Why is Linux not RTOS?

Why is Linux not RTOS?

[Rick Brown]

Hi,

Why is Linux kernel considered a (hard) realtime OS? I already
understand the basic reason is that the linux kernel does not
guarantee that a task will be completed on time. But I would
appreciate answers in terms of more of kernel jargons.

What stops us from classifying kernel as hard RTOS? Is it because at
times the kernel is non-preemptive (for e.g. while holding spinlocks)?
Has it got something to do with interrupt latency / scheduling latency
etc?

Is the behaviour of the kernel (when it is preemptive) similar to hard
real time OS ??

Thanks,

Rick
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[John Anthony Kazos Jr.]

> Why is Linux kernel considered a (hard) realtime OS? I already
> understand the basic reason is that the linux kernel does not
> guarantee that a task will be completed on time. But I would
> appreciate answers in terms of more of kernel jargons.
> 
> What stops us from classifying kernel as hard RTOS? Is it because at
> times the kernel is non-preemptive (for e.g. while holding spinlocks)?
> Has it got something to do with interrupt latency / scheduling latency
> etc?
> 
> Is the behaviour of the kernel (when it is preemptive) similar to hard
> real time OS ??

From what I've read, the preemption in the kernel is mostly a hack because 
nobody's found a way to reduce the latency of certain long functions yet.

And there is a separate RTOS version called Real-Time Linux or RTL...I 
think. Try googling. It's used in life-support machinery and so forth.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Rick Brown]

On 4/4/07, John Anthony Kazos Jr. <jakj@j-a-k-j.com> wrote:
> > Why is Linux kernel considered a (hard) realtime OS? I already
> > understand the basic reason is that the linux kernel does not
> > guarantee that a task will be completed on time. But I would
> > appreciate answers in terms of more of kernel jargons.
> >
> > What stops us from classifying kernel as hard RTOS? Is it because at
> > times the kernel is non-preemptive (for e.g. while holding spinlocks)?
> > Has it got something to do with interrupt latency / scheduling latency
> > etc?
> >
> > Is the behaviour of the kernel (when it is preemptive) similar to hard
> > real time OS ??
>
> From what I've read, the preemption in the kernel is mostly a hack because
> nobody's found a way to reduce the latency of certain long functions yet.
>
> And there is a separate RTOS version called Real-Time Linux or RTL...I
> think. Try googling. It's used in life-support machinery and so forth.

Yes, I'm aware that there are quite a lot of different patches /
projects aimed at giving better RTOS behaviour ... but my question was
aimed at vanilla kernel.

Thanks,

Rick
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Daniel Cheng]

Rick Brown wrote:
> Yes, I'm aware that there are quite a lot of different patches /
> projects aimed at giving better RTOS behaviour ... but my question was
> aimed at vanilla kernel.

An OS consider RealTime or not, does not depends on
how low the latency is. It depends on how predictable
the latency is.

> 
> Thanks,
> 
> Rick


-- 
This space was intended to be left blank.


--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@nl.linux.org
Please read the FAQ at http://kernelnewbies.org/FAQ

Re: Why is Linux not RTOS?

[Tzahi Fadida]

On Wednesday 04 April 2007 18:21:16 Daniel Cheng wrote:
> Rick Brown wrote:
> > Yes, I'm aware that there are quite a lot of different patches /
> > projects aimed at giving better RTOS behaviour ... but my question was
> > aimed at vanilla kernel.
>
> An OS consider RealTime or not, does not depends on
> how low the latency is. It depends on how predictable
> the latency is.

What is this i hear about PREEMT_RT from VirtualLogix:
http://www.wirelessiq.com/content/newsfeed/9867.html

Perhaps this is what the poster seeks...



-- 
Regards,
        Tzahi.
--
Tzahi Fadida
Blog: http://tzahi.blogsite.org | Home Site: http://tzahi.webhop.info
WARNING TO SPAMMERS:  see at 
http://members.lycos.co.uk/my2nis/spamwarning.html
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[sandeep lahane]

