August 2014 - kernelnewbies@kernelnewbies.org

Re: Reading Code
by Nick Krause Aug. 8, 2014

Aug. 8, 2014

On Fri, Aug 8, 2014 at 3:40 PM, Pol <poleyschen(a)hotmail.com> wrote: > Well, seems like you have to also learn to google, since I found the answer > to all your questions within seconds. > > Pol > > > > On 08.8.14 21:16, Nick Krause wrote: >> >> For the last few days , I have been reading the code for do_fork and >> other parts of the scheduler code. >> As most of the developers feel this is in my best interest. I do >> however have a few questions about the >> code for the function, schedule. >> 1. In task struct we have this one defined if needed >> 1481 #ifdef CONFIG_BLOCK >> 1482 /* stack plugging */ >> 1483 struct blk_plug *plug; >> 1484 #endif >> I am curious is this used for block IO or something else as we are >> checking for it in schedule() with this line, >> sched_submit_work(tsk); which in turn points to sched_submit_work and >> in turn checks for blk_plug >> 2814 static inline void sched_submit_work(struct task_struct *tsk) >> 2815 { >> 2816 if (!tsk->state || tsk_is_pi_blocked(tsk)) >> 2817 return; >> 2818 /* >> 2819 * If we are going to sleep and we have plugged IO queued, >> 2820 * make sure to submit it to avoid deadlocks. >> 2821 */ >> 2822 if (blk_needs_flush_plug(tsk)) >> 2823 blk_schedule_flush_plug(tsk); >> 2824 } >> Since schedule returns this and after searching using lxr, can't find >> much on this, what is this type actually defined as when compiled? >> 368 #define __sched __attribute__((__section__(".sched.text"))) >> 3. context_switch(rq, prev, next); /* unlocks the rq */ >> I am getting confused by this function after tracing it, if someone >> wants to explain it to me that would be great. >> >> In addition I am not not trying to turn this into a self help just for >> me, but my questions about git I goggled for 2 hours and didn't >> >> find any answers, sorry about that guys. In addition I hope theses >> questions are valid. >> Cheers Nick >> > Pol, I think perhaps I was goggling the wrong things. If someone wants to help me answer my questions that would be great. Sorry Pol

1 0

Reading Code
by Nick Krause Aug. 8, 2014

Aug. 8, 2014

For the last few days , I have been reading the code for do_fork and other parts of the scheduler code. As most of the developers feel this is in my best interest. I do however have a few questions about the code for the function, schedule. 1. In task struct we have this one defined if needed 1481 #ifdef CONFIG_BLOCK 1482 /* stack plugging */ 1483 struct blk_plug *plug; 1484 #endif I am curious is this used for block IO or something else as we are checking for it in schedule() with this line, sched_submit_work(tsk); which in turn points to sched_submit_work and in turn checks for blk_plug 2814 static inline void sched_submit_work(struct task_struct *tsk) 2815 { 2816 if (!tsk->state || tsk_is_pi_blocked(tsk)) 2817 return; 2818 /* 2819 * If we are going to sleep and we have plugged IO queued, 2820 * make sure to submit it to avoid deadlocks. 2821 */ 2822 if (blk_needs_flush_plug(tsk)) 2823 blk_schedule_flush_plug(tsk); 2824 } Since schedule returns this and after searching using lxr, can't find much on this, what is this type actually defined as when compiled? 368 #define __sched __attribute__((__section__(".sched.text"))) 3. context_switch(rq, prev, next); /* unlocks the rq */ I am getting confused by this function after tracing it, if someone wants to explain it to me that would be great. In addition I am not not trying to turn this into a self help just for me, but my questions about git I goggled for 2 hours and didn't find any answers, sorry about that guys. In addition I hope theses questions are valid. Cheers Nick

2 2

Re: Help with git
by Kristofer Hallin Aug. 8, 2014

Aug. 8, 2014

Thinking about the same thing. This list is not interesting any more. On 8 Aug 2014 20:28, "Robert P. J. Day" <rpjday(a)crashcourse.ca> wrote: > On Fri, 8 Aug 2014, Kristofer Hallin wrote: > > > This list is now discussing Nick's problems all the time and nothing > > else. > > more specifically, the list is now teaching nick how to use git. so, > on that note, i'm unsubscribing until this idiocy goes away. catch you > on the flip side, as they say. > > rday > > -- > > ======================================================================== > Robert P. J. Day Ottawa, Ontario, CANADA > http://crashcourse.ca > > Twitter: http://twitter.com/rpjday > LinkedIn: http://ca.linkedin.com/in/rpjday > ======================================================================== > >

