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

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

June 4, 2015

I mean how can we use the pid in the scenario of finding the process that is not used. On 4 Jun 2015 16:00, "nick" <xerofoify(a)gmail.com> wrote: > > > On 2015-06-04 08:52 AM, Mustafa Hussain wrote: > > Ok, everything is clear except one thing what we will do exactly with > (pid) > > s? > > On 4 Jun 2015 15:40, "nick" <xerofoify(a)gmail.com> wrote: > > > Its used in task struct to create a doubly linked list with init/systemd > being the head or first process and the others linked in other of their > hierarchy. > Nick > >> > >> > >> On 2015-06-04 08:34 AM, Mustafa Hussain wrote: > >>> great, How can I check if this running task is used or not.. I mean by > >> not > >>> used that the task is running but not used by the user > >>> > >>> > >> Here is the issue through in order to find out what tasks the user is > >> running we > >> need to known the exact pid(s) each time. This is next to impossible to > do > >> without > >> writing a syscall and that’s a little beyond you if your asking about > the > >> scheduler. > >> However, there may be a debugging feature in perf or other tool that > does > >> this, you > >> can google to see if something wrote a tool or kernel module for this. > >> Nick > >>> On June 4, 2015 8:22:03 AM EDT, Mustafa Hussain < > >> mustafa.hussain93(a)gmail.com> > >>> wrote: > >>>> So I was just thinking about if i did this i can close apps that is > >>>> running > >>>> and user don't use them.. > >>>> On 4 Jun 2015 14:59, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > >>>> > >>> That's not a idle task, that's a task in the > >>> TASK_INTERRUPTIBLE or TASK_RUNNING phase. > >>> Nick > >>>>> > >>>>> > >>>>> On June 4, 2015 7:56:30 AM EDT, Mustafa Hussain < > >>>>> mustafa.hussain93(a)gmail.com> wrote: > >>>>>> All i am trying to do is to detect idle task and remove it from the > >>>>>> running > >>>>>> queue or deactivate it. > >>>>> There is no reason for that. The idle tasks are only ever scheduled > >>>> when > >>>>> there is no other processes able to > >>>>> run. So trying to remove them is a > >>>>> bad idea. > >>>>> Nick > >>>>>> Thank you for your patience :) > >>>>>> On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > >>>>>> > >>>>>>> > >>>>>>> > >>>>>>> On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < > >>>>>>> mustafa.hussain93(a)gmail.com> wrote: > >>>>>>>> System crashes, system can not start > >>>>>>>> > >>>>>>> I was not thinking and this schedules the idle thread. What are > >>>> you > >>>>>> trying > >>>>>>> to accomplish through. > >>>>>>> Nick > >>>>>>>> > >>>>>>>> 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. > >>>>>>> > >>>>>>> -- > >>>>>>> Sent from my Android device with K-9 Mail. Please excuse my > >>>> brevity. > >>>>>>> > >>>>> > >>>>> -- > >>>>> Sent from my Android device with K-9 Mail. Please excuse my brevity. > >>>>> > >>> > >>> -- > >>> Sent from my Android device with K-9 Mail. Please excuse my brevity. > >>> > >> > > >

2 1

HID Quirks
by Abhishek bist June 4, 2015

June 4, 2015

What is quirks in HID and why they are used there. I tried it but didn't get a clear image of it and it's usage implementation.

1 0

rcu_sched stall in a dual-core dedicated server on a datacenter
by miguel.telleria＠iteisa.com June 4, 2015

