Linux-Development-Sys Digest #641, Volume #6 Wed, 21 Apr 99 17:14:34 EDT
Contents:
Re: Kernel 2.2.6: irda does not compile - Help (mlw)
Re: Building pgcc-1.1.2 (Piniek aka Piotr Ingling)
Ip Aliases - sending packets from aliases? (Geoff Wong)
Re: Can device driver use the filing system? (mlw)
glibc2 and stat function (Andre Oliveira da Costa)
physical Memory ([EMAIL PROTECTED])
Re: EXT2 Sponsor File system (Tin Siladin)
Sound] How do I stop recording ? (R�khar�ur Egilsson)
high availability products for Linux? ([EMAIL PROTECTED])
Problem with File System Check (Varghese Panicker K K)
Re: "Permission denied" ("Bill Zimmerly")
Re: Development guide for HP CRet & PhotoRetII (Andreas Dilger)
Beowulf, Can we use it? ("Clint Byrum")
----------------------------------------------------------------------------
From: mlw <[EMAIL PROTECTED]>
Subject: Re: Kernel 2.2.6: irda does not compile - Help
Date: Wed, 21 Apr 1999 12:09:44 +0000
Peter Samuelson wrote:
>
> [Thomas Jaeger <[EMAIL PROTECTED]>]
> > i just tried to compile the latest kernel 2.2.6 with irda (all
> > options enabled for a modular kernel)
>
> > i got following error:
> > iriap.c:31: linux/irda.h: No such file or directory
> > iriap.c: In function `iriap_getvaluebyclass_confirm':
> > iriap.c:458: `CS_ASCII' undeclared (first use this function)
> > iriap.c:458: (Each undeclared identifier is reported only once
> > iriap.c:458: for each function it appears in.)
> > iriap.c:470: warning: unreachable code at beginning of switch statement
>
> Right. Richard Gooch doesn't have this in his Kernel Newsflash page
> yet (http://www.atnf.csiro.au/~rgooch/linux/docs/kernel-newsflash.html)
> but I imagine he will soon. Browsing through patch-2.2.6.gz, looks
> like:
>
> - a lot of material was removed from include/net/irda/irmod.h
> including `#define CS_ASCII 0x00'
> - `#include <linux/irda.h>' was added to five or six files
> - include/linux/irda.h did not and still does not exist
>
> Draw your own conclusions. You might try creating a dummy
> include/linux/irda.h with `#define CS_ASCII 0x00'. Dunno if that would
> work or not, haven't tried it. Alternatively, roll back the irda
> updates in 2.2.6.
I saw the same things. It is not just CS_ASCII, but also other defines
and data strutures. I created a irda.h and started putting things in it,
but, it was getting big. Being somewhat lazy, and not needing irda, I
decided to disable irda and move on.
--
Mohawk Software
Windows 95, Windows NT, UNIX, Linux. Applications, drivers, support.
Take the Mohawk Software Computer Survey at: www.mohawksoft.com
------------------------------
From: [EMAIL PROTECTED] (Piniek aka Piotr Ingling)
Subject: Re: Building pgcc-1.1.2
Date: Tue, 20 Apr 1999 13:23:19 GMT
Reply-To: [EMAIL PROTECTED]
Dnia Mon, 19 Apr 1999 03:00:43 GMT, Alexander Rodyukov <[EMAIL PROTECTED]>
napisa�(a):
>Hello, Friends.
>
>I'm running into pretty strange problem... while trying to compile
>pgcc-1.1.2 i'm getting following results:
[...]
>haifa-sched.c:8259: Internal compiler error in function update_flow_info
[...]
>$ gcc -v
>Reading specs from /usr/lib/gcc-lib/i486-linux/2.7.2.3/specs
>gcc version 2.7.2.3
>
Is the 2.7.2.3 really supported - I'm afraid you need some newer one. Have you
checked the documentation?
Piotr Ingling
e-mail: [EMAIL PROTECTED]
------------------------------
From: Geoff Wong <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.networking
Subject: Ip Aliases - sending packets from aliases?
