headers
Zhu Tom-a2289c | 1 Sep 2006 02:37

Is this a problem of glibc to support NPTL?

Hi,

When testing NPTL on armlinux, (2.6.9+glibc 2.3.3)
we found that in glibc, there is no clone.S under
Nptl/sysdeps/unix/sysv/linux/arm,
so, clone.S in sysdeps/unix/sysv/linux/arm/ is used, 

But the RESET_PID is not defined, so, clone() will not invalidate the
pid in tls, which is 
copied from parent.

Is this a bug?
Should it be like:
There is a clone.S under Nptl/sysdeps/unix/sysv/linux/arm/ which defines
the RESET_PID and include sysdeps/unix/sysv/linux/arm/clone.S 

Thanks,
Tom

 
This is the code of clone.

ENTRY(__clone)
  <at>  sanity check args
 cmp r0, #0
 cmpne r1, #0
 moveq r0, #-EINVAL
 beq PLTJMP(syscall_error)

  <at>  insert the args onto the new stack
(Continue reading)

Permalink | Reply | 1 comment |
headers
Daniel Jacobowitz | 1 Sep 2006 03:52
Picon
Favicon

Re: Is this a problem of glibc to support NPTL?

On Fri, Sep 01, 2006 at 08:37:40AM +0800, Zhu Tom-a2289c wrote:
> Hi,
>  
> When testing NPTL on armlinux, (2.6.9+glibc 2.3.3)
> we found that in glibc, there is no clone.S under
> Nptl/sysdeps/unix/sysv/linux/arm,
> so, clone.S in sysdeps/unix/sysv/linux/arm/ is used, 
>  
> But the RESET_PID is not defined, so, clone() will not invalidate the
> pid in tls, which is 
> copied from parent.
>  
> Is this a bug?

You should ask whoever produced your backport of NPTL to that old glibc
version; FSF glibc did not support it until 2.4.

--

-- 
Daniel Jacobowitz
CodeSourcery

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 0 comments |
headers
kabbin | 1 Sep 2006 05:07
Picon
Favicon

eth0: Invalid ethernet MAC address

Dear All
  I faild to mount NFS, but TFTPis fine in U-boot.

//**************************************************************
eth0: SMC91C94 (rev 9) at 0xc4858300 IRQ 188 [nowait]
eth0: Invalid ethernet MAC address.  Please set using ifconfig
//**************************************************************

I have read the FAQ, so maybe there are some problems on linux kernel side.
http://www.denx.de/wiki/view/DULG/EthernetDoesNotWorkInLinux
As Wolfgang Denk said,this problem has been discussed before.
http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-July/023579.html
But I still have no idea how to deal with it. 
ps. I counld not find the answer in the mail list, after all, it is 2004!
Could someone help me please?

Following is partially my boot log:

//**********************************************************************************
Linux version 2.6.10-omap (yanming <at> localhost.localdomain) (gcc version 3.4.1) #2
2 Fri Sep 1 10:25:20 CST 2006
CPU: TI OMAP2420 (ARM1136JF-S) [4107b362] revision 2 (ARMv6TEJ)
CPU: D VIPT write-back cache
CPU: I cache: 32768 bytes, associativity 4, 32 byte lines, 256 sets
CPU: D cache: 32768 bytes, associativity 4, 32 byte lines, 256 sets
Machine: OMAP24xx H4
Memory policy: ECC disabled, Data cache writeback
Power Management for TI OMAP.
Clocking rate (xtal/DPLL/MPU): 12.0/532/266 MHz
Kernel command line: mem=64M console=ttyS0,115200n8 noinitrd root=/dev/nfs ip=14
(Continue reading)

