Linux-Development-Sys Digest #133, Volume #8 Sat, 9 Sep 00 19:13:13 EDT
Contents:
Re: Caching files from CD---problem when playing MP3s on CD ("Les Mikesell")
Re: Linux server to hold thousands of tcp connections? ("Les Mikesell")
Re: Wish for a writable ISO-9660 compatible filsystem ("Les Mikesell")
Re: Installing Win98, Win2000 and Linux on one PC?! ("Gene Hill")
newbie: developping inetd subserver (Pierre Wieser)
# of maximum outstanding disk IO's ("Hog Rider")
Re: # of maximum outstanding disk IO's ("Peter T. Breuer")
Re: Wish for a writable ISO-9660 compatible filsystem (brian moore)
Re: # of maximum outstanding disk IO's (Kaz Kylheku)
Re: OFF_T and 64bit Qustation?? (Philip Armstrong)
Re: which kernel version support AGP? (Philip Armstrong)
URGENT help required - iproute2+tc Quality of Service ("Steve Cox")
Re: newbie: developping inetd subserver (Juergen Heinzl)
----------------------------------------------------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc
Subject: Re: Caching files from CD---problem when playing MP3s on CD
Date: Sat, 09 Sep 2000 16:13:52 GMT
"Bruce Stephens" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> For convenience, I've stuck lots of MP3 files onto a CD. However,
> when I try playing them using freeamp (say), I find the CD is pretty
> constantly active, which is annoying and wasteful---especially on a
> laptop.
>
> What's going wrong? Why isn't the 5M or so file just buffered? Hmm,
> or does file buffering work per block or something?
>
> More importantly, what might I try to avoid this behaviour? Is there
> some setting I can change to alter the behaviour of this?
Stdio only performs a small amount of read-ahead and buffering doesn't
help until something reads the data. If you did a 'cat file >/dev/null'
you would force the contents to be read into the buffer and the disk
would not be accessed on subsequent reads until something else
forces it out of the buffer. If you really want to control physical
access, create a ramdisk and copy there first.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.networking,comp.protocols.tcp-ip
Subject: Re: Linux server to hold thousands of tcp connections?
Date: Sat, 09 Sep 2000 16:33:42 GMT
"Dan Kegel" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
>
> > What are the max # of
> > BSD sockets that could probably be held open at the same time on LINUX?
>
> Somewhere between 10,000 and 100,000 with the 2.4 kernel.
> I suggest starting out with a single-threaded model using
> poll(). You may never need to add additional threads; since
> you won't be doing any disk I/O or slow computations, nothing
> should block long enough for you to benefit from multiple threads.
Does someone have a reasonable estimate of the memory required per
connection to survive a routing glitch or a set of extremely slow
client connections where most of your connections will have sent
a full TCP window of data without getting an acknowledgment. Also,
what happens in Linux when this requirement exceeds available
RAM/swap? I think I have seen this situation crash an otherwise robust
freebsd box.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "Les Mikesell" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc
Subject: Re: Wish for a writable ISO-9660 compatible filsystem
Date: Sat, 09 Sep 2000 16:47:55 GMT
"Otto Wyss" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> While I was reorganizing my backup (using a CD-writer), I had the idea
> of using an ISO-9660 image file mounted through the loopback device. I
> soon had to learn it won't work the way I liked since ISO-9660 is simply
> readonly.
>
> I think it's time Linux gets the ability to use writable image files, so
> I'm going to make the following proposition:
>
> 1. Creating a writable ISO-9660 compatible fs (I'm calling it isorw)
> - mounted on a writable device, writing/deleting/etc is possible
> - written on a CD, it should be readable as if it were ISO-9660
> - capable of Joliet and/or RockRigde extension
[...]
> Q: Why ISO-9660 compatible
> A: For archiving puroses (storing and retrieving on CD's for serveral
> years) there aren't many alternatives. Vfat is one but not very
> attractive to me. UDF might be another in a few year but not now since
> not many computers can read UDF.
Perhaps 'not many' in the sense of not many different OS types, but since
most Windows computers that have a CDRW also have the directcd
software included, there are in fact a large number in an absolute
sense that can do UDF.
> What are you thinking about my proposition? Could this be done or are
> there obstacles I don't see. Is it alltogether not necessary, because
> there's a much better solution?
I assume that the UDF format was designed to be more efficient than
ISO9660 for r/w access, so I think it would be better to support
UDF in the long run.
