Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Frantisek Rysánek]

Dear Mr. Tosatti,

I've been asked by Mr. Francois Romieu, the author of
drivers/net/wan/dscc4.c,
to give you a summary of my user-space experience with recent versions
of dscc4.c and the Generic HDLC Layer (drivers/net/wan/hdlc_*.c, hdlc.o).

The message is: it works, to the extent that it can be configured and
TCP/IP works. I've tried DSCC4 hardware back-to-back and against
a Cisco, in a few different configurations. I've tried the three basic
encapsulations: cisco-hdlc, itu-hdlc and sync ppp.
I don't know enough FR and X.25 to be able to come up with a workable
setup (didn't even try).

I was using linux 2.4.21-rc1 and 2.4.21 (release). This works, but that
version of dscc4.c suffers from a hardware quirk in the DSCC4 hardware
- the port/link cannot be reconfigured without a cold reboot.
The effects of this quirk are alleviated by a patch to dscc4.c that has been
submitted by F.Romieu for approval to the linux kernel and that reportedly
has made i to the -ac series, but not to the vanilla 2.4.21.
I have not touched the sources of the Generic HDLC Layer (hdlc*.c) that is
present in 2.4.21-rc1 to 2.4.21. I was using the vanilla source without
modification. I don't know about any differences in the hdlc*.c in the -ac
series - I didn't investigate that flavor.

As for the userspace sethdlc.c by Krzystof Halasa: I was using ver.1.12,
modified by Mr. Romieu, who has "cut some non-compiling functionality."
The current version from Krzysztof Halasa is 1.15.

I barely had the time to test basic workability and document my test bench.
Unfortunately I did not do any stress/long-term/feature-crosscompatibility
testing.
Specifically, I was using uniprocessor machines only (no SMP).
I did not test all the features available in the Generic HDLC Layer, I don't
even
fully understand all the technologies covered - I was only interested in
basic
point-to-point IP over sync serial.

I've written a short mini-HOWTO - the chapter on test results in terms of
transfer rates and ping round trips is at
http://sweb.cz/Frantisek.Rysanek/sync/dscc4+HDLC-Mini-HOWTO.html#Drivers.ope
n.tests

I hope that I was not completely wasting your time :-)
If you have further questions, ask me.
Thanks for the great job that you're doing.

Sincerely yours

Frank Rysanek



----- Original Message -----
From: "Francois Romieu" <romieu@fr.zoreil.com>
To: "Frantisek Rysánek" <rysanek@fccps.cz>
Sent: Friday, July 11, 2003 9:25 PM
Subject: [marcelo@conectiva.com.br: Re: Why is generic hldc beig ignored?
RE:Linux 2.4.22-pre4]


> Imho l-k is waiting for a summary of your experiences with post
> 2.4.21-pre linux hdlc.
>
> Btw latest dscc4 updates are available in Alan Cox's -ac series.
>
> ----- Forwarded message from Marcelo Tosatti
<marcelo@conectiva.com.br> -----
>
> From: Marcelo Tosatti <marcelo@conectiva.com.br>
> Date: Fri, 11 Jul 2003 14:18:10 -0300 (BRT)
> Subject: Re: Why is generic hldc beig ignored?   RE:Linux 2.4.22-pre4
> X-X-Sender: marcelo@freak.distro.conectiva
> To: Henrique Oliveira <henrique2.gobbi@cyclades.com>
> Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>,
>    Kevin Curtis <kevin.curtis@farsite.co.uk>,
>    lkml <linux-kernel@vger.kernel.org>
> References: <7C078C66B7752B438B88E11E5E20E72E25C978@GENERAL.farsite.co.uk>
>  <Pine.LNX.4.55L.0307101410570.25103@freak.distro.conectiva>
>  <003101c34712$a9b8f480$602fa8c0@henrique>
<1057914760.8028.27.camel@dhcp22.swansea.linux.org.uk>
>  <013901c347cd$44586f60$602fa8c0@henrique>
> X-Spam-Status: No, hits=-28.7 required=4.0
> tests=EMAIL_ATTRIBUTION,IN_REP_TO,REFERENCES,REPLY_WITH_QUOTES,
>       USER_AGENT_PINE,X_MAILING_LIST
> version=2.53-fr_1
> X-Spam-Level:
> X-Spam-Checker-Version: SpamAssassin 2.53-fr_1 (1.174.2.15-2003-03-30-exp)
>
>
>
> On Fri, 11 Jul 2003, Henrique Oliveira wrote:
>
> > Hello !!!
> > No offence here but the generic hdlc layer's been always confusing. I
will
> > write here what I believe is going on.
> > Until version 2.4.20 we had the old hdlc layer, with only one source
file
> > hdlc.c. People that wanted to use the new hdlc layer had to apply a
patch
> > provided by Halasa or by the WAN cards vendors. The kernel 2.4.21 came
out
> > with the new hdlc layer, that includes a couple of files:
hdlc_generic.c,
> > hdlc_fr.c, hdlc_ppp.c, hdlc_raw.c, etc. I don't know the status of
the -ac
> > tree but I can say that the kernel 2.4.21 has the new code.
> > The new hdlc layer really needs new tools. This new tool can
(supposedly) be
> > found at Halasa's web site. I don't know if someone has tested these
tools
> > but I am about to run a test this afternoon. If someone is interested on
the
> > results, just drop me a line.
>
> I am.
>
> Please report the results of your tests and CC lkml.
>
> And Alan, what the -ac tree hdlc changes are about ?
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
> ----- End forwarded message -----

