[PATCH] scsi_mid_low_api.txt in lk 2.5.72+

[PATCH] scsi_mid_low_api.txt in lk 2.5.72+

[Douglas Gilbert]

[-- Attachment #1: Type: text/plain, Size: 1102 bytes --]

The scsi_mid_low_api.txt file describes the interface
between the SCSI mid level and a low level driver,
seen from the latter's perspective. It is found in
the kernel source in the Documentation/scsi directory.

The attachment incorporates a patch sent to me by Christoph
and ideas in Mike Anderson's refcounting document (url?).

It anticipates various patches that Christoph has sent to
this list being accepted. These include:
   - addition of scsi_host_alloc()
   - renaming of struct SHT to struct scsi_host_template

The attachment (gzipped due to its size) is a diff against
lk 2.5.72 . A gzipped version of the document can be found at:
   http://www.torque.net/scsi/scsi_mid_low_api.txt.gz


Idea: perhaps scsi_add_host() could have an extra argument
to inhibit the scsi device scan (or alternatively a new field
in struct scsi_host_template and struct Scsi_Host). Then
LLDs that know about their devices hot (un)plugging can use
scsi_add_device() and scsi_remove_device() at their leisure.
[Actually ieee1394/sbp2.c is already doing this ...]

Comments and corrections welcome.

Doug Gilbert

[-- Attachment #2: scsi_mid_low_api.txt2572dg.diff.gz --]
[-- Type: application/x-gzip, Size: 15358 bytes --]

Re: [PATCH] scsi_mid_low_api.txt in lk 2.5.72+

[Christoph Hellwig]

On Wed, Jun 18, 2003 at 10:37:42PM +1000, Douglas Gilbert wrote:
> It anticipates various patches that Christoph has sent to
> this list being accepted. These include:
>    - addition of scsi_host_alloc()
>    - renaming of struct SHT to struct scsi_host_template

This seems to be in now.  I'll also send you another update
once the header reorgansization is finished.

> Idea: perhaps scsi_add_host() could have an extra argument
> to inhibit the scsi device scan (or alternatively a new field
> in struct scsi_host_template and struct Scsi_Host). Then
> LLDs that know about their devices hot (un)plugging can use
> scsi_add_device() and scsi_remove_device() at their leisure.
> [Actually ieee1394/sbp2.c is already doing this ...]

I though about this already but maybe the cleaner interface
would be to remove the call to scsi_scan_host from scsi_add_host
and let drivers that want it do it themselves.  (and the same
for scsi_forget_host in scsi_remove_host)