Date: 21 Apr 1999 09:01:38 GMT
Hi,
The following question relates to IP aliases.
I'm not sure if the groups I've posted to are the correct
place - but I couldn't find a really correct group.
Bear with me if you've any expertise in programming IP Aliases.
If I have the following IP aliases configuration
(IP numbers changed to protect the innocent :-):
eth0 Link encap:Ethernet HWaddr 00:48:54:01:28:61
inet addr:10.0.0.172 Bcast:10.0.0.175 Mask:255.255.255.248
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:921326 errors:0 dropped:0 overruns:0 frame:0
TX packets:92310 errors:4 dropped:0 overruns:0 carrier:10
collisions:8167
Interrupt:10 Base address:0xd000
eth0:1 Link encap:Ethernet HWaddr 00:48:54:01:28:61
inet addr:10.0.0.173 Mask:255.255.255.255
UP RUNNING MTU:1500 Metric:1
RX packets:1 errors:0 dropped:0 overruns:0 frame:0
TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
collisions:0
eth0:2 Link encap:Ethernet HWaddr 00:48:54:01:28:61
inet addr:10.0.0.174 Mask:255.255.255.255
UP RUNNING MTU:1500 Metric:1
RX packets:2 errors:0 dropped:0 overruns:0 frame:0
TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
collisions:0
It's relatively easy to write code that will listen()
on each of these interfaces. I've been going mad trying
to get the "reverse" to work.
Is it possible to write code under Linux that allows outgoing
packets (after a connect(), etc) to come from the aliased
interfaces (eth0:1, etc) rather than simply eth0?
If so; could someone point me at the correct ioctl()
or the socket options or even a snippet of code.
Please remove the "eatspam" from my reply address to
send to me if you reply directly.
thanks,
Geoff
------------------------------
From: mlw <[EMAIL PROTECTED]>
Subject: Re: Can device driver use the filing system?
Date: Wed, 21 Apr 1999 13:01:12 +0000
Philip Boucherat wrote:
>
> Hi,
>
> According to people on the RedHat mailing list, I am strongly advised
> not to try and use the filing system from a device driver I'm developing
> for RedHat V5.2.
>
> I would like someone out there to tell me it *is* OK to read/write files
> from a driver (and preferably how to do it). The file access is only
> during driver load and unload - I need to read firmware from a file and
> download it to the device - and I would have thought the driver would be
> called in some sort of user context at these times so it shouldn't be
> too much of a problem.
>
> I can do this in Windows NT and 95. For NetWare I've got a separate
> daemon-type module that is there solely to access the file system, and
> when the driver wants a file, it sends a message to the module via
> shared memory and the daemon fetches the file and sends it back to the
> driver via the same shared memory - would this sort of arrangement be
> possible in Linux - i.e. can a user process and a driver communicate via
> shared memory, or any other way, given that the driver needs to initiate
> the communication?
>
A while back I tried to start a Windows NT driver model emulator project
for Linux. Either I could not inspire people well enough, or the group
lacked enthusiasm. Either it went nowhere. Anyway this was one of the
issues I had.
In essence, my solution was a user space process that used signals and
shared memory to manage user-ish things the kernel could do under
Windows NT. I guess this is similar to how Netware does it.
As for file access from the kernel. I don't understand why this is not a
priority, but, it does not seem to be and does not seem to be without
work-arounds. Good luck.
--
Mohawk Software
Windows 95, Windows NT, UNIX, Linux. Applications, drivers, support.
Take the Mohawk Software Computer Survey at: www.mohawksoft.com
------------------------------
From: Andre Oliveira da Costa <[EMAIL PROTECTED]>
Subject: glibc2 and stat function
Date: Wed, 21 Apr 1999 09:57:45 -0300
Reply-To: [EMAIL PROTECTED]
Hi there,
I don't know if this is the most adequate place to ask this; it it
isn't, please forgive me.
