Message-Id: <200812162135.27902.bzolnier@gmail.com>
Date:	Tue, 16 Dec 2008 21:35:27 +0100
From:	Bartlomiej Zolnierkiewicz <bzolnier@...il.com>
To:	Borislav Petkov <petkovbb@...glemail.com>
Cc:	linux-kernel@...r.kernel.org, linux-ide@...r.kernel.org,
	Borislav Petkov <petkovbb@...il.com>
Subject: Re: [PATCH 0/9] ide-atapi: remove ide_atapi_pc from the irq handler


Hi,

On Tuesday 16 December 2008, Borislav Petkov wrote:
> Hi Bart,
> 
> here's a first attempt at removing all references to ide_atapi_pc in the ATAPI
> IRQ handler. I've moved some of the members to the drive struct and will deal
> with them later :). The next step is to add an ide_atapi_queue_pc() routine

Hmm.  I recall discussing this before and having some concerns about struct
ide_atapi_pc removal idea probably being premature...

Looking at the patchset itself -- I'm very sorry but I just cannot apply it
with all these command specific fields going into ide_drive_t:

        unsigned long pc_flags;
        int pc_req_xfer;
        int pc_buf_size;
        int pc_error;

While they may look similar to the existing pc_* fields:

        /* callback for packet commands */
        void (*pc_callback)(struct ide_drive_s *, int);

        void (*pc_update_buffers)(struct ide_drive_s *, struct ide_atapi_pc *);
        int  (*pc_io_buffers)(struct ide_drive_s *, struct ide_atapi_pc *,
                              unsigned int, int);

they are clearly not.  Existing pc_* callbacks are per-device type (they
should have been probably named drv_*) and new ones are per-command ones.

I also don't see a way for code to differentiate between pc and failed_pc.

Moreover if we would like to ever add support for multiple queued commands
(actually I think that in a very limited way ide-tape does it already with
DSC) we would have to completely revert these changes.

Re-assuming: this is one-step forward / one-step back approach and not
exactly the kind of design that we would like to advocate.

> similar to ide_cd_queue_pc() and then rewrite all functions in the drivers to
> use struct requests and local buffers instead of pc->buf and then, after that
> works reliably, finally get rid of ide_atapi_pc completely.

Why not take care of low hanging fruits first?  There is still plenty...

- ide-cd uses private struct request for REQUEST SENSE (=> ide_drive_t's)

- it should be possible to port ide-cd to use ide_{issue,transfer}_pc()
  _without_ dealing with struct ide_atapi_pc issue at all

- why not investigate the possibility of porting ide-cd to using sg-s for
  PIO transfers first -- this should get rid of ide_cd_restore_request()
  and most of other rq fields manipulations

...and this is from taking not so long look at the current driver.

Why attack the "hard" problem first (struct ide_atapi_pc one) while there
are easier ones to deal with?  It may happen that once these easier ones
are fixed we will be able to view the "hard" one from a completely different
perspective (because code's complexity will decrease a lot in the meantime
it no longer will be a "hard" problem) and apply a better solution.

Besides I'm not fully convinced that struct ide_atapi_pc has to go first in
order to port ide-cd over ide-atapi -- we can just instead port ide-cd to
use struct ide_atapi_pc (however probably only after converting the driver
to use sg-s for PIO transfers -- otherwise it becomes another "hard" problem)
and then deal with struct ide_atapi_pc for all ATAPI code (be using struct
request fields directly or by merging struct ide_atapi_pc into ide_task_t
and always using the latter structure for ATA[PI] commands -- which is the
option that I personally came to prefer lately)...

Thanks,
Bart
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists