From: Boaz Harrosh <bharrosh@panasas.com>
To: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: James.Bottomley@suse.de, jaxboe@fusionio.com,
linux-scsi@vger.kernel.org, Sathya.Prakash@lsi.com
Subject: Re: [PATCH] scsi: fix scsi_get_lba helper function for pc command
Date: Wed, 02 Jun 2010 12:23:01 +0300 [thread overview]
Message-ID: <4C0622F5.3040903@panasas.com> (raw)
In-Reply-To: <20100602180515J.fujita.tomonori@lab.ntt.co.jp>
On 06/02/2010 12:05 PM, FUJITA Tomonori wrote:
> On Wed, 02 Jun 2010 11:54:25 +0300
> Boaz Harrosh <bharrosh@panasas.com> wrote:
>
>>> Some LLDs already use scsi_get_lba(). They don't know scsi command
>>> from pc or fs so they can't use scsi_get_lba() now. Well, they can see
>>> scsi_cmnd->request to know pc or fs but it's layer violation. They
>>> shouldn't access to the block layer stuff.
>>>
>>
>> right, so hide it all under an API that is implemented at the right
>> level, and protect them. The driver should call a single simple
>> API the magic will be done at the generic level
>>
>> if (blk_fs_rq(req))
>> return blk_rq_pos(req);
>> else
>> return scsi_somthing();
>
> The first version that I posted yesterday is something like that. But
> I don't think that it's clean.
>
I can't find it. I'd like to see
>
>>>> Only drivers/code that actually cares should call this member.
>>>>
>>>> It should be easy enough to search for __sector and change them
>>>> to a function call.
>>>
>>> I don't think that calling such function in LLDs is a clean design. We
>>> set request->__sector for fs requests. Why not for pc requests? Then
>>> we can use blk_rq_pos() cleanly.
>>>
>>
>> Because "fs requests" have a defined blk_rq_pos(). But "pc requests"
>> don't necessarily have one. Pretending to invent one is stupid and
>> dangerous. Better keep a -1 for all "pc requests".
>
> I don't see your logic, why using blk_rq_pos() for pc is dangerous?
I already explained. because you must know the device type before
you assume the presence of lba.
>
>> scsi LLDs can be pass-through for disks, as well as lots of other
>> type of devices. They should not assume the role of a disk. The use
>> of blk_rq_pos() at LLDs is probably wrong in the first place. Unless
>> it is some kind of scsi emulation driver like scsi_debug.
>
> Not only scsi_debug. Have you seen how scsi_get_lba() is used? LLDs
> sometime needs to know lba in scsi commands.
>
> How about calling this function in sd_prep_fn, sr_prep_fn instead of
> scsi_setup_blk_pc_cmnd? Then it doesn't affect OSD.
That's fine because these know that they are disk and ROM/RAM device types.
Watch out with sr because it supports other command sets. there might
be more too it.
Boaz
next prev parent reply other threads:[~2010-06-02 9:23 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-06-02 7:33 [PATCH] scsi: fix scsi_get_lba helper function for pc command FUJITA Tomonori
2010-06-02 8:14 ` Boaz Harrosh
2010-06-02 8:39 ` FUJITA Tomonori
2010-06-02 8:54 ` Boaz Harrosh
2010-06-02 9:05 ` FUJITA Tomonori
2010-06-02 9:23 ` Boaz Harrosh [this message]
2010-06-02 13:13 ` Martin K. Petersen
2010-06-03 10:59 ` FUJITA Tomonori
2010-06-03 11:03 ` Prakash, Sathya
2010-06-02 14:07 ` James Bottomley
2010-06-02 14:55 ` Douglas Gilbert
2010-06-03 11:18 ` FUJITA Tomonori
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4C0622F5.3040903@panasas.com \
--to=bharrosh@panasas.com \
--cc=James.Bottomley@suse.de \
--cc=Sathya.Prakash@lsi.com \
--cc=fujita.tomonori@lab.ntt.co.jp \
--cc=jaxboe@fusionio.com \
--cc=linux-scsi@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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.