[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20170213120912.GA24412@verge.net.au>
Date: Mon, 13 Feb 2017 13:09:13 +0100
From: Simon Horman <horms@...ge.net.au>
To: Geert Uytterhoeven <geert@...ux-m68k.org>
Cc: Sergei Shtylyov <sergei.shtylyov@...entembedded.com>,
Magnus Damm <magnus.damm@...il.com>,
Netdev <netdev@...r.kernel.org>,
Linux-Renesas <linux-renesas-soc@...r.kernel.org>
Subject: Re: [PATCH/RFC v3 net] ravb: unmap descriptors when freeing rings
On Fri, Feb 10, 2017 at 02:52:50PM +0100, Geert Uytterhoeven wrote:
> On Wed, Jan 25, 2017 at 5:18 PM, Sergei Shtylyov
> <sergei.shtylyov@...entembedded.com> wrote:
> > On 01/24/2017 09:21 PM, Simon Horman wrote:
> >
> >> From: Kazuya Mizuguchi <kazuya.mizuguchi.ks@...esas.com>
> >>
> >> "swiotlb buffer is full" errors occur after repeated initialisation of a
> >> device - f.e. suspend/resume or ip link set up/down. This is because
> >> memory
> >> mapped using dma_map_single() in ravb_ring_format() and ravb_start_xmit()
> >> is not released. Resolve this problem by unmapping descriptors when
> >> freeing rings.
> >
> >
> > Could you look into the sh_eth driver which seems to have the same issue?
>
> Indeed, after a few suspend/resume cycles on r8a7791/koelsch:
>
> WARNING: CPU: 1 PID: 1699 at lib/dma-debug.c:517 add_dma_entry+0xfc/0x148
> DMA-API: exceeded 7 overlapping mappings of cacheline 0x0000000001a827e3
Thanks for confirming that. It matches my expectation after
reading of the sh_eth code.
Powered by blists - more mailing lists