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Francois Romieu]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=unknown-8bit, Size: 587 bytes --]

Frantisek Rysánek <rysanek@fccps.cz> :
[...]
> As for the userspace sethdlc.c by Krzystof Halasa: I was using ver.1.12,
> modified by Mr. Romieu, who has "cut some non-compiling functionality."
> The current version from Krzysztof Halasa is 1.15.

To be fair to Krzysztof, the "non-compiling functionality" I cut simply
came from the fact that latest sethdlc utility contained code related to
CONFIG_HDLC_RAW_ETH option which wasn't available in vanilla 2.4.21-pre.
Please note that this specific code exists in 2.4.22-pre3-ac but not in
plain Marcelo's 2.4.22-pre. 

Regards

--
Ueimor

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Krzysztof Halasa]

"Frantisek Rysánek" <rysanek@fccps.cz> writes:

> As for the userspace sethdlc.c by Krzystof Halasa: I was using ver.1.12,
> modified by Mr. Romieu, who has "cut some non-compiling functionality."

It was a q&d version to support Ethernet emulation and did not compile
without it. Was in the README in fact.

> The current version from Krzysztof Halasa is 1.15.

This one doesn't have this problem, Ethernet emulation doesn't get
compiled if the kernel has no support for it.

> Specifically, I was using uniprocessor machines only (no SMP).

This is a common problem. From time to time I perform SMP tests, but they
are neither extensive or long-term :-(

> I've written a short mini-HOWTO - the chapter on test results in terms of
> transfer rates and ping round trips is at
> http://sweb.cz/Frantisek.Rysanek/sync/dscc4+HDLC-Mini-HOWTO.html#Drivers.ope
> n.tests

Looks nice, that's something I always wanted to have... I'll send you
an errata.
Great job.
-- 
Krzysztof Halasa
Network Administrator

RE: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Kevin Curtis]

My apologies, it is already there in 2.4.21, however I have been monitoring
the change logs, and I can't see it mentioned there anywhere.

Thanks very much for pointing this out.


Kevin

-----Original Message-----
From: Henrique Oliveira [mailto:henrique2.gobbi@cyclades.com] 
Sent: 10 July 2003 19:40
To: Marcelo Tosatti; Kevin Curtis
Cc: lkml
Subject: Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4


Hi,
The patch for the generic HDLC layer was included on the kernel 2.4.21. Thus
this layer is already on the main tree (unless, of course, someone has
removed it, I havent checked 2.4.22 yet). This layer provides data link
protocol (ppp, hdlc, raw-hdlc, x25, frame-relay, cisco-hdlc) for the kernel.
It's mainly used by synchronous cards drivers (Cyclades, Moxa, SDL, Farsite,
etc, etc, etc). regards Henrique

----- Original Message -----
From: "Marcelo Tosatti" <marcelo@conectiva.com.br>
To: "Kevin Curtis" <kevin.curtis@farsite.co.uk>
Cc: "lkml" <linux-kernel@vger.kernel.org>
Sent: Thursday, July 10, 2003 10:11 AM
Subject: Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4


>
>
> On Thu, 10 Jul 2003, Kevin Curtis wrote:
>
> > The usual request for generic hdlc (please).
> > Why are requests for it's inclusion being ignored?
> >
> >
> > Kevin Curtis
> > Linux Development
> > FarSite Communications Ltd
> > www.farsite.co.uk
> > tel:  +44 1256 330461
> > fax:  +44 1256 854931
>
>
> Where is the patch and why do you want it in?
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-kernel" in the body of a message to majordomo@vger.kernel.org 
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Kevin Curtis]

The usual request for generic hdlc (please).
Why are requests for it's inclusion being ignored?


