Re: Kernel 2.6.32.60 64bit Crash/Hung
Hi Giridhara, http://ark.intel.com/products/64622/Intel-Xeon-Processor-E5-4650-20M-Cache-2_70-GHz-8_00-GTs-Intel-QPI Show it has only 8 core of CPU. Why are you setting the CONFIG_NR_CPUS=64. certainly it will have a issue in the kernel. -Anand Moon On Tuesday, January 27, 2015 10:27 AM, Giridhara RP (grp) wrote: -Original Message- From: valdis.kletni...@vt.edu [mailto:valdis.kletni...@vt.edu] Sent: Tuesday, January 27, 2015 10:15 AM To: Giridhara RP (grp) Cc: Greg KH; kernelnewbies@kernelnewbies.org Subject: Re: Kernel 2.6.32.60 64bit Crash/Hung On Tue, 27 Jan 2015 04:06:00 +, "Giridhara RP (grp)" said: > Which kernel version should I use to solve this crash/panic? That will depend on exactly why you're still running a 2.6.32 kernel from 6 years ago. Remember - the fact it's all the way up to 2.6.32.65 doesn't mean you have all the bugfixes. It only means you have all the bugfixes that qualified for the -stable side of the fence (which is a small fraction of all the fixes that have happened in the past 6 years). --- Hi Valdis, We were running 2.6.32.60 assuming it's a stable build and my assumption was wrong :(. Should I go ahead with stable 3.18.3 [2015-01-16 ]. Thanks Giri ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
7 swappers and one process with name doe:handler
In pawapper processes aswell, when i see the backtrace i see some
mem aborts but couldnot get to the place which is holding the lock
On Tuesday, January 27, 2015, Arun KS wrote:
> Hi Manty,
>
> On Tue, Jan 27, 2015 at 11:52 AM, manty kuma > wrote:
> > Dear Arun,
> >
> > typedef struct raw_spinlock {
> > arch_spinlock_t raw_lock;
> > #ifdef CONFIG_GENERIC_LOCKBREAK
> > unsigned int break_lock;
> > #endif
> > #ifdef CONFIG_DEBUG_SPINLOCK
> > unsigned int magic, owner_cpu;
> > void *owner;
> > #endif
> > #ifdef CONFIG_DEBUG_LOCK_ALLOC
> > struct lockdep_map dep_map;
> > #endif
> > } raw_spinlock_t;
> >
> >
> > Unfortunately, CONFIG_DEBUG_SPINLOCK is disabled on my sw. So i do not
> have
> > this information. :(
> >
> > Any other way?
> What is the output from your runq command?
>
>
> Thanks,
> Arun
> >
> > On Tue, Jan 27, 2015 at 3:12 PM, Arun KS > wrote:
> >>
> >> Hello Manty,
> >>
> >> On Tue, Jan 27, 2015 at 11:27 AM, manty kuma > wrote:
> >> >
> >> > Hi All,
> >> >
> >> > Currently I am looking into an issue which is holding the printk
> >> > lock(logbuf_lock)
> >> >
> >> > logbuf_lock = {
> >> > raw_lock = {
> >> > owner = 0,
> >> > next = 4
> >> > }
> >> > }
> >> >
> >> > I read about ticketing spinlocks to understand the above output. I
> >> > understood that 4 cpu's are waiting for the lock while someone
> already has
> >> > it.
> >> >
> >> > I would like to know who has it? Is there a way I can know which
> process
> >> > is acquiring a given spinlock? Is the information not stored in it?
> >> >
> >> > I am using crash tool to analyse the dump
> >>
> >>
> >> crash> raw_spinlock_t
> >> typedef struct raw_spinlock {
> >> arch_spinlock_t raw_lock;
> >> unsigned int magic;
> >> unsigned int owner_cpu;
> >> void *owner;
> >> } raw_spinlock_t;
> >> SIZE: 16
> >>
> >> The owner field points to task holding the lock, if locked.