The problem I'm facing is: I have a RedHat 5.2 system (glibc2 v2.0.7),
and I'm trying to compile (with gcc) a library that uses the stat
function (declared on <sys/stat.h>) to get file status. The problem is:
if I turn on debug information (-g), the dynamic linker is unable to
find 'stat' in any of the system libraries; the lib compiles without a
hitch, but 'stat' remains undefined:
[sisley:~/prj/iup/lib/Linux] nm -B libiupmot.so | grep stat
U stat
This doesn't happen if I turn off debug information and turn on inline
functions optimization (-O3 or -finline-functions):
[sisley:~/prj/iup/lib/Linux] nm -B libiupmot.so | grep stat
U __xstat
Searching libc.so for these symbols, this is what I find:
[sisley:/lib] nm -B libc.so.6 | grep -w "_*stat"
(no output)
[sisley:/lib] nm -B libc.so.6 | grep -w "_*xstat"
000690e0 T __xstat
000690e0 W _xstat
Looking into stat.h, I noticed there are different definitions for stat:
one inlined and one not inlined, conditioned by some constants defined
elsewhere. The inlined version is actually a wrapper for other functions
(__xstat, __lxstat, __fxstat etc.) which, as we could see from the
output above, _are_ present on libc.
To make a long story short: shouldn't there be a stat function in some
syslib, for the case where I am not using the inlined version of stat?
Did I miss something here?
Just for the record: I compile the same lib on a libc5 system
(Slackware) flawlessly -- i.e. 'stat' _is_ indeed included into the lib:
[duchamp:~/prj/iup/lib/Linux] nm -B libiupmot.so | grep stat
U _xstat
0001ac00 t stat
00027f00 t stat
Any help would be much appreciated. I would be thankful if the answers
could be posted to my email address also, since I don't check this
newsgroup very often.
Regards,
--
Andre Oliveira da Costa
([EMAIL PROTECTED])
------------------------------
From: [EMAIL PROTECTED]
Subject: physical Memory
Date: Wed, 21 Apr 1999 14:21:08 GMT
Moin,
I have a PCI-Card with busmaster-DMA. I programed the address with the
following lines:
unsigned long *dmabuf;
dmabuf = (unsigned long *) malloc(4096);
SIOREG(0x08) = (unsigned log) dmabuf;
Because it do not work, I think, that the address from the malloc() function
is not a physical address.
How can I get the reel address of an page in unser-memory and how can I made
that this page will not be swapped.
Thanks
Christian Mardm�ller
============= Posted via Deja News, The Discussion Network ============
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
------------------------------
From: Tin Siladin <[EMAIL PROTECTED]>
Subject: Re: EXT2 Sponsor File system
Date: Wed, 21 Apr 1999 12:43:38 -0400
"Erik N." wrote:
>
> Hello Tin (or anybody else)!
>
> Thank you for you interest in my posting and your very interesting answer.
> There is maybe something I should have mentioned in my first posting. The
> goal of this project isn't to fully implement an all working sponsoring file
> system, but simply a stripped down file system with the basic operations
> working. The file system isn't going to be distributed to the linux
> community or something like that, it is just a part of my examination.
Good. If you believe that a sponsoring file system can be useful, just
show
that it's possible and why it is useful (write a killer-app :) ).
> So for now (at least), I will leave the really tricky issues aside. I will
> not modify the fsck or something like that...
> I just need some guidance on the implementation and where in the source code
> I will make the changes.
>
> This is how I think I will proceed from here:
>
> 1. Make a new partition to hold the sp-fs.
> I believe this will involve modifying the source code for mke2fs (the
> tool for making a ext2 file system). And the default definitions of the
> ext2 directory structure.
Yes, just copy the mke2fs source tree and replace all occurrences of
ext2 to extsp (or whatever you want to call it) and then deal with the
structures you will define in the extsp_fs.h, extsp_fs_i.h and
extsp_fs_sb.h files in the /usr/src/linux/include/linux directory. You
should also copy the /usr/src/linux/fs/ext2 tree to
/usr/src/linux/fs/extsp and do the global replace there.