Kevin Curtis
Linux Development
FarSite Communications Ltd
www.farsite.co.uk
tel:  +44 1256 330461
fax:  +44 1256 854931


-----Original Message-----
From: Marcelo Tosatti [mailto:marcelo@conectiva.com.br] 
Sent: 09 July 2003 23:25
To: lkml
Subject: Linux 2.4.22-pre4



Hi,

Here goes -pre4. It contains a lot of updates and fixes.

We decided to include this new code quota code which allows usage of quotas
with 32bit UID/GIDs.

Most Toshibas should work now due to an important ACPI fix.

Please help and test.


Summary of changes from v2.4.22-pre3 to v2.4.22-pre4
============================================

<dtor_core@ameritech.net>:
  o [NET] Attach inner qdiscs to TBF

<lethal@linux-sh.org>:
  o sh64: Add FIOQSIZE definition
  o sh64: Fixup Cayman IRQ reporting
  o sh64: SH-5 PCI updates
  o sh64: Fix privileged insn handling
  o sh64: IDE support

<lode_leroy@hotmail.com>:
  o [IPV4] display bootserver in /proc/net/pnp

<lunz@falooley.org>:
  o [NET] Fix refcounting of dev->promiscuity for af_packet

<m.c.p@wolk-project.de>:
  o [RESEND 5th] Fix oom killer braindamage

<stelian@popies.net>:
  o Export 'acpi_disabled' symbol to modules

Alan Cox <alan@lxorguk.ukuu.org.uk>:
  o Make ACPI work on lots more boxes
  o config for new Nvidia AGP
  o parisc sync up (resend of resend of resend ... 8))
  o AGP update - new intel, add nvidia
  o ebda check in ibm hotplug is insufficient
  o update mpt fusion driver
  o fix the eexpress
  o move sdla to mod_timer
  o add code for missing c7000 driver
  o resend - fix security bits in binfmt_exec/som
  o re-fix printk level for buffer cachehash
  o exec part of security fix
  o fix inverted dnotify
  o fix definition of boot_DT
  o add the new agp modes to the headers
  o kernel/fork helper for exec security fix
  o S/390 CLAW bits
  o fix up z85230 queue wake logic

Benjamin Herrenschmidt <benh@kernel.crashing.org>:
  o ppc32: Fix races in low level adb drivers

Christoph Hellwig <hch@infradead.org>:
  o quota patch breaks kernel build

Christoph Hellwig <hch@lst.de>:
  o new quota code
  o fix Q_SYNC for dev == 0

David S. Miller <davem@nuts.ninka.net>:
  o [SPARC64]: sys_sparc32.c needs linux/quotacompat.h
  o [FS]: Provide unshare_files() declaration and export to modules
  o [SPARC]: SEMTIMEDOP for both Sparc ports

Marcelo Tosatti <marcelo@freak.distro.conectiva>:
  o Cset exclude: rusty@rustcorp.com.au|ChangeSet|20030707182325|08049
  o Remove bogus diff from drivers/char/Config.in
  o Changed EXTRAVERSION to -pre4
  o Cset exclude: hannal@us.ibm.com|ChangeSet|20030707180059|08076
  o Fixes ext3 quota/truncate oops
  o CRIS architecture update
  o Cset exclude: Remove NFS direct IO patches Cset exclude:
trond.myklebust@fys.uio.no|ChangeSet|20030708095239|55752
  o Cset exclude: remove NFS direct IO patches Cset exclude:
trond.myklebust@fys.uio.no|ChangeSet|20030706143259|16957
  o Add missing fs/quota_v2.c file
  o Comment out VIA_APOLLO_P4X400 handling in drm_agpsupport.h: Alan will
fix that up later

Matthew Wilcox <willy@debian.org>:
  o pci_name()

Mikael Pettersson <mikpe@csd.uu.se>:
  o i386 cpufeature.h cleanup + comment

Paul Mackerras <paulus@samba.org>:
  o PPC32: Minor updates to comments and processor register definitions
  o PPC32: Minor boot wrapper cleanups
  o PPC32: Define screen_info if CONFIG_FB is set for the sake of vesafb
  o PPC32: Make __kernel_ino_t be unsigned long like on other architectures

