Read-only bind mounts - will they be included in 2.4.x kernel?

Read-only bind mounts - will they be included in 2.4.x kernel?

[Frank Myhr]

Hi,

There was some discussion 6 months ago about including Herbert
Poetzl's Bind Mount Extensions http://www.13thfloor.at/patches/ in a
2.4.x (x = 23 at the time) kernel. Can someone please give an update?

The Bind Mount Extensions allow bind-mounting part of a rw filesystem
read-only, which is very useful to people running vservers and other
chrooted stuff.

Thanks,
Frank

how to detect udp packets drops ?

[KyoungSoo Park]

Hi,

I'm not sure if my question is appropriate, but is there any way to 
directly detect
UDP packet drops in linux 2.4.x ?
I'd like to know how many UDP packets get actually dropped by the kernel 
for the overloaded time period.
Any suggestions?

Thanks,
KyoungSoo

Re: how to detect udp packets drops ?

[Måns Rullgård]

KyoungSoo Park <kyoungso@cs.princeton.edu> writes:

> Hi,
>
> I'm not sure if my question is appropriate, but is there any way to
> directly detect UDP packet drops in linux 2.4.x ?  I'd like to know
> how many UDP packets get actually dropped by the kernel for the
> overloaded time period.  Any suggestions?

The thing about UDP is that you can't tell if the packet got through,
that's what unreliable means.  However, if you just want to test a
link you could send a stream of packets with known data in them, such
as an increasing number sequence.  The receiving end will then be able
to tell if there were any drops.  Remember that out of order delivery
is allowed for UDP, so you might want to check for that too.

-- 
Måns Rullgård
mru@kth.se

Re: how to detect udp packets drops ?

[Måns Rullgård]

KyoungSoo Park <kyoungso@cs.princeton.edu> writes:

> Thank you for your kind response.  But, if I follow your test
> senario, I cannot know which entity drops the packets.  Routers on
> the path also can get congested and may drop the packets.

That is true.

> What I really want is to know is if the "kernel"(whch is supposed to
> receive packets) drops packets.

Do you mean you want to know if the kernel's receive buffer overflows?
That could be possible in theory, but I don't know whether it's
possible with the current Linux kernels.

> Also I don't want to contribute to the congestion by sending test UDP
> packets, because that may worsen the situation.
> Isn't there any way to monitor the packet drops at the end host by
> just looking at the "packet drop counter"?

If the packet reaches the end host the only reason for dropping it
would be a filled receive buffer.  The size of the buffer can be set
using setsockopt().

-- 
Måns Rullgård
mru@kth.se