Les Mikesell
[EMAIL PROTECTED]
------------------------------
From: "Gene Hill" <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux,comp.os.linux.setup,microsoft.public.win2000.applications
Subject: Re: Installing Win98, Win2000 and Linux on one PC?!
Date: Sat, 9 Sep 2000 10:22:42 -0700
What you are proposing won't be hard. The only thing to mention is that you
must install win2k with FAT32 file system. If you use NTFS, win98 won't be
able to see that partition.
Gene
"Shicheng" <[EMAIL PROTECTED]> wrote in message
news:8pdk29$pii$[EMAIL PROTECTED]...
> Hello there,
> We would like to install Win98, Win2000 and linux (redhat 6.0)
> three OSs onto our PC. The PC has a 30 GB hard disk, 128 MB memory
> and a 700 MHz CPU.
>
> We would like to have the above three OSs installed; after the
> installation, we could select one of the three OSs during the boot time,
> otherwise, the PC will be booted automatically to the default
> OS (Win 2000 is the default one). Each OS would use one partition,
> so the three OSs would need three partitions.
> Apart from these three OS partitions, we may also need to create
> two more partitions using the remaining space of the disk:
> one such a partition would be for the storage of linux's data and the
> other one would be for the data storage for both the Win98 and Win2000
> OSs; so the last data partition needs to be seen by both the 98 and the
> 2000 OSs.
>
> We would be grateful you could give us some advice on the above.
>
> Thanks,
>
> Shicheng
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
------------------------------
From: Pierre Wieser <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: newbie: developping inetd subserver
Date: Sat, 09 Sep 2000 18:51:43 GMT
Hi,
As I have small time left, I'm currently trying to write a small C/S
appplication. The server-side should be a subserver of inetd...
I've understood that the server receives its messages on STDIN. So, I
can send and log a simple message to my server.
But I didn't found any responses on two points:
- how can my server get informations about the client which is sending
the message ? (or how can I get a socket when the connection is handled
by inetd ?)
- how can my server response to the client ? Should it open a new
connection or can it response via inetd ?
Thanks in advance for your help
Regards
Pierre
------------------------------
From: "Hog Rider" <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc
Subject: # of maximum outstanding disk IO's
Date: Sat, 9 Sep 2000 11:46:11 -0700
We are doing some performance testing of disk storage on different
platforms, and I was wondering if Linux has a maximum number of IO's that
can be outstanding at any given time. Translated, if the storage device is
busy and the h/w queue is full, how many IO's will Linux queue up waiting
for the storage device to be available again.
Thanks in advance!
Rider
------------------------------
From: "Peter T. Breuer" <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc
Subject: Re: # of maximum outstanding disk IO's
Date: 9 Sep 2000 19:18:56 GMT
In comp.os.linux.misc Hog Rider <[EMAIL PROTECTED]> wrote:
: We are doing some performance testing of disk storage on different
: platforms, and I was wondering if Linux has a maximum number of IO's that
: can be outstanding at any given time. Translated, if the storage device is
: busy and the h/w queue is full, how many IO's will Linux queue up waiting
And what do you mean by "an io".
: for the storage device to be available again.
Your question is vague. Why not use the correct terms instead of baby
talk!
The kernel has a hard limit NR_REQUEST of how many block requests it
will queue on a block device at a time. There is a 2-1 split preserved
in favour of reads against writes on that queue too. But each driver
code independently can take requests off that kernel queue as fast as it
likes, and pass it to the driven device as fast as it likes. And the
kernel limit is changeable by simply changing the constant value and
recompiling (as I recall, the default kernel limit is 64, but go check).
And the size of each block request is variable too. The kernel can and
does aggregate adjacent - or even nonadjacent - requests into one
according to rules defined by the individual drivers. This happens as
requests are passed to the driver queue.
So I don't grok what you mean. Look at ll_rw_blk.c. All will become
plain. Or at least part will ...
Peter
------------------------------
From: [EMAIL PROTECTED] (brian moore)
Crossposted-To: comp.os.linux.misc
Subject: Re: Wish for a writable ISO-9660 compatible filsystem
Date: Sat, 09 Sep 2000 20:34:46 GMT
On Fri, 8 Sep 2000 23:33:40 +0200,
Otto Wyss <[EMAIL PROTECTED]> wrote:
> While I was reorganizing my backup (using a CD-writer), I had the idea
> of using an ISO-9660 image file mounted through the loopback device. I
> soon had to learn it won't work the way I liked since ISO-9660 is simply
> readonly.
