Linux-Development-Sys Digest #226, Volume #8 Thu, 19 Oct 00 12:13:08 EDT
Contents:
Re: possible tcpd bug ("D. Stimits")
Concurrent service clients number problem in Apache ("Choon Young, Cho")
Re: possible tcpd bug (Kaz Kylheku)
Re: low-level I/O details with memory-mapped I/O on Intel architecture ("Brian")
Re: System Halt from Kernel Space (Josef Moellers)
Re: ftp performence over wireless lan (Michel Talon)
Re: K62-550 3DNow! (Michel Talon)
Re: memory utilities (Gert Wollny)
Re: possible tcpd bug ([EMAIL PROTECTED])
nanosleep (Stephen Lohning)
Re: possible tcpd bug (Matthew Patterson)
Help needed with mmap (Buffer passed from Kernel to User) (lsnover)
C++ compiler tools ("David Steinhoff")
How to control ide bus with linux? ("NTUST News")
Re: How to control ide bus with linux? (Tony Lawrence)
Re: low-level I/O details with memory-mapped I/O on Intel architecture ("Steve
Schefter")
Re: lp under kernel 2.4.0test9 (Thomas Huber)
----------------------------------------------------------------------------
Date: Wed, 18 Oct 2000 18:26:52 -0600
From: "D. Stimits" <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: Re: possible tcpd bug
Matthew Patterson wrote:
>
> i was configuring tcpwrappers on a redhat 6.2 system the other day, and it
> didn't perform like the man pages said it should. according to the man pages, it
> checks hosts.allow, then hosts.deny, if it isn't in either, it allows the
> connection. here's the configuration that i had (different IPs, but you get the
> idea):
>
> hosts.allow:
> in.telnetd:192.168.0.
>
> hosts.deny
> ALL:ALL
>
> in this state, if i try to telnet to the machine, it wouldn't allow the
> connection. however, if i comment the "ALL:ALL" line in hosts.deny, it allows
> the connection. I consider this a bug, some may consider it a feature. i also
> don't know if it is something redhat-specific, as i haven't had a chance to try
> another distro yet.
>
> thanks for thoughts
>
> MHP
Did you restart inetd after you made changes? Try killall -HUP inetd
between changes.
------------------------------
From: "Choon Young, Cho" <[EMAIL PROTECTED]>
Subject: Concurrent service clients number problem in Apache
Date: Thu, 19 Oct 2000 10:41:37 +0900
Hi there,
My company use Compaq Proliant 8500 with linux radhat 6.2 and Apache
webserver.
When concurrent clients number is increase around 200, httpd servers are
reduce to 0 and can't service to clients eventhought memory and process are
under 50% usage.
After few minutes httpds are appeared again.
I think the problem is out of concurrent clients number limits.
It's true?
How can i increas the concurrent service clients number above 256 ?
Some books advice to modify HARD_SERVER_LIMIT in "/src/include/httpd.h"
file.
It works? Is there anyone who have similar expericences?
Please help me.
thanks in advance
Cho
------------------------------
From: [EMAIL PROTECTED] (Kaz Kylheku)
Subject: Re: possible tcpd bug
Reply-To: [EMAIL PROTECTED]
Date: Thu, 19 Oct 2000 04:06:27 GMT
On Wed, 18 Oct 2000 18:26:52 -0600, D. Stimits <[EMAIL PROTECTED]> wrote:
>Did you restart inetd after you made changes? Try killall -HUP inetd
>between changes.
These /etc/hosts.{allow,deny} configuration files are read by instances of tcpd
which are spawned from inetd as wrappers around the actual service executables
(hence ``tcp wrappers''). The configuration changes go into effect immediately,
the next time some connection attempt is made; nothing needs to be restarted.
You should killall -HUP inetd when you make changes to /etc/inetd.conf .
--
Any hyperlinks appearing in this article were inserted by the unscrupulous
operators of a Usenet-to-web gateway, without obtaining the proper permission
of the author, who does not endorse any of the linked-to products or services.