On 4/5/07, Tzahi Fadida <Tzahi.ML2@gmail.com> wrote:
> On Wednesday 04 April 2007 18:21:16 Daniel Cheng wrote:
> > Rick Brown wrote:
> > > Yes, I'm aware that there are quite a lot of different patches /
> > > projects aimed at giving better RTOS behaviour ... but my question was
> > > aimed at vanilla kernel.
> >
> > An OS consider RealTime or not, does not depends on
> > how low the latency is. It depends on how predictable
> > the latency is.
>
> What is this i hear about PREEMT_RT from VirtualLogix:
> http://www.wirelessiq.com/content/newsfeed/9867.html
>
> Perhaps this is what the poster seeks...
>
>
>
> --
> Regards,
> Tzahi.
> --
> Tzahi Fadida
> Blog: http://tzahi.blogsite.org | Home Site: http://tzahi.webhop.info
> WARNING TO SPAMMERS: see at
> http://members.lycos.co.uk/my2nis/spamwarning.html
>
> --
> To unsubscribe from this list: send an email with
> "unsubscribe kernelnewbies" to ecartis@nl.linux.org
> Please read the FAQ at http://kernelnewbies.org/FAQ
>
>

I think what virtuallogix is offering is a real time virtualization
software. They have a paravirtualization based approach using which
guest OSes like RTOS or other rich OSes can be run simultaneously on
an embedded platform. These guest OSes can communicate using inter OS
communication mechanisms. They are partitioning resources which can be
partitioned like system RAM and resources like CPU, MMU and interrupt
controller are virtualized since they can't be partitioned. So
basically, what they are doing is almost totally irrelevant with this
question, since they are not trying to make Linux a RTOS, rather they
are making Linux and other guest OSes co-exist with RTOSes
simultaneously. Please CMIIW.

-- 

Regards,
Sandeep.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Raseel Bhagat]

Hi Sandeep,

On 4/5/07, sandeep lahane <sandeep.lahane@gmail.com> wrote:

>They have a paravirtualization based approach using which
> guest OSes like RTOS or other rich OSes can be run simultaneously on
> an embedded platform. These guest OSes can communicate using inter OS
> communication mechanisms. They are partitioning resources which can be
> partitioned like system RAM and resources like CPU, MMU and interrupt
> controller are virtualized since they can't be partitioned. So
> basically, what they are doing is almost totally irrelevant with this
> question, since they are not trying to make Linux a RTOS, rather they
> are making Linux and other guest OSes co-exist with RTOSes
> simultaneously. Please CMIIW.
>

I completely concur with you. And it makes lot of sense too.
For example RTLinux (Real time Linux) from FSMLabs is another such approach.
They have a micro-kernel , which is basically a core real tie\me
kernel, which sits on top of the vanilla linux kernel. This way, all
the real time tasks are handled by the Microkernel during whcih time
Linux kernel runs as an idle process. Only when no RT tasks are
present, the vanilla Linux kernel executes all the non-RT tasks.
This way, RT behaviour is accomplished without having to modify the
core Linux kernel.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[sandeep lahane]

On 4/5/07, Raseel Bhagat <raseelbhagat@gmail.com> wrote:
> Hi Sandeep,
>
> On 4/5/07, sandeep lahane <sandeep.lahane@gmail.com> wrote:
>
> >They have a paravirtualization based approach using which
> > guest OSes like RTOS or other rich OSes can be run simultaneously on
> > an embedded platform. These guest OSes can communicate using inter OS
> > communication mechanisms. They are partitioning resources which can be
> > partitioned like system RAM and resources like CPU, MMU and interrupt
> > controller are virtualized since they can't be partitioned. So
> > basically, what they are doing is almost totally irrelevant with this
> > question, since they are not trying to make Linux a RTOS, rather they
> > are making Linux and other guest OSes co-exist with RTOSes
> > simultaneously. Please CMIIW.
> >
>
> I completely concur with you. And it makes lot of sense too.
> For example RTLinux (Real time Linux) from FSMLabs is another such approach.
> They have a micro-kernel , which is basically a core real tie\me
> kernel, which sits on top of the vanilla linux kernel. This way, all
> the real time tasks are handled by the Microkernel during whcih time
> Linux kernel runs as an idle process. Only when no RT tasks are
> present, the vanilla Linux kernel executes all the non-RT tasks.
> This way, RT behaviour is accomplished without having to modify the
> core Linux kernel.
>
> --
> To unsubscribe from this list: send an email with
> "unsubscribe kernelnewbies" to ecartis@nl.linux.org
> Please read the FAQ at http://kernelnewbies.org/FAQ
>
>

