linux-kernel - Re: sky2 panic in 2.6.32.1 under load (new oops)

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: <4B3A4C8A.6080001@majjas.com>
Date:	Tue, 29 Dec 2009 13:38:02 -0500
From:	Michael Breuer <mbreuer@...jas.com>
To:	Stephen Hemminger <shemminger@...ux-foundation.org>
Cc:	Andrew Morton <akpm@...ux-foundation.org>,
	"Berck E. Nash" <flyboy@...il.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	netdev@...r.kernel.org
Subject: Re: sky2 panic in 2.6.32.1 under load (new oops)

The frequency of the errors is reduced, but in place I'm seeing a large 
number of DHCPINFORM/DHCPACK to/from the system from which I'm 
generating load.

It appears that it's now taking a higher throughput to cause the 
interrupt status errors than without this patch, but they're still 
occurring. Just for fun, I'll try without this patch and with VT-D 
disabled later today.


On 12/29/2009 12:30 PM, Stephen Hemminger wrote:
> There could be DMA problems, either because chip isn't remembering
> upper address, or because of lack of wires from chip to memory controller;
> there have been problems on motherboards where I/O devices couldn't access
> all of memory.
>
> Try this, it forces each transmit to have full 64 bit address. It drops an
> optimization that was done to reduce the number of transmit descriptors
> used.
>
> --- a/drivers/net/sky2.c	2009-12-29 09:23:37.114074275 -0800
> +++ b/drivers/net/sky2.c	2009-12-29 09:26:38.699912035 -0800
> @@ -1038,10 +1038,11 @@ static void tx_init(struct sky2_port *sk
>   	sky2->tx_tcpsum = 0;
>   	sky2->tx_last_mss = 0;
>
> -	le = get_tx_le(sky2,&sky2->tx_prod);
> -	le->addr = 0;
> -	le->opcode = OP_ADDR64 | HW_OWNER;
> -	sky2->tx_last_upper = 0;
> +	if (sizeof(dma_addr_t) == sizeof(u32)) {
> +		le = get_tx_le(sky2,&sky2->tx_prod);
> +		le->addr = 0;
> +		le->opcode = OP_ADDR64 | HW_OWNER;
> +	}
>   }
>
>   /* Update chip's next pointer */
> @@ -1669,10 +1670,9 @@ static netdev_tx_t sky2_xmit_frame(struc
>
>   	/* Send high bits if needed */
>   	upper = upper_32_bits(mapping);
> -	if (upper != sky2->tx_last_upper) {
> +	if (sizeof(mapping)>  sizeof(u32)) {
>   		le = get_tx_le(sky2,&slot);
>   		le->addr = cpu_to_le32(upper);
> -		sky2->tx_last_upper = upper;
>   		le->opcode = OP_ADDR64 | HW_OWNER;
>   	}
>
> @@ -1762,10 +1762,9 @@ static netdev_tx_t sky2_xmit_frame(struc
>   			goto mapping_unwind;
>
>   		upper = upper_32_bits(mapping);
> -		if (upper != sky2->tx_last_upper) {
> +		if (sizeof(mapping)>  sizeof(u32)) {
>   			le = get_tx_le(sky2,&slot);
>   			le->addr = cpu_to_le32(upper);
> -			sky2->tx_last_upper = upper;
>   			le->opcode = OP_ADDR64 | HW_OWNER;
>   		}
>
> --- a/drivers/net/sky2.h	2009-12-29 09:23:37.164072886 -0800
> +++ b/drivers/net/sky2.h	2009-12-29 09:25:15.302197993 -0800
> @@ -2191,7 +2191,6 @@ struct sky2_port {
>
>   	u16		     tx_pending;
>   	u16		     tx_last_mss;
> -	u32		     tx_last_upper;
>   	u32		     tx_tcpsum;
>
>   	struct rx_ring_info  *rx_ring ____cacheline_aligned_in_smp;
>    

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/