June 4, 2015

Dear kernelnewbies, In one of the dedicated servers that our company has for shared web hosting we have had a series of log lines: Jun 3 00:33:53 hostname kernel: INFO: rcu_sched self-detected stall on CPU { 0} (t=15001 jiffies g=66854079 c=66854078 q=686736) Jun 3 00:33:53 hostname kernel: INFO: rcu_sched detected stalls on CPUs/tasks: { 0} (detected by 1, t=15002 jiffies, g=66854079, c=66854078, q=686736) The lines appeared every 3 minutes from 00:33:54 to 05:12:33 with the last line being: Jun 3 05:12:33 ns25 kernel: INFO: rcu_sched self-detected stall on CPU { 0} (t=4155459 jiffies g=66854079 c=66854078 q=10628601) Jun 3 05:12:33 ns25 kernel: INFO: rcu_sched detected stalls on CPUs/tasks: { 0} (detected by 1, t=4155460 jiffies, g=66854079, c=66854078, q=10628601) with all of them there are traces of CPU states with the kernel function trace, CPU registers, etc. Some more info about this: - The problem ended silently as it started, no server restart took place, uptime still shows plenty of time. - The value in t=nnnnnn seems to be the time elapsed in this state in Jiffies. 4155460 = 16621 sg = 4h 36min, roughly the time difference between 05:12 and 00:33 - Kernel 3.10 from our datacenter. Dumping /proc/config.gz I obtain the following info: # RCU Subsystem CONFIG_TREE_RCU=y # CONFIG_PREEMPT_RCU is not set CONFIG_RCU_STALL_COMMON=y # CONFIG_RCU_USER_QS is not set CONFIG_RCU_FANOUT=64 CONFIG_RCU_FANOUT_LEAF=16 # CONFIG_RCU_FANOUT_EXACT is not set # CONFIG_RCU_FAST_NO_HZ is not set # CONFIG_TREE_RCU_TRACE is not set # CONFIG_RCU_NOCB_CPU is not set # RCU Debugging # CONFIG_SPARSE_RCU_POINTER is not set CONFIG_RCU_CPU_STALL_TIMEOUT=60 CONFIG_HAVE_UNSTABLE_SCHED_CLOCK=y # CONFIG_CGROUP_SCHED is not set # CONFIG_SCHED_AUTOGROUP is not set CONFIG_DEFAULT_IOSCHED="deadline" CONFIG_SCHED_OMIT_FRAME_POINTER=y CONFIG_SCHED_SMT=y CONFIG_SCHED_MC=y CONFIG_SCHED_HRTICK=y CONFIG_NET_SCHED=y # Queueing/Scheduling # CONFIG_USB_EHCI_TT_NEWSCHED is not set CONFIG_PREEMPT_NOTIFIERS=y CONFIG_PREEMPT_NONE=y # CONFIG_PREEMPT_VOLUNTARY is not set # CONFIG_PREEMPT is not set My questions are ---------------- 1. Am I right assuming that the whole CPU was stalled for all that period? We have proof that the machine was still working (log info among it), but maybe it was only working with one core. 2. The task that was running on the stalled core, was it trapped without running the whole time? With the log lines above I have always had this line: kernel: CPU: 0 PID: 1288 Comm: php-cgi Not tainted 3.10.23-xxxx-std-ipv6-64 #1 3. Can there be other tasks trapped? AFAIK modern Linux 3.10 does not keep separate run-queues for each core, right? 4. What are the meaning of the t=, g= and c= values in the rcu_sched log? t=4155459 jiffies g=66854079 c=66854078 q=10628601 AFAICT, t= represents the time elapsed since this situation started. g= and c= are constant values and q= is an increasing figure. 5. Any ideas of where can this come from? Here is the trace if it helps: [<ffffffff8160806c>] __const_udelay+0x2c/0x30 [<ffffffff8106338a>] arch_trigger_all_cpu_backtrace+0x6a/0xa0 [<ffffffff81115cdd>] rcu_check_callbacks+0x2ed/0x560 [<ffffffff81117c17>] ? acct_account_cputime+0x17/0x20 [<ffffffff810e97df>] ? account_system_time+0xcf/0x180 [<ffffffff810ca2c3>] update_process_times+0x43/0x80 [<ffffffff810f95c1>] tick_sched_handle.isra.12+0x31/0x40 [<ffffffff810f9704>] tick_sched_timer+0x44/0x70 [<ffffffff810df07a>] __run_hrtimer.isra.29+0x4a/0xd0 [<ffffffff810df9b5>] hrtimer_interrupt+0xf5/0x230 [<ffffffff810626c4>] smp_apic_timer_interrupt+0x64/0xa0 [<ffffffff81d421ca>] apic_timer_interrupt+0x6a/0x70 <EOI> [<ffffffff81183e80>] ? switch_names+0xc0/0xc0 [<ffffffff81d409ee>] ? _raw_spin_lock+0xe/0x30 [<ffffffff81184eb3>] shrink_dentry_list+0x43/0x1f0 [<ffffffff811854b7>] prune_dcache_sb+0x137/0x160 [<ffffffff81173ab8>] prune_super+0x168/0x1a0 [<ffffffff81131119>] shrink_slab+0x129/0x1e0 [<ffffffff811331f8>] ? shrink_zone+0xa8/0xd0 [<ffffffff811337af>] do_try_to_free_pages+0x31f/0x410 [<ffffffff811339dd>] try_to_free_pages+0x9d/0xc0 [<ffffffff8112b96e>] __alloc_pages_nodemask+0x4de/0x840 [<ffffffff81159560>] alloc_pages_vma+0xa0/0x1a0 [<ffffffff81142b67>] handle_pte_fault+0x7b7/0x9f0 [<ffffffff81146906>] ? vma_gap_update+0x26/0x30 [<ffffffff8113e293>] ? anon_vma_interval_tree_insert+0x83/0x90 [<ffffffff81143047>] handle_mm_fault+0x147/0x1e0 [<ffffffff8106c391>] __do_page_fault+0x151/0x4b0 [<ffffffff811478ff>] ? vma_merge+0x1ff/0x320 [<ffffffff81148a4b>] ? do_brk+0x1bb/0x320 [<ffffffff8106c719>] do_page_fault+0x9/0x10 [<ffffffff81d41172>] page_fault+0x22/0x30 Any ideas will be welcome.

