October 2012 - kernelnewbies@kernelnewbies.org

how many "contexts" are there?
by Robert P. J. Day Oct. 8, 2012

Oct. 8, 2012

the standard explanation of context related to linux is that there are three "contexts" one can be in at any time: * user context * kernel, process context * kernel, interrupt context but that's clearly(?) an incomplete (or not refined enough) list, since it doesn't include kernel threads, and a quick google showed this: http://stackoverflow.com/questions/9389688/in-what-context-kernel-thread-ru… so is there a more refined or up-to-date list of contexts which explains them fairly well, including the subtle distinctions? thanks. rday -- ======================================================================== Robert P. J. Day Ottawa, Ontario, CANADA http://crashcourse.ca Twitter: http://twitter.com/rpjday LinkedIn: http://ca.linkedin.com/in/rpjday ========================================================================

2 1

enabling config_wireless_ext
by captain Oct. 7, 2012

Oct. 7, 2012

Sorry Confused. I'm building from the latest 3.6 kernel from git in order to be able to submit patches. Well if I get that far. I'm running the built on my netbook but I'm having problems building my wifi drivers against the 3.6.0 kernel. Basically I've got an error: error: 'struct net_device' has no member named 'wireless_handlers' Which is fair enough if you look in the kernel source include/linux/netdevice.h there's a "'#ifdef CONFIG_WIRELESS_EXT" So this is easy all I have to do is enable that switch and rebuild the kernel. You might not believe it but that's were I come unstuck. I did a search in "make menuconfig" for that switch and it does provide some info but normally when I do a search the help tells me where to find it. (Which is a God send.) So searching for it is finding it but not giving me suitable help on where it is in the menu structure. I then decided that I'd search for it in the .config file. Can't find it in there either. This is a mailing list for newbies and I know this is a stupid question but can anybody advise me on where that switch is in make menuconfig? I did find this page out there but that's pointing at Kconfig in source code not where it is in the menus: http://cateee.net/lkddb/web-lkddb/WIRELESS_EXT.html

2 2

Question on initrd and initramfs
by Ezequiel Garcia Oct. 7, 2012

Oct. 7, 2012

Hi, This is *not* another question about initrd vs initramfs (I already read Documentation/...). That said, I'm doing some experiments with: * qemu * a very small kernel * busybox static-linked rootfs I can create an initramfs using this: $ cd initramfs $ find . | cpio -H newc -o > ../initramfs.cpio Then I use start qemu like this: $ qemu -kernel bzImage -initrd initramfs.cpio This works. Now to my first question: The kernel is compiled WITHOUT any fs driver... So how the f@#ck did he mounted initramfs? How are the files formatted? And second question: I want to repeat the above procedure but using initrd instead of initramfs, is this still supported? I'd like to use romfs or squashfs... anyone knows how to create this initrd and how to give it to qemulated kernel? This seems trivial, but I spent all morning and had no luck... :-( FWIW: My current kernel size (compressed) is around 500k! This is the smaller kernel I managed to boot. It has nothing (not even /proc) so it's pretty useless but at least it can run a shell :-) The question I'm trying to answer is: how little can it be, while still remaining useful? Thanks! Ezequiel.

4 12

scsi adapter module over block device - need help
by Dmitry Filippov Oct. 6, 2012

Oct. 6, 2012

Hi all, i'm just learning kernel and not experienced in writing code for it. I'm going to write kernel module in studying purposes and i hope someone from you could help me with it and in best case guide me through overall process. I want to write module that implements SCSI-disk logic, keeping its data on block device. For example partition of existing disk. So i want to pass the name of block device as parameter to my module while loading to kernel. So it will be some kind of virtual(pseudo) scsi device that appears in /dev as scsi disk and in fact some block device as backend. I want it to be as simple as possible... keeping just essential parts... and it is not intended to be some kind of production driver... i want it to be just as simple as sbull from LDD))) As from my point of view there are 2 major parts, with which i'm completely not confident and doesn't know from where to start... 1. scsi subsystem... so how to write my module to interact with scsi subsystem in this case and to not reinvent the wheel. using scsi kernel API as much as possible. i have found interesting thing in kernel source scsi_debug module... which represent scsi disk and have ram as backend (so some kind of sbull but with scsi). However i got overwhelmed with its code. It is very difficult for me to sort out with it by my own... maybe someone can help? 2. module code interaction with existing block device... that is i completely don't understand how to achieve. ok.. i have device name in kernel space... whats then? how i can access this device? so how i can write some adapter for existing device and transit some writes/reads and other fops from my device to existing... I saw same questions on forums... and the only answer is to use sys_write... or vfs_write... to existing block device node... if it is most simple/correct way? i hope someone could help me sort out in those things eventually. Really need your help! thanks in advance, -Dmitry Filippov

2 1

Re: Gpio interrupt shall pass the signal to application
by jeshwanth Kumar N K Oct. 6, 2012

