June 2015 - kernelnewbies@kernelnewbies.org - Kernelnewbies Mailing Lists

Re: idle task check
by Mustafa Hussain June 4, 2015

June 4, 2015

System crashes, system can not start On June 3, 2015 9:41:52 PM EDT, Mustafa Hussain <mustafa.hussain93(a)gmail.com> wrote: >i want to dequeue the idle task how can i do this ? Why there is no point. Clearly your asking questions in order to learn the scheduler. If your interested in learning it I can help but, you need to think about what you trying to accomplish first. Nick >On Thu, Jun 4, 2015 at 3:40 AM, Mustafa Hussain ><mustafa.hussain93(a)gmail.com >> wrote: > >> Hi nick, >> i applied your suggested edit and i got "bad: scheduling from the >idle >> thread!" >> how can i solve this ? >> >> On Wed, Jun 3, 2015 at 12:29 AM, nick <xerofoify(a)gmail.com> wrote: >> >>> >>> >>> On 2015-06-02 06:25 PM, Valdis.Kletnieks(a)vt.edu wrote: >>> > On Tue, 02 Jun 2015 23:38:48 +0200, Mustafa Hussain said: >>> > >>> >>> /*Check if the pointer pointing to the idle class is equal to >prev's >>> >>> sched_class*/ >>> >>> if(prev->sched_class == idle) >>> >>> After this condition you can just: >>> >>> printk(KERN_INFO "Prev is equal to idle_sched_class,now running >the >>> idle >>> >>> sched_class\n"); >>> > >>> > Hopefully, you didn't take Nick's advice without thinking about >it.... >>> > >>> > As I type this, powertop tells me: >>> > >>> > Summary: 821.8 wakeups/second, 0.0 GPU ops/seconds, 0.0 VFS >ops/sec >>> and 18.8% CPU use >>> > >>> > That printk is going to spam your dmesg pretty hard. >>> > >>> > A better question is: >>> > >>> > If prev is about to go idle, *what do you want to do*? (Hint: >newer >>> > kernels already do a bunch of stuff when a cpu/core goes idle, you >>> > probably want to make sure you're not working against something >here...) >>> > >>> I didn't account for rate limiting the debug messages, forgot about >that >>> . :) >>> I do agree his question is not the best but he wanted a answer so I >>> decided >>> to just give him a answer that works for his learning. >>> Nick >>> >> >> -- Sent from my Android device with K-9 Mail. Please excuse my brevity.

1 0

unlocking the recv
by Grzegorz Dwornicki June 3, 2015

June 3, 2015

Hello I am wondering with pure theoretical task. Is there a posibility to unlock the recv function without making the socket nonblocking? Lets assume that we have our server and client: server <---network ---> client Server callend the recv function and waits for the data from client soo its blocked in wait_queue. The socket is in nonblocking state. Can this proces/thread be unblocked on demand?

3 2

Fwd: How flowi4 variable fl->u.ip4.fl4_gre_key variable updates in kernel code for GRE
by Harsh Jain June 3, 2015

June 3, 2015

Hi All, In which function kernel fills the GRE key value in "fl->u.ip4.fl4_gre_key" struct flowi4.I am trying to apply IPSEC encrytion policy based on GRE key in packet. The function "xfrm_flowi_sport" prints uli->gre_key = 0 for GRE packets. Is it a bug in kernel code?. Regards Harsh Jain ---------- Forwarded message ---------- From: Harsh Jain <harshjain.prof(a)gmail.com> Date: Mon, Jun 1, 2015 at 1:57 AM Subject: Packet encryption based on gre key value with ip xfrm command To: Kernelnewbies <kernelnewbies(a)kernelnewbies.org> Hi, I am trying to encrypt Gre packet have specific key values in GRE header with following command ip xfrm policy add src 192.168.1.9 dst 192.168.1.5 proto gre key 3 dir in tmpl src 192.168.1.9 dst 192.168.1.5 proto esp reqid 16387 mode transport But it is not working. If I remove the "key 3" from above system encrypt all GRE packets. I tried with kernel version 3.18 and iproute2 version.2.4. I got iproute2 patch file having changes to support filtering based on keys but didn't find corresponding kernel patch. How to encrypt Packets based on GRE key value.? Regards Harsh Jain

