linux-kernel - Re: block/bsg.c

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <1184766854.3464.4.camel@localhost.localdomain>
Date:	Wed, 18 Jul 2007 08:54:14 -0500
From:	James Bottomley <James.Bottomley@...elEye.com>
To:	FUJITA Tomonori <tomof@....org>
Cc:	jens.axboe@...cle.com, fujita.tomonori@....ntt.co.jp,
	akpm@...ux-foundation.org, linux-kernel@...r.kernel.org,
	linux-scsi@...r.kernel.org
Subject: Re: block/bsg.c

On Wed, 2007-07-18 at 09:20 +0900, FUJITA Tomonori wrote:
> From: James Bottomley <James.Bottomley@...elEye.com>
> Subject: Re: block/bsg.c
> Date: Tue, 17 Jul 2007 13:53:54 -0500
> 
> > On Tue, 2007-07-17 at 12:19 +0200, Jens Axboe wrote:
> > > > > Since Linus is happily snoring by now, could you test and see if the
> > > > > tree works for you?
> > > > 
> > > > It works for me. I'll submit some minor patches against your bsg
> > > > branch to scsi-ml. Can you push them together?
> > > 
> > > Certainly, I'll pull them into the bsg branch.
> > 
> > While you're at it, here's a patch to separate BSG and SCSI again (so
> > SCSI can be built modular).  The way I did it was simply to move the
> > SCSI specific logic into SCSI.  When you come up with a generic way to
> > register the bsg requiring drivers, then we can move it out again.
> 
> Thanks, looks nice.

You're welcome ... although there's still a problem for modular builds.
This is what my /sys/class/bsg looks like:

So you see the if (rq->kobj.parent) is causing confusing naming.  The
reason the first one shows up as 0:0:0:0 is because in an initrd
scsi_mod is loaded first (which is when bsg binds) followed by sd_mod
(which is what gives the device the ULD binding and hence the name).  I
don't see any way around this, so I'd advocate simply using the sdev
name rather than the block device name and dumping the if.

> > --- a/drivers/scsi/scsi_sysfs.c
> > +++ b/drivers/scsi/scsi_sysfs.c
> > @@ -714,6 +714,7 @@ static int attr_add(struct device *dev, struct device_attribute *attr)
> >  int scsi_sysfs_add_sdev(struct scsi_device *sdev)
> >  {
> >  	int error, i;
> > +	struct request_queue *rq = sdev->request_queue;
> >  
> >  	if ((error = scsi_device_set_state(sdev, SDEV_RUNNING)) != 0)
> >  		return error;
> > @@ -733,6 +734,16 @@ int scsi_sysfs_add_sdev(struct scsi_device *sdev)
> >  	/* take a reference for the sdev_classdev; this is
> >  	 * released by the sdev_class .release */
> >  	get_device(&sdev->sdev_gendev);
> > +
> > +	