Date:	Thu, 19 Jul 2007 10:57:53 +0200 (CEST)
From:	Geert Uytterhoeven <Geert.Uytterhoeven@...ycom.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
cc:	Andy Whitcroft <apw@...dowen.org>, Jens Axboe <axboe@...nel.dk>,
	"James E.J. Bottomley" <James.Bottomley@...elEye.com>,
	linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org,
	Alessandro Rubini <rubini@...ion.unipv.it>,
	linuxppc-dev@...abs.org, Paul Mackerras <paulus@...ba.org>,
	Jens Axboe <jens.axboe@...cle.com>
Subject: Re: [patch 1/3] ps3: Disk Storage Driver

	Hi Andrew,

On Wed, 18 Jul 2007, Andrew Morton wrote:
> On Mon, 16 Jul 2007 18:15:40 +0200
> Geert Uytterhoeven <Geert.Uytterhoeven@...ycom.com> wrote:
> 
> > From: Geert Uytterhoeven <Geert.Uytterhoeven@...ycom.com>
> > 
> > Add a Disk Storage Driver for the PS3:
> 
> Your patchset significantly hits powerpc, scsi and block.  So who gets to
> merge this?  Jens?  James?  Paul?
> 
> Me, I guess ;)

As Paul is on holidays, please take it. The PS3 storage driver core support is
already in mainline, but the actual drivers aren't, as Paul was waiting for
acks from the maintainers.

BTW, do you prefer incremental patches for the comments below, or an update of
the full patchset?

> > +#define KERNEL_SECTOR_SIZE	512
> 
> Sigh.  We have at least ten separate definitions of SECTOR_SIZE< none of
> them in the right place.  Cleanup opportunity for someone.

Will replace by 512 resp. >> 9, as per Jens' suggestion.

> > +struct ps3disk_private {
> > +	spinlock_t lock;		/* Request queue spinlock */
> > +	struct request_queue *queue;
> > +	struct gendisk *gendisk;
> > +	unsigned int blocking_factor;
> > +	struct request *req;
> > +	u64 raw_capacity;
> > +	unsigned char model[ATA_ID_PROD_LEN+1];
> > +};
> > +#define ps3disk_priv(dev)	((dev)->sbd.core.driver_data)
> 
> hm, this code has "ata" all over it and actually uses a libata #define (at
> least) but there is no Kconfig dependency on ATA.  Fair enough, I guess,
> but a bit funny-looking.

I needed the definitions just for drive identification.

> > +static void ps3disk_scatter_gather(struct ps3_storage_device *dev,
> > +				   struct request *req, int gather)
> > +{
> > +	unsigned int sectors = 0, offset = 0;

> Local variable `sectors' doesn't do anything.

Good catch!

> > +static int ps3disk_submit_request_sg(struct ps3_storage_device *dev,
> > +				     struct request *req)
> > +{
> > +	struct ps3disk_private *priv = ps3disk_priv(dev);
> > +	int write = rq_data_dir(req), res;
> > +	const char *op = write ? "write" : "read";
> > +	u64 start_sector, sectors;
> > +	unsigned int region_id = dev->regions[dev->region_idx].id;
> 
> So we're ignoring the sector_t stuff.  I guess it's 64-bit only.  Still, it
> might be nice to use sector_t for consistency, readability and possible
> future 32-bitness?

I used u64 as they're directly passed to hypervisor calls and that's what the
hypervisor calls take.

BTW, Cell will never support a 32-bit kernel, as lots of on-chip stuff lies
outside the 32-bit address space.

> > +#ifdef DEBUG
> > +	unsigned int n = 0;
> > +	struct bio *bio;
> > +	rq_for_each_bio(bio, req)
> > +		n++;
> 
> I'm surprised that the block core doesn't have a helper to count the number
> of bios in a request.
> 
> Please prefer to put a blank line between end-of-locals and start-of-code.

Done.

> > +	dev_dbg(&dev->sbd.core,
> > +		"%s:%u: %s req has %u bios for %lu sectors %lu hard sectors\n",
> > +		__func__, __LINE__, op, n, req->nr_sectors,
> > +		req->hard_nr_sectors);
> > +#endif
> > +
> > +	start_sector = req->sector*priv->blocking_factor;
> > +	sectors = req->nr_sectors*priv->blocking_factor;
> 
> s/*/ * /.  checkpatch missed this.

Will change (yes, checkpatch missed it).

> I suspect you didn't run cehckpatch anyway.

I did ;-)

> Please run checkpatch.

All it complains about is (bogus) <asm/firmware.h> and a single too long line I
don't want to break.

> > +/* ATA helpers copied from drivers/ata/libata-core.c */
> 
> ooh, bad person.

I didn't have much choice, as most of it was static and I don't need the full
libata core anyway.

If I would factor it out, any good suggestion where to put the factored out
code?

> > +	ps3disk_identify(dev);
> 
> ps3disk_identify() can return an error?

Sending storage commands may fail, but being unable to identify the disk isn't
fatal, as it's just for informational purposes.

> > +static int ps3disk_remove(struct ps3_system_bus_device *_dev)
> > +{
> > +	struct ps3_storage_device *dev = to_ps3_storage_device(&_dev->core);
> > +	struct ps3disk_private *priv = ps3disk_priv(dev);
> > +
> > +	__clear_bit(priv->gendisk->first_minor / PS3DISK_MINORS,
> > +		    &ps3disk_mask);
> 
> I see no locking here which makes this __clear_bit and the above __set_bit
> non-racy?

Were .probe()/.remove() made concurrent again? I thought that idea was dropped
because it caused too many problems?

> > +	kfree(priv);
> > +	ps3disk_priv(dev) = NULL;
> 
> I suspect this nulling here will just hide any bugs?  If we're going to
> write anything there, probably 0xdeadbeef would be better?

Hmm, earlier reviewers explicitly asked for putting it there...

Thanks for your comments!

With kind regards,
 
Geert Uytterhoeven
Software Architect

Sony Network and Software Technology Center Europe
The Corporate Village · Da Vincilaan 7-D1 · B-1935 Zaventem · Belgium
 
Phone:    +32 (0)2 700 8453	
Fax:      +32 (0)2 700 8622	
E-mail:   Geert.Uytterhoeven@...ycom.com	
Internet: http://www.sony-europe.com/
 	
Sony Network and Software Technology Center Europe	
A division of Sony Service Centre (Europe) N.V.	
Registered office: Technologielaan 7 · B-1840 Londerzeel · Belgium	
VAT BE 0413.825.160 · RPR Brussels	
Fortis Bank Zaventem · Swift GEBABEBB08A · IBAN BE39001382358619

Powered by blists - more mailing lists