Yes, may be the poster is looking for RTLinux kind of thing.
-- 

Regards,
Sandeep.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Mark Hounschell]

sandeep lahane wrote:
> On 4/5/07, Raseel Bhagat <raseelbhagat@gmail.com> wrote:
>> Hi Sandeep,
>>
>> On 4/5/07, sandeep lahane <sandeep.lahane@gmail.com> wrote:
>>
>> >They have a paravirtualization based approach using which
>> > guest OSes like RTOS or other rich OSes can be run simultaneously on
>> > an embedded platform. These guest OSes can communicate using inter OS
>> > communication mechanisms. They are partitioning resources which can be
>> > partitioned like system RAM and resources like CPU, MMU and interrupt
>> > controller are virtualized since they can't be partitioned. So
>> > basically, what they are doing is almost totally irrelevant with this
>> > question, since they are not trying to make Linux a RTOS, rather they
>> > are making Linux and other guest OSes co-exist with RTOSes
>> > simultaneously. Please CMIIW.
>> >
>>
>> I completely concur with you. And it makes lot of sense too.
>> For example RTLinux (Real time Linux) from FSMLabs is another such
>> approach.
>> They have a micro-kernel , which is basically a core real tie\me
>> kernel, which sits on top of the vanilla linux kernel. This way, all
>> the real time tasks are handled by the Microkernel during whcih time
>> Linux kernel runs as an idle process. Only when no RT tasks are
>> present, the vanilla Linux kernel executes all the non-RT tasks.
>> This way, RT behaviour is accomplished without having to modify the
>> core Linux kernel.
>>
>> -- 
>> To unsubscribe from this list: send an email with
>> "unsubscribe kernelnewbies" to ecartis@nl.linux.org
>> Please read the FAQ at http://kernelnewbies.org/FAQ
>>
>>
> 
> Yes, may be the poster is looking for RTLinux kind of thing.

The vanilla Linux kernel can and is already being used in RT environments.
A dual processor box when configured properly can provide a very
deterministic env for a properly written RT application. The trick is to
realize that the 'box' must be dedicated to that application and that application
alone.

With Ingos work in progress at http://people.redhat.com/mingo/realtime-preempt/
that is becoming of less importance however.

Mark
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Raseel Bhagat]

Hi Tzahi,

On 4/5/07, Tzahi Fadida <Tzahi.ML2@gmail.com> wrote:
> What is this i hear about PREEMT_RT from VirtualLogix:
> http://www.wirelessiq.com/content/newsfeed/9867.html

I think the author is trying to mention the different innovative
solutions to the latency issue in the GPOS Linux Kernel. PREEMPT_RT ,
IIRC  was a Real Time patch by Ingo Molar.
VirtualLogix's Virtual OS is another such solution, etc.

Thanks,
Raseel
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Rodrigo Rubira Branco]

[-- Attachment #1: Type: text/plain, Size: 1638 bytes --]

Ok, let's clarify somethings ;)


First of all I really recommend this links:
http://www.linuxjournal.com/article/9361
http://rt.wiki.kernel.org/

The real-time implementations previously commented in the list relies on
"interrupt shielding" (forcing interrupts to run on a reserved CPU) and
"CPU pinning" (forcing real-time tasks to only run on a CPU which
doesn't run any other tasks), so we can't say it's a 'real' real-time ;)
(it does not take the best advantage of multiple cpu's for example and
needs more than just one in anyway).

Legal notice: It's just my personal contribution.


P.S.:  Sorry my delay to answer that ;)


On Thu, 2007-04-05 at 15:37 +0530, Raseel Bhagat wrote:
> Hi Tzahi,
> 
> On 4/5/07, Tzahi Fadida <Tzahi.ML2@gmail.com> wrote:
> > What is this i hear about PREEMT_RT from VirtualLogix:
> > http://www.wirelessiq.com/content/newsfeed/9867.html
> 
> I think the author is trying to mention the different innovative
> solutions to the latency issue in the GPOS Linux Kernel. PREEMPT_RT ,
> IIRC  was a Real Time patch by Ingo Molar.
> VirtualLogix's Virtual OS is another such solution, etc.
> 
> Thanks,
> Raseel
> -
> To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.linux-learn.org/faqs
-- 
Rodrigo Rubira Branco
Software Engineer 
Advanced Linux Response Team (ALRT) / Linux on Power Toolchain
IBM Linux Technology Center (IBM/LTC)
rrbranco@br.ibm.com

GPG KeyID: 1FCEDEA1

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Why is Linux not RTOS?

[K.R. Foley]

Rodrigo Rubira Branco wrote:
> Ok, let's clarify somethings ;)

Please don't top post.