------------------------------
From: "Brian" <[EMAIL PROTECTED]>
Subject: Re: low-level I/O details with memory-mapped I/O on Intel architecture
Date: 19 Oct 2000 05:52:42 GMT
"John J. Willey" <[EMAIL PROTECTED]> ?????
news:[EMAIL PROTECTED]...
> I understand that the port I/O instructions IN, OUT, raise
> an additional signal on the address bus to indicate that
> the address refers to I/O space rather than memory space.
>
> However, with memory-mapped I/O I have not been able to
> locate information about the mechanism whereby 'foreign'
> memory, residing in ROM or peripheral devices, is
> distinguished from system RAM.
>
> That is, when memory-referencing instructions operate on
> non-system memory that has been configured through memory-
> mapping initialization via the undocumented system BIOS,
> what control or signal redirects the data transfer to the
> foreign memory/registers in ROM or, for example, a PCI-
> connected peripheral?
>
> How is the foreign memory "tagged" to distinguish it from
> a simple erroneous attempt to reach nonexistent system RAM?
>
> How does the 'foreign' memory announce itself electronically
> and thus make itself available to the CPU's memory-manipulating
> instructions?
>
> If this is the wrong forum for such a basic question, might someone be
> able to direct me to a more appropriate group?
As far as I know, the chipset (usually the north bridge) does the tricks. It
contains programmable registers that hold all the memory address decoding
information. When the CPU issues a memory address, the chipset looks up the
registers and decides which way it should go. For example, if a register
contains the information "address 0xc0000 to 0xdffff should be directed to
PCI side", then a memory reference to 0xc0000 would be passed down to PCI
bus for further decoding, and the system ram with the same address would not
be touched. For most normal memory access, no registers would claim it, and
thus is directed to the system ram. All these registers are programmed by
the system BIOS druing POST.
Brian
------------------------------
From: Josef Moellers <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.embedded
Subject: Re: System Halt from Kernel Space
Date: Thu, 19 Oct 2000 08:10:11 +0200
Erik Hensema wrote:
[ ... ]
> Just disable interrupts and go to an endless loop. This will effectivel=
y
> lock up the machine, which is what you want.
Unless, of course, the machine is an SMP box ...
-- =
Josef M=F6llers (Pinguinpfleger bei FSC)
If failure had no penalty success would not be a prize (T. Pratchett)
------------------------------
From: Michel Talon <[EMAIL PROTECTED]>
Subject: Re: ftp performence over wireless lan
Date: Thu, 19 Oct 2000 10:11:30 +0200
Jerome Corre <[EMAIL PROTECTED]> wrote:
> thanks for the advice i have installed ethereal and i am looking at it
> now, I have notice something else which might help to find the problem.
> When i start an ftp session (using 'ftp server'), I get the message
> 'Connected to server' but then i take a good 1-2 minute before i get
> another line of text with the name of the server, the port the type of
> server etc.. and a line asking me for a login name.
> Are the two problem related or is this just something else?
> i'll post the result from using ethereal later
One possibility is that you have DNS problems. For example the distant server
my be protected by tcpwrappers and try to resolve your IP adress to find your
name. If reverse DNS is badly configured, this can take some time, or fail
and you will be rejected.
--
Michel Talon
------------------------------
From: Michel Talon <[EMAIL PROTECTED]>
Subject: Re: K62-550 3DNow!
Date: Thu, 19 Oct 2000 10:17:16 +0200
Edward Lee <[EMAIL PROTECTED]> wrote:
> jwk wrote:
>> On Tue, 17 Oct 2000 16:01:14 GMT, John Hall
>> <[EMAIL PROTECTED]> wrote:
>> >How's the support for this cpu? I'm purchasing a laptop within the next week
>> >and the one I have my eye on has one of these. I know Linux will run fine on
>> >it, but i'm curious about performance. I know this chip emulates MMX to a
>> >certain degree, but that you really need to specifically take advantage of
>> >the 3DNOW! instruction set to see the bigger benefit. Does anyone have any
>> >comment? Specifically i'm curious if 3DNow acceleration has been implemented
>> >in the kernel or in any of the X servers, or if I can compile my own X
>> >server with 3dNow support.
>> >
> Nop, Pentium/Celeron will have illegal instructions with 3DNow. I don't think X
> even supports MMX, since X is supposed to run on many other none x86 chips.
