netdev - Re: [PATCH 1/6] net: davinci_emac: Fix hangs with interrupts

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

Message-ID: <20150113193634.GQ16533@saruman>
Date:	Tue, 13 Jan 2015 13:36:34 -0600
From:	Felipe Balbi <balbi@...com>
To:	Tony Lindgren <tony@...mide.com>
CC:	David Miller <davem@...emloft.net>, <netdev@...r.kernel.org>,
	<linux-omap@...r.kernel.org>,
	Brian Hutchinson <b.hutchman@...il.com>,
	Felipe Balbi <balbi@...com>
Subject: Re: [PATCH 1/6] net: davinci_emac: Fix hangs with interrupts

On Tue, Jan 13, 2015 at 11:29:23AM -0800, Tony Lindgren wrote:
> On davinci_emac, we have pulse interrupts. This means that we need to
> clear the EOI bits when disabling interrupts as otherwise the interrupts
> keep happening. And we also need to not clear the EOI bits again when
> enabling the interrupts as otherwise we will get tons of:
> 
> unexpected IRQ trap at vector 00
> 
> These errors almost certainly mean that the omap-intc.c is signaling
> a spurious interrupt with the reserved irq 127 as we've seen earlier
> on omap3.
> 
> Let's fix the issue by clearing the EOI bits when disabling the
> interrupts. Let's also keep the comment for "Rx Threshold and Misc
> interrupts are not enabled" for both enable and disable so people
> are aware of this when potentially adding more support.
> 
> Note that eventually we should handle the RX and TX interrupts
> separately like cpsw is now doing. However, so far I have not seen
> any issues with this based on my testing, so it seems to behave a
> little different compared to the cpsw that had a similar issue.
> 
> Cc: Brian Hutchinson <b.hutchman@...il.com>
> Cc: Felipe Balbi <balbi@...com>

pretty much the same thing that happens with CPSW, I think that a future
patch might want to change things so that we only write EOI to the IRQ
that actually fires, though.

Reviewed-by: Felipe Balbi <balbi@...com>

> Signed-off-by: Tony Lindgren <tony@...mide.com>
> ---
>  drivers/net/ethernet/ti/davinci_emac.c | 19 ++++++++++---------
>  1 file changed, 10 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/net/ethernet/ti/davinci_emac.c b/drivers/net/ethernet/ti/davinci_emac.c
> index ea71251..383ed52 100644
> --- a/drivers/net/ethernet/ti/davinci_emac.c
> +++ b/drivers/net/ethernet/ti/davinci_emac.c
> @@ -922,6 +922,16 @@ static void emac_int_disable(struct emac_priv *priv)
>  		if (priv->int_disable)
>  			priv->int_disable();
>  
> +		/* NOTE: Rx Threshold and Misc interrupts are not enabled */
> +
> +		/* ack rxen only then a new pulse will be generated */
> +		emac_write(EMAC_DM646X_MACEOIVECTOR,
> +			EMAC_DM646X_MAC_EOI_C0_RXEN);
> +
> +		/* ack txen- only then a new pulse will be generated */
> +		emac_write(EMAC_DM646X_MACEOIVECTOR,
> +			EMAC_DM646X_MAC_EOI_C0_TXEN);
> +
>  		local_irq_restore(flags);
>  
>  	} else {
> @@ -951,15 +961,6 @@ static void emac_int_enable(struct emac_priv *priv)
>  		 * register */
>  
>  		/* NOTE: Rx Threshold and Misc interrupts are not enabled */
> -
> -		/* ack rxen only then a new pulse will be generated */
> -		emac_write(EMAC_DM646X_MACEOIVECTOR,
> -			EMAC_DM646X_MAC_EOI_C0_RXEN);
> -
> -		/* ack txen- only then a new pulse will be generated */
> -		emac_write(EMAC_DM646X_MACEOIVECTOR,
> -			EMAC_DM646X_MAC_EOI_C0_TXEN);
> -
>  	} else {
>  		/* Set DM644x control registers for interrupt control */
>  		emac_ctrl_write(EMAC_CTRL_EWCTL, 0x1);
> -- 
> 2.1.4
> 

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)