1 0

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

June 4, 2015

Ok everything is clear now thanks for your time, i will search a while and try to code that syscall :) Thanks again. On 4 Jun 2015 16:18, "nick" <xerofoify(a)gmail.com> wrote: > > > On 2015-06-04 09:16 AM, Mustafa Hussain wrote: > > I mean how can we use the pid in the scenario of finding the process that > > is not used. > Just check if the process is in the TASK_RUNNING state if it is then > it's in a runnable state, waiting to run on the queue. Otherwise > it's idling waiting for something or has completed its work. > Nick > > On 4 Jun 2015 16:00, "nick" <xerofoify(a)gmail.com> wrote: > > > >> > >> > >> On 2015-06-04 08:52 AM, Mustafa Hussain wrote: > >>> Ok, everything is clear except one thing what we will do exactly with > >> (pid) > >>> s? > >>> On 4 Jun 2015 15:40, "nick" <xerofoify(a)gmail.com> wrote: > >>> > >> Its used in task struct to create a doubly linked list with init/systemd > >> being the head or first process and the others linked in other of their > >> hierarchy. > >> Nick > >>>> > >>>> > >>>> On 2015-06-04 08:34 AM, Mustafa Hussain wrote: > >>>>> great, How can I check if this running task is used or not.. I mean > by > >>>> not > >>>>> used that the task is running but not used by the user > >>>>> > >>>>> > >>>> Here is the issue through in order to find out what tasks the user is > >>>> running we > >>>> need to known the exact pid(s) each time. This is next to impossible > to > >> do > >>>> without > >>>> writing a syscall and that’s a little beyond you if your asking about > >> the > >>>> scheduler. > >>>> However, there may be a debugging feature in perf or other tool that > >> does > >>>> this, you > >>>> can google to see if something wrote a tool or kernel module for this. > >>>> Nick > >>>>> On June 4, 2015 8:22:03 AM EDT, Mustafa Hussain < > >>>> mustafa.hussain93(a)gmail.com> > >>>>> wrote: > >>>>>> So I was just thinking about if i did this i can close apps that is > >>>>>> running > >>>>>> and user don't use them.. > >>>>>> On 4 Jun 2015 14:59, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > >>>>>> > >>>>> That's not a idle task, that's a task in the > >>>>> TASK_INTERRUPTIBLE or TASK_RUNNING phase. > >>>>> Nick > >>>>>>> > >>>>>>> > >>>>>>> On June 4, 2015 7:56:30 AM EDT, Mustafa Hussain < > >>>>>>> mustafa.hussain93(a)gmail.com> wrote: > >>>>>>>> All i am trying to do is to detect idle task and remove it from > the > >>>>>>>> running > >>>>>>>> queue or deactivate it. > >>>>>>> There is no reason for that. The idle tasks are only ever > scheduled > >>>>>> when > >>>>>>> there is no other processes able to > >>>>>>> run. So trying to remove them is a > >>>>>>> bad idea. > >>>>>>> Nick > >>>>>>>> Thank you for your patience :) > >>>>>>>> On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> > wrote: > >>>>>>>> > >>>>>>>>> > >>>>>>>>> > >>>>>>>>> On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < > >>>>>>>>> mustafa.hussain93(a)gmail.com> wrote: > >>>>>>>>>> System crashes, system can not start > >>>>>>>>>> > >>>>>>>>> I was not thinking and this schedules the idle thread. What are > >>>>>> you > >>>>>>>> trying > >>>>>>>>> to accomplish through. > >>>>>>>>> Nick > >>>>>>>>>> > >>>>>>>>>> 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. > >>>>>>>>> > >>>>>>>>> -- > >>>>>>>>> Sent from my Android device with K-9 Mail. Please excuse my > >>>>>> brevity. > >>>>>>>>> > >>>>>>> > >>>>>>> -- > >>>>>>> Sent from my Android device with K-9 Mail. Please excuse my > brevity. > >>>>>>> > >>>>> > >>>>> -- > >>>>> Sent from my Android device with K-9 Mail. Please excuse my brevity. > >>>>> > >>>> > >>> > >> > > >