> >>
> >> Here is what i have in one of my dumps,
> >> crash> logbuf_lock
> >> logbuf_lock = $6 = {
> >> raw_lock = {
> >> {
> >> slock = 464001960,
> >> tickets = {
> >> owner = 7080,
> >> next = 7080
> >> }
> >> }
> >> },
> >> magic = 3735899821,
> >> owner_cpu = 4294967295,
> >> owner = 0x
> >> }
> >>
> >> since lock is not acquired, it shows 0x.
> >> Otherwise point to a task_struct *.
> >>
> >> thanks,
> >> Arun
> >>
> >> >
> >> >
> >> > Regards,
> >> > Manty
> >> >
> >> > ___
> >> > Kernelnewbies mailing list
> >> > Kernelnewbies@kernelnewbies.org
> >> > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
> >> >
> >
> >
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
Dear Arun,
typedef struct raw_spinlock {
arch_spinlock_t raw_lock;
#ifdef CONFIG_GENERIC_LOCKBREAK
unsigned int break_lock;
#endif
#ifdef CONFIG_DEBUG_SPINLOCK
unsigned int magic, owner_cpu;
void *owner;
#endif
#ifdef CONFIG_DEBUG_LOCK_ALLOC
struct lockdep_map dep_map;
#endif
} raw_spinlock_t;
Unfortunately, CONFIG_DEBUG_SPINLOCK is disabled on my sw. So i do not have
this information. :(
Any other way?
On Tue, Jan 27, 2015 at 3:12 PM, Arun KS wrote:
> Hello Manty,
>
> On Tue, Jan 27, 2015 at 11:27 AM, manty kuma wrote:
> >
> > Hi All,
> >
> > Currently I am looking into an issue which is holding the printk
> lock(logbuf_lock)
> >
> > logbuf_lock = {
> > raw_lock = {
> > owner = 0,
> > next = 4
> > }
> > }
> >
> > I read about ticketing spinlocks to understand the above output. I
> understood that 4 cpu's are waiting for the lock while someone already has
> it.
> >
> > I would like to know who has it? Is there a way I can know which process
> is acquiring a given spinlock? Is the information not stored in it?
> >
> > I am using crash tool to analyse the dump
>
>
> crash> raw_spinlock_t
> typedef struct raw_spinlock {
> arch_spinlock_t raw_lock;
> unsigned int magic;
> unsigned int owner_cpu;
> void *owner;
> } raw_spinlock_t;
> SIZE: 16
>
> The owner field points to task holding the lock, if locked.
>
> Here is what i have in one of my dumps,
> crash> logbuf_lock
> logbuf_lock = $6 = {
> raw_lock = {
> {
> slock = 464001960,
> tickets = {
> owner = 7080,
> next = 7080
> }
> }
> },
> magic = 3735899821,
> owner_cpu = 4294967295,
> owner = 0x
> }
>
> since lock is not acquired, it shows 0x.
> Otherwise point to a task_struct *.
>
> thanks,
> Arun
>
> >
> >
> > Regards,
> > Manty
> >
> > ___
> > Kernelnewbies mailing list
> > Kernelnewbies@kernelnewbies.org
> > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
> >
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
Hi Manty,
On Tue, Jan 27, 2015 at 11:52 AM, manty kuma wrote:
> Dear Arun,
>
> typedef struct raw_spinlock {
> arch_spinlock_t raw_lock;
> #ifdef CONFIG_GENERIC_LOCKBREAK
> unsigned int break_lock;
> #endif
> #ifdef CONFIG_DEBUG_SPINLOCK
> unsigned int magic, owner_cpu;
> void *owner;
> #endif
> #ifdef CONFIG_DEBUG_LOCK_ALLOC
> struct lockdep_map dep_map;
> #endif
> } raw_spinlock_t;
>
>
> Unfortunately, CONFIG_DEBUG_SPINLOCK is disabled on my sw. So i do not have
> this information. :(
>
> Any other way?
What is the output from your runq command?
