* [PATCH 0/2] delay initializing queue for request-based dm
@ 2009-08-08 4:55 ` Nikanth Karthikesan
0 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-08 4:55 UTC (permalink / raw)
To: Jens Axboe, Alasdair G Kergon; +Cc: Kiyoshi Ueda, dm-devel, linux-kernel
Hi
Currently for device-mapper devices, we initialize the mempool for requests
and elevator at device creation time. But only request-based devices use the
elevator and mempool. These patches change it to allocate the mempool and
initialize the elevator only for request-based devices. This is done by
allocating a queue at device creation time, but delaying the initialization of
elevator and allocation of mempool till the first table load time, when the
type of device would be known.
[PATCH 1/2] Allow delaying initialization of queue after allocation
[PATCH 2/2] Initialize mempool and elevator only for request-based dm devices
Another approach would be to free the mempool and detach elevator, if it would
be a bio-based device. But it is better to avoid initialization itself.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 0/2] delay initializing queue for request-based dm
@ 2009-08-08 4:55 ` Nikanth Karthikesan
0 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-08 4:55 UTC (permalink / raw)
To: Jens Axboe, Alasdair G Kergon
Cc: Kiyoshi Ueda, Hannes Reinecke, dm-devel, linux-kernel
Hi
Currently for device-mapper devices, we initialize the mempool for requests
and elevator at device creation time. But only request-based devices use the
elevator and mempool. These patches change it to allocate the mempool and
initialize the elevator only for request-based devices. This is done by
allocating a queue at device creation time, but delaying the initialization of
elevator and allocation of mempool till the first table load time, when the
type of device would be known.
[PATCH 1/2] Allow delaying initialization of queue after allocation
[PATCH 2/2] Initialize mempool and elevator only for request-based dm devices
Another approach would be to free the mempool and detach elevator, if it would
be a bio-based device. But it is better to avoid initialization itself.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH-v2 0/2] delay initializing queue for request-based dm
2009-08-08 4:55 ` Nikanth Karthikesan
@ 2009-08-10 10:48 ` Nikanth Karthikesan
-1 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-10 10:48 UTC (permalink / raw)
To: Jens Axboe
Cc: Kiyoshi Ueda, Mike Snitzer, linux-kernel, dm-devel,
Alasdair G Kergon
On Saturday 08 August 2009 10:25:53 Nikanth Karthikesan wrote:
> Hi
>
> Currently for device-mapper devices, we initialize the mempool for requests
> and elevator at device creation time. But only request-based devices use
> the elevator and mempool. These patches change it to allocate the mempool
> and initialize the elevator only for request-based devices. This is done by
> allocating a queue at device creation time, but delaying the initialization
> of elevator and allocation of mempool till the first table load time, when
> the type of device would be known.
>
> [PATCH 1/2] Allow delaying initialization of queue after allocation
>
> [PATCH 2/2] Initialize mempool and elevator only for request-based dm
> devices
>
> Another approach would be to free the mempool and detach elevator, if it
> would be a bio-based device. But it is better to avoid initialization
> itself.
>
Changes since v1:
1. Refreshed on top of a4e7d46407d73f35d217013b363b79a8f8eafcaa
2. Added comments suggested by Mike Snitzer.
3. Cleaned up the second patch as suggested by Mike Snitzer.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH-v2 0/2] delay initializing queue for request-based dm
@ 2009-08-10 10:48 ` Nikanth Karthikesan
0 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-10 10:48 UTC (permalink / raw)
To: Jens Axboe
Cc: Alasdair G Kergon, Kiyoshi Ueda, Hannes Reinecke, dm-devel,
linux-kernel, Mike Snitzer
On Saturday 08 August 2009 10:25:53 Nikanth Karthikesan wrote:
> Hi
>
> Currently for device-mapper devices, we initialize the mempool for requests
> and elevator at device creation time. But only request-based devices use
> the elevator and mempool. These patches change it to allocate the mempool
> and initialize the elevator only for request-based devices. This is done by
> allocating a queue at device creation time, but delaying the initialization
> of elevator and allocation of mempool till the first table load time, when
> the type of device would be known.
>
> [PATCH 1/2] Allow delaying initialization of queue after allocation
>
> [PATCH 2/2] Initialize mempool and elevator only for request-based dm
> devices
>
> Another approach would be to free the mempool and detach elevator, if it
> would be a bio-based device. But it is better to avoid initialization
> itself.
