On 06Mar23, Peter Wemm allegedly wrote:
> (~/.ssh/config with ServerAlive* probes) for different reasons. The
> (overloaded) router would drop connections that seemed idle. Sending
> probes helped prevent that - or at least making the router drop somebody
> else's instead.
Probably explains w
On 19Jan16, Jilles Tjoelker allegedly wrote:
> I think the recv.2 and send.2 man pages are long enough as they are, and
> separate recvmmsg.3 and sendmmsg.3 pages will be clearer. This is also
> because recvmmsg/sendmmsg can be ignored when performance is good enough
> without them. This differs f
On 08Jan16, Adrian Chadd allegedly wrote:
> On 7 January 2016 at 23:58, Mark Delany wrote:
> I'm there, on 16 threads.
That's intriquing. On CURRENT?
You must be doing smarter than 16 * recvmsg() or 16 * select(). What's
the thread structure?
> I'd rather we do it
On 08Jan16, Bruce Evans allegedly wrote:
> If the NIC can't reach line rate
> Network stack overheads are also enormous.
Bruce makes some excellent points.
I challenge anyone to get line rate UDP out of FBSD (or Linux) for a
1G NIC yet alone a 10G NIC listening to a single port. It was exactly
m
On 07Jan16, Luigi Rizzo allegedly wrote:
> On Thu, Jan 7, 2016 at 11:28 AM, Konstantin Belousov
> Regarding patching the application(s): of course it is not scalable
> if there are many applications that will refuse to compile if
> the *mmsg() functions are absent. However I expect this set of
> a
> You just repeat arguments for the text in my messages, which you removed
> on reply.
My goal is to get you to scruitinize.
Thank you for helping.
Mark.
___
freebsd-net@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-net
T
> Why is a signal lost in the scenario you described ?
Because the return can only indicate a signal/error *or* a batch of
messages but not both and the semantics of recvmsg() means that both
could occur.
Don't just consider signals, consider any -1/errno return from
recvmsg() such as -1/EAGAIN o
On 04Jan16, Konstantin Belousov allegedly wrote:
> On Sun, Jan 03, 2016 at 09:47:20PM +0000, Mark Delany wrote:
> > Eg, if a signal arrives after
> > more than zero packets have been processed by recvmmsg() what is the
> > correct return value? -1 or the count of messages r
On 03Jan16, Adrian Chadd allegedly wrote:
> It doesn't help at low connection rates. It helps at high connection /
> concurrency rates as the time going in/out of the kernel and getting
> back to steady state execution changes.
This is what I found when I did a comparison at $dayjob. You obviously
On 17Jul14, Daniel Corbe allegedly wrote:
> From the perspective of totally wrecking the performance of the host
> network stack: how much more overhead am I really introducing by looking
> at every packet inside of the netmap framework and going "am I really
> interested in this? Or should I sim
> > While good for performance, it unfortunately appears that NS_FORWARD
> > does not work in this mode - presumably because NR_REG_ONE_NIC doesn't
> > include the host ring.
> >
>
> correct, this is not a supported mode at the moment.
> If you want to implement it you should do it into netmap_pol
(Very recent fbsd10)
My application is taking advantage of NR_REG_ONE_NIC to register
separate handlers for each h/w ring. (Pro tip, you must re-open
/dev/netmap each time, a dup() fd doesn't work).
While good for performance, it unfortunately appears that NS_FORWARD
does not work in this mode -
> 2) the manpage refers to NR_RING_NIC_SW when I think it means
>NR_REG_NIC_SW.
Found another.
2a) manpage refers to NR_REG_SW_NIC when the include file has NR_REG_SW
To summarize:
manpageinclude
NR_REG_ALL_NIC
NR_REG_SW_NIC NR_REG_SW
NR_RING_NIC_SW NR_
Subject line says it all. I don't know what the convention is, but I
presume everything should be declared const whenever possible, thus
the appended patch.
Mark.
*** /usr/include/net/netmap_user.h Sun Mar 16 12:01:36 2014
--- /tmp/./netmap_user.hFri Mar 21 07:39:16 2014
*
(Luigi's page suggests posting here.)
Very recent freebsd10 (r263256)
1) the manpage says "SEE TRANSPARENT MODE" but no such section
exists.
2) the manpage refers to NR_RING_NIC_SW when I think it means
NR_REG_NIC_SW.
3) No mention is made of access control. I think earlier documentation
On 29Jan13, John Baldwin allegedly wrote:
> A common use case I have at work is to find a busted connection using netstat
> -n or sockstat and then want to tcpdrop it. However, tcpdrop requires spaces
> between the address and port so I can't simply cut and paste from one
> terminal
> window i
On Thu, Apr 18, 2002 at 12:49:45AM -0500, Mike Silbersack allegedly wrote:
>
> On Wed, 17 Apr 2002, Mark Delany wrote:
>
> > Are we discussing what happens when the number of pending connections
> > exceeds the backlog? If the suggestion is to leave such connections
On Wed, Apr 17, 2002 at 09:32:34PM -0500, Mike Silbersack allegedly wrote:
> > This is wrong too; it should silently drop the ACK and leave the
> > connection in the pending queue.
> How do the apps which try to rate-limit connections (OpenSSH, sendmail) do
> it? Would that behavior be defeated
18 matches
Mail list logo