I have compiled XFree 4.01 recently, and have observed that a lot of -DUSE_MMX
are present. So i assume that MMX are assembly coded somewhere in the server.
In general, gcc will not output MMX or 3DNow instructions. An example of
program who uses them is the mpeg encoder gogo (see the ports). Here they are
compiled through nasm which has support for them.
>> There are some routines in the kernel that use 3DNow! (checksumming,
>> some raid computations). The difference won't be that great. A certain
>> mp3-encoder has a 3DNow! enhanced version. That could be a bigger
>> difference, but not shocking.
> 3DNow is not much faster than equivalent MMX instructions. It allows more SIMD
> instructions and better coding, relatively speaking, in terms of assembly
> codes. However, unless Intel licenses the 3DNow instructions, it is too much
> troubles to go beyond MMX. Pentium/Celeron runs MMX better than K6-2, for same
> clock speed.
I have a Duron and have Linux and FreeBSD on it. I have observed that Linux
kernel tests use of MMX, 3DNow etc. and concludes that he will use MMX which
is faster.
--
Michel Talon
------------------------------
From: Gert Wollny <[EMAIL PROTECTED]>
Subject: Re: memory utilities
Date: Thu, 19 Oct 2000 11:42:58 +0200
Weiguang Shi wrote:
> Which tool can I use to check the memory(physical&virtual) sizes under linux?
> If you have more tools to recommend to monitor/deal with memory,
> could you please let me know too?
Try http://reality.sgi.com/raju/software.html#gmemusage
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: possible tcpd bug
Date: Thu, 19 Oct 2000 11:48:06 -0000
In article <[EMAIL PROTECTED]>,
Matthew Patterson <[EMAIL PROTECTED]> wrote:
>i was configuring tcpwrappers on a redhat 6.2 system the other day, and it
>didn't perform like the man pages said it should. according to the man pages, it
>checks hosts.allow, then hosts.deny, if it isn't in either, it allows the
>connection. here's the configuration that i had (different IPs, but you get the
>idea):
>
>hosts.allow:
>in.telnetd:192.168.0.
>
>hosts.deny
>ALL:ALL
>
>in this state, if i try to telnet to the machine, it wouldn't allow the
>connection. however, if i comment the "ALL:ALL" line in hosts.deny, it allows
>the connection. I consider this a bug, some may consider it a feature. i also
>don't know if it is something redhat-specific, as i haven't had a chance to try
>another distro yet.
Where were you telneting from? If from the same machine, you'll probably
be showing up as 127.0.0.1 instead of the 192.168 address. In which case
you'll need a localhost entry in hosts.allow.
--
http://www.spinics.net/linux
------------------------------
From: Stephen Lohning <[EMAIL PROTECTED]>
Subject: nanosleep
Date: Thu, 19 Oct 2000 22:13:40 +1000
has anybody got an example of how to use nanosleep function ?
I assume that it will only be of use in a kernal driver as that sort of
resolution is
not possible in user space programs.
------------------------------
From: Matthew Patterson <[EMAIL PROTECTED]>
Subject: Re: possible tcpd bug
Date: Thu, 19 Oct 2000 07:35:39 -0500
Matthew Patterson wrote:
>
> i was configuring tcpwrappers on a redhat 6.2 system the other day, and it
> didn't perform like the man pages said it should. according to the man pages, it
> checks hosts.allow, then hosts.deny, if it isn't in either, it allows the
> connection. here's the configuration that i had (different IPs, but you get the
> idea):
>
> hosts.allow:
> in.telnetd:192.168.0.
>
> hosts.deny
> ALL:ALL
>
> in this state, if i try to telnet to the machine, it wouldn't allow the
> connection. however, if i comment the "ALL:ALL" line in hosts.deny, it allows
> the connection. I consider this a bug, some may consider it a feature. i also
> don't know if it is something redhat-specific, as i haven't had a chance to try
> another distro yet.
>
> thanks for thoughts
>
> MHP
There seems to be some confusion, mostly because of my error. the hosts.allow
shows:
in.telnetd:127.
in.telnetd:192.168.0.
and hosts.deny is:
ALL:ALL
my concern is because i was making the telnet connection from an address within
192.168.0. Again, the IP's have been changed to protect my company.
Sorry for the lack of full information.