Randy Dunlap <rddunlap@osdl.org>:
  o make profile= doc. clearer

Rusty Russell <rusty@rustcorp.com.au>:
  o Configure.help Polish translation location update
  o unreachable code in drivers_media_video_cpia_pp.c
  o 2.4 drivers_char_random.c fix sample shellscripts
  o trivial patch
  o fix sound doc typos
  o fs_bfs_dir.c unused variables
  o Decision PCCOM4_PCCOM8 serial support for 2.4.19
  o Re: setrlimit incorrectly allows hard limits to exceed
  o fix linewrap in Documentation_ia64_efirtc.txt
  o fix linewrap in Documentation_arm_SA1100_CERF
  o fix linewrap in Documentation_filesystems_befs.txt
  o [2.5 patch] two small MTD fixes
  o 2.4 patch for more debug safety
  o esssolo1.c doesn't free resources correctly

Trond Myklebust <trond.myklebust@fys.uio.no>:
  o Back out some congestion control changes that were causing trouble among
other things for the "soft" mount option.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org More majordomo info at
http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Marcelo Tosatti]

On Thu, 10 Jul 2003, Kevin Curtis wrote:

> The usual request for generic hdlc (please).
> Why are requests for it's inclusion being ignored?
>
>
> Kevin Curtis
> Linux Development
> FarSite Communications Ltd
> www.farsite.co.uk
> tel:  +44 1256 330461
> fax:  +44 1256 854931


Where is the patch and why do you want it in?

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Henrique Oliveira]

Hi,
The patch for the generic HDLC layer was included on the kernel 2.4.21. Thus
this layer is already on the main tree (unless, of course, someone has
removed it, I havent checked 2.4.22 yet). This layer provides data link
protocol (ppp, hdlc, raw-hdlc, x25, frame-relay, cisco-hdlc) for the kernel.
It's mainly used by synchronous cards drivers (Cyclades, Moxa, SDL, Farsite,
etc, etc, etc).
regards
Henrique

----- Original Message -----
From: "Marcelo Tosatti" <marcelo@conectiva.com.br>
To: "Kevin Curtis" <kevin.curtis@farsite.co.uk>
Cc: "lkml" <linux-kernel@vger.kernel.org>
Sent: Thursday, July 10, 2003 10:11 AM
Subject: Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4


>
>
> On Thu, 10 Jul 2003, Kevin Curtis wrote:
>
> > The usual request for generic hdlc (please).
> > Why are requests for it's inclusion being ignored?
> >
> >
> > Kevin Curtis
> > Linux Development
> > FarSite Communications Ltd
> > www.farsite.co.uk
> > tel:  +44 1256 330461
> > fax:  +44 1256 854931
>
>
> Where is the patch and why do you want it in?
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Alan Cox]

On Iau, 2003-07-10 at 19:39, Henrique Oliveira wrote:
> Hi,
> The patch for the generic HDLC layer was included on the kernel 2.4.21. Thus
> this layer is already on the main tree (unless, of course, someone has
> removed it, I havent checked 2.4.22 yet). This layer provides data link
> protocol (ppp, hdlc, raw-hdlc, x25, frame-relay, cisco-hdlc) for the kernel.
> It's mainly used by synchronous cards drivers (Cyclades, Moxa, SDL, Farsite,
> etc, etc, etc).

2.4.21 has much older code than the current stuff (which has been in -ac
for a while). As I understand it the new hdlc layer needs newer tools ?

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Henrique Oliveira]

Hello !!!
No offence here but the generic hdlc layer's been always confusing. I will
write here what I believe is going on.
Until version 2.4.20 we had the old hdlc layer, with only one source file
hdlc.c. People that wanted to use the new hdlc layer had to apply a patch
provided by Halasa or by the WAN cards vendors. The kernel 2.4.21 came out
with the new hdlc layer, that includes a couple of files: hdlc_generic.c,
hdlc_fr.c, hdlc_ppp.c, hdlc_raw.c, etc. I don't know the status of the -ac
tree but I can say that the kernel 2.4.21 has the new code.
The new hdlc layer really needs new tools. This new tool can (supposedly) be
found at Halasa's web site. I don't know if someone has tested these tools
but I am about to run a test this afternoon. If someone is interested on the
results, just drop me a line.
regards
Henrique