1 0

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

June 4, 2015

Ok, everything is clear except one thing what we will do exactly with (pid) s? On 4 Jun 2015 15:40, "nick" <xerofoify(a)gmail.com> wrote: > > > On 2015-06-04 08:34 AM, Mustafa Hussain wrote: > > great, How can I check if this running task is used or not.. I mean by > not > > used that the task is running but not used by the user > > > > > Here is the issue through in order to find out what tasks the user is > running we > need to known the exact pid(s) each time. This is next to impossible to do > without > writing a syscall and that’s a little beyond you if your asking about the > scheduler. > However, there may be a debugging feature in perf or other tool that does > this, you > can google to see if something wrote a tool or kernel module for this. > Nick > > On June 4, 2015 8:22:03 AM EDT, Mustafa Hussain < > mustafa.hussain93(a)gmail.com> > > wrote: > >> So I was just thinking about if i did this i can close apps that is > >> running > >> and user don't use them.. > >> On 4 Jun 2015 14:59, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > >> > > That's not a idle task, that's a task in the > > TASK_INTERRUPTIBLE or TASK_RUNNING phase. > > Nick > >>> > >>> > >>> On June 4, 2015 7:56:30 AM EDT, Mustafa Hussain < > >>> mustafa.hussain93(a)gmail.com> wrote: > >>>> All i am trying to do is to detect idle task and remove it from the > >>>> running > >>>> queue or deactivate it. > >>> There is no reason for that. The idle tasks are only ever scheduled > >> when > >>> there is no other processes able to > >>> run. So trying to remove them is a > >>> bad idea. > >>> Nick > >>>> Thank you for your patience :) > >>>> On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > >>>> > >>>>> > >>>>> > >>>>> On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < > >>>>> mustafa.hussain93(a)gmail.com> wrote: > >>>>>> System crashes, system can not start > >>>>>> > >>>>> I was not thinking and this schedules the idle thread. What are > >> you > >>>> trying > >>>>> to accomplish through. > >>>>> Nick > >>>>>> > >>>>>> 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. > >>>>> > >>>>> -- > >>>>> Sent from my Android device with K-9 Mail. Please excuse my > >> brevity. > >>>>> > >>> > >>> -- > >>> Sent from my Android device with K-9 Mail. Please excuse my brevity. > >>> > > > > -- > > Sent from my Android device with K-9 Mail. Please excuse my brevity. > > >