MHP
------------------------------
From: lsnover <[EMAIL PROTECTED]>
Subject: Help needed with mmap (Buffer passed from Kernel to User)
Date: Thu, 19 Oct 2000 12:36:08 GMT
I'm trying to map a large memory buffer (16M) from kernel to user. I
am using bigphysarea to allocate the memory. I am able to read & write
to it in kernel space, but when I try to use the memory address
returned to user space with mmap, I get 0xfffffff (-1). The data
written to it in kernel space is not there and I cannot write to the
memory in my user program. I believe bigphysarea is allocating using
kmalloc.
Any suggestions on what I am doing wrong? Below are excerpts from my
code.
Note: The sample code below is getting 4K of memory. This was changed
from 16M for debug purposes.
User code:
****************************************************************
.......
bpadrvfd = open("/dev/bpadrv",O_RDWR);
.......
#define IMAGE_PAGES 1
int test_image_alloc()
{
void * mmapAddr;
void * physaddr;
mmapAddr = mmap(NULL, IMAGE_PAGES*PAGE_SIZE,
PROT_READ|PROT_WRITE,
MAP_SHARED,
bpadrvfd, 0);
printf("retval for mmap = 0x%x\n",mmapAddr);
return ((int)mmapAddr);
}
Kernel module:
****************************************************************
static int bpamod_mmap( struct file* filep,
struct vm_area_struct *vma)
{
unsigned long size, start, page;
unsigned char * addr;
int off;
printk("bpamod_mmap\n");
printk("start=0x%x, end=0x%x, offset=0x%x flags=0x%x\n",
(int)vma->vm_start, (int)vma->vm_end, (int)vma-
>vm_offset,vma->vm_flags);
size = vma->vm_end-vma->vm_start;
if (size>MAX_BPABUF)
return -EINVAL;
addr = bigphysarea_alloc_pages(size / PAGE_SIZE, 1, GFP_KERNEL);
if (addr == 0)
return -EINVAL;
bpasaveaddr[addrcount++] = addr;
memset(addr, 0, size);
for (off=0; off < (vma->vm_end-vma->vm_start); off+=PAGE_SIZE)
mem_map_reserve(MAP_NR(__va(addr+off)));
*((unsigned*)addr) = 0x12345678;
size = (unsigned long)vma->vm_end-vma->vm_start;
start = (unsigned long)vma->vm_start;
page = (unsigned long)addr;
while (size > 0)
{
if (remap_page_range(start, page, PAGE_SIZE, PAGE_SHARED))
return -EAGAIN;
start+=PAGE_SIZE;
page+=PAGE_SIZE;
size-=PAGE_SIZE;
}
return 0;
}
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: "David Steinhoff" <[EMAIL PROTECTED]>
Crossposted-To:
alt.linux,alt.linux.redhat,alt.os.linux,alt.os.linux.redhat,comp.linux,comp.linux.os,comp.os.linux.development,linux.redhat,linux.redhat.devel,linux.redhat.development
Subject: C++ compiler tools
Date: Thu, 19 Oct 2000 13:51:27 GMT
We are beginning to port our C++ applications from NT to Linux (Redhat) and
are looking for opinions/recommendations for compilers. One important
consideration is that the compiler must be close to the C++ standard in its
language support. Our products use the STL heavily, so the implementation
it works with must be supported and robust. We have a lot of code and would
like to find a fast compiler that comes with a good debugger and generates
fast code. Lastly, some of our 3rd party code does not support gcc, so it
is not an option for us.
Thanks,
Dave
=================================================
David Steinhoff
Vice President, Software Development
Clareos, Inc.
209 S. Ashley St.
Ann Arbor, MI 48104
[EMAIL PROTECTED]
O: (734)827-1254
CP: (734)649-1304
------------------------------
From: "NTUST News" <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc,comp.os.linux.networking,tw.bbs.comp.linux
Subject: How to control ide bus with linux?
Date: Thu, 19 Oct 2000 22:23:57 +0800
hello all
I wanna control one chip with linux.