1 0

Help with git
by Nick Krause Aug. 8, 2014

Aug. 8, 2014

I can't seem to apply one of my created patches after trying for the last hour. This is what is popping up, if anybody can tell me is wrong that would be great. sed: can't read /home/nick/linux-next/.git/rebase-apply/info: No such file or directory sed: can't read /home/nick/linux-next/.git/rebase-apply/info: No such file or directory sed: can't read /home/nick/linux-next/.git/rebase-apply/info: No such file or directory Patch does not have a valid e-mail address. Nick

10 28

Usb module loading
by selva Aug. 8, 2014

Aug. 8, 2014

Hi All I have written a usb module which will be linked if any keyboard is hot plug. I follow the step below to make it work 1) rmmod usbhid 2) insmod hello.ko 3) hot plug the keyboard when the keyboard is hot plugged my kernel module gets loaded and in syslog the messages regarding the detection of the keyboard are present. My problem is i have to always remove "usbhid", if i donot that then "usbhid" module is always linked with the keyboard. How to make make sure that my module usb_kbd.ko gets linked to the keyboard with out removing "usbhid" ? My code is present in the following link https://github.com/selvagit/little_penguin/blob/master/task_05/hello.c The syslog is present in the following path https://github.com/selvagit/little_penguin/blob/master/task_05/syslog I had looked at udev rules, but what i have understood is we can set attributes for the modules but we can not determine which kernel module should be loaded for any device. Thanks, Selvaraj

5 6

Contributing to kernel release pages
by Timothée Ravier Aug. 8, 2014

Aug. 8, 2014

Hi, I'd like to contribute to the Linux kernel release pages. Could someone add me to the EditorsGroup please? I've asked on IRC but got no answer. My user name is TimotheeRavier. Thanks, -- Timothée Ravier

1 0

[RFC] TCP ISN access from userlevel
by Aleksey Midenkov Aug. 8, 2014

Aug. 8, 2014

Hello! Expansion of Linux Kernel API for getting from userlevel TCP ISN by socket descriptor. Would such improvement deserve merge? The use case is to differ hosts behind NAT. The general plan of changes: 1. copy ISN into the tcp_sock structure when the TCP session is created; 2. add ioctl for getting ISN; or 2. copy that value into the tcp_info structure returned by getsockopt(). I see that 'snt_isn' and 'rcv_isn' are stored in 'tcp_request_sock'. Is 'tcp_request_sock' have the lifetime of 'tcp_sock'? I.e. can it be always accessed via 'fastopen_rsk'? Please, CC me on reply.

1 0

Questions about fork and scheduler
by Nick Krause Aug. 8, 2014

Aug. 8, 2014

Would someone who works the scheduler and process subsystems a lot send me a few messages tomorrow to help me check my understanding of these subsystems in order to certify my knowledge first. Regards Nick

1 0

Re: Delivery Status Notification (Failure)
by Nick Krause Aug. 7, 2014

Aug. 7, 2014

Does anybody have some work in the scheduler subsystem I can work on that is a good top dipping for a kernel newbie. Regards Nick

4 6

Questions about struct task_struct
by Nick Krause Aug. 7, 2014

Aug. 7, 2014

I am new to kernel programming and have a lot questions about this structure. Also if after you guys reply I have a few \ questions for Greg Kroah Hartman about usb development and stable trees. I am very excited about learning this properly now and if anyone notices something wrong in my understanding please let me known :). Thanks So Much, Nick Questions about task_struct 1.volatile long state; Is this line like jiffies for timers with regards to the keyword volatile? 2.void *stack; Is this a pointer to the kernel stack for the process related to the process that the structure is keeping data for and is void to make it an opaque data type? Is this normally an int or some other data type when called with most processes? 3.unsigned int flags; /* per process flags, defined below */, My only question about this is what flags are there and are these the ifdef statements in the struct? 4.Why are we making the usage count atomic?(pretty obvious but real world reasons would be great) 5.I assume int ptrace is for ptrace or debugged processes? If I wrong please let me known too. 6. What is wakkee_flips and wakkee_flips_decay used for? 7.This is probably a stupid question but are these the scheduler cases for CFS? 1240 int on_rq; 1241 1242 int prio, static_prio, normal_prio; 1243 unsigned int rt_priority; 1244 const struct sched_class *sched_class; 1245 struct sched_entity se; 1246 struct sched_rt_entity rt; 1247 #ifdef CONFIG_CGROUP_SCHED 1248 struct task_group *sched_task_group; 1249 #endif 1250 struct sched_dl_entity dl; 8.What are preempt notifiers and what are they for?(another stupid question) 9. Are these lines for the policy used and to block certain processes to only a few processors? 1261 unsigned int policy; 1262 int nr_cpus_allowed; 1263 cpumask_t cpus_allowed; 8.What are the options like CONFIG_PREEMPT_RCU used for in terms of workloads or locking as I heard a lot about them on the threads on the lkml? P.S. Sorry about some many questions,. Answer then whenever you guys have time and if there stupid and easily found somewhere let me known.

4 6