Linux-Development-Sys Digest #729, Volume #6 Thu, 20 May 99 09:14:14 EDT
Contents:
truncating mapped memory gives tight loop (Steve Peltz)
Re: i386 ENTER instruction problem (Georg Acher)
Re: never reboot to upgrade ? (Tony Smith)
Linux - Windows killer? (Greg Crockart)
Re: Hostile Takeover of Linux (John Hughes)
SCSI tape module problem ("David A. Wilson")
Re: invalid ELF header in glibc (Andreas Jaeger)
Re: mapping user space and kernel space (Robert Kaiser)
Re: Changes in signals (kernel > 2.2.1) ? (Andreas Schwab)
Re: Pop Authentication for Sendmail (Gary Mills)
Re: Programming PCI devices (Rajarshi Bandyopadhyay)
i386 ENTER instruction problem (Pierre Muller)
Re: how to dump core ??? (Emile van bergen)
never reboot to upgrade ? (Mike Rushford)
A growing online shopping mall (Hyper)
Re: Programming PCI devices ("Rajarshi Bandyopadyay(STR99)")
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Steve Peltz)
Subject: truncating mapped memory gives tight loop
Date: 19 May 1999 16:23:01 GMT
Alpha SX164 running 2.2.9; if I have a shared map on a file, and the
file gets truncated, I get a tight loop (which clears if the file gets
extended again).
I realize that accessing mapped memory past the end of a file is
unspecified behavior, but it used to do what you'd expect, give a SIGSEGV
(or, possibly, SIGBUS). The current behavior, while maybe conforming,
is certainly not very useful (and I found out about it with a job that
had a high priority; took a while to figure out what was going on,
as the whole system immediately hung).
This changed fairly recently, I'm almost positive since 2.2.0...
Truncating a file can interact oddly with a locked mapped region; Digital
Unix blocks the truncate until the memory is unlocked or unmapped, which
is also not particulary nice, but certainly guarantees that you can't get
a fault for accessing locked memory. I'd actually prefer that it allow
the truncate, and simply keep the memory mapped until it is no longer
locked (after all, if it is locked, it doesn't need any backing storage).
This could even be extended to allowing re-locking the region, which
would re-map pages (presumably zero filled), which could be shared with
any other process that locks the same pages (even if one process has a
page locked, another process that doesn't have it locked should get a
fault based on the current end-of-file).
------------------------------
From: [EMAIL PROTECTED] (Georg Acher)
Subject: Re: i386 ENTER instruction problem
Date: 20 May 1999 09:13:28 GMT
In article <7hutm2$1fc$[EMAIL PROTECTED]>, [EMAIL PROTECTED] (Pierre Muller)
writes:
|>
|> Using the ENTER instuction seems to create
|> problems when the amount %esp must be decreased
|> makes it change page !
|>
|> Does the page mecanism not know how to handle ENTER instruction ?
|> Is this a known i386 Linux problem ?
|> Is it solved in recent version ?
|>
|> Or is ENTER simply consider as an invalid instruction in Linux ?
|> This was tested on v2.0.33
One warning (found out on an P2): At least ENTER 0 (ie. just pushing ebp and
and setting it to esp) is more than 500-1000% SLOWER * than
'puhsl %ebp; movl %esp,%ebp'. It seems that this instruction has a bug and/or
is executed in microcode. On the other hand, LEAVE is slightly faster than the
'dicrete' commands.
*) recursive Fibonacci(40) is twice the time with ENTER as without it, LEAVE
gains about 3%.
--
Bye
Georg Acher, [EMAIL PROTECTED]
http://www.in.tum.de/~acher/
"Oh no, not again !" The bowl of petunias
------------------------------
From: Tony Smith <[EMAIL PROTECTED]>
Subject: Re: never reboot to upgrade ?
Date: Thu, 20 May 1999 10:07:35 +0100
Reply-To: [EMAIL PROTECTED]
This is a multi-part message in MIME format.
==============50FADE92080CB74258F62A33
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Mike Rushford wrote:
> I have enjoyed the stability of Linux for years and have only had to
> shut down my Linux boxes to upgrade the next Kernel version.
>
> Is there a way of building a kernel so that it takes the place of a
> currently running one without ever shutting down?
>
> Same question for Libraries in memory can we force library upgrades by
> just putting the new libs and sym links on disk and then doing something
>
> like sync and ldconfig to force a relink to the newly installed ones?
