();
> > +
>
> Memory barriers should always have a comment explaining why
> they are needed. I think checkpatch would complain here.
I don't see any checkpatch warning, but you are right. I will document
all DMA memory barriers.
Gerhard
Powered by blists - more mailing lists