gmane.linux.kernel.kernelnewbies

RE: Kernel 64bit

Giridhara RP (grp — 2013-05-23T10:08:50

I did not see any file (vmlinuz-<version>.x86_64) under /boot. I was only looking at bzImage found under <arch>/boot/ . Am I missing something?


[root< at >grpRHEL2 linux-2.6.32.60]# pwd
/root/lin_2/linux-2.6.32.60
[root< at >grpRHEL2 linux-2.6.32.60]# find . | grep -i vmlinuz
[root< at >grpRHEL2 linux-2.6.32.60]# 
[root< at >grpRHEL2 linux-2.6.32.60]# 
[root< at >grpRHEL2 linux-2.6.32.60]# file vmlinux
vmlinux: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped

[root< at >grpRHEL2 linux-2.6.32.60]# file vmlinux.o
vmlinux.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped

[root< at >grpRHEL2 linux-2.6.32.60]# file arch/x86/boot/bzImage 
arch/x86/boot/bzImage: Linux kernel x86 boot executable bzImage, version 2.6.32.60 (root< at >grpRHEL2) #1 SM, RO-rootFS, swap_dev 0x3, Normal VGA
[root< at >grpRHEL2 linux-2.6.32.60]# file arch/x86_64/boot/bzImage 
arch/x86_64/boot/bzImage: symbolic link to `../../x86/boot/bzImage'

-----Original Message-----
From: amit mehta [mailto:gmate.amit< at >gmail.com] 
Sent: T

Re: Kernel 64bit

amit mehta — 2013-05-23T10:01:05

Generally, After building a new kernel, I issue 'update-grub' command
on my ubuntu box to add proper entries for the newly built kernel in
the grub configuration file, but I think 'make install' will take care of
this as well. However, as you've confirmed that you've successfully
booted from the newly built kernel, which is still being reported as
a 32 bit kernel, therefore we can safely reject the case of incorrect
and or missing grub entry.

Do you see compressed x86_64 bit kernel image(vmlinuz-<version>.x86_64)
under /boot ?

RE: Kernel 64bit

Giridhara RP (grp — 2013-05-23T09:15:15

Hi Amit

 Looks like I copied wrong output in my previous email. Here is the output. Also, what should I do with 'grub'? 

 5. After boot 'uname -a' , shows this.
 Linux 2.6.32-60 #1 SMP Mon Mar 25 16:20:30 IST 2013 i686 GNU/Linux

-----Original Message-----
From: amit mehta [mailto:gmate.amit< at >gmail.com] 
Sent: Thursday, May 23, 2013 2:40 PM
To: Giridhara RP (grp)
Cc: kernelnewbies< at >kernelnewbies.org
Subject: Re: Kernel 64bit

On Thu, May 23, 2013 at 10:16 AM, Giridhara RP (grp) <grp< at >cisco.com> wrote:

Are you sure, you are booting from the newly built 64 bit kernel, because 'uname' is still reporting the older kernel (2.6.29.6 instead of 2.6.32.60). You might have missed updating grub after building the new kernel (version: 2.6.32.60)

-Amit

Re: Kernel 64bit

amit mehta — 2013-05-23T09:10:00

Are you sure, you are booting from the newly built 64 bit kernel,
because 'uname' is still reporting the older kernel (2.6.29.6 instead of
2.6.32.60). You might have missed updating grub after building the
new kernel (version: 2.6.32.60)

-Amit

Kernel 64bit

Giridhara RP (grp — 2013-05-23T04:46:47

Hi,

I am trying to build 64bit kernel (linux kernel 2.6.32-60). But, once I boot the image (ISO), uname -a still shows as i686. It does not show as 'X86_64'.  My vmlinuz and vmlinuz.o shows 64bit. Any pointers much appreciated. I have followed below steps. 

1. make X86_64_defconfig
2. make V=1 (comiler shows that -m64 is getting used).
3. [root< at >grpRHEL2 linux-2.6.32.60]# file vmlinux vmlinux.o 
vmlinux:   ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped
vmlinux.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped
4.  bzImage does not say x86_64 boot.
 root< at >grpRHEL2 boot]# pwd
/root/lin_2/linux-2.6.32.60/arch/x86_64/boot
[root< at >grpRHEL2 boot]# file bzImage 
bzImage: symbolic link to `../../x86/boot/bzImage'
[root< at >grpRHEL2 boot]# file ../../x86/boot/bzImage 
../../x86/boot/bzImage: Linux kernel x86 boot executable bzImage, version 2.6.32.60 (root< at >grpRHEL2) #1 SM, RO-rootFS, swap_dev 0x3, Normal VGA

5. After boot 'uname -a' , shows this.
Linux 2.6.29.6 #1 S

android older sdk version

Ben Wu — 2013-05-23T02:11:22

Dear All:
  I now  new to android development, and found andorid4.2 
source code is too large, I want study older verison SDK about android 
such as 1.5 or 1.6. where can i download it ? many thanks_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

android old source code

Ben Wu — 2013-05-23T02:09:27

Dear All:
  I now  new to android development, and found andorid4.2 source code is too large, I want study older verison SDK about android such as 1.5 or 1.6. where can i download it ? many thanks_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Reference material for x86-64 assembly on Linux

Kumar amit mehta — 2013-05-22T05:30:42

Thank you for suggesting me this book, I'll have a look, though this book
seem to be covering the IA32 platform only.