Thanks,
Arun
>
> On Tue, Jan 27, 2015 at 3:12 PM, Arun KS wrote:
>>
>> Hello Manty,
>>
>> On Tue, Jan 27, 2015 at 11:27 AM, manty kuma wrote:
>> >
>> > Hi All,
>> >
>> > Currently I am looking into an issue which is holding the printk
>> > lock(logbuf_lock)
>> >
>> > logbuf_lock = {
>> > raw_lock = {
>> > owner = 0,
>> > next = 4
>> > }
>> > }
>> >
>> > I read about ticketing spinlocks to understand the above output. I
>> > understood that 4 cpu's are waiting for the lock while someone already has
>> > it.
>> >
>> > I would like to know who has it? Is there a way I can know which process
>> > is acquiring a given spinlock? Is the information not stored in it?
>> >
>> > I am using crash tool to analyse the dump
>>
>>
>> crash> raw_spinlock_t
>> typedef struct raw_spinlock {
>> arch_spinlock_t raw_lock;
>> unsigned int magic;
>> unsigned int owner_cpu;
>> void *owner;
>> } raw_spinlock_t;
>> SIZE: 16
>>
>> The owner field points to task holding the lock, if locked.
>>
>> Here is what i have in one of my dumps,
>> crash> logbuf_lock
>> logbuf_lock = $6 = {
>> raw_lock = {
>> {
>> slock = 464001960,
>> tickets = {
>> owner = 7080,
>> next = 7080
>> }
>> }
>> },
>> magic = 3735899821,
>> owner_cpu = 4294967295,
>> owner = 0x
>> }
>>
>> since lock is not acquired, it shows 0x.
>> Otherwise point to a task_struct *.
>>
>> thanks,
>> Arun
>>
>> >
>> >
>> > Regards,
>> > Manty
>> >
>> > ___
>> > Kernelnewbies mailing list
>> > Kernelnewbies@kernelnewbies.org
>> > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>> >
>
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
Thank you Sanjeev,
But as of now i am not sure on how to reproduce the issue. So i have to
deal with the dump.
Regards,
Manty
On Tue, Jan 27, 2015 at 3:17 PM, sanjeev sharma wrote:
> Hello
>
> you can try to print running task backtrack information in kernel/lockdep.c
>
> Regards
> Sanjeev Sharma
>
> On Tue, Jan 27, 2015 at 11:27 AM, manty kuma wrote:
>
>> Hi All,
>>
>> Currently I am looking into an issue which is holding the printk
>> lock(logbuf_lock)
>>
>> logbuf_lock = {
>> raw_lock = {
>> owner = 0,
>> next = 4
>> }
>> }
>>
>> I read about ticketing spinlocks to understand the above output. I
>> understood that 4 cpu's are waiting for the lock while someone already has
>> it.
>>
>> I would like to know who has it? Is there a way I can know which process
>> is acquiring a given spinlock? Is the information not stored in it?
>>
>> I am using crash tool to analyse the dump
>>
>> Regards,
>> Manty
>>
>> ___
>> Kernelnewbies mailing list
>> Kernelnewbies@kernelnewbies.org
>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>>
>>
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
Hello
you can try to print running task backtrack information in kernel/lockdep.c
Regards
Sanjeev Sharma
On Tue, Jan 27, 2015 at 11:27 AM, manty kuma wrote:
> Hi All,
>
> Currently I am looking into an issue which is holding the printk
> lock(logbuf_lock)
>
> logbuf_lock = {
> raw_lock = {
> owner = 0,
> next = 4
> }
> }
>
> I read about ticketing spinlocks to understand the above output. I
> understood that 4 cpu's are waiting for the lock while someone already has
> it.
>
> I would like to know who has it? Is there a way I can know which process
> is acquiring a given spinlock? Is the information not stored in it?
>
> I am using crash tool to analyse the dump
>
> Regards,
> Manty
>
> ___
> Kernelnewbies mailing list
> Kernelnewbies@kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Ticketing spinlocks - How to get to the owner who locked the spinlock?
