- kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

The effect of nice value in CFS scheduler
by Anh Le March 21, 2014

March 21, 2014

Hi everyone, I'm studying the CFS scheduler. I'm using ubuntu 13.10 and I assume that it's using CFS in it kernel. I try to see the effect of nice value by running 2 instances of the same program (just a while loop of calculations), putting them on the same CPU and setting different nice values to them (one is 19 and one is -20). I include in the program some printf to see how much it has executed. According to my understanding, the process with the small nice value should get more processor time than the one with the big value. However they seem to be on par with each other (they execute approximately the same number of loops in the same duration), so I'm really confused. I'd appreciate any help.

2 1

how to get PFN of page cache?
by pravin jadav March 21, 2014

March 21, 2014

i want to develop linux kernel module that will get PFN of pagecache pages. please help me how to develop this kind of kernel module. -- pravin jadav M.Tech (IT pursuing )

5 9

Adding custom system calls
by arun kumar March 20, 2014

March 20, 2014

How do i add a Custom system call to my Linux Kernel. I want to know what files to modify and where to make the entries. Arch : x86 64bit Kernel: 3.12.14

3 2

Fwd: Adding custom system calls
by arun kumar March 20, 2014

March 20, 2014

yeah it does seems like a homework question.. I remember doing it on KernelVersion: 2.6 by modifying some files including unistd.h writing a wrapper for it and writing the code for it. This included some tweaking inside the /arch subtree of the kernel. my intentions are entirely academic. Currently working on kernelVer: 3.12.14 I am somewhat struggling to comprehend the changes (steep learning curve), if you could guide me on what files to look for, i am sure i can find my way from there. and Thanks Greg. --big fan :) On Thu, Mar 20, 2014 at 11:11 PM, Greg KH <greg(a)kroah.com> wrote: > On Thu, Mar 20, 2014 at 10:55:42PM +0530, arun kumar wrote: > > How do i add a Custom system call to my Linux Kernel. > > I want to know what files to modify and where to make the entries. > > That really sounds like a homework question... > > Why do you want to do this, and what have you tried to do that has > failed? >

2 1

is there any faster way to download using git?
by Rahul Garg March 20, 2014

March 20, 2014

Hi All, I am trying to download following git branch git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git But it is downloading very slow with speed of 10-40 KB/s. So I was wondering is there any faster way to download this branch. Regards Rahul

3 2

Unable to find /sbin/hotplug in my running kernel
by Rahul Garg March 20, 2014

March 20, 2014

Hi I am unable to find /sbin/hotplug in my running kernel. I checked my config file, CONFIG_HOTPLUG_CPU=y is enabled. Also one more doubt which I have is, with out this file how hotplug event is working. My running kernel version is 3.14.0-rc6. Any leads will be appreciated. Regards Rahul

3 2

RE: Please help.. trying to use USB keyboard to open a shell on LCD display
by Chan Kim March 20, 2014

