When & how is the SCSI strategy routine called?

Hi list,

Tracing the kernel 2.6.8 code I found that scsi_mod maintains separate
request queues for each SCSI device. It uses the block layer queuing
facility to do this. What I could not find out was that once a request is
queued into a queue (for a particular device), WHEN and HOW is the strategy
routine "scsi_request_fn()" called ?

All I could find on the net was that "The kernel calls the strategy routine
when ever it believes that it is appropriate to invoke it."

Please help ... Any pointers will be highly appreciated.

TIA,

Rajat


Disclaimer:

This message and any attachment(s) contained here are information that is
confidential,proprietary to HCL Technologies and its customers, privileged
or otherwise protected by law.The information is solely intended for the
individual or the entity it is addressed to. If you are not the intended
recipient of this message, you are not authorized to read, forward,
print,retain, copy or disseminate this message or any part of it. If you
have received this e-mail in error, please notify the sender immediately by
return e-mail and delete it from your computer.


-
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:When & how is the SCSI strategy routine called?

This is my understanding.
scsi_request_fn is the scsi's request function which performs the actual
data
transfer. clean up the queue and loop back to fetch next request.

About the Question:WHEN and HOW is the strategy routine "scsi_request_fn()"
called ?

Every request queue have a few members to decide whether to call
scsi_request_fn() or not. scsi_request_fn is called after the queue is
unplugged.

request_queue_t->unplug_thresh
Whenever request queue is about to get full. which is the number of
requests after which the queue should be unplugged. unplugging of request
will in turn call scsi_request_fn().

request_queue_t->unplug_timer
After this much time the queue will be unplugged to rip the IO in turn
it will call scsi_request_fn().

Also in case of Async Direct IO the block layer will unplug the queue
after
submission of bios to queue by calling blk_run_address_space function.

scsi_request_fn will be called whenever the block layer calls queue unplug
or
queue is about to get full or after a certain unplug timer value.

Hope it helps.


-----Original Message-----
From: linux-scsi-owner@vger.kernel.org
[mailto:linux-scsi-owner@vger.kernel.org]On Behalf Of Rajat Jain, Noida
Sent: Tuesday, March 22, 2005 8:31 AM
To: linux-scsi@vger.kernel.org; linux-newbie@vger.kernel.org
Cc: Rajat Jain, Noida
Subject: When & how is the SCSI strategy routine called?


Hi list,

Tracing the kernel 2.6.8 code I found that scsi_mod maintains separate
request queues for each SCSI device. It uses the block layer queuing
facility to do this. What I could not find out was that once a request is
queued into a queue (for a particular device), WHEN and HOW is the strategy
routine "scsi_request_fn()" called ?

All I could find on the net was that "The kernel calls the strategy routine
when ever it believes that it is appropriate to invoke it."

Please help ... Any pointers will be highly appreciated.

TIA,

Rajat



-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

RE: When & how is the SCSI strategy routine called?

Thanks lk.

It cleared a lot of things.

How ever, what I am wondering is:

- Does each device queue have a corresponding thread that is woken up when
it is time to call the request function? And the request function is called
in the context of this thread?

- I explored and found that there is some "worker thread" kblockd that does
this task of monitoring the queue and calling the request function. I want
to know whether there is a separate thread for each queue (device) , or just
one for all the block devices?

TIA,

Rajat


-----Original Message-----
From: lk [mailto:linux_kernel@patni.com]
Sent: Wednesday, March 23, 2005 1:29 PM
To: linux-scsi@vger.kernel.org; rajatj@noida.hcltech.com;
linux-newbie@vger.kernel.org
Subject: Re:When & how is the SCSI strategy routine called?

This is my understanding.
scsi_request_fn is the scsi's request function which performs the actual
data transfer. clean up the queue and loop back to fetch next request.

About the Question:WHEN and HOW is the strategy routine "scsi_request_fn()"
called ?

Every request queue have a few members to decide whether to call
scsi_request_fn() or not. scsi_request_fn is called after the queue is
unplugged.

request_queue_t->unplug_thresh
Whenever request queue is about to get full. which is the number of
requests after which the queue should be unplugged. unplugging of request
will in turn call scsi_request_fn().

request_queue_t->unplug_timer
After this much time the queue will be unplugged to rip the IO in turn
it will call scsi_request_fn().

Also in case of Async Direct IO the block layer will unplug the queue
after
submission of bios to queue by calling blk_run_address_space function.

scsi_request_fn will be called whenever the block layer calls queue unplug
or
queue is about to get full or after a certain unplug timer value.

Hope it helps.


-----Original Message-----
From: linux-scsi-owner@vger.kernel.org
[mailto:linux-scsi-owner@vger.kernel.org]On Behalf Of Rajat Jain, Noida
Sent: Tuesday, March 22, 2005 8:31 AM
To: linux-scsi@vger.kernel.org; linux-newbie@vger.kernel.org
Cc: Rajat Jain, Noida
Subject: When & how is the SCSI strategy routine called?


Hi list,

Tracing the kernel 2.6.8 code I found that scsi_mod maintains separate
request queues for each SCSI device. It uses the block layer queuing
facility to do this. What I could not find out was that once a request is
queued into a queue (for a particular device), WHEN and HOW is the strategy
routine "scsi_request_fn()" called ?

All I could find on the net was that "The kernel calls the strategy routine
when ever it believes that it is appropriate to invoke it."