>
> It would be cool to claim linux could be up for ever with only hardware
> upgrades forcing a shutdown!
>
> --
> Mike [EMAIL PROTECTED] http://www.hooked.net/~tvs/eyes/
I doubt very much if you'll ever get a dynamic kernel upgrade on a single
machine (e.g. because of the interdependencies that would need to be
resolved between the kernel and the modules). What I think you want (and I
too would like this) is something akin to what Compaq (ex Tandem) provide
with their XC and S4000 products: Single System Image. i.e. Two (or more)
machines which look like one machine to all users. i.e. one process table,
all shared memory accessible from any node, all disk devices accessible from
any node. This allows you to take a node out of the cluster, upgrade it, and
plug it back in with zero downtime. Processes can be interrupted and moved
to a different node *between* instructions for load balancing or fault
tolerance.
I don't know what work, if any, is being done on this, but if we get that
functionality on Linux, it would make justifications for its use in
commercial solutions much easier (currently I am encountering a LOT of
resistance to the idea, and this sort of HA clustering would really tip the
balance).
Can anyone offer further insight into this area of development?
Tony
==============50FADE92080CB74258F62A33
Content-Type: text/x-vcard; charset=us-ascii;
name="tony.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Tony Smith
Content-Disposition: attachment;
filename="tony.vcf"
begin:vcard
n:Smith;Tony
x-mozilla-html:FALSE
adr:;;;;;;
version:2.1
email;internet:[EMAIL PROTECTED]
fn:Tony Smith
end:vcard
==============50FADE92080CB74258F62A33==
------------------------------
From: Greg Crockart <[EMAIL PROTECTED]>
Subject: Linux - Windows killer?
Date: Wed, 19 May 1999 17:41:00 +0100
Reply-To: [EMAIL PROTECTED]
If you're interested in Linux, this is quite cool.
>From Wednesday 19th Silicon.com is taking an in depth look at the future
of Linux - http://www.silicon.com/linuxweek.
In addition to all of the latest Linux and Open Source news, you can
watch exclusive interviews with leading lights from the Linux community
including Miguel de Icaza, father of GNOME and Eric Raymond, who's
stepping out of the Linux limelight. You can also check out the results
of the Silicon.com poll on whether Linux is ready for corporate use at
http://www.silicon.com/linuxweek.
Silicon.com will also be interviewing corporates that already have moved
over to Linux and will be debating key issues including:
- Will developer in-fighting damage Linux?
- Can Microsoft quash Linux?
- Why isn't Linux taking off in the corporate environment now that it
has decent services and support?
If you've got anything you want to see covered you can email the Silicon
editorial team at [EMAIL PROTECTED]
Thanks and enjoy the special!
------------------------------
From: John Hughes <[EMAIL PROTECTED]>
Subject: Re: Hostile Takeover of Linux
Date: 20 May 1999 11:26:13 +0200
"Ross Vandegrift" <[EMAIL PROTECTED]> writes:
> (This message is in UTF-8 format, due to the nature of an example provided
> below. I imagine that some servers and readers won't be able to read it
> properly, because of this. Sorry.)
You may be sorry but your message isn't UTF-8. It's plain old
ISO-8859-1.
> >Huh? What's 8-bit ASCII? There is no such beast. 256 is the *WRONG*
> >answer, always has been and always will be.
>
> No, 256 could be a correct answer, though 8-bit ASCII is properly called
> Extended ASCII. It is charset ISO Latin-1. It is 8 bits wide, with the
> extra characters used for non-english letters.
ISO 8859-1 contains the printable characters of ISO 646 as a subset.
So does ISO 8859-2,3,4,5,... and IBM code page 436, 850, ... and HP
Roman-8 and the Mac character set and ...
They could all be called "extended ASCII" if you felt like it.
Using the correct names avoids confusion (see below):
> >If "8-bit is the norm" what is the character with (decimal) code 128?
>
> The charecter `C' with a squiggle under it: (�).
Wrong. Capital C Cedilla is 12/07 (hex C7, decimal 199) in
ISO-8859-1. 128 is � in various IBM code pages, CP 437 for example.
128 is not a character in ISO 8859-1. The "Latin" sets don't define
the control characters. Latin-1 only has 191 characters.
> A quick web search on Extended ASCII turned up this decent looking page:
> http://www.johnco.cc.ks.us/~mfoster/ascii.html
It looks decent, but it makes the same mistake you do. There is no
one "extended ASCII", there are a multitude.
