netdev
[Top] [All Lists]

[PATCH 0/3] e100: Enable receiving bogus packets and saving FCS

Subject: [PATCH 0/3] e100: Enable receiving bogus packets and saving FCS
From: Ben Greear <greearb@xxxxxxxxxxxxxxx>
Date: Mon, 24 Nov 2003 23:53:11 -0800
Cc: netdev@xxxxxxxxxxx
In-reply-to: <20031124173330.096c0751.davem@redhat.com>
Organization: Candela Technologies
References: <3FC2931B.3070903@candelatech.com> <20031124152933.1420f6cf.davem@redhat.com> <3FC298C9.8080302@candelatech.com> <20031124173330.096c0751.davem@redhat.com>
Sender: netdev-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.5) Gecko/20031007
David S. Miller wrote:
On Mon, 24 Nov 2003 15:48:25 -0800
Ben Greear <greearb@xxxxxxxxxxxxxxx> wrote:


So, RX-ALL can be one flag, another for RX-FCS, and the TX-CUSTOM-FCS
can be a separate patch altogether?


That is the basic idea, yes.

Here are three patches. These were extracted manually from my big patchset. Where flags are defined, I left in the pktgen receive flag and the send-to-self flag from my larger patch set. If this is unacceptable, I can shuffle the flags around....

rx_all_ethtool.patch is the ethtool glue to turn on/off the RX-ALL and SAVE-FCS
flags.  I have not tested any of the ethtool-ops code branches as e100 and e1000
does not appear to use it.

rx_all_headers.patch should take care of defining the flags in if.h and 
netdevice.h

rx_all_e100.patch converts the e100 driver to support these features.  I've
ripped out the tx-custom-fcs code, but the rx-all and save-fcs code is there.
The default behaviour is virtually identical, because while the old code
was enabling receiving a lot of bad packets when in PROMISC mode, the check
in e100_main rx code was throwing away anything that was not marked as good
anyway...

However, one thing I want to make absolutely clear is that I do not
want multiple ways to get the same information out of the kernel.
We decided to build netlink into the kernel always because this means
it is present in everyone's kernel.  And therefore, we don't need to
add another way to obtain the same config information already
available via netlink.  The existing cases where an ioctl() style
call exists to obtain the same info available via netlink is merely
for compatability with older BSD tools.  No new such things will be
added.

Ok, I won't ask again. I attempted to get rid of all the 'ndstats' code from these patches. If some is still in there, it's an accident and I'll remove it.

Thanks,
Ben


-- Ben Greear <greearb@xxxxxxxxxxxxxxx> Candela Technologies Inc http://www.candelatech.com



<Prev in Thread] Current Thread [Next in Thread>