This chip supports 8 bit IDE bus
So....If I wanna program in linux.
How can I do that?
Please give me some solutions .....thanks....
------------------------------
From: Tony Lawrence <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc,comp.os.linux.networking,tw.bbs.comp.linux
Subject: Re: How to control ide bus with linux?
Date: Thu, 19 Oct 2000 10:59:39 -0400
NTUST News wrote:
>
> hello all
>
> I wanna control one chip with linux.
> This chip supports 8 bit IDE bus
> So....If I wanna program in linux.
> How can I do that?
> Please give me some solutions .....thanks....
Linux Device Drivers, by Alessandro Rubin
http://www.aplawrence.com/Books/lindd.shtml
And then ask more specific questions..
--
Tony Lawrence ([EMAIL PROTECTED])
SCO/Linux articles, help, book reviews, tests,
job listings and more : http://www.pcunix.com
------------------------------
Reply-To: <[EMAIL PROTECTED]>
From: "Steve Schefter" <[EMAIL PROTECTED]>
Subject: Re: low-level I/O details with memory-mapped I/O on Intel architecture
Date: Thu, 19 Oct 2000 15:46:11 GMT
>"John J. Willey" <[EMAIL PROTECTED]> ?????
>news:[EMAIL PROTECTED]...
>> How is the foreign memory "tagged" to distinguish it from
>> a simple erroneous attempt to reach nonexistent system RAM?
>>
>> How does the 'foreign' memory announce itself electronically
>> and thus make itself available to the CPU's memory-manipulating
>> instructions?
The short answer is that it does not, at least not ahead of the actual
read/write.
Unfortunately, how the host bus-to-PCI bridge works is going to
be specific to the particular chip you have, and that will be specific
to the type of processor you have. As I design only ISA and PCI
cards, I don't get into the host bus itself, just the peripheral side.
On the PCI bus, each device just looks at the address bus and the
command bits to see if it is being spoken to. The command bits
can be memory read, memory write, I/O read, I/O write, configuration
space read, ... In the case of PCI bus, there are not different lines
to indicate memory versus I/O versus configuration, just different
combinations of the same 4 lines. If a device sees that it is being
spoken to (based on the address it was assigned by the BIOS
at boot time -- a slot specific function), it asserts a line that says
it is the target of the instruction. It can then present or capture
the data and indicate that it is done. Or it can say, I'm busy now,
try again shortly in which case the CPU will be held in wait state
until the host bus-to-PCI bridge chip tries again.
On systems that have ISA slots as well, there is a PCI-to-ISA
bridge chip too. If no other device on the PCI bus identifies itself
as the target of the command in time, this bridge chip says it is and
puts the instruction on the ISA bus. Here, there are four mutually
exclusive lines to indicate the operation: memory read, memory
write, I/O read and I/O write. Again, the ISA devices respond
if they see the proper operation and their own address (as set
by DIP switches, driver software or the ISA PnP BIOS). But
they just respond by capturing/presenting the data, not by setting
any signal to say "it was me you were looking for and I've
responded". So if you try to read data from an invalid address,
the data bits just float and you read "air". Usually, the bits float
to a logical 1 so that it comes back as all ffs. Try this with the
DOS debug program if you have it.
Steve Schefter
The Software Group Limited
www.wanware.com
------------------------------
Reply-To: [EMAIL PROTECTED]
From: [EMAIL PROTECTED] (Thomas Huber)
Subject: Re: lp under kernel 2.4.0test9
Date: 19 Oct 2000 18:03:20 +0100
In article <N8xE5.33237$[EMAIL PROTECTED]>,
Chirok Han <[EMAIL PROTECTED]> writes:
> Hi.
>
> So far, lp.o has been successfully loaded under my 2.2.17 kernel.
> But when I upgraded the kernel to 2.4.0-test9,
> it suddenly stopped working. I get the following message
>
> kernel: lp: driver loaded but no devices found
I had that when the parallel port was not activated in the
BIOS. On 2.2.x, it worked all the same, but on 2.4.0-testx,
I had to activate it in the BIOS to work.
Thomas
------------------------------
** 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
******************************