----- Original Message -----
From: "Alan Cox" <alan@lxorguk.ukuu.org.uk>
To: "Henrique Oliveira" <henrique2.gobbi@cyclades.com>
Cc: "Marcelo Tosatti" <marcelo@conectiva.com.br>; "Kevin Curtis"
<kevin.curtis@farsite.co.uk>; "lkml" <linux-kernel@vger.kernel.org>
Sent: Friday, July 11, 2003 2:12 AM
Subject: Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4


> On Iau, 2003-07-10 at 19:39, Henrique Oliveira wrote:
> > Hi,
> > The patch for the generic HDLC layer was included on the kernel 2.4.21.
Thus
> > this layer is already on the main tree (unless, of course, someone has
> > removed it, I havent checked 2.4.22 yet). This layer provides data link
> > protocol (ppp, hdlc, raw-hdlc, x25, frame-relay, cisco-hdlc) for the
kernel.
> > It's mainly used by synchronous cards drivers (Cyclades, Moxa, SDL,
Farsite,
> > etc, etc, etc).
>
> 2.4.21 has much older code than the current stuff (which has been in -ac
> for a while). As I understand it the new hdlc layer needs newer tools ?

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Marcelo Tosatti]

On Fri, 11 Jul 2003, Henrique Oliveira wrote:

> Hello !!!
> No offence here but the generic hdlc layer's been always confusing. I will
> write here what I believe is going on.
> Until version 2.4.20 we had the old hdlc layer, with only one source file
> hdlc.c. People that wanted to use the new hdlc layer had to apply a patch
> provided by Halasa or by the WAN cards vendors. The kernel 2.4.21 came out
> with the new hdlc layer, that includes a couple of files: hdlc_generic.c,
> hdlc_fr.c, hdlc_ppp.c, hdlc_raw.c, etc. I don't know the status of the -ac
> tree but I can say that the kernel 2.4.21 has the new code.
> The new hdlc layer really needs new tools. This new tool can (supposedly) be
> found at Halasa's web site. I don't know if someone has tested these tools
> but I am about to run a test this afternoon. If someone is interested on the
> results, just drop me a line.

I am.

Please report the results of your tests and CC lkml.

And Alan, what the -ac tree hdlc changes are about ?

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Jeff Garzik]

On Fri, Jul 11, 2003 at 02:18:10PM -0300, Marcelo Tosatti wrote:
> I am.
> 
> Please report the results of your tests and CC lkml.
> 
> And Alan, what the -ac tree hdlc changes are about ?

IMO hdlc is a "raidtools-like situation":

New HDLC stuff (already in 2.5) was also developed for 2.4.  It changes
the 2.4 HDLC userland ABI... but pretty much everybody who still
cares about HDLC is using the new (changed) ABI.

	Jeff

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Krzysztof Halasa]

Jeff Garzik <jgarzik@pobox.com> writes:

> New HDLC stuff (already in 2.5) was also developed for 2.4.  It changes
> the 2.4 HDLC userland ABI... but pretty much everybody who still
> cares about HDLC is using the new (changed) ABI.

Precisely. Still, the old HDLC is now a history, 2.4.21 contains the new
ABI/API. It's just missing some regular updates (not breaking the ABI).
-- 
Krzysztof Halasa
Network Administrator

Re: Why is generic hldc beig ignored? RE:Linux 2.4.22-pre4

[Krzysztof Halasa]

Alan Cox <alan@lxorguk.ukuu.org.uk> writes:

> 2.4.21 has much older code than the current stuff (which has been in -ac
> for a while). As I understand it the new hdlc layer needs newer tools ?

Yes, this is the same new tool that is required for plain 2.4.21 and -ac
and for the last patch as well (not yet applied by you, BTW it's
http://hq.pm.waw.pl/pub/linux/hdlc/hdlc-2.4.21pre5-ac3-1.14.patch
or ftp:// the same, production-tested = you can apply it safely).
2.4.20 used older sethdlc tool but in practice all users were already
using the newer tool + kernel patch.

The last patch (1.14) contains few small bug-fixes and new Ethernet framing
for HDLC and FR. There are no incompatible ABI changes (i.e. older
sethdlc binary works with new kernels, but you of course need new tool for
new features) after the "big" switch in 2.4.21pre.
-- 
Krzysztof Halasa
Network Administrator