--
John Hughes <[EMAIL PROTECTED]>,
Atlantic Technologies Inc. Tel: +33-1-4313-3131
66 rue du Moulin de la Pointe, Fax: +33-1-4313-3139
75013 PARIS.
------------------------------
From: "David A. Wilson" <[EMAIL PROTECTED]>
Subject: SCSI tape module problem
Date: Wed, 19 May 1999 16:18:12 GMT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=
------------------------------
From: Andreas Jaeger <[EMAIL PROTECTED]>
Subject: Re: invalid ELF header in glibc
Date: 20 May 1999 09:42:28 +0200
>>>>> arjan de boer writes:
> Hello,
> I have a RedHat 5.0 / i586 OS and now and then updated some libaries. No
> have the following problem: the gnu assembler and loader (from the
> binutils package) complain about the libc.so.6 having an invalid ELF
> header. To be more precisely, 'ld' says:
>> ld: error in loading shared libraries
>> libc.so.6: invalid ELF header
> 'as' gives similar output. Other programs like 'ls' which also use
> 'libc.so.6' don't complain. Changing back to the original 'binutils' and
> 'glibc' RPM's that came with the RH5.0 distribution didn't change the
> problem.
> What might be the problem - and solution?
It seems that you've broken somehow libc.so.6, I'd advise to reinstall
the glibc package.
You can try to analyze libc.so.6 with objdump.
Andreas
--
Andreas Jaeger [EMAIL PROTECTED] [EMAIL PROTECTED]
for pgp-key finger [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Robert Kaiser)
Subject: Re: mapping user space and kernel space
Date: 20 May 1999 08:49:30 GMT
In article <[EMAIL PROTECTED]>,
Nitin Malik <[EMAIL PROTECTED]> writes:
> I need to map the user buffer to kernel memory to prevent copying
> overheads.
If you want to access a user buffer in kernel space without copying
overhead, you have to modify the kernel (see ftp://ftp.sysgo.de/pub/Linux
for some hints). However, another common method is to allocate a kernel
buffer and make it user-accessible through mmap(). Which is more appropriate
depends on what you want to do.
> Can some one give me an outline of how the locking is to be done?
Under the above mentioned URL, you'll find a kernel patch that demonstrates
how this can be done by (mis)using the mlock() system call.
Hope this helps
Rob
================================================================
Robert Kaiser email: rkaiser AT sysgo DOT de
SYSGO RTS GmbH
Mainz / Germany
------------------------------
From: Andreas Schwab <[EMAIL PROTECTED]>
Subject: Re: Changes in signals (kernel > 2.2.1) ?
Date: 20 May 1999 11:25:08 +0200
[EMAIL PROTECTED] (Jason Neudorf) writes:
|> >
|> >Not sure it's clear from Andreas' message, but designated initializers
|> >aren't portable either; they'll work with GCC and they'll be in the next
|> >C standard, but most C compilers don't support them yet.
|> >
|> >Just initialize using normal assignment; it's The Right Thing To Do.
|> >
|>
|> My attempts to initialize using normal assignments don't seem to work.
"Doesn't work" is a useless description. Why don't you just tell us what
you did?
--
Andreas Schwab "And now for something
[EMAIL PROTECTED] completely different"
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Gary Mills)
Crossposted-To:
alt.comp.linux.xxx,alt.comp.linux.isp,alt.os.linux,comp.os.linux.development.apps,comp.os.linux.misc,comp.os.linux.networking,comp.os.linux.security,comp.os.linux.setup,comp.os.linux.x,computer42.mail2news.linux-alert,comp.mail.sendmail
Subject: Re: Pop Authentication for Sendmail
Date: 19 May 1999 17:29:51 GMT
In <[EMAIL PROTECTED]> Roland Nadeau <[EMAIL PROTECTED]> writes:
> This is a problem for possable spam relaying. Can someone point me to
>a fix to use POP to authenticate or
> any other possable fix.
One solution for roaming users is described at...
http://mail.cc.umanitoba.ca/drac/index.html
--
Don't mailto:[EMAIL PROTECTED]
Don't mailto:[EMAIL PROTECTED]
Don't mailto:[EMAIL PROTECTED]
--
-Gary Mills- -Unix Support- -U of M Academic Computing and Networking-
------------------------------
From: Rajarshi Bandyopadhyay <[EMAIL PROTECTED]>
Subject: Re: Programming PCI devices
Date: Thu, 20 May 1999 15:33:23 +0530
Rajarshi Bandyopadhyay wrote:
> Hello Linux World,
>
> I have got down to my first kernel-hacking, a driver for a PCI data
> acquisition card. Can anyone help me with the specifics of PCI? I have a
> lot of src code as samples with me, but all are for ISA cards. Any help,
> links, docs etc.