Oct. 6, 2012

Actually what I am doing is, I have 3 inputs to beaglebone from BLDC motor hall sensor. If there is any change in any of 3 inputs, the kernel module has to send signal to my beaglebone. From the interrupt my application reads the input from sysfs and pass the next sequence to pwm. PWM I am using userspace only. So is my design right for handling fast input change, I mean 50 micro seconds ? Any suggestion ? I am new to device driver concept, please suggest. Thanks :) Sent from my HTC ----- Reply message ----- From: "Jonathan Neuschäfer" <j.neuschaefer(a)gmx.net> Date: Sat, Oct 6, 2012 10:29 pm Subject: Gpio interrupt shall pass the signal to application To: "jeshwanth Kumar N K" <jeshkumar555(a)gmail.com> Cc: "kernelnewbies" <kernelnewbies(a)kernelnewbies.org> On Sat, Oct 06, 2012 at 01:50:55PM +0530, jeshwanth Kumar N K wrote: > Hello All, > > I am writing a kernel module that shall pass the signal to application running in user space. So which signal i should use? And how to achieve this? Thank you. What kind of information need to be transmitted from your device to the application? Jonathan Neuschäfer

1 0

Gpio interrupt shall pass the signal to application
by jeshwanth Kumar N K Oct. 6, 2012

Oct. 6, 2012

Hello All, I am writing a kernel module that shall pass the signal to application running in user space. So which signal i should use? And how to achieve this? Thank you. Sent from my HTC

4 3

Re: Gpio interrupt shall pass the signal to application
by jeshwanth Kumar N K Oct. 6, 2012

Oct. 6, 2012

Hi Fabio, Thanks for the reply. Ya I got your point, and I did the same thing by exporting the sysfs gpio and poll for value attribute It works fine but the thing is time. Actually my interrupts may be every 50 micro seconds. So I thought of dealing in kernel mode itself. What's your suggestion ? Sent from my HTC ----- Reply message ----- From: "Fabio Pozzi" <pozzi.fabio(a)gmail.com> Date: Sat, Oct 6, 2012 2:29 pm Subject: Gpio interrupt shall pass the signal to application To: "jeshwanth Kumar N K" <jeshkumar555(a)gmail.com> Cc: "kernelnewbies" <kernelnewbies(a)kernelnewbies.org> I think you could use sysfs_notify to send the notification and use select() or poll() to wait for changes in the sysfs file in the interrupt handler routine. You can find an example of this in the gpio-fan driver. Hope it helps -- Saluti, Fabio Pozzi

2 1

how to process data recvd from a device that wont give interrupts
by devendra.aaru Oct. 5, 2012

Oct. 5, 2012

Hi all, I found that you can use a kernel timer and poll for the hardware interrupt registers whether the interrupt flag is set or not, but this will take atleast some good amount of CPU. are there any ways other than using the kernel timer? thanks,

3 5

IPSec/IPv6 test gets unexpected ENOBUFS on connect on linux 3.2
by Bhushan Kumar Oct. 5, 2012

Oct. 5, 2012

HI, I am getting the below mentioned error for Linux kernel 3.2.18.i am looking for patch for the same. If anyone having suggestion\patch is much appreciated . And I will be thankful for ASAP response. " TCP/IPv6 test doing repeated heavy sending and reconnecting, connect()s start to fail with ENOBUFS on the Linux 3.2.18. " Thanks and Regards, Bhushan Information transmitted by this e-mail is proprietary to MphasiS, its associated companies and/ or its customers and is intended for use only by the individual or entity to which it is addressed, and may contain information that is privileged, confidential or exempt from disclosure under applicable law. If you are not the intended recipient or it appears that this mail has been forwarded to you without proper authority, you are notified that any use or dissemination of this information in any manner is strictly prohibited. In such cases, please notify us immediately at mailmaster(a)mphasis.com and delete this mail from your records.

2 1

How does I/O ports work in linux kernel
by Vineel Reddy Oct. 5, 2012

Oct. 5, 2012

Hi All, I have read about I/O ports recently from many recommended books for linux. Each of the book effectively taught me only two things 1. Reserving the I/O ports using request_region(...) function 2. Then access the I/O ports using inb, outb functions Apart from repeated reading and googling I still have the following queries Question 1: The one thing that the books haven't sufficiently answered is 'how does linux know to which device it should write or read from when we use inb and outb functions'? Say for example we are dealing with parallel port for instance. How does linux know that the parallel port is registered at 0x378 ? Question 2: Who registers the parallel port at 0x378 ? I have read somewhere that all the mapped ports can be found in /proc/ioports. Who generates this map. BIOS? or the Kernel? Question 3: How is */proc/ioports* different from */proc/iomem*(System Memory Map). Is there any overlap between them(because 0x378 is also an address, so it must be some where in the system map. right?) ? Question 4: Is 0x378 the physical address in the system memory map generated by BIOS as stated here http://www.pcguide.com/ref/mbsys/res/ioSummary-c.html Or a logical address from above table remapped inside the RAM which is in turn mapped at *0x00100000* physical I am banging my head with all of these bits and pieces of information without having a clue to stitch all of them together. Please give me some pointers to understand them. Thanks in advance. Regards Vineel Kumar Reddy Kovvuri

2 1