>
> I think it's time Linux gets the ability to use writable image files, so
> I'm going to make the following proposition:
>
> 1. Creating a writable ISO-9660 compatible fs (I'm calling it isorw)
> - mounted on a writable device, writing/deleting/etc is possible
> - written on a CD, it should be readable as if it were ISO-9660
> - capable of Joliet and/or RockRigde extension
>
> 2. Creating tools to handle isorw
> - kernel device drivers, readable part might be identical to ISO-9660
> - mkimage for creating image files
> - possible extension for CD writing programms
>
> Q: Why a new filesystem
> A: Sofar there is no fs for Linux which is writable on a disk and still
> ISO-9660 combatible on a CD. Of course a writable image could be made
> with ext2fs but it always needs special treatment if written to CD.
Which it always will need be for ISO-9660.
> Q: Why ISO-9660 compatible
> A: For archiving puroses (storing and retrieving on CD's for serveral
> years) there aren't many alternatives. Vfat is one but not very
> attractive to me. UDF might be another in a few year but not now since
> not many computers can read UDF.
But you haven't looked at the structure of ISO-9660. If you add a
single block to a file at the start of your 'image' you have to move
EVERYTHING after it (or waste the space -- which is not nice for trying
to fit as much as you can on an 650MB CDR).
> Q: How complicated is isorw
> A: It depends on the way it's implemented. The reading part could
> possible taken from the ISO driver. The writing part might be implementd
> along the line of ISO-9660 Level 3 (as Rod Smith kindly suggested). IMHO
> although I haven't read yet it I think it's doable.
>
> Q: Why not simply changing the current ISO implementation.
> A: This shouldn't be changed since lot of people needs it working.
> Anyway if isorw is working it might be merged in later on.
>
> Q: Could isorw be used on other systems, i.e. Windows
> A: Of course but only readonly as an ordinary ISO-9660 file unless isorw
> is implemented there as well.
>
> Q: How about Joliet format
> A: There shouldn't be a problem with Joliet and/or RockRidge extensions,
> it should be handled as usual.
>
> What are you thinking about my proposition? Could this be done or are
> there obstacles I don't see. Is it alltogether not necessary, because
> there's a much better solution?
>
> Is there anyone who is interested in isorw as well and likes to help me.
> I'm currently not a Linux/Unix developer, I mostly programmed on TANDEM
> computers.
If you did it, it would be unbearably slow. "Hey, how come when I
update it a file it takes MINUTES before the 'cp' is done??"
Moving around hundreds of megabytes, even on a relatively fast medium
like a hard drive is not exactly quick.
ISO-9660 is highly optimized (by design) for read-only access. You
would need to rebuild the image far too often to make it usable.
--
Brian Moore | Of course vi is God's editor.
Sysadmin, C/Perl Hacker | If He used Emacs, He'd still be waiting
Usenet Vandal | for it to load on the seventh day.
Netscum, Bane of Elves.
------------------------------
From: [EMAIL PROTECTED] (Kaz Kylheku)
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc
Subject: Re: # of maximum outstanding disk IO's
Reply-To: [EMAIL PROTECTED]
Date: Sat, 09 Sep 2000 20:50:26 GMT
On Sat, 9 Sep 2000 11:46:11 -0700, Hog Rider <[EMAIL PROTECTED]> wrote:
>We are doing some performance testing of disk storage on different
>platforms, and I was wondering if Linux has a maximum number of IO's that
>can be outstanding at any given time. Translated, if the storage device is
>busy and the h/w queue is full, how many IO's will Linux queue up waiting
>for the storage device to be available again.
Linux has a dynamic buffer cache that uses all available memory. Buffers that
are modified are called dirty; they are periodically flushed to disk.
------------------------------
From: [EMAIL PROTECTED] (Philip Armstrong)
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.portable
Subject: Re: OFF_T and 64bit Qustation??
Date: 9 Sep 2000 21:59:31 +0100
In article <%bau5.22$[EMAIL PROTECTED]>,
Sean Oh <[EMAIL PROTECTED]> wrote:
>Then, is there patch for making 64bit LFS support??
Check out http://www.scyld.com/ for patches
or try a 2.4.0 prerelease kernel :) (ext2 has 64 bit file support in
2.4)
Note that you *might* need to upgrade your libc to get full 64bit
support in userland apps. See the details on the above web pages for
details.
cheers,
Phil
--
http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt
------------------------------
From: [EMAIL PROTECTED] (Philip Armstrong)
Subject: Re: which kernel version support AGP?