You will also want to add:
struct extsp_inode_info extsp_i;
to the union u in struct inode in /usr/src/linux/include/linux/fs.h
>
> Since the structure of a directory holds a table containing all the
> inodes/files in the directory, I will modify this table to also hold a
> pointer to a default account for the file, or a pointer to a place where
> information about sponsoring users accounts are stored. I am not
> really clear about how or where the account information will be stored
> yet. Maybe in a file (bad - can be deleted) or in the directory
> structure
Perhaps you should examine how quotas are implemented. I'm not sure how
linux does it, but however they do it, you may be able to do something
similar.
> itself by modifying the rec_len. I don't know much about the rec_len so
> I would appriciate any explaination about that.
"Linux Kernel Internals" by Beck, et al. is a very useful book.
> 2. Modify the kernel specific part of the source code for the ext2 file
> system so the linux kernel will be able to read the sponsor file system
> (also do other stuff). This will be implemented as a loadable kernel
> module.
>
> 3. Write a program that will be an "on mount" running process that
> continuously walks through the filesystem and charges the users for
> their files. They will cost accordingly to the disk space they occupy.
> The program will remove files that cannot be payed for.
> There will always be a default account for each file. Third party
> sponsors can connect their accounts to the default account and thereby
> sponsor the file.
I wasn't even thinking about a daemon process, but I guess it makes
sense.
You'll have to decide what functionality the daemon should have, and
what you
need in the kernel. If you're Linus you'll say "it can all be done in
the daemon". If you're Bill you'll say : "I'll put it all in the kernel
(and I won't let anybody see it)". Oops, that means that I think like
Bill. (Scary ?).
>
> So... any suggestions or comments about my "plan"? Anything that sounds
> really crazy or wrong? Please tell me anything...
1. If I OWN file x, and its size goes against my quota, and you SPONSOR
file x, and then I decide to delete file x, the ownership should change
to yours and go against your QUOTA. Preferably atomically :)
2. Reading up on ACL's may be useful.
Of course that's just my opinion, I may be wrong.
Tin Siladin
CrosStor Software
------------------------------
From: [EMAIL PROTECTED]�encroute.fr (R�khar�ur Egilsson)
Crossposted-To: comp.os.linux.development.apps
Subject: Sound] How do I stop recording ?
Date: 21 Apr 1999 14:48:31 GMT
Reply-To: =?iso-8859-1?Q?[EMAIL PROTECTED]=E9-en-croute.fr?=
Has anybody any experience with threads and sound programming ??
I need a way to make the sound driver stop recording without closing
the file descriptor.
What is the official (correct) method of stopping recording from
/dev/dsp while doing a read(2) ?
That is, if you want to stop immeadiately but the process needs to do
some processing/cleanup afterwards and thus can't be "just" killed.
Scenerio:
===========
In my application I create a new thread that starts recording from
/dev/dsp.
The thread records at 8000Hz, 8 bits, it doesn't matter if it
records in stereo or not.
When I want to stop recording the parent process does:
ioctl(fd_dsp, SNDCTL_DSP_RESET, 0);
The recording process stops recording, but does not return from the
read(2) until 7-8 seconds later. It returns -1 as expected.
After the ioctl the kernel logs (immediately after the ioctl):
kernel: Sound: DMA (input) timed out - IRQ/DRQ config error?
The ioctl(2) and read(2) are done in seperate threads on the same fd.
The ioctl reports no error (0 returned).
My problem is that my application can do nothing but hang there and
wait 7-8 seconds for the recording process finish and clean up.
Is this normal behaviour ??
--
RIKHARDUR EGILSSON - Ing�nieur Syst�me
echo '[q]sa[ln0=aln80%Pln80/snlbx]16isb15CB32EF3AF9C0E5D7272C3AF4F2snlbxq'|dc
------------------------------
From: [EMAIL PROTECTED]
Subject: high availability products for Linux?
