[PATCH 0/10] remove scsi_request usage from sg and st

[PATCH 0/10] remove scsi_request usage from sg and st

[Mike Christie]

The following patches, made against scsi-misc, remove the scsi_request
usage from st and sg, do some tiny cleanup on sg, and make sg and st
always perform scatterlist IO.

I have tested against libata and scsi_debug for disks. There are slight
changes, so someone with a tape device should test the patches, but Kai
since Kai did all the debugging last round I have not made many changes.
We will probably hit many problems with LLDs using low max_hw_sectors,
because they have not been converted to export their real values on the
host template so I think these patches should sit in scsi-misc for a
good deal of time.

Re: [PATCH 0/10] remove scsi_request usage from sg and st

[Mike Christie]

Mike Christie wrote:
> The following patches, made against scsi-misc, remove the scsi_request
> usage from st and sg, do some tiny cleanup on sg, and make sg and st
> always perform scatterlist IO.
> 
> I have tested against libata and scsi_debug for disks. There are slight
> changes, so someone with a tape device should test the patches, but Kai
> since Kai did all the debugging last round I have not made many changes.
> We will probably hit many problems with LLDs using low max_hw_sectors,
> because they have not been converted to export their real values on the
> host template so I think these patches should sit in scsi-misc for a
> good deal of time.
> 

I for got to cc the st and sg maintainers. In case you guys do not 
follow linux-scsi closely and do not want all those patches sent to you 
I put them here
http://www.cs.wisc.edu/~michaelc/block/use-sg/v8/
I can email them again too if you want.

Re: [PATCH 0/10] remove scsi_request usage from sg and st

[Mike Christie]

Mike Christie wrote:
> We will probably hit many problems with LLDs using low max_hw_sectors,
> because they have not been converted to export their real values on the
> host template so I think these patches should sit in scsi-misc for a
> good deal of time.
> 

oh yeah just to clarify, this patch [PATCH 2/10] seperate max sectors 
and max hw sectors, makes it so the scsi host template's max_sectors is 
passed up as q->max_hw_sectors. I did not know the correct limit for any 
SCSI LLD's sht->max_sectors so I could not convert them if they were 
previously setting a conservtaive value for reasons like libata.

Re: [PATCH 0/10] remove scsi_request usage from sg and st

[Christoph Hellwig]

On Tue, Nov 08, 2005 at 04:06:29AM -0600, Mike Christie wrote:
> The following patches, made against scsi-misc, remove the scsi_request
> usage from st and sg, do some tiny cleanup on sg, and make sg and st
> always perform scatterlist IO.
> 
> I have tested against libata and scsi_debug for disks. There are slight
> changes, so someone with a tape device should test the patches, but Kai
> since Kai did all the debugging last round I have not made many changes.
> We will probably hit many problems with LLDs using low max_hw_sectors,
> because they have not been converted to export their real values on the
> host template so I think these patches should sit in scsi-misc for a
> good deal of time.

looks like linus is closing the big merge window for 2.6.15, so we won't
be able to get this in fully.  What do you thing about getting in the
infrastructure at least, that is patches: 1,3,6,7 for 2.6.15 and the rest
afterwards, after the max_sectors thing has settled a little?

Re: [PATCH 0/10] remove scsi_request usage from sg and st

[Mike Christie]

Christoph Hellwig wrote:
> On Tue, Nov 08, 2005 at 04:06:29AM -0600, Mike Christie wrote:
> 
>>The following patches, made against scsi-misc, remove the scsi_request
>>usage from st and sg, do some tiny cleanup on sg, and make sg and st
>>always perform scatterlist IO.
>>
>>I have tested against libata and scsi_debug for disks. There are slight
>>changes, so someone with a tape device should test the patches, but Kai
>>since Kai did all the debugging last round I have not made many changes.
>>We will probably hit many problems with LLDs using low max_hw_sectors,
>>because they have not been converted to export their real values on the
>>host template so I think these patches should sit in scsi-misc for a
>>good deal of time.
> 
> 
> looks like linus is closing the big merge window for 2.6.15, so we won't
> be able to get this in fully.  What do you thing about getting in the
> infrastructure at least, that is patches: 1,3,6,7 for 2.6.15 and the rest
> afterwards, after the max_sectors thing has settled a little?
> 

If you merge #3 you need the late #11 
http://marc.theaimsgroup.com/?l=linux-scsi&m=113149038422831&w=2

It is up to Jens though since I am really digging deep into the block 
layer code in #3 and #6 may need to be fixed for SG_IO sr commands. I 
can rediff them if you they do not merge correctly. I noticed the block 
layer move code is now merge in scsi-misc.

Oh yeah was my last max_sectors close
http://marc.theaimsgroup.com/?l=linux-scsi&m=113147482216587&w=2
or did it need work, or do I just need to get LLD writers to set their 
sht->max_sectors correctly?

Re: [PATCH 0/10] remove scsi_request usage from sg and st

[Christoph Hellwig]

On Thu, Nov 10, 2005 at 06:51:21AM -0600, Mike Christie wrote:
> Oh yeah was my last max_sectors close
> http://marc.theaimsgroup.com/?l=linux-scsi&m=113147482216587&w=2
> or did it need work, or do I just need to get LLD writers to set their 
> sht->max_sectors correctly?

Looks good to me.