Hello Manty,
On Tue, Jan 27, 2015 at 11:27 AM, manty kuma wrote:
>
> Hi All,
>
> Currently I am looking into an issue which is holding the printk
> lock(logbuf_lock)
>
> logbuf_lock = {
> raw_lock = {
> owner = 0,
> next = 4
> }
> }
>
> I read about ticketing spinlocks to understand the above output. I understood
> that 4 cpu's are waiting for the lock while someone already has it.
>
> I would like to know who has it? Is there a way I can know which process is
> acquiring a given spinlock? Is the information not stored in it?
>
> I am using crash tool to analyse the dump
crash> raw_spinlock_t
typedef struct raw_spinlock {
arch_spinlock_t raw_lock;
unsigned int magic;
unsigned int owner_cpu;
void *owner;
} raw_spinlock_t;
SIZE: 16
The owner field points to task holding the lock, if locked.
Here is what i have in one of my dumps,
crash> logbuf_lock
logbuf_lock = $6 = {
raw_lock = {
{
slock = 464001960,
tickets = {
owner = 7080,
next = 7080
}
}
},
magic = 3735899821,
owner_cpu = 4294967295,
owner = 0x
}
since lock is not acquired, it shows 0x.
Otherwise point to a task_struct *.
thanks,
Arun
>
>
> Regards,
> Manty
>
> ___
> Kernelnewbies mailing list
> Kernelnewbies@kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Ticketing spinlocks - How to get to the owner who locked the spinlock?
Hi All,
Currently I am looking into an issue which is holding the printk
lock(logbuf_lock)
logbuf_lock = {
raw_lock = {
owner = 0,
next = 4
}
}
I read about ticketing spinlocks to understand the above output. I
understood that 4 cpu's are waiting for the lock while someone already has
it.
I would like to know who has it? Is there a way I can know which process is
acquiring a given spinlock? Is the information not stored in it?
I am using crash tool to analyse the dump
Regards,
Manty
___
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Block Device Driver Tutorial
Hi Gunjan On Sun, Jan 25, 2015 at 5:13 AM, Gunjan Mehta wrote: > Any online links for study and understanding? > Or any other institute? > > On Wed, Jan 21, 2015 at 9:42 PM, Saket Sinha > wrote: >> >> Hi, >> >> I would refer Anil Pugalia from Sysplay(http://sysplay.in/) . >> >> You can find his articles at >> http://www.opensourceforu.com/author/anil-kumar-pugalia/ >> >> On Thu, Jan 22, 2015 at 10:20 AM, Gunjan Mehta >> wrote: >>> >>> Hi All, >>> >>> I am looking for some good tutorials for block device driver. >>> It would be great someone knows best training center in >>> India(Bangalore,Pune,Mumbai) >>> >>> Regards >>> >>> ___ >>> Kernelnewbies mailing list >>> Kernelnewbies@kernelnewbies.org >>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies >>> >> > > > ___ > Kernelnewbies mailing list > Kernelnewbies@kernelnewbies.org > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > Care to look at https://github.com/pranjas/block_driver -- ---P.K.S ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Block Device Driver Tutorial
Hello, First of all you need to decide whether you would like to do by yourself or like to get training from Training center.I am assuming that you have fair understanding of character device driver and Linux kernel internals because you are seeking help in block device driver. you should prefer Linux Device driver book by Alessandro *Rubini*, and Greg Kroah-Hartman. Regards Sanjeeev Sharma On Sun, Jan 25, 2015 at 5:43 PM, Gunjan Mehta wrote: > Any online links for study and understanding? > Or any other institute? > > On Wed, Jan 21, 2015 at 9:42 PM, Saket Sinha > wrote: > >> Hi, >> >> I would