From: Michael Durket <hidden> Date: 2011-02-17 13:51:08
I'm trying to fix a problem with packet drops and frame errors (really rxbds_empty and rx_discards (or rx_fw_discards depending on whether or not its a tg3 supported chip or a bnx2 chip)). We have this happening on pretty much all the RedHat 5.x systems we're running these drivers on. In investigating what remedies to employ to fix this, I'm trying to determine if new versions of this driver are supplied by Broadcom, or by people within the Linux community so
I can find the right place to ask questions about it. I'm pretty sure this isn't the right list for these types of questions, but if someone could tell me whether I should be contacting Broadcom directly, or the name of some other Linux mailing list or forum that might help, I'd appreciate it.
From: Ben Hutchings <hidden> Date: 2011-02-17 14:16:35
On Thu, 2011-02-17 at 05:24 -0800, Michael Durket wrote:
I'm trying to fix a problem with packet drops and frame errors (really
rxbds_empty and rx_discards (or rx_fw_discards depending on whether or
not its a tg3 supported chip or a bnx2 chip)). We have this happening
on pretty much all the RedHat 5.x systems we're running these drivers
on. In investigating what remedies to employ to fix this, I'm trying
to determine if new versions of this driver are supplied by Broadcom,
or by people within the Linux community so
I can find the right place to ask questions about it. I'm pretty sure
this isn't the right list for these types of questions, but if someone
could tell me whether I should be contacting Broadcom directly, or the
name of some other Linux mailing list or forum that might help, I'd
appreciate it.
If you received the drivers as part of RHEL, you should ask Red Hat.
However, in general, RH takes drivers from Linus's tree - either as part
of a complete release, or as a backport. Most development on network
drivers for current production seems to be done by developers employed
or contracted by the corresponding manufacturers, but it is subject to
review by the Linux community as represented here on netdev.
The MAINTAINERS file in the Linux source tree lists the primary
maintainer(s) for each driver:
BROADCOM BNX2 GIGABIT ETHERNET DRIVER
M: Michael Chan [off-list ref]
L: netdev@vger.kernel.org
S: Supported
F: drivers/net/bnx2.*
F: drivers/net/bnx2_*
BROADCOM TG3 GIGABIT ETHERNET DRIVER
M: Matt Carlson [off-list ref]
M: Michael Chan [off-list ref]
L: netdev@vger.kernel.org
S: Supported
F: drivers/net/tg3.*
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
From: Eric Dumazet <hidden> Date: 2011-02-17 14:17:42
Le jeudi 17 février 2011 à 05:24 -0800, Michael Durket a écrit :
I'm trying to fix a problem with packet drops and frame errors (really
rxbds_empty and rx_discards (or rx_fw_discards depending on whether or
not its a tg3 supported chip or a bnx2 chip)). We have this happening
on pretty much all the RedHat 5.x systems we're running these drivers
on. In investigating what remedies to employ to fix this, I'm trying
to determine if new versions of this driver are supplied by Broadcom,
or by people within the Linux community so
I can find the right place to ask questions about it. I'm pretty sure
this isn't the right list for these types of questions, but if someone
could tell me whether I should be contacting Broadcom directly, or the
name of some other Linux mailing list or forum that might help, I'd
appreciate it.
One possible cause of packet drops is when softirqs are disabled for too
long periods, even if NIC has a big RX ring (check ethtool -g eth0)
Why aren't the softirqs converted to workqueues? Wouldn't that cut
dependencies to other softirq users and improve latency?
Probably a stupid question, thanks.
Micha
From: Eric Dumazet <hidden> Date: 2011-02-20 19:17:09
Le dimanche 20 février 2011 à 19:19 +0100, Micha Nelissen a écrit :
Eric Dumazet wrote:
quoted
One possible cause of packet drops is when softirqs are disabled for too
long periods, even if NIC has a big RX ring (check ethtool -g eth0)
Why aren't the softirqs converted to workqueues? Wouldn't that cut
dependencies to other softirq users and improve latency?
Because it was done like that in the old days.
Its a bit less important these days, now typical machines have 8+ cpus.
Each device interrupt can be handled by its own cpu :)
From: Stephen Hemminger <hidden> Date: 2011-02-20 20:41:20
On Sun, 20 Feb 2011 20:17:00 +0100
Eric Dumazet [off-list ref] wrote:
Le dimanche 20 février 2011 à 19:19 +0100, Micha Nelissen a écrit :
quoted
Eric Dumazet wrote:
quoted
One possible cause of packet drops is when softirqs are disabled for too
long periods, even if NIC has a big RX ring (check ethtool -g eth0)
Why aren't the softirqs converted to workqueues? Wouldn't that cut
dependencies to other softirq users and improve latency?
Because it was done like that in the old days.
Its a bit less important these days, now typical machines have 8+ cpus.
Each device interrupt can be handled by its own cpu :)
The latency to schedule a work queue is still much higher
than the latency to do a softirq. Last time I played around with it,
things like loopback performance dropped 10% if using work queue.