Permalink | Reply | 5 comments |
headers
James Steward | 1 Sep 2006 06:00
Picon

Re: eth0: Invalid ethernet MAC address

On Fri, 2006-09-01 at 11:07 +0800, kabbin wrote:
> Dear All
>   I faild to mount NFS, but TFTPis fine in U-boot.
> 
> //**************************************************************
> eth0: SMC91C94 (rev 9) at 0xc4858300 IRQ 188 [nowait]
> eth0: Invalid ethernet MAC address.  Please set using ifconfig
> //**************************************************************
> 
> I have read the FAQ, so maybe there are some problems on linux kernel side.
> http://www.denx.de/wiki/view/DULG/EthernetDoesNotWorkInLinux
> As Wolfgang Denk said,this problem has been discussed before.
> http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-July/023579.html
> But I still have no idea how to deal with it. 
> ps. I counld not find the answer in the mail list, after all, it is 2004!
> Could someone help me please?

The mailing list of 2004 has the answers you're looking for.  Did you
read this part?
http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2004-July/023261.html

I think what you might need is an initramfs that sets the MAC addr using
ifconfig.

Read about early user space in the kernel sources directory;
Documentation/early-userspace/README

JS.

-------------------------------------------------------------------
(Continue reading)

Permalink | Reply | 1 comment |
headers
Anthony Sharobaiko | 1 Sep 2006 08:46
Favicon

Re: PXA255: Using STD_UART for initial debug output

Ole Reinhardt wrote:
> I would like to port Linux to a PXA255 based board where I don't have
> access to the FFUART. The only UART I can directly access is the
> STD_UART. Can you please give me a hint how I have to modify
> debug-macro.S to support this uart for the very early debugging?
> 
> I would just be very happy if I could see the decompression output :-)

Decompression proc uses putstr() from include/asm-arm/arch-pxa/uncompress.h

UART address is #defined there.

--

-- 
Anthony Sharobaiko,
Engineer Programmer

FIORD Co., Ltd
Russia, 199034, St.Petersburg, V.O. 17-Liniya, 4-6
tel/fax: +7 (812) 323 6212, 323 6253

E-mail: anthony <at> fiord.com
http://www.fiord.com
http://www.isagraf.ru

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
(Continue reading)

Permalink | Reply | 0 comments |
headers
Peter Menzebach | 1 Sep 2006 09:30
Picon

Re: AW: BDI2000 and kernel 2.6

Erdmann, Thomas wrote:
>...
> And there i am looking for an easyer and more reliable way to do a Remote Debugging than with gdbserver/gdb/ddd.

Hi Thomas,
I am working here with threaded applications on an armv4 target with
gdbserver/gdb/ddd. It works reliably, but I have to care about the
following in ddd:

-> set "Prefix for loading for absolute shared library symbol files" to
a non existing directory
-> set "Search path for non absolute shared library symbol files" to a
directory containing the libs
-> work carefully with "set scheduler-locking on/off"
-> Sometimes I cannot step over a library function. Then I have to use
"stepi" 3 times, and can then leave the function with "finish". I think,
even this problem can be solved somehow, but I don not know the solution..

I am using ddd 3.3.10, gdb 6.4, gcc 3.4.x, glibc 2.3.x

Best regards
Peter

--

-- 
Peter Menzebach
Menzebach und Wolff IT-Consulting GbR
Phone +49 751 355 387 1

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
(Continue reading)