refer Anil Pugalia from Sysplay(http://sysplay.in/) . >> >> You can find his articles at >> http://www.opensourceforu.com/author/anil-kumar-pugalia/ >> >> On Thu, Jan 22, 2015 at 10:20 AM, Gunjan Mehta >> wrote: >> >>> Hi All, >>> >>> I am looking for some good tutorials for block device driver. >>> It would be great someone knows best training center in >>> India(Bangalore,Pune,Mumbai) >>> >>> Regards >>> >>> ___ >>> Kernelnewbies mailing list >>> Kernelnewbies@kernelnewbies.org >>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies >>> >>> >> > > ___ > Kernelnewbies mailing list > Kernelnewbies@kernelnewbies.org > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies > > ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
RE: Kernel 2.6.32.60 64bit Crash/Hung
-Original Message- From: valdis.kletni...@vt.edu [mailto:valdis.kletni...@vt.edu] Sent: Tuesday, January 27, 2015 10:15 AM To: Giridhara RP (grp) Cc: Greg KH; kernelnewbies@kernelnewbies.org Subject: Re: Kernel 2.6.32.60 64bit Crash/Hung On Tue, 27 Jan 2015 04:06:00 +, "Giridhara RP (grp)" said: > Which kernel version should I use to solve this crash/panic? That will depend on exactly why you're still running a 2.6.32 kernel from 6 years ago. Remember - the fact it's all the way up to 2.6.32.65 doesn't mean you have all the bugfixes. It only means you have all the bugfixes that qualified for the -stable side of the fence (which is a small fraction of all the fixes that have happened in the past 6 years). --- Hi Valdis, We were running 2.6.32.60 assuming it's a stable build and my assumption was wrong :(. Should I go ahead with stable 3.18.3 [2015-01-16 ]. Thanks Giri ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Kernel 2.6.32.60 64bit Crash/Hung
On Tue, 27 Jan 2015 04:06:00 +, "Giridhara RP (grp)" said: > Which kernel version should I use to solve this crash/panic? That will depend on exactly why you're still running a 2.6.32 kernel from 6 years ago. Remember - the fact it's all the way up to 2.6.32.65 doesn't mean you have all the bugfixes. It only means you have all the bugfixes that qualified for the -stable side of the fence (which is a small fraction of all the fixes that have happened in the past 6 years). pgpYEnC9YEo_G.pgp Description: PGP signature ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
RE: Kernel 2.6.32.60 64bit Crash/Hung
-Original Message- From: valdis.kletni...@vt.edu [mailto:valdis.kletni...@vt.edu] Sent: Tuesday, January 27, 2015 9:32 AM To: Giridhara RP (grp) Cc: Greg KH; kernelnewbies@kernelnewbies.org Subject: Re: Kernel 2.6.32.60 64bit Crash/Hung On Tue, 27 Jan 2015 03:52:05 +, "Giridhara RP (grp)" said: > Thanks for the quick turnaround. Which is the stable version? Kernel > 2.6.32.60 was used in RHEL 6.2 and we thought this kernel version is stable. It's stable enough that RedHat can afford to keep supporting its paying customers. You probably will be quite disillusioned if you download the .src.rpm for the kernel and start counting %patch lines in the .SPEC file.:) --- Hi Valdis, Which kernel version should I use to solve this crash/panic? Thanks Giri ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Kernel 2.6.32.60 64bit Crash/Hung
On Tue, 27 Jan 2015 03:52:05 +, "Giridhara RP (grp)" said: > Thanks for the quick turnaround. Which is the stable version? Kernel > 2.6.32.60 was used in RHEL 6.2 and we thought this kernel version is stable. It's stable enough that RedHat can afford to keep supporting its paying customers. You probably will be quite disillusioned if you download the .src.rpm for the kernel and start counting %patch lines in the .SPEC file.:) pgpn0uQ2k0Fxt.pgp Description: PGP signature ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
