static void req_c= rypt_dec_pending_encrypt(struct req_dm_crypt_io *io)=0A= {=0A= int error =3D 0;=0A= struct request *clone =3D NULL;=0A= =0A= if (io) {=0A= error =3D io->error;=0A= if (io->cloned_request) {=0A= clone =3D io->cloned_request;=0A= } else {=0A= ....

Here I think the = request is being sent to the underlying device below the target (I think th= e .map_rq would be returning DM_MAPIO_REMAPPED for this IO). Now wit= h "dm_dispatch_request()" deprecated, how would this be done now?=

Also for b= io based targets, there is a facility within target to allocate some space = using the "per_bio_data_size", and then use that in the "map= " callback using the "dm_per_bio_data()", do we have an equi= valent mechanism for request based targets. I saw some patch where there ar= e there is something about letting request based targets to use this ("= ;per_io_data_size"), but I did not see anything the version of kernel = that I have - Ubuntu 16.04 (4.4.0-92).

Sorry, here is a correction to my last comment;

From: dm-devel-bounces@redh=
at.com <dm-devel-bounces@redhat.com> on behalf of Tariq Mustafa <t=
mustafa@hotmail.com>

Sent: Wednesday, November 1, 2017 11:10:58 AM

To: Mike Snitzer

Cc: dm-devel@redhat.com

Subject: Re: [dm-devel] Replacement for dm_dispatch_request for olde=
r request based targets

Thanks Mike. I was looking at the request based crypt, and saw the follo=
wing code;

static void req_c=
rypt_dec_pending_encrypt(struct req_dm_crypt_io *io)=0A=
{=0A=
	int error =3D 0;=0A=
	struct request *clone =3D NULL;=0A=
=0A=
	if (io) {=0A=
		error =3D io->error;=0A=
		if (io->cloned_request) {=0A=
			clone =3D io->cloned_request;=0A=
		} else {=0A=
                ....
        	}=0A=
	} else {
                .=
...=0A=
	}=0A=
=0A=
	atomic_dec(&io->pending);=0A=
=0A=
	if (error < 0)=0A=
		dm_kill_unmapped_request(clone, error);=0A=
	else=0A=
		dm_dispatch_request(clone);    <----=0A=
}

Here I think the =
request is being sent to the underlying device below the target (I think th=
e .map_rq would be returning DM_MAPIO_REMAPPED for this IO). Now wit=
h "dm_dispatch_request()" deprecated, how would this be done now?=

Additional=
 question:
Also for b=
io based targets, there is a facility within target to allocate some space =
using the "per_bio_data_size", and then use that in the "map=
" callback using the "dm_per_bio_data()", do we have an equi=
valent mechanism for request based targets. I saw some patch where there ar=
e there is something about letting request based targets to use this ("=
;per_io_data_size"), but I did not see anything the version of kernel =
that I have - Ubuntu 16.04 (4.4.0-92).

Thank you!=

From: Mike Snitzer <sn=
itzer@redhat.com>

Sent: Wednesday, November 1, 2017 9:08:57 AM

To: Tariq Mustafa

Cc: dm-devel@redhat.com

Subject: Re: Replacement for dm_dispatch_request for older request b=
ased targets

On Wed, Nov 01 2017 at  2:23am -0400,

Tariq Mustafa <tmustafa@hotmail.com> wrote:

>    Hello,

> 

>    I am trying to figure out what would be the appropri=
ate replacement for

>    "dm_dispatch_request()" on the newer kerne=
l? This export has been removed

>    (by commit - [1][dbf978]), what would be an equivale=
nt call to submit the

>    request.

dm_dispatch_request was renamed to dm_dispatch_clone_request (IIRC).

But are you asking what is the new export?  There isn't one.  The=
re

should be no need for anything other than DM core (dm-rq.c) to call

these interfaces that are no longer exported.

Please explain what you're trying to do if you need more help.

Thanks,

Mike

--_000_BY2PR18MB0344A93463B7C7A3EDEAA878C25F0BY2PR18MB0344namp_--

--===============4688717293992642863==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

--===============4688717293992642863==--