> 
> 
> First of all I really recommend this links:
> http://www.linuxjournal.com/article/9361
> http://rt.wiki.kernel.org/
> 
> The real-time implementations previously commented in the list relies on
> "interrupt shielding" (forcing interrupts to run on a reserved CPU) and
> "CPU pinning" (forcing real-time tasks to only run on a CPU which
> doesn't run any other tasks), so we can't say it's a 'real' real-time ;)
> (it does not take the best advantage of multiple cpu's for example and
> needs more than just one in anyway).

The above description is inaccurate when talking about about Ingo
Molnar's PREEMPT_RT or Real-Time patch. The -rt patchset, as it is also
called, takes full advantage of SMP capabilities for both interrupts and
Real-Time tasks.

> 
> Legal notice: It's just my personal contribution.
> 
> 
> P.S.:  Sorry my delay to answer that ;)
> 
> 
> On Thu, 2007-04-05 at 15:37 +0530, Raseel Bhagat wrote:
>> Hi Tzahi,
>>
>> On 4/5/07, Tzahi Fadida <Tzahi.ML2@gmail.com> wrote:
>>> What is this i hear about PREEMT_RT from VirtualLogix:
>>> http://www.wirelessiq.com/content/newsfeed/9867.html
>> I think the author is trying to mention the different innovative
>> solutions to the latency issue in the GPOS Linux Kernel. PREEMPT_RT ,
>> IIRC  was a Real Time patch by Ingo Molar.
>> VirtualLogix's Virtual OS is another such solution, etc.
>>
>> Thanks,
>> Raseel
>> -
>> To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://www.linux-learn.org/faqs


-- 
	kr

--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@nl.linux.org
Please read the FAQ at http://kernelnewbies.org/FAQ

Re: Why is Linux not RTOS?

[Pradeep]

> Hi,
>
> Why is Linux kernel considered a (hard) realtime OS? I already

Linux kernel is never considered a real time OS.
A real time OS should be predictable.The general linux kernel is not 
predictable because of many reasons.
A few are it uses virtual memory. when there is virtual memory there will be 
page swaps which are un predictable.
And the linux scheduler is not a deterministic one.
There is a scheduler project which provides a deternistic behaviour to the 
scheduler.This scheduler is called O(1) scheduler.
This gives a constant scheduling latency. This is used in many embedded 
linux to make it a real time OS. Ex: motavista

> understand the basic reason is that the linux kernel does not
> guarantee that a task will be completed on time. But I would
> appreciate answers in terms of more of kernel jargons.
>
> What stops us from classifying kernel as hard RTOS? Is it because at

An RTOS is just a facilitator to build Real time systems. A Real time system 
developer should analyse a RTOS wether is will be useful for his system.
If some body is holding spin lock and making it non-preemptive its a bad 
design. So even a RTOS linux has spin locks. Its the designer who should 
take care.

> times the kernel is non-preemptive (for e.g. while holding spinlocks)?
> Has it got something to do with interrupt latency / scheduling latency
> etc?
>
> Is the behaviour of the kernel (when it is preemptive) similar to hard
> real time OS ??
>
> Thanks,
>
> Rick
> -
> To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.linux-learn.org/faqs 

-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Raseel Bhagat]

Hi

On 4/5/07, Pradeep <pradeep.annavarapu@moschip.com> wrote:
>
> > Hi,
> >
> > Why is Linux kernel considered a (hard) realtime OS? I already
>
> Linux kernel is never considered a real time OS.
> A real time OS should be predictable.The general linux kernel is not
> predictable because of many reasons.

To ellaborate more, any RTOS HAS TO be deterministic , i.e, it should
have a deterministic latecy, a predictable jitter which will never
cross its threshold even during heavy loads.


> > What stops us from classifying kernel as hard RTOS? Is it because at
>
> An RTOS is just a facilitator to build Real time systems. A Real time system
> developer should analyse a RTOS wether is will be useful for his system.
> If some body is holding spin lock and making it non-preemptive its a bad
> design. So even a RTOS linux has spin locks. Its the designer who should
> take care.

Basically , a GPOS (General Purpose OS) is build to optimize the
average-response time. Whereas, an RTOS is designed to optimize the
worst-case response time. Hence non-RT tasks, in case of an RTOS , are
given the least prioirity where as the deadlines for RT tasks are not
to be crossed under ANY circumstances.

Thanks,
Raseel.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Why is Linux not RTOS?

[Pharaoh .]