RE: Kernel 2.6.32.60 64bit Crash/Hung
Hi Greg, Thanks for the quick turnaround. Which is the stable version? Kernel 2.6.32.60 was used in RHEL 6.2 and we thought this kernel version is stable. Regards Giri -Original Message- From: Greg KH [mailto:g...@kroah.com] Sent: Tuesday, January 27, 2015 8:19 AM To: Giridhara RP (grp) Cc: kernelnewbies@kernelnewbies.org Subject: Re: Kernel 2.6.32.60 64bit Crash/Hung On Tue, Jan 27, 2015 at 12:06:56AM +, Giridhara RP (grp) wrote: > Hi, > > > > I am facing crash/hang with 2.6.32.60 Kernel (64bit). Note this is a _very_ old and obsolete kernel, please use a modern kernel release if you wish to get help from the community. If you are being forced to use this old version, then get support from the company you are paying to provide this kernel, as that is what their job is. Best of luck, greg k-h ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: Kernel 2.6.32.60 64bit Crash/Hung
On Tue, Jan 27, 2015 at 12:06:56AM +, Giridhara RP (grp) wrote: > Hi, > > > > I am facing crash/hang with 2.6.32.60 Kernel (64bit). Note this is a _very_ old and obsolete kernel, please use a modern kernel release if you wish to get help from the community. If you are being forced to use this old version, then get support from the company you are paying to provide this kernel, as that is what their job is. Best of luck, greg k-h ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
RE: Kernel 2.6.32.60 64bit Crash/Hung
Hello, Please note my CPU is this. Intel(R) Xeon(R) CPU E5-4650 v2 @ 2.40GHz stepping 04 From: Giridhara RP (grp) Sent: Tuesday, January 27, 2015 5:37 AM To: kernelnewbies@kernelnewbies.org Subject: Kernel 2.6.32.60 64bit Crash/Hung Hi, I am facing crash/hang with 2.6.32.60 Kernel (64bit). My maxcpus (CONFIG_NR_CPUS) count is 64 and below is the call trace. I see log saying "BUG: spinlock lockup on CPU#74". Is it a known bug? If its fixed let me know the patch. Any pointers really appreciated. Please note, if I change maxcpus=8 kernel boots successfully. Thanks Giri ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Kernel 2.6.32.60 64bit Crash/Hung
Hi, I am facing crash/hang with 2.6.32.60 Kernel (64bit). My maxcpus (CONFIG_NR_CPUS) count is 64 and below is the call trace. I see log saying "BUG: spinlock lockup on CPU#74". Is it a known bug? If its fixed let me know the patch. Any pointers really appreciated. Please note, if I change maxcpus=8 kernel boots successfully. Thanks Giri [Jan 26 06:41:35.439] [ 67.606000] Non-volatile memory driver v1.3^M [ 67.616000] Linux agpgart interface v0.103^M [ 67.637000] [drm] Initialized drm 1.1.0 20060810^M [ 67.652000] Serial: 8250/16550 driver, 4 ports, IRQ sharing enabled^M [ 68.092000] serial8250: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A^M [Jan 26 06:41:36.679] [ 68.846000] serial8250: ttyS1 at I/O 0x2f8 (irq = 3) is a 16550A^M [ 68.91] 00:03: ttyS0 at I/O 0x3f8 (irq = 4) is a 16550A^M [Jan 26 06:41:37.149] [ 69.316000] brd: module loaded^M [Jan 26 06:41:41.015] [ 73.183000] loop: module loaded^M [ 73.191000] lkdtm : Invalid command^M [ 73.20] kgdb: Registered I/O driver kgdbts.