>
Changes since v1:
1. Refreshed on top of a4e7d46407d73f35d217013b363b79a8f8eafcaa
2. Added comments suggested by Mike Snitzer.
3. Cleaned up the second patch as suggested by Mike Snitzer.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH-v3 0/2] delay initializing queue for request-based dm
2009-08-10 10:48 ` Nikanth Karthikesan
@ 2009-08-11 9:32 ` Nikanth Karthikesan
-1 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-11 9:32 UTC (permalink / raw)
To: Jens Axboe
Cc: Kiyoshi Ueda, Mike Snitzer, linux-kernel, dm-devel,
Alasdair G Kergon
On Monday 10 August 2009 16:18:12 Nikanth Karthikesan wrote:
> On Saturday 08 August 2009 10:25:53 Nikanth Karthikesan wrote:
> > Hi
> >
> > Currently for device-mapper devices, we initialize the mempool for
> > requests and elevator at device creation time. But only request-based
> > devices use the elevator and mempool. These patches change it to allocate
> > the mempool and initialize the elevator only for request-based devices.
> > This is done by allocating a queue at device creation time, but delaying
> > the initialization of elevator and allocation of mempool till the first
> > table load time, when the type of device would be known.
> >
> > [PATCH 1/2] Allow delaying initialization of queue after allocation
> >
> > [PATCH 2/2] Initialize mempool and elevator only for request-based dm
> > devices
> >
> > Another approach would be to free the mempool and detach elevator, if it
> > would be a bio-based device. But it is better to avoid initialization
> > itself.
>
> Changes since v1:
> 1. Refreshed on top of a4e7d46407d73f35d217013b363b79a8f8eafcaa
> 2. Added comments suggested by Mike Snitzer.
> 3. Cleaned up the second patch as suggested by Mike Snitzer.
>
Changes since v2:
1. Add comments suggested by Kiyoshi Ueda
2. Export symbol elv_register_queue() to modules.
3. Call elv_register_queue() after initializing the queue to use the elevator.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH-v3 0/2] delay initializing queue for request-based dm
@ 2009-08-11 9:32 ` Nikanth Karthikesan
0 siblings, 0 replies; 6+ messages in thread
From: Nikanth Karthikesan @ 2009-08-11 9:32 UTC (permalink / raw)
To: Jens Axboe
Cc: Alasdair G Kergon, Kiyoshi Ueda, Hannes Reinecke, dm-devel,
linux-kernel, Mike Snitzer
On Monday 10 August 2009 16:18:12 Nikanth Karthikesan wrote:
> On Saturday 08 August 2009 10:25:53 Nikanth Karthikesan wrote:
> > Hi
> >
> > Currently for device-mapper devices, we initialize the mempool for
> > requests and elevator at device creation time. But only request-based
> > devices use the elevator and mempool. These patches change it to allocate
> > the mempool and initialize the elevator only for request-based devices.
> > This is done by allocating a queue at device creation time, but delaying
> > the initialization of elevator and allocation of mempool till the first
> > table load time, when the type of device would be known.
> >
> > [PATCH 1/2] Allow delaying initialization of queue after allocation
> >
> > [PATCH 2/2] Initialize mempool and elevator only for request-based dm
> > devices
> >
> > Another approach would be to free the mempool and detach elevator, if it
> > would be a bio-based device. But it is better to avoid initialization
> > itself.
>
> Changes since v1:
> 1. Refreshed on top of a4e7d46407d73f35d217013b363b79a8f8eafcaa
> 2. Added comments suggested by Mike Snitzer.
> 3. Cleaned up the second patch as suggested by Mike Snitzer.
>
Changes since v2:
1. Add comments suggested by Kiyoshi Ueda
2. Export symbol elv_register_queue() to modules.
3. Call elv_register_queue() after initializing the queue to use the elevator.
Thanks
Nikanth
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2009-08-11 11:56 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-08-08 4:55 [PATCH 0/2] delay initializing queue for request-based dm Nikanth Karthikesan
2009-08-08 4:55 ` Nikanth Karthikesan
2009-08-10 10:48 ` [PATCH-v2 " Nikanth Karthikesan
2009-08-10 10:48 ` Nikanth Karthikesan
2009-08-11 9:32 ` [PATCH-v3 " Nikanth Karthikesan
2009-08-11 9:32 ` Nikanth Karthikesan
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.