Please help ... Any pointers will be highly appreciated.

TIA,

Rajat




Disclaimer:

This message and any attachment(s) contained here are information that is
confidential,proprietary to HCL Technologies and its customers, privileged
or otherwise protected by law.The information is solely intended for the
individual or the entity it is addressed to. If you are not the intended
recipient of this message, you are not authorized to read, forward,
print,retain, copy or disseminate this message or any part of it. If you
have received this e-mail in error, please notify the sender immediately by
return e-mail and delete it from your computer.


-
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: When & how is the SCSI strategy routine called?

There would be a single work queue (kblockd) for all block
devices on which deffered work of different block
device queues will be queued after the unplug_timer.


----- Original Message -----
From: "Rajat Jain, Noida"
To: "lk" ; ;
;
Sent: Wednesday, March 23, 2005 12:49 AM
Subject: RE: When & how is the SCSI strategy routine called?


>
>
> Thanks lk.
>
> It cleared a lot of things.
>
> How ever, what I am wondering is:
>
> - Does each device queue have a corresponding thread that is woken up when
> it is time to call the request function? And the request function is
called
> in the context of this thread?
>
> - I explored and found that there is some "worker thread" kblockd that
does
> this task of monitoring the queue and calling the request function. I want
> to know whether there is a separate thread for each queue (device) , or
just
> one for all the block devices?
>
> TIA,
>
> Rajat
>
>
> -----Original Message-----
> From: lk [mailto:linux_kernel@patni.com]
> Sent: Wednesday, March 23, 2005 1:29 PM
> To: linux-scsi@vger.kernel.org; rajatj@noida.hcltech.com;
> linux-newbie@vger.kernel.org
> Subject: Re:When & how is the SCSI strategy routine called?
>
> This is my understanding.
> scsi_request_fn is the scsi's request function which performs the actual
> data transfer. clean up the queue and loop back to fetch next request.
>
> About the Question:WHEN and HOW is the strategy routine
"scsi_request_fn()"
> called ?
>
> Every request queue have a few members to decide whether to call
> scsi_request_fn() or not. scsi_request_fn is called after the queue is
> unplugged.
>
> request_queue_t->unplug_thresh
> Whenever request queue is about to get full. which is the number of
> requests after which the queue should be unplugged. unplugging of
request
> will in turn call scsi_request_fn().
>
> request_queue_t->unplug_timer
> After this much time the queue will be unplugged to rip the IO in turn
> it will call scsi_request_fn().
>
> Also in case of Async Direct IO the block layer will unplug the queue
> after
> submission of bios to queue by calling blk_run_address_space function.
>
> scsi_request_fn will be called whenever the block layer calls queue
unplug
> or
> queue is about to get full or after a certain unplug timer value.
>
> Hope it helps.
>
>
> -----Original Message-----
> From: linux-scsi-owner@vger.kernel.org
> [mailto:linux-scsi-owner@vger.kernel.org]On Behalf Of Rajat Jain, Noida
> Sent: Tuesday, March 22, 2005 8:31 AM
> To: linux-scsi@vger.kernel.org; linux-newbie@vger.kernel.org
> Cc: Rajat Jain, Noida
> Subject: When & how is the SCSI strategy routine called?
>
>
> Hi list,
>
> Tracing the kernel 2.6.8 code I found that scsi_mod maintains separate
> request queues for each SCSI device. It uses the block layer queuing
> facility to do this. What I could not find out was that once a request is
> queued into a queue (for a particular device), WHEN and HOW is the
strategy
> routine "scsi_request_fn()" called ?
>
> All I could find on the net was that "The kernel calls the strategy
routine
> when ever it believes that it is appropriate to invoke it."
>
> Please help ... Any pointers will be highly appreciated.
>
> TIA,
>
> Rajat
>
>
>
>
> Disclaimer:
>
> This message and any attachment(s) contained here are information that is
> confidential,proprietary to HCL Technologies and its customers, privileged
> or otherwise protected by law.The information is solely intended for the
> individual or the entity it is addressed to. If you are not the intended
> recipient of this message, you are not authorized to read, forward,
> print,retain, copy or disseminate this message or any part of it. If you
> have received this e-mail in error, please notify the sender immediately
by
> return e-mail and delete it from your computer.
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html


-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: When & how is the SCSI strategy routine called?

On Wed, Mar 23 2005, Rajat Jain, Noida wrote:
>
>
> Thanks lk.
>
> It cleared a lot of things.
>
> How ever, what I am wondering is:
>
> - Does each device queue have a corresponding thread that is woken up when
> it is time to call the request function? And the request function is called
> in the context of this thread?
>
> - I explored and found that there is some "worker thread" kblockd that does
> this task of monitoring the queue and calling the request function. I want
> to know whether there is a separate thread for each queue (device) , or just
> one for all the block devices?

There is on kblockd thread per cpu for all queues. But note that the
request function is only invoked through that, if you get the unplug
threshold kicking in _or_ the io scheduler uses it to awake the queue.
The normal path for unplugs (this is what invokes the request handler)
is exceeding unplug_thresh or somebody calling blk_run_backing_dev()
either directly or from blk_run_address_space().

Normally drivers will re-invoke the request handler themselves, when
requests complete. So once the unplug has happened, if there is a
continual influx of io for that queue, the queue will not get replugged
and it is the drivers responsibility to reinvoke the request handler.

--
Jens Axboe

-
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