^M [Jan 26 06:42:03.435] [ 73.552000] BUG: spinlock lockup on CPU#74, swapper/1, 880c9c1d2ac0^M <=== [ 73.552000] Pid: 1, comm: swapper Not tainted 2.6.32.60 #2^M [ 73.552000] Call Trace:^M [ 73.552000] [] _raw_spin_lock+0x103/0x12b^M [ 73.552000] [] _spin_lock+0x4f/0x62^M [ 73.552000] [] ? scheduler_tick+0x49/0x166^M [ 73.552000] [] scheduler_tick+0x49/0x166^M [ 73.552000] [] update_process_times+0x42/0x52^M [ 73.552000] [] tick_periodic+0x63/0x65^M [ 73.552000] [] tick_handle_periodic+0x1e/0x6b^M [ 73.552000] [] smp_apic_timer_interrupt+0x84/0x97^M [ 73.552000] [] apic_timer_interrupt+0x13/0x20^M [ 73.552000] [] ? kgdbts_break_test+0x1/0x22^M [ 73.552000] [] ? kgdbts_break_test+0x0/0x22^M [ 73.552000] [] ? configure_kgdbts+0x1ef/0x4bb^M [ 73.552000] [] ? blk_register_region+0x28/0x2a^M [ 73.552000] [] ? init_kgdbts+0x0/0x16^M [ 73.552000] [] ? init_kgdbts+0x14/0x16^M [ 73.552000] [] ? do_one_initcall+0x57/0x157^M [ 73.552000] [] ? kernel_init+0x1f0/0x241^M [ 73.552000] [] ? child_rip+0xa/0x20^M [ 73.552000] [] ? restore_args+0x0/0x30^M [ 73.552000] [] ? kernel_init+0x0/0x241^M [ 73.552000] [] ? child_rip+0x0/0x20^M [ 73.552000] sending NMI to all CPUs:^M [ 73.552000] NMI backtrace for cpu 74^M [ 73.552000] CPU 74:^M [ 73.552000] Modules linked in:^M [ 73.552000] Pid: 1, comm: swapper Not tainted 2.6.32.60 #2 UCSB-B420-M3^M [ 73.552000] RIP: 0010:[] [] delay_loop+0x20/0x2a^M [ 73.552000] RSP: 0018:880c9c003da8 EFLAGS: 0002^M [ 73.552000] RAX: 0001d185 RBX: RCX: 0030^M [ 73.552000] RDX: 0001d19a RSI: 0010 RDI: 0001d19b^M [ 73.552000] RBP: 880c9c003da8 R08: 000b R09: 81947ad0^M [ 73.552000] R10: b4a0 R11: 81878da0 R12: 1000^M [ 73.552000] R13: 0092 R14: 0002 R15: 81947ac0^M [ 73.552000] FS: () GS:880c9c00() knlGS:^M [ 73.552000] CS: 0010 DS: 0018 ES: 0018 CR0: 8005003b^M [ 73.552000] CR2: CR3: 01001000 CR4: 000406e0^M [ 73.552000] DR0: DR1: DR2: [Jan 26 06:42:04.041] ^M [ 73.552000] DR3: DR6: 0ff0 DR7: 0400^M [ 73.552000] Call Trace:^M [ 73.552000] <#DB[1]> <> Pid: 1, comm: swapper Not tainted 2.6.32.60 #2^M [ 73.552000] Call Trace:^M [ 73.552000] [] nmi_watchdog_tick+0x111/0x24d^M [ 73.552000] [] do_nmi+0xbd/0x269^M [ 73.552000] [] nmi+0x1a/0x2c^M [ 73.552000] [] ? delay_loop+0x20/0x2a^M [ 73.552000] <> [] __delay+0xa/0xc^M [ 73.552000] [] __const_udelay+0x3c/0x3e^M [ 73.552000] [] native_safe_apic_wait_icr_idle+0x31/0x44^M [ 73.552000] [] default_send_IPI_mask_sequence_phys+0x49/0x110^M [ 73.552000] [] physflat_send_IPI_mask+0x9/0xb^M [ 73.552000] [] physflat_send_IPI_all+0x14/0x16^M [ 73.552000] [] arch_trigger_all_cpu_backtrace+0x4e/0x7c^M [ 73.552000] [] _raw_spin_lock+0x108/0x12b^M [ 73.552000] [] _spin_lock+0x4f/0x62^M [ 73.552000] [] ? scheduler_tick+0x49/0x166^M [ 73.552000] [] scheduler_tick+0x49/0x166^M [ 73.552000] [] update_process_times+0x42/0x52^M [ 73.552000] [] tick_periodic+0x63/0x65^M [ 73.552000] [] tick_handle_periodic+0x1e/0x6b^M [ 73.552000] [] smp_apic_timer_interrupt+0x84/0x97^M [ 73.552000] [] apic_timer_interrupt+0x13/0x20^M [ 73.552000] [] ? kgdbts_break_test+0x1/0x22^M [ 73.552000] [] ? kgdbts_break_test+0x0/0x22^M [ 73.552000] [] ? configure_kgdbts+0x1ef/0x4bb^M [ 73.552000] [] ? blk_register_region+0x28/0x2a^M [ 73.552000] [] ? init_kgdbts+0x0/0x16^M [ 73.552000] [] ? init_kgdbts+0x14/0x16^M [ 73.552000] [] ? do_one_initcall+0x57/0x157^M [ 73.552000] [] ? kernel_init+0x1f0/0x241^M [ 73.552000] [] ? child_rip+0xa/0x20^M [ 73.552000] [] ? restore_args+0x0/0x30^M [ 73.552000] [] ? kernel_init+0x0/0x241^M [ 73.552000] [] ? child_rip+0x0/0x20^M ___ Kernelnewbies mailing list Kernelnewbies@kernel
Re: New member