Date: 21 Apr 1999 13:27:45 GMT
I'm trying to come up with a list of commercial high availability
products for Linux, and would appreciate pointers to companies doing
that.
I'm developing a proposal for a server, which has to be up all the
time. If anything fails on it, a backup needs to take over.
Thanks
Joe
------------------------------
From: Varghese Panicker K K <[EMAIL PROTECTED]>
Subject: Problem with File System Check
Date: Wed, 21 Apr 1999 16:27:29 +0530
HI All
I've used LINUX and a variety of UNIX flavors during
the academics. I have noticed one thing, LINUX systems are
more prone to crash after an improper shutdown (may be a
power failure which used to happen vey often). In one case I
found that while booting, it is asking to do fsck manually
and asks the root passwd. But when I give the root passwd,
the system just reboots and gives the earlier message again
(Infinite loop). At this point I need a floppy to disk to
reboot the system.
However, most of the other UNIX flavors like Solaris,
IRIX, AIX etc. very rarely (when compared to LINUX) gave such
errors. Even the very old Sun OS 4.0.x. (running on SUN3
machines) was more stable.
Considering the popularity that LINUX has achieved
as a good (or is it stable ?) OS, I think this is an important
issue. I was wondering, whether there is any thing specific
to LINUX filesystems (or filesystem checks) which makes it
behave 'badly'. Any comments ?
regards
varghese
------------------------------
From: "Bill Zimmerly" <[EMAIL PROTECTED]>
Subject: Re: "Permission denied"
Date: Wed, 21 Apr 1999 09:05:05 -0500
Peter Samuelson <[EMAIL PROTECTED]> wrote in message
news:7fjjcu$q70$[EMAIL PROTECTED]...
> [snip]
> I don't remember if you said which distribution you were running
> ... anyway, for Debian, the pid file is in /var/run, so:
It's RedHat 5.1 Peter, and I'll take a look at /var/run. Thanks!
------------------------------
From: [EMAIL PROTECTED] (Andreas Dilger)
Crossposted-To:
comp.os.ms-windows.programmer.drivers,comp.os.linux.misc,comp.os.linux.development.apps
Subject: Re: Development guide for HP CRet & PhotoRetII
Date: 21 Apr 1999 15:53:14 GMT
In article <[EMAIL PROTECTED]>, Anders Staehr <[EMAIL PROTECTED]> wrote:
>Has anyone tried to develop a driver for the HP DeskJet 890, 1100 or
>1120 printers utilizing the HP proprietary CRet or PhotoRetII mode, and
>where did you find the necessary information?
The first thing I would check is the latest version of GhostScript, as this
is as close to a "printer driver" as you get with Linux (barring proprietary
stuff like in WordPerfect). If these printers already handle PostScript
directly, then there's not really any reason for doing more, is there? If
they don't handle PS, then I'd suggest enhancing GS starting with one of the
other HP color PCL drivers, which may have most of the functionality already.
Cheers, Andreas
--
Andreas Dilger University of Calgary \"If a man ate a pound of pasta and
Micronet Research Group \ a pound of antipasto, would they
Dept of Electrical & Computer Engineering \ cancel out, leaving him still
http://www-mddsp.enel.ucalgary.ca/People/adilger/ hungry?" -- Dogbert
------------------------------
From: "Clint Byrum" <[EMAIL PROTECTED]>
Subject: Beowulf, Can we use it?
Date: Wed, 21 Apr 1999 13:29:39 -0700
I'm trying to see if we can use Beowulf to provide bigger and/or more robust
linux systems to run our business apps on. Currently, we can put(based on
benchmarks) 200 users on a 4 Processor, intel SC450NX Board w/ 1GB of RAM, a
Mylex HW Raid Controller, running on the 2.2.2 kernel. What I want to know
is, If we wanted to do 300 users, could we just use beowulf to "cluster" in
another machine of similar caliber(or maybe half). Would it be better just
to get a Gigbit ethernet connect between them and have one use nfs to access
the other one(half the users would log into the second machine)? Please
advise! Thanks in advance!
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