On 4/5/07, Raseel Bhagat <raseelbhagat@gmail.com> wrote:
> Hi
>
> On 4/5/07, Pradeep <pradeep.annavarapu@moschip.com> wrote:
> >
> > > Hi,
> > >
> > > Why is Linux kernel considered a (hard) realtime OS? I already
> >
> > Linux kernel is never considered a real time OS.
> > A real time OS should be predictable.The general linux kernel is not
> > predictable because of many reasons.
>
> To ellaborate more, any RTOS HAS TO be deterministic , i.e, it should
> have a deterministic latecy, a predictable jitter which will never
> cross its threshold even during heavy loads.
>
>
> > > What stops us from classifying kernel as hard RTOS? Is it because at
> >
> > An RTOS is just a facilitator to build Real time systems. A Real time
> system
> > developer should analyse a RTOS wether is will be useful for his system.
> > If some body is holding spin lock and making it non-preemptive its a bad
> > design. So even a RTOS linux has spin locks. Its the designer who should
> > take care.
>
> Basically , a GPOS (General Purpose OS) is build to optimize the
> average-response time. Whereas, an RTOS is designed to optimize the
> worst-case response time. Hence non-RT tasks, in case of an RTOS , are
> given the least prioirity where as the deadlines for RT tasks are not
> to be crossed under ANY circumstances.
>
> Thanks,
> Raseel.
>
> --
> To unsubscribe from this list: send an email with
> "unsubscribe kernelnewbies" to ecartis@nl.linux.org
> Please read the FAQ at http://kernelnewbies.org/FAQ
>
>


Why Linux is not RTOS?

It was not designed to be a RTOS, thats why.

-Pharaoh.
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Not Initialize the shrinker->list after kmalloc() in mm/vmscan.c(V2.6.12)?

[qingxiaoming]

Dear all:
I am reading the function set_shrinker() in mm/vmscan.c of V2.6.12, and I
have a question about the not initialization of list_head,

/*
 * Add a shrinker callback to be called from the vm
 */
struct shrinker *set_shrinker(int seeks, shrinker_t theshrinker)
{
        struct shrinker *shrinker;

        shrinker = kmalloc(sizeof(*shrinker), GFP_KERNEL);
        if (shrinker) {
	        shrinker->shrinker = theshrinker;
	        shrinker->seeks = seeks;
	        shrinker->nr = 0;
	        down_write(&shrinker_rwsem);
	        list_add_tail(&shrinker->list, &shrinker_list);
	        up_write(&shrinker_rwsem);
	}
	return shrinker;
}

As above, the shrinker is allocated from kmalloc, coming from slab
allocator,the list in shrinker is not initialized, directly list_add_tail()
to shrinker_list, don't need to INIT_LIST_HEAD(shrinker->list)?

Thanks,
Best Regards!
Xiaoming.Qing


-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs

Re: Not Initialize the shrinker->list after kmalloc() in mm/vmscan.c(V2.6.12)?

[Rajat Jain]

Hi Qing,

On 4/5/07, qingxiaoming <qing_xiaoming@anyka.com> wrote:
> Dear all:
> I am reading the function set_shrinker() in mm/vmscan.c of V2.6.12, and I
> have a question about the not initialization of list_head,
>
> /*
>  * Add a shrinker callback to be called from the vm
>  */
> struct shrinker *set_shrinker(int seeks, shrinker_t theshrinker)
> {
>        struct shrinker *shrinker;
>
>        shrinker = kmalloc(sizeof(*shrinker), GFP_KERNEL);
>        if (shrinker) {
>                shrinker->shrinker = theshrinker;
>                shrinker->seeks = seeks;
>                shrinker->nr = 0;
>                down_write(&shrinker_rwsem);
>                list_add_tail(&shrinker->list, &shrinker_list);
>                up_write(&shrinker_rwsem);
>        }
>        return shrinker;
> }
>
> As above, the shrinker is allocated from kmalloc, coming from slab
> allocator,the list in shrinker is not initialized, directly list_add_tail()
> to shrinker_list, don't need to INIT_LIST_HEAD(shrinker->list)?
>

IMHO, the linked list is not headed by shrinker->list, rather by
shrinker_list. So the statement

list_add_tail(&shrinker->list, &shrinker_list);

is actually adding shrinker to the list shrinker_list. The shrinker->
list is merely a pointer to help in atttaching the node to the list.

Thanks,

Rajat
-
To unsubscribe from this list: send the line "unsubscribe linux-newbie" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.linux-learn.org/faqs