Re: [PATCH] 2.6.1 tg3 DMA engine test failure

[Grant Grundler]

On Fri, Jan 23, 2004 at 11:32:41PM -0800, David S. Miller wrote:
>    From: Grant Grundler <grundler@xxxxxxxxxxxxxxxx>
>    Date: Sat, 24 Jan 2004 00:30:32 -0700
> 
>    My gut feeling is if linux aligns or pads things nicely for any reason,
>    then the bye enables don't get used or clobber padding.
>    
> If the packet data length is an odd number of bytes, there is nothing
> we can do about this, and the newer tigon3 chips are going to use a
> cacheline burst for the end of the packet with the trailing byte
> enables turned off.  I've seen this myself and sparc64 PCI controllers
> generate a streaming byte hole error interrupt when it occurs and I
> get messages logged in dmesg :)

Ugh. Ok.

I was concerned about about the buffer not ending on a well aligned
boundary and someone elses data getting clobbered. But RX buffers are
managed by the OS and I expect them to be well aligned.

>    Maybe keep a shorter note about the bit changed meaning in later models
>    just to document the issues.
>    
> We can "document it" by having the setting of this bit be protected by
> chip version numbers.  I'd happily accept such a patch.

Well, you pointed out it needs to be set on bcm5700/01 for it's 
intended purpose.  And it needs to be set on 5703/04 to enable
PCI-X bug workaround.  BE bit "always" needs to be set.
Did you intend to alias the BE constant to another name?
(and then use respective chip version for each constant)

BTW, I don't know what the story is with bcm5705.

thanks,
grant
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html