Date: 9 Sep 2000 22:24:33 +0100
In article <[EMAIL PROTECTED]>,
Toby Haynes <[EMAIL PROTECTED]> wrote:
>!! "Benny" == Benny Li <[EMAIL PROTECTED]> writes:
>
> Benny> hi, I want to use Voodoo3 3000 AGP, but I heard that earlier kernels
> Benny> don't support AGP yet. Can somebody please tell me which kernel
> Benny> version(s) support AGP?
>
>I've seen agpgart patches for 2.2.13 and later - there may be some in earlier
>kernels. Or you could back port the patches ...
Aren't these only required for agp cards that require access to main
system memory for caching textures and the like?
Most video cards have abandoned that feature of AGP as not being worth
the trouble as far as I can see...certainly my AGP Matrox Millennium
II works fine without an agpgart module.
Phil
--
http://www.kantaka.co.uk/ .oOo. public key: http://www.kantaka.co.uk/gpg.txt
------------------------------
From: "Steve Cox" <[EMAIL PROTECTED]>
Subject: URGENT help required - iproute2+tc Quality of Service
Date: Sat, 09 Sep 2000 23:20:31 +0000
Crossposted-To: comp.os.linux.networking
Hi,
I'm in urgent need of help with 'Quality of Service' facilities of iproute2 -
specifically the tc command. I have been trying to set up a linux box
with 2 NICs to act as a bandwidth throttling router. This is urgently
required for work to test bandwidth functionality but I've been
running inot problems
The router is up and running fine but I cannot get the tc command to work.
I always seem to get the same errors. I've been esperimenting with
differnet commands but my first attempt - see example - is copied out
of the iproute2 HOWTO:
./tc class add dev eth0 parent 10:0 classid 10:1 cbq bandwidth \
10Mbit rate 10Mbit allot 1514 weight 1Mbit prio 8 maxburst 20 avpkt 1000
./tc class add dev eth0 parent 10:1 classid 10:100 cbq bandwidth 10Mbit \
rate 1Mbit allot 1514 weight 800Kbit prio 5 maxburst 20 avpkt 1000 bounded
./tc qdisc add dev eth0 parent 10:100 sfq quantum 1514b perturb 15
./tc filter add dev eth0 parent 10:0 protocol ip prio 100 u32 match ip \
dst 172.16.0.0/16 flowid 10:100
and generates:
RTNETLINK answers: No such file or directory
RTNETLINK answers: No such file or directory
RTNETLINK answers: No such file or directory
RTNETLINK answers: Invalid argument
I first tried the with the standard Mandrake 2.2.4-15 kernel - and the
latest rpm download of iproutes2. Next I donwloaded the iproute2
'patches' for the kernel and recompiled - also downloading the latest
sources for iproute2 and patching those. Finally, I have tried compiling
the latest 2.4 kernel with all the appropraite modules as requested in
most documentation - I have even tried compiling in (rather than loading
moduled) nearly any network related function I can find that could possibly
be linked - but still I get the same errors. Whether they are being
generated by the same problem I don't know (I have seen documentation
saying that there are very limited error messaging for tc). It must
be something else I'm doing - or haven't done.
I'm going to try 'shaper' Sunday - just in case - but it looks as if
it will only run when modularized (and I compiled it straight in the last kernel
build) so its Sunday in the office too :-( )
Any thoughts, possible tests, suggestions on getting to the bottom of
this will be very greatfully received. Also, any pointers to other forums to
try would be helpful.
Many thanks
Steve C
stevec@canaries_DOT_co_DOT_uk
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Crossposted-To: comp.os.linux.development.apps
Subject: Re: newbie: developping inetd subserver
Date: 9 Sep 2000 22:54:53 GMT
In article <[EMAIL PROTECTED]>, Pierre Wieser wrote:
>Hi,
>
>As I have small time left, I'm currently trying to write a small C/S
>appplication. The server-side should be a subserver of inetd...
>
>I've understood that the server receives its messages on STDIN. So, I
>can send and log a simple message to my server.
>
>But I didn't found any responses on two points:
>
>- how can my server get informations about the client which is sending
>the message ? (or how can I get a socket when the connection is handled
>by inetd ?)
[-]
You don't. inetd gets the socket for you.
>- how can my server response to the client ? Should it open a new
>connection or can it response via inetd ?
[-]
You do not communicate via inetd but again via the standard file
descriptors.
Cheers,
Juergen
--
\ Real name : J�rgen Heinzl \ no flames /
\ EMail Private : [EMAIL PROTECTED] \ send money instead /
------------------------------
** 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
******************************