Soft CPU Affinity

Arlie Stephens — 2013-05-22T02:54:00

Hi Folks,

I'm totally failing to come up with the right terms to google, so I'm
hoping the collective wisdom of this list can help me once again....

What I'm trying to understand is how the linux kernel decides where to
run things, in the absence of calls to explicitly set cpu affinity. 

Here's the scenario - 
- 16 core system (not a huge number; not small either)
- hard irq that's been assigned to a particular cpu by
  irqbalance. Let's say it's running on cpu 8.
- the hard irq triggers (probably) some kind of softirq activity,
  which runs ... where? ...  [by the looks of it preferentially on cpus
  8 and 9]
- this results in some task(s) becoming ready to run, which run
  ... where?... [by the looks of it, preferentially on cpus 8-11]

There are NUMA considerations - cpus 0-7 are on one socket, with cpus
8-15 on another. AFAIK, 8 and 9 or 8 and 10 are no "closer" together
than 8 and 15. Due to BIOS issues, the kernel does not have a fully
accurate picture of the NUMA layout, but it appears to know wher

Re: Bootloader sequence with time prefix

2013-05-20T14:40:21

On Mon, 20 May 2013 15:07:33 +0530, Dhyan said:


The bootloader is probably not where you're going to find places to optimize.
It's got two jobs to do - load the kernel and load the initrd. And it's usually
pretty optimized to do that. About all you can do there is minimize the size of
the kernel and initrd so there's less to load (and move to SSD if you're really
hardcore).

The only other place to really win in the boot loader is to set the "hit any
key to go to the boot menu" timeout to 0 or 1 seconds.  Of course, this means
you have little (or zero) chance to enter an alternate kernel or commandline.

The *big* places for improving boot time:

1) Turn off services you don't need.  If you don't need NFS, don't waste
the 0.75 seconds or whatever it takes to load.

2) Use an 'init' process that can multithread better than the old sysvinit
system that runs one startup script at a time.  systemd (for all its other
faults, including a huge learning curve) does this, as does upstart (I'm told).
I've even seen

CPU usage of particular process

Rahul Bedarkar — 2013-05-20T10:05:28

Hi,

I am trying to write small program to get CPU usage for a particular
process. My logic is to read user, nice, system, idle values from
/proc/stat. Then read user and system time from /proc/<pid>/stat. Then wait
for some time and again read same values. After that I am doing following
calculations
CPU% = (no_of_processor * (process_cpu_usage2-process_cpu_usage1) * 100 ) /
(total_cpu_usage2-total_cpu_usage1)

When I compare my output with top, it looks like correct, but sometimes it
is more than 100. Is my logic correct here ? Please correct me if I am
worng.

Regards,
Rahul
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Reference material for x86-64 assembly on Linux

wannabehacker wb — 2013-05-20T09:13:26

"The art of assembly language" by randall hyde is an excellent reference.
 Intel website has x86 reference manual which is also a very good in-depth
reference

http://www.intel.com/content/www/us/en/processors/architectures-software-developer-manuals.html

Cheers,
wbhack3r


On Mon, May 20, 2013 at 11:43 AM, amit mehta <gmate.amit< at >gmail.com> wrote:

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Bootloader sequence with time prefix

Dhyan — 2013-05-20T09:37:33

Thank you Valdis !!!

I was working on to measure boottime for  bootime optimization.




On Mon, May 20, 2013 at 2:40 PM, <Valdis.Kletnieks< at >vt.edu> wrote:

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Reference material for x86-64 assembly on Linux

Rajat Sharma — 2013-05-20T09:26:26

Following book describes GNU assembler in a very simple manner:
http://www.wrox.com/WileyCDA/WroxTitle/productCd-0764579010.html


On Mon, May 20, 2013 at 2:43 PM, wannabehacker wb <wbhack3r< at >gmail.com>wrote:

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Bootloader sequence with time prefix

2013-05-20T09:10:51

On Mon, 20 May 2013 13:55:25 +0530, Dhyan said:


That will require fixing your bootloader, which may be more than slightly
interesting, as there's usually no timer services initialized by the
boot loader (so all you'll get is a bunch of [0.0000] headers).