1 0

Accessing Ram on PCI card
by Robert Clove June 4, 2015

June 4, 2015

Hi, I have a pci card that has some amount of ram and flash memory on it. I have found the bar address range for the memory. I want to transfer a file to that memory and read from that memory. Any pointers(in terms of sample code) how can i do that? Regards

2 6

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

June 4, 2015

great, How can I check if this running task is used or not.. I mean by not used that the task is running but not used by the user On June 4, 2015 8:22:03 AM EDT, Mustafa Hussain <mustafa.hussain93(a)gmail.com> wrote: >So I was just thinking about if i did this i can close apps that is >running >and user don't use them.. >On 4 Jun 2015 14:59, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > That's not a idle task, that's a task in the TASK_INTERRUPTIBLE or TASK_RUNNING phase. Nick >> >> >> On June 4, 2015 7:56:30 AM EDT, Mustafa Hussain < >> mustafa.hussain93(a)gmail.com> wrote: >> >All i am trying to do is to detect idle task and remove it from the >> >running >> >queue or deactivate it. >> There is no reason for that. The idle tasks are only ever scheduled >when >> there is no other processes able to >> run. So trying to remove them is a >> bad idea. >> Nick >> >Thank you for your patience :) >> >On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: >> > >> >> >> >> >> >> On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < >> >> mustafa.hussain93(a)gmail.com> wrote: >> >> >System crashes, system can not start >> >> > >> >> I was not thinking and this schedules the idle thread. What are >you >> >trying >> >> to accomplish through. >> >> Nick >> >> > >> >> >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. >> >> >> >> -- >> >> Sent from my Android device with K-9 Mail. Please excuse my >brevity. >> >> >> >> -- >> Sent from my Android device with K-9 Mail. Please excuse my brevity. >> -- Sent from my Android device with K-9 Mail. Please excuse my brevity.

1 0

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

June 4, 2015

So I was just thinking about if i did this i can close apps that is running and user don't use them.. On 4 Jun 2015 14:59, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > > > On June 4, 2015 7:56:30 AM EDT, Mustafa Hussain < > mustafa.hussain93(a)gmail.com> wrote: > >All i am trying to do is to detect idle task and remove it from the > >running > >queue or deactivate it. > There is no reason for that. The idle tasks are only ever scheduled when > there is no other processes able to > run. So trying to remove them is a > bad idea. > Nick > >Thank you for your patience :) > >On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > > > >> > >> > >> On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < > >> mustafa.hussain93(a)gmail.com> wrote: > >> >System crashes, system can not start > >> > > >> I was not thinking and this schedules the idle thread. What are you > >trying > >> to accomplish through. > >> Nick > >> > > >> >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. > >> > >> -- > >> Sent from my Android device with K-9 Mail. Please excuse my brevity. > >> > > -- > Sent from my Android device with K-9 Mail. Please excuse my brevity. >

1 0

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

June 4, 2015

All i am trying to do is to detect idle task and remove it from the running queue or deactivate it. Thank you for your patience :) On 4 Jun 2015 14:51, "Nicholas Krause" <xerofoify(a)gmail.com> wrote: > > > On June 4, 2015 3:35:25 AM EDT, Mustafa Hussain < > mustafa.hussain93(a)gmail.com> wrote: > >System crashes, system can not start > > > I was not thinking and this schedules the idle thread. What are you trying > to accomplish through. > Nick > > > >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. > > -- > Sent from my Android device with K-9 Mail. Please excuse my brevity. >

1 0

signing kernel modules on RHEL 7
by Chakradhar thota June 4, 2015

June 4, 2015

Hello Everyone, I have compiled kernel module on RHEL7 but when I insert the module, I got following warning "module verification failed: signature and/or required key missing - tainting kernel". I tried signing the module on custom kernel and find it working. How can we sign the module for a target system with standard RHEL distribution? where can we find keys for signing the module on standard kernel? Regards, Chakradhar

4 5