March 20, 2014

Hi, lovely linux people, I just found that the printk called from insde the kernel and the printf from the busybox is ultimately using the same routine. (When we use printf from busybox, it eventually calls the system call sys_write which eventually calls some time of bit blitting functions. The printk in the kernel also uses the same bit blitting functions. bit blitting functions writes pixel data line by line to draw the characters on the LCD) For example, to hide the cursor, which happens both in printk and printf cases when a new data is written, in printk case : the hid_cursor(vc) is called from inside vt_console_print function. in printf case : the hid_cursor(vc) is called from inside do_con_write function. * vc is of type struct vc_data *. Since the hid_cursor function is the same, it follows the same path after that. To name the main functions in this case .. From the hid_cursor, vc->vc_sw->con_cursor(vc, CM_ERASE) is called which is fbcon_cursor function. Then fbcon_cursor calls ops->cursor(vc, info, mode, y, get_color(vc, info, c, 1), get_color(vc, info, c, 0)) which is bit_cursor. Then bit_cursor calls soft_cursor(info, &cursor). Then soft_cursor calls info->fbops->fb_imageblit(info, image) which is cfb_bitblit function. Then cfb_bitblit calls fast_imageblit(image, p, dst1, fgcolor, bgcolor). Then finally fast_imageblit repeatedly calls FB_WRITEL((end_mask & eorx)^bgx, dst++) which is actually writing the pixcel data (combined pixels in 32 bits) to the framebuffer memory. My problem is : the same function path works for printk (texts are written on the LCD ok), but the same function path doesn't work for the printf case. It just stops at the first pixel data write (FB_WRITE) in printf case. I guess the ack is not coming from that memory. I've printed the address of the frame buffer and it looked not much different (0x69xxxxxx). I'll check if the address range is really in the frame buffer range allocated before (during the framebuffer probe initialization). I'll also look into the MMU page table setup.. But I thought there maybe someone who has had the same experience like the one I have now. Has anybody have similar situation before? (printk prints ok but printf doesn't print to the LCD using frame buffer) Any hint would be appreciated. Thank you! Chan ________________________________ From : "Chan Kim" <ckim(a)etri.re.kr> Sent : 2014-03-17 09:14:08 ( +09:00 ) To : kernelnewbies(a)kernelnewbies.org <kernelnewbies(a)kernelnewbies.org> Cc : Subject : RE: Please help.. trying to use USB keyboard to open a shell on LCD display There is no reponse yet to this question.. any help/hint would be appreciated. I found that I may have to use getty in the /etc/inittab file to get tty for a vc. For busybox, the example in init/init.c suggests entries like tty4::respawn:/sbin/getty 38400 tty4 I earlier found the keyboard seems to be connected to tty1 and the printk output is switched to tty1 towards the end during the initialization. So I tried in the inittab tty1::respawn:/sbin/getty 38400 tty1 and 1::respawn:/sbin/getty 38400 tty1 But didn't work. I know how to start a shell on tty1 (using inittab). But How do I map tty1 to the vc and to the LCD using boot argument or busybox setting? This seems to be the key question. Thansk, Chan ________________________________ From : "Kim Chan" <ckim(a)etri.re.kr> Sent : 2014-03-12 15:31:17 ( +09:00 ) To : kernelnewbies(a)kernelnewbies.org <kernelnewbies(a)kernelnewbies.org> Cc : Subject : Please help.. trying to use USB keyboard to open a shell on LCD display Hello, folks, It is a comfort to me that I have people to ask questions here and I hope I could be answering to some hard quetions someday. I've beeing trying to attach a USB keyboard to our embedded system to open a shell on the LCD display using framebuffer console. The USB host controller is from the opencore(USB1.1) and it works for USB memory. Some of the boot arguments are (I've tried some variants) "console=tty1 " "fbcon=map:0000,font:SUN8x16,vc:1-3 " "root=/dev/ram " The problem is, after the final sys_execve() which executes /sbin/init of the ramdisk, I cannot see any printf results frome the busybox init_main function not to mention the shell. So I cannot use shell commands like getty. Using uart as the console I can see printf results in the busybox init_main routine. (I use "console=ttyS,mmio,0xff003000,115200n8 " in that case) Using some prints on separate uart print routine, I can see from the kbd_event function that the key value is going to current vc which is using tty1. The key value is then handed to put_queue function where the character is somehow copied to tty buffer. Flush_to_ldisc is scheduled for it and it will probably get passed on to the virtual console. By the way, I guess the keyboard data is delivered to current fg_console (vc 0 which is using tty1). Added to this, there is a notifier_call_chain and the notifier block is registered by vcs_poll_data_get function which I don't know weather is necessary or not. So one of my questions is if vcs is necessary for my case. (I guess not. There is no other notifier blocks registered) I gathered that without inittab (I don't use inittab which should be ok for most cases) the assumed default inittab tty2::askfirst:-/bin/sh tty3::askfirst:-/bin/sh tty4::askfirst:-/bin/sh is being applied by the busybox so there maybe 3 shells on tty2,tty3, and tty4 running somewhere. But even if there are shells running I don't know how to switch to those ttys. I tried using Ctrl-Alt F2 (or F3, F4) using the USB keyboard to switch to tty2,3,4 to see if any shell comes up on the LCD but couldn't see it. I tried using inittab with no avail before. And I don't know how the printf in busybox is connected to the kernel. I guess busybox talks with stdio(stdout,stdin..) and the file is connected to the kernel anyhow. I've digged up how vc driver is connected to the console switch functions and to the bit blitting functions and how the printk is handled in the kernel. But these console, tty, framebuffer console.. things are very complex to me. and I hope somebody could point me to what should be checked. Any hint will be deeply appreciated. Best Regards, Chan _______________________________________________ Kernelnewbies mailing list Kernelnewbies(a)kernelnewbies.org<mailto:Kernelnewbies@kernelnewbies.org> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

1 0

correct response to -ERESTARTSYS from wait_event_interruptible_timeout()
by Jeff Haran March 20, 2014

March 20, 2014

Hi, I am struggling a bit to understand how a driver ioctl() handler should respond when its call to wait_event_interruptible_timeout() returns -ERESTARTSYS. My understanding is that -ERESTARTSYS means the process caught a signal during the wait. Comments in include/linux/errno.h indicate that -ERESTARTSYS should never be returned to user space. Seems that when this happens, -EINTR should be returned instead. However, man 7 signal indicates that if a signal handler is registered using SA_RESTART, the system call should be restarted. I've tried to follow the ioctl() system call logic from SYSCALL_DEFINE3(ioctl, ...) in fs/ioctl.c down to the driver ioctl() functions and I haven't seen any code that appears to implement this restarting behavior. I suppose this could be something that glibc implements. So my question is "Restarted by whom?" To be more specific, is it up to the device driver ioctl handler to implement this restarting behavior by somehow inspecting the flags of the signal handlers registered with the process and call into wait_event_interrupible_timeout() again when SA_RESTART is set? Or is the restarting behavior implemented farther up the call tree from the driver ioctl handler and if so, what should the ioctl handler return to ensure that the system call returns to user space with EINTR or gets restarted according to SA_RESTART? Thanks, Jeff Haran

1 0

thread context switching
by Nada Saif March 19, 2014

March 19, 2014

Hi, I want to measure thread context switching in c, How I can do that? Thanks, N.A.S

7 7

Is demand paging enabled in linux kernels of Android?
by manty kuma March 19, 2014

March 19, 2014

Hi, Suppose I have an executable of size 60MB and I want to execute it on android(ARM), will 60MB be allocated to load the code onto the RAM? I remember studying about demand paging in Linux kernel. Is it relevant here? If yes, is it implemented in Linux kernels used in Android? Also, is there a config with which i can enable or disable the configs. I posted the same on SO. Answer it either place. Thank you. http://stackoverflow.com/questions/22499031/is-demand-paging-implemented-in… BR, Sandeep Kumar

3 5