Re: [PATCH] [Target_Core_Mod/pSCSI]: Add optional legacy scsi_execute_async() usage for Linux/SCSI passthrough

[Boaz Harrosh <bharrosh@panasas.com>]

On 04/20/2009 11:36 PM, Nicholas A. Bellinger wrote:
> On Sun, 2009-04-19 at 13:56 +0300, Boaz Harrosh wrote:
>> Are you aware that scsi_execute_async() has gone in 2.6.30-rc1?
>>
>> I'm not sure what would be the best alternative for you. I would say
>> a bio, but it is still being debated. Your current options are:
>>
>> 1. bio_alloc then loop () bio_add_pc_page, and finally blk_rq_append_bio
>>    (Which block people don't like)
>> 2. sglist => page-pointers-array translation and blk_rq_map_user with 
>>    struct rq_map_data mode. (not possible with all kind of sglists)
>> 2. sglist => iovec translation and blk_rq_map_user_iov()
>>    (Very very ugly mapping of pages to virtual pointers)
>>
>> I have a similar situation with my OSD code.
>>
> 
> Hey,
> 
> Thanks for the pointers on this..  Yeah, I knew it was going away soon,
> but needed it for some legacy drivers..:-)
> 
> Anyways, I will drop this patch for scsi_execute_async() when I branch
> for v2.6.30-rcX and look and updating target_core_mod/pSCSI to one of
> the listed methods to get up and running.  From the looks of it however
> we might want something slightly better than blk_rq_map_user() for the
> kernel space pages attached to struct scatterlist and page_links..
> 
> Just FYI, the target_core_mod struct scatterlist memory will be in
> struct list_head with struct page for allocated se_cmd_t (received CDB),
> so it will be pretty easy from the target_core_mod+subsystem plugin side
> side to put something that along the lines of list =>
> page-pointers-array translation and blk_rq_map_user() into a new
> function..

Be ware that page-pointers-array mean that only first/last page can be
incomplete and all middle pages must be full PAGE_SIZE. As I recall this
might not be the case for network sglists.

> 
> lio-core-2.6.git/drivers/target/target_core_transport.c is using
> include/linux/scatterlist.h macros when interacting with struct
> scatterlist allocated for each subsystem plugin (from the linked list
> struct page memory)  I will have a look and see what would need to be
> down to allow all kinds of sglists with something new..
> 

right, in the general sglist model, only linked-bios will fit one-to-one.

You should look into something like Tejun's RFCed blk_rq_map_kern_sgl()
(Which was not yet accepted)

> Thanks for your comments!
> 
> --nab
> 
> 

We'll see what comes up
Boaz