1 0

Re: idle task check
by Mustafa Hussain June 2, 2015

June 2, 2015

this what i was talking you about :) On Sat, May 9, 2015 at 9:29 PM, nick <xerofoify(a)gmail.com> wrote: > > > On 2015-05-09 12:44 PM, Mustafa Hussain wrote: > > Hi all, > > > > i am working with __sched __schedule() > > > > and i want to add some thing i want to check whether the prev will become > > idle or not how can i do this. > > > > thanks in advance > > > > - > > - > Mustafa, > After looking around in pick_next_task there is a scheduler class called > idle_sched_class for idling when tasks need to be run. Something like > /*declare sched_class pointer to idle_sched_class address*/ > const struct sched_class *idle=&idle_sched_class; > /*Check if the pointer pointing to the idle class is equal to prev's > sched_class*/ > if(prev->sched_class == idle) > After this condition you can just: > printk(KERN_INFO "Prev is equal to idle_sched_class,now running the idle > sched_class\n"); > If you have another question about this please let me known. > Cheers, > Nick > > > > > > > > _______________________________________________ > > Kernelnewbies mailing list > > Kernelnewbies(a)kernelnewbies.org > > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > > >

2 1

Re: idle task check
by Mustafa Hussain June 2, 2015

June 2, 2015

yes On Tue, Jun 2, 2015 at 11:40 PM, nick <xerofoify(a)gmail.com> wrote: > > > On 2015-06-02 05:38 PM, Mustafa Hussain wrote: > > this what i was talking you about :) > > > > On Sat, May 9, 2015 at 9:29 PM, nick <xerofoify(a)gmail.com> wrote: > > > >> > >> > >> On 2015-05-09 12:44 PM, Mustafa Hussain wrote: > >>> Hi all, > >>> > >>> i am working with __sched __schedule() > >>> > >>> and i want to add some thing i want to check whether the prev will > become > >>> idle or not how can i do this. > >>> > >>> thanks in advance > >>> > >>> - > >>> - > >> Mustafa, > >> After looking around in pick_next_task there is a scheduler class called > >> idle_sched_class for idling when tasks need to be run. Something like > >> /*declare sched_class pointer to idle_sched_class address*/ > >> const struct sched_class *idle=&idle_sched_class; > >> /*Check if the pointer pointing to the idle class is equal to prev's > >> sched_class*/ > >> if(prev->sched_class == idle) > >> After this condition you can just: > >> printk(KERN_INFO "Prev is equal to idle_sched_class,now running the idle > >> sched_class\n"); > >> If you have another question about this please let me known. > >> Cheers, > >> Nick > >>> > First do you known what scheduling classes/ task structures are? > Nick > >>> > >>> > >>> _______________________________________________ > >>> Kernelnewbies mailing list > >>> Kernelnewbies(a)kernelnewbies.org > >>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > >>> > >> > > >

1 0

How .suspend and .resume are invoked
by Anand Moon June 1, 2015

June 1, 2015

hi All, I would like to know how .suspend and .resume in device driver invoked where ACPI is not used. What I meant is suppose you have USB not connected then what is the state of the device driver. Is the .probe function running continuously in the background. -Anand Moon