>
> Thanx ,
> Raj
In case I haven't been very clear in the prev. post, I would like to know
how I should get the Base Addr and IRQ of my card from the system ( since
in case of PCI it is s'psed to be set on its own).
TIA
Raj
------------------------------
From: [EMAIL PROTECTED] (Pierre Muller)
Subject: i386 ENTER instruction problem
Date: 19 May 1999 17:50:58 GMT
Using the ENTER instuction seems to create
problems when the amount %esp must be decreased
makes it change page !
Does the page mecanism not know how to handle ENTER instruction ?
Is this a known i386 Linux problem ?
Is it solved in recent version ?
Or is ENTER simply consider as an invalid instruction in Linux ?
This was tested on v2.0.33
------------------------------
From: Emile van bergen <[EMAIL PROTECTED]>
Subject: Re: how to dump core ???
Date: Wed, 19 May 1999 20:15:53 +0200
On Wed, 19 May 1999, Zsolt Zsoldos wrote:
[SNIP]
>find out more about the cause of the assert failure. Now, this works
>fine on Irix, but on Linux, I do not get the core dump, just says
>"Aborted" and exits without generating the core file. I also tried
>to use "kill( getpid(), SIGABRT );" - same result.
Tried ulimit -c unlimited? Without it, the max. size of a core dump is
zero, effectively preventing any core dumps.
--
M.vr.gr. / Best regards,
Emile van Bergen (e-mail address: [EMAIL PROTECTED])
This e-mail message is 100% electronically degradeable and produced
on a GNU/Linux system.
------------------------------
From: Mike Rushford <[EMAIL PROTECTED]>
Crossposted-To: revue.linux-net
Subject: never reboot to upgrade ?
Date: 20 May 1999 13:10:25 +0200
I have enjoyed the stability of Linux for years and have only had to
shut down my Linux boxes to upgrade the next Kernel version.
Is there a way of building a kernel so that it takes the place of a
currently running one without ever shutting down?
Same question for Libraries in memory can we force library upgrades by
just putting the new libs and sym links on disk and then doing something
like sync and ldconfig to force a relink to the newly installed ones?
It would be cool to claim linux could be up for ever with only hardware
upgrades forcing a shutdown!
--
Mike [EMAIL PROTECTED] http://www.hooked.net/~tvs/eyes/
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
------------------------------
From: Hyper <[EMAIL PROTECTED]>
Subject: A growing online shopping mall
Date: 19 May 99 14:11:47 -0400
Reply-To: [EMAIL PROTECTED]
For links to the best shopping on the internet go to the Hyper-Lynx
homepage at http://www.hyper-lynx.com
------------------------------
From: "Rajarshi Bandyopadyay(STR99)" <[EMAIL PROTECTED]>
Subject: Re: Programming PCI devices
Date: Thu, 20 May 1999 17:15:13 +0530
On Thu, 20 May 1999 [EMAIL PROTECTED] wrote:
> In article <[EMAIL PROTECTED]> you wrote:
> : Hello Linux World,
> What kind of card is it ? I just finished writing the driver
Hi Vlad,
The card is a Nudaq (www.nudaq.com) PCI-9118 HG data acquistion card.
Salient features: 1K on-board FIFO memory, 330KHz sampling rate etc.
(Details in manual, also on website)
> To find IRQ of your PCI devices look at pci_find_device. It returns
> data structure with Irq there......
Thanx for the IRQ pointer.
Another thing. I am making the driver for an extremely real-time
application, and my boss wants to squeeze as much out of this card as
possible ( its quite costly, apparently). So , I am planning to
use DMA transfer ( understandably ). But I want to know the maximum size
of the DMA buffer that I can use safely without losing any samples.
How can I find that out?
Other than that, can u please make any suggestion to improve speeds(for
example, good usage of on-board FIFO etc) ?
I would love to have ur source code (I guess most daq cards have a
lot of things in common). Thanx a lot for that.
Thanx for all ur help.
Regards,
Raj
------------------------------
** 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
******************************