What problem are you trying to solve by adding those timestamps? Maybe
there's a different way to solve your issue.
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Bootloader sequence with time prefix

Dhyan — 2013-05-20T08:25:25

Dear* *Valdis,

Thank you for your quick reply, but what i need is "*time in bootloader
print output*". Kernel i already enabled and i am getting the time,

So i need time information like
*
*
[  Time XXXX  ] *U-Boot 1.1.4-gedeced79 (Feb  6 2012 - 09:27:11)*
*
*
[  Time XXX2  ] *Load address: 0x80e80000*
[  Time XXX3  ] *DRAM:  1024 MB*

*Hit any key to stop autoboot:  0 *
*kernel   < at > 80008000 (3728760)*
*ramdisk  < at > 81000000 (295902)*
*I2C read: I/O error*
*
*
*Starting kernel ...*
*
*
*Uncompressing Linux... done, booting the kernel.*

[    0.000000] Initializing cgroup subsys cpu   ===> Kernel I am already
getting timing information with CONFIG_PRINTK_TIME=y

Thanks & Regards
Dhyan



On Mon, May 20, 2013 at 1:19 PM, <Valdis.Kletnieks< at >vt.edu> wrote:

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Re: Bootloader sequence with time prefix

2013-05-20T07:49:12

On Mon, 20 May 2013 12:37:25 +0530, Dhyan said:


I think you meant CONFIG_PRINTK_TIME there - and if that kernel option is
set, all your printks will show up with that, even stuff coming out of
early_printk().

Getting userspace to do it as well is a userspace problem, not a kernel
problem.  You'll have to fix multiple pieces - everything from your
syslog daemon (whichever one you use) to sysvinit/systemd/upstart or
whatever you boot with to output the right stuff on the console.  There's
almost certainly other stuff lurking as well, but your syslog and init
packages are going to be about 90% of it.
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Bootloader sequence with time prefix

Dhyan — 2013-05-20T07:07:25

Dear all,

Is there any uboot config option in to print boot sequence with time
prefixes similar to kernel option CONFIG_PRINTK ?

Thanks & Regards
Dhyan
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Reference material for x86-64 assembly on Linux

amit mehta — 2013-05-20T06:13:28

Folks, Can you please suggest me some nice books or tutorials
that concentrates more on the x86_64 assembly programming
using GNU assembler. Unfortunately, I haven't done any assembly
programming for the last seven years, but my current job requires
me to analyse kernel crashes and a lot of them (probably due to
widespread use of x86-64 architecture) originate on x86-64 machines
and quite often disassembling is the last resort to inspect the function
parameters, stack frames etc.

There seem to tons of books, tutorials, assemblers available over the
internet, but  I'm looking for something that can give me jumpstart on
x86_64 assembly, specially in Linux environment.

Recently, While browsing, I've found these two:
1: x86_64 ABI (System V Application Binary Interface,
AMD64 Architecture Processor Supplement) -
2: x86-64 Machine-Level Programming - Randal and David

Appreciate a lot, If you can recommend me your favourite text book
on x86-64 assembly or any such reference material.

Thanks,
-Amit

Re: Analyzing Kernel call traces.

Peter Teoh — 2013-05-18T04:24:38


sorry , u mean "backtrace" call trace?   or kernel oops?

http://www.linuxforu.com/2011/01/understanding-a-kernel-oops/

and here is another trace:

http://elinux.org/Kernel_Function_Trace

which depends on the instrumentation method:

http://elinux.org/images/6/68/Kfiboot-9.lst

http://elinux.org/Kernel_Instrumentation

http://elinux.org/Instrumentation_API

many of these traces, simply depends on the concept of call frames, or a
range of memory  addresses allocated on the stack used by the functions.

above page also mentioned the use of gcc -pg, and not mentioned are other
features of gcc (man gcc):

           -finstrument-functions
           -finstrument-functions-exclude-function-list=sym,sym,...
           -finstrument-functions-exclude-file-list=file,file,...

Beware though, sometimes compilation will explicitly remove the use of
frame pointer:

-fomit-frame-pointer

without the "ebp" and "esp" to demarcate the start and end of a frame,
there is no way to know the beginning and end of a call frame

Re: select call takes more time than the given timeout

Victor Buciuc — 2013-05-17T12:49:38


As the manual page for select specifies (see below) you should expect some
delays due to kernel scheduling. If you need more precision you might wanna
try something else.

The *timeout* argument specifies the minimum interval that *select*()
should block waiting for a file descriptor to become ready. (This interval
will be rounded up to the system clock granularity, and kernel scheduling
delays mean that the blocking interval may overrun by a small amount.) If
both fields of the *timeval* structure are zero, then *select*() returns
immediately. (This is useful for polling.) If *timeout* is NULL (no
timeout), *select*() can block indefinitely.

Regards,
Victor.
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies< at >kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies