misc documentation changes (partially to reflect new kernel releases)

This commit is contained in:
he 1999-08-28 16:34:52 +00:00
parent 5a5ca755bf
commit 10497c391e
3 changed files with 59 additions and 27 deletions

View File

@ -1,16 +1,14 @@
Installation instructions for eftp4linux
First, you need to compile 2.2.x kernel with x25 support enabled.
You also need a recent version of the isdn4k-utils compiled against
2.2.x kernel header files and the x25 utility programmes.
Kernel
======
You need a recent kernel 2.2.x kernel that supports some special
features. 2.0.x kernels won't work. When configuring the kernel for
compilation, (i.e. by `make menuconfig') enable the following:
features. 2.0.x kernels won't work. As erly 2.2.x kernels contain
outdated isdn code, it is recommended to use kernel 2.2.12 or newer.
When configuring the kernel for compilation, (i.e. by `make
menuconfig') enable the following:
Code maturity level options --->
@ -59,14 +57,12 @@ ix25-2.1.128.patch fixes a bug in the Linus X.25 PLP layer. This is
at the very end of include/net/x25.h
before compiling the kernel.
isdn4linux CVS snapshot has been reported to work more reliably with
2.2.x-kernels. It also provides for a
feature useful for eftd to reliably determine
the peer's isdn No.
Kernels 2.3.4 and newer now contain the new
isdn4linux-code. You can try this (but be
aware that 2.3.x are the non-stable kernels).
isdn4linux CVS snapshot can be used to replace your kernel's isdn
subsystem whith a more recent versions.
As Kernels 2.2.12 and 2.3.14 now contain
up-to-date isdn versions, this should no longer
be necessary.
isdn4k-utils

View File

@ -1,4 +1,4 @@
$Id: README,v 1.3 1999/07/26 22:04:26 he Exp $
$Id: README,v 1.4 1999/08/28 16:34:53 he Exp $
README file for eftp4linux
==========================
@ -94,17 +94,9 @@ Details on this test distribution
In addition to the files contained in this distribution, you will need
- A recent Linux 2.2.x kernel (2.2.1 or newer is recommended).
- A recent Linux 2.2.x kernel (2.2.12 or newer is recommended).
It is also possible to replace the isdn subsystem by a recent snapshot
from the isdn4linux CVS server. This is not necessary, but the
CVS version has been reported to work more reliably with recent
kernels. The CVS versions as well as kernels >= 2.3.4 also supports
a new IOCTL which can be used
by eftd for reliably identifying the isdn peer by means of the
remote phone number.
There is also a patch ix25-2.1.128.diff that deals with a low layer
- There is also a patch ix25-2.1.128.diff that deals with a low layer
(X.25) interworking problem observed with some clients
(CSD isdn toolbox and maybe also some RVS clients). They send
misformatted X.25 call request packets that confuse the linux X.25 layer.

View File

@ -1,4 +1,4 @@
$Id: FAQ,v 1.2 1999/07/01 18:00:20 he Exp $
$Id: FAQ,v 1.3 1999/08/28 16:35:05 he Exp $
QUESTION: How can I control the amount of debugging output?
=========
@ -22,6 +22,7 @@ To remove isdnlog output to stderr, you can also comment out the
QUESTION: I don't want anonymous users to dial in without a MSN
=========
If a client without a MSN dials in, the ISDN number 0 is assumed.
With the line "guestservers 1* 2* 3* 4* 5* 6* 7* 8* 9*" in
/etc/isdn/eftaccess connections with ISDN numbers beginning with
@ -29,3 +30,46 @@ With the line "guestservers 1* 2* 3* 4* 5* 6* 7* 8* 9*" in
the file /etc/isdn/eftaccess used.
(What's about international connections then? Do they begin with "0"?
I dont'n know, never had one...)
QUESTION: Why do I not get full 64 kBit/s transfer speed.
=========
First note that an isdn B channel supports 64*1000 Bit/s,
not 64*1024 Bit/s.
Second, note that eftp4linux is very careful not to overestimate
the measured transfer speed (certain ftp clients don't properley
measure data in OS-internal queues and falsly report transfer speeds
larger than the network bandwidth for short files. Be aware of this if
you do comparisions)
With standard protocol configuration, data is transferred in chunks of
around 1000 Bytes. The headers of the 3 protocol layers
X.75/LapB (2 bytes), X.25 DTE-DTE (3 bytes), and the EUROFILE layer
itsself (typically 6--7 bytes for large files) add slightly
more than 1% overhead in total (which is significiently lower than
a tcp/ip/ppp stack).
The hdlc framing additionally adds 2 bytes for the CRC and 2 flags to
each frame, the hdlc bit stuffing adds around 3% extra bits for
random data. (The exact value depends on the data and might be
between 0 and 20 %. For comparisions: as syncppp uses the same hdlc
framing, the same overhead applies here.)
All in all, there is an expected protocol overhead of around 4.5%,
which should allow to transfer around 7640 bytes/s.
Note that the current eftp4linux is not especially tuned for
speed. It is compiled without optimization. If debugging output of
eftp4linux isself or the drivers (e.g. by means of hisaxctrl,
x25trace) is activated, additional computation time is required.
If certain protocol parameters (in particular X.25 or eurofile window
sizes of one) are in effect, transfer speed might also decrease.
Other additional delays, such as waiting for disk or NFS IO or
waiting for beeing schedules on a heavy loaded system, might also
decrease the transfer speed.