Permalink | Reply | 0 comments |
headers
Erdmann, Thomas | 1 Sep 2006 09:55

AW: BDI2000 and kernel 2.6

> -----Ursprüngliche Nachricht-----
> Von: Peter Menzebach [mailto:pm-armlist <at> mw-itcon.de] 
> Gesendet: Freitag, 1. September 2006 09:31
> An: Erdmann, Thomas
> Cc: linux-arm-kernel <at> lists.arm.linux.org.uk
> Betreff: Re: AW: BDI2000 and kernel 2.6
> 
> Erdmann, Thomas wrote:
> >...
> > And there i am looking for an easyer and more reliable way 
> to do a Remote Debugging than with gdbserver/gdb/ddd.
> 
> Hi Thomas,
> I am working here with threaded applications on an armv4 
> target with gdbserver/gdb/ddd. It works reliably, but I have 
> to care about the following in ddd:
> 
> -> set "Prefix for loading for absolute shared library symbol 
> files" to
> a non existing directory
> -> set "Search path for non absolute shared library symbol files" to a
> directory containing the libs
> -> work carefully with "set scheduler-locking on/off"
> -> Sometimes I cannot step over a library function. Then I have to use
> "stepi" 3 times, and can then leave the function with 
> "finish". I think, even this problem can be solved somehow, 
> but I don not know the solution..
> 
> 
> I am using ddd 3.3.10, gdb 6.4, gcc 3.4.x, glibc 2.3.x
(Continue reading)

Permalink | Reply | 1 comment |
headers
Michael Schnell | 1 Sep 2006 10:27
Picon

Re: AW: BDI2000 and kernel 2.6

Erdmann, Thomas wrote:
> I think i end this Discussion now, it isn't helpful for me.
>   
That's really bad manners. The results of the discussion on what to do 
if GDB fails in some way might be _very_ helpful for others !

-Michael

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm-kernel
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php
Permalink | Reply | 0 comments |
headers
satpal parmar | 1 Sep 2006 11:29
Picon

change i serial interface i linux 2.6.17

Hi every one ;

I am compiling linux kernel for porting on arm 926 board. In my previous
project i have used linux kernel 2.6.11 andour own uart code.Now i want to
use 2.6.17  kernel and driver cove that i have used ealier.

The problem i am facing is that serail port interface in 2/6.27 . Can
anybody face any probem with serail interafce. (tty.h  strucutre tty_struct
is totally replaced by two new structures).CAn any body help be out to
iunderstand the flow of code in serial interfece of linux kernel.hwo to

here is fragment of code :  linux 2.6.17

55 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L55>struct
tty_buffer <http://www.gelato.unsw.edu.au/lxr/ident?a=arm;i=tty_buffer> {
56 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L56>
struct tty_buffer<http://www.gelato.unsw.edu.au/lxr/ident?a=arm;i=tty_buffer>*
next <http://www.gelato.unsw.edu.au/lxr/ident?a=arm;i=next>;
57 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L57>
char *char_buf_ptr;
58 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L58>
unsigned char *flag_buf_ptr;
59 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L59>
int used;
60 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L60>
int size <http://www.gelato.unsw.edu.au/lxr/ident?a=arm;i=size>;
61 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L61>
int active;
62 <http://www.gelato.unsw.edu.au/lxr/source/include/linux/tty.h?a=arm#L62>
int commit <http://www.gelato.unsw.edu.au/lxr/ident?a=arm;i=commit>;
(Continue reading)

Permalink | Reply | 0 comments |
headers
Lothar Wassmann | 1 Sep 2006 11:51
Picon
Favicon

Re: pxa255 lcd underrun due to flash read/write

Hi,

> > PXA255 before trying to implement your design. 640x480x16 is not practical on 
> > this part for the reason you've discovered: there simply isn't enough system 
> > bandwidth to support a display of this size.  
> that's the very first thought i had when i discovered the problem.
> unfortunately i'm having hard time to explain this as the same hardware
> works fine with windows ce 4.2.
> 
Did you check the performance of flash rom accesses under WinCE?

> dd if=/dev/mtdblock2 of=/dev/null
> is how i can easily reproduce the problem. / is mounted on mtdblock2.
> similarly using urandom as source and writing some file also reproduces
> the problem.
> 
Since kernel version 2.6.14 I have also observed LCD disturbance
during flash rom accesses but at the same time the mtd accesses
experienced a performance boost by a factor of 3.

2.6.11:
time dd if=/dev/mtdblock1 of=/dev/null bs=4096 count=1024
1024+0 records in
1024+0 records out

real    0m2.150s
user    0m0.000s
sys     0m0.120s

2.6.17:
(Continue reading)

Permalink | Reply | 0 comments |

Gmane