Sahan,
I've had a chance to look at this configuration, finally. The crash looks to
have nothing to do with devices. I bet that a local packet emitted into your
configuration by FromHost, is being sent back into your machine by ToHost.
And it looks like we did not sufficiently clean the packet.
Perhaps a diff like the attached will help. Give it a try if you haven't yet
given up.
Best of luck
E
Sahan Gamage wrote:
Hi,
I upgraded to latest kernel version supported by click (2.6.24.7) and
used the latest click from CVS. Also I got the e1000e driver and
applied Joonwoo's patches. Intially the setup (along with my old
config file) seemed to work fine. But after several iperf runs, the
kernel locked up and I had to reset the system. When I examined the
kernel log (/var/log/messages) the crash log looked much similar to
the previous one.
Any pointers to start debugging the problem is appreciated.
Thanks
-sahan
On Tue, Dec 15, 2009 at 1:32 PM, Roman Chertov <[email protected]> wrote:
Sahan Gamage wrote:
Hi,
I am using click1-7.0rc1 with Linux kernel version 2.6.19.2. I am
I recommend pulling latest click from git as there were quite a bit of
changes after rc1 was released. Also, I suggest to upgrade to 2.6.24.7
kernel.
using the e1000 driver downloaded from Intel web site because the
version supplied with click does not recognize my ethernet card.
You might want to try the e1000e module that Joonwoo posted on this
list. You can grep the driver code for your model number to see if it
supports it or not.
I want to use click to simulate a long fat network (change just the
delay of network). I am using click module (using click-install) with
the attached configuration file (some minor tweaks to the supplied
localdelay.click). But when I install click module and run some iperf
Good to see that people are looking at samples instead of just asking
how to use the FromHost element.
tests the kernel locks up. The out put of dmesg is attached also. (the
message repeated many times).
If you upgrade to later kernel/Click versions this might go away as
there were several patches to ToDevice to address various lockup issues.
Roman
If anybody ran into the same problem and have any idea to solve the
problem that would be very helpful.
Thanks
-sahan
------------------------------------------------------------------------
_______________________________________________
click mailing list
[email protected]
https://amsterdam.lcs.mit.edu/mailman/listinfo/click
_______________________________________________
click mailing list
[email protected]
https://amsterdam.lcs.mit.edu/mailman/listinfo/click
diff --git a/elements/linuxmodule/fromhost.cc b/elements/linuxmodule/fromhost.cc
index 5cff771..ae5634a 100644
--- a/elements/linuxmodule/fromhost.cc
+++ b/elements/linuxmodule/fromhost.cc
@@ -36,6 +36,7 @@ CLICK_CXX_PROTECT
#include <linux/inetdevice.h>
#include <linux/if_arp.h>
#include <net/route.h>
+#include <net/dst.h>
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 24)
# include <net/net_namespace.h>
#endif
@@ -403,8 +404,9 @@ FromHost::fl_tx(struct sk_buff *skb, net_device *dev)
int next = fl->next_i(fl->_tail);
if (likely(next != fl->_head)) {
Packet **q = (fl->_capacity <= smq_size ? fl->_q.smq : fl->_q.lgq);
+ skb_dst_drop(skb);
Packet *p = Packet::make(skb);
- p->set_timestamp_anno(Timestamp::now());
+ p->timestamp_anno().assign_now();
if (fl->_clear_anno)
p->clear_annotations(false);
fl->_stats.tx_packets++;
_______________________________________________
click mailing list
[email protected]
https://amsterdam.lcs.mit.edu/mailman/listinfo/click