On Mon, Jan 26, 2015 at 01:46:26AM +0530, srinivas bakki wrote: > Anuz, > Sorry about that. But am also sorry that you could find a problem > as silly as that. We have a greater problem at hand. > I don't think top posting is a silly problem. I think it makes it hard to follow the thread and the archive. ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: New member
On Sun, Jan 25, 2015 at 08:10:15PM +, Anuz Pratap Singh Tomar wrote: > On Sun, Jan 25, 2015 at 7:38 PM, srinivas bakki > wrote: > > > Valdis, > >You got to be polite with people.Not everybody is as smart as > > you, but they would like to contribute. Just keep in in mind that there's > > no future for linux without such people. You cannot keep bullying everybody > > like this. > > > > Talking about polite, you are Top posting, how rude is that? I thought the post was very helpfull and not rude at all. I guess we all perceive things differently. ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: New member
On Mon, Jan 26, 2015 at 1:07 AM, Satwantjit Kaur <99satw...@gmail.com> wrote: > Hi, > I am a final year B.Tech (CSE) student from NIT Jalandhar. I like > programming and I know C and C++ programming languages. I have worked > on IPC and socket programming in C/C++. I wish to take up a project in > Linux Kernel development and contribute to it. Can somebody guide me > further? > > -- > Satwantjit Kaur > Dear all This is an interesting discussion, because it always comes up frequently. Oh and btw, thank you for mentioning my name as contributor although I am rarely joining in lately. here's my view: I whole heartedly agree with Valdis, Anuz and Robert, that to get into kernel development, you need to have "vision" first. However, as someone who grew interest at kernel development while digging into clustering, I can tell you that sometimes (or most of the times) the "vision" isn't always within us from day one. So, my trick is, (hopefully it will work with you too): think about your own OS installation (at home, office, etc_ and what bugs you most and how you want it to make it better? Maybe you want it to feel snappier when you click mouse? then perhaps you need to jump into scheduler. Maybe you feel awful when you see your OS chews memory too much (although you're not sure whether to blame OS, your web browser, your office suite etc), then maybe memory management is the way to go. Or maybe you're sick your lovely ipod isn't detected properly? then go ahead into device driver, possibly into USB (cmiiw people). But above all, you need to be careful not to feed kernel something that can be done in user space e.g security policy is mostly user space domain. Evaluate your need properly PS: guys, i am always monitoring this list passively, so I am still in :) -- regards, Mulyadi Santosa Freelance Linux trainer and consultant blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Re: New member
Am 2015-01-26 04:45, schrieb valdis.kletni...@vt.edu: > On Sun, 25 Jan 2015 16:29:00 -0800, "Robert P. J. Day" said: > >> what's wrong with ftrace? but seriously, you know what newcomers >> might get some value out of doing if they want to contribute? writing >> some really, really good documentation or tutorials. > > If somebody were to do a really good "Top 25 perf/ftrace > tricks/recipes" > for the best incantations for common use cases, they'd have a hard time > buying themselves a beer at a gathering of Linux geeks. ;) Actually, I was interested in writing something along these lines but in practice, Brendan Gregg (who is amazing) probably has you covered. http://www.brendangregg.com/linuxperf.html Cheers, Silvan ___ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