2 2

Re: Stop usbhid from claiming usb device on hotplug
by Armin Moradi June 1, 2015

June 1, 2015

>>> Basically what your doing wrong is not understanding the hierarchy of USB devices on your system's bus. >>> What is happening is your trying to register a USB device where the Id marco is not the correct one in the >>> hierarchy and therefore is over ruled by the one from usbhid. >>> If you send your complete code to the list I can try and see if there are other issues but that's the most likely. >>> Nick >> >> Thanks a lot! I have posted the complete code previously on this >> thread, but it doesn't have much. I'll see what I can find out about >> the usb subsystem hierarchy. Any recommended resources I could use to >> learn more about the usb (and other) subsystems (e.g. pci, etc.) other >> than LDD? >> > I used Essential Linux Device Drivers as my resource. Also look in the driver source code for usb keyboard driver > and read through it. If you can't find it I will point it out to you but that should get you started. One piece > of advice is when writing drivers, try to find one in the drivers folder with similar hardware/requirements > to base your driver off of that your writing. Also don't ask about challenges like this on the list as there > meant to be self taught. > Hope that Helps, > Nick Thanks, I'll see what I can find. Again, my question has nothing to do with the challenge. -- Armin Moradi

1 0

Re: Stop usbhid from claiming usb device on hotplug
by Armin Moradi June 1, 2015

June 1, 2015

> Basically what your doing wrong is not understanding the hierarchy of USB devices on your system's bus. > What is happening is your trying to register a USB device where the Id marco is not the correct one in the > hierarchy and therefore is over ruled by the one from usbhid. > If you send your complete code to the list I can try and see if there are other issues but that's the most likely. > Nick Thanks a lot! I have posted the complete code previously on this thread, but it doesn't have much. I'll see what I can find out about the usb subsystem hierarchy. Any recommended resources I could use to learn more about the usb (and other) subsystems (e.g. pci, etc.) other than LDD? >>> Your right, questions about eucalyptus based tasks should be asked >>off list >>> to avoid people getting easy answers through. >> >>No, the rules are to do it *on your own*, without asking anybody, on or >>off >>list. >> >>> However I think a lot of this could be avoided if we taught people >>how to do >>> kernel programming research better. >> >>Well, why don't you make suggestions from your own experience how we >>can improve >>that? > Thats fine for me it was just not listening and pulling in effort, also lack of confidence . However I am noticing a lot of people having difficulty > understanding drivers and the driver core. The issue is that most people don't know > where to look. I would recommend putting a list on kernel newbies of resources on a page for websites, books, tools that are useful. Also it would be good to explain on their how to use crags/cscope for kernel code searching. > Nick I think I should clarify the eudyptula situation. I started doing the challenge, and then I got to task05, which is the first task that tells you to read LDD. I started reading the chapter that was instructed (I'm really trying to avoid spoilers), but then also read other chapters which looked interesting. Before I knew it, I was doing something not related to my task at all (as far as I understood the task, again, no spoilers). That's when I asked the question about usbhid here and promptly got kicked. So, my question has nothing to do with the eudyptula challenge and I'm not aware of any of the tasks being related to this question. Sidenote: unfortunately, I feel like the eudyptula challenge has patented basic kernel questions as I can't even ask anything because someone, somewhere, is doing something vaguely similar. I appreciate the challenge, but maybe the Matasano's new challenge format <http://cryptopals.com> would do more to help interested people learn rather than create an exclusive tunnel. > The issue is that most people don't know where to look. That is the issue I'm having as the header files, while pretty good in describing what's going on, don't explain the higher level stuff that's needed before diving into header docs. Although I may have jumped in unprepared. -- Armin Moradi

1 0

Re: Stop usbhid from claiming usb device on hotplug
by Valdis.Kletnieks＠vt.edu June 1, 2015

June 1, 2015

On Sun, 31 May 2015 21:31:20 -0400, Nicholas Krause said: > Your right, questions about eucalyptus based tasks should be asked off list > to avoid people getting easy answers through. No, the rules are to do it *on your own*, without asking anybody, on or off list. > However I think a lot of this could be avoided if we taught people how to do > kernel programming research better. Well, why don't you make suggestions from your own experience how we can improve that?

1 0

Re: Stop usbhid from claiming usb device on hotplug
by Valdis.Kletnieks＠vt.edu June 1, 2015

June 1, 2015

On Sun, 31 May 2015 20:39:06 -0400, Nicholas Krause said: > If you send your complete code to the list I can try and see if there are other issues but that's the most likely. Actually, he shouldn't send it to the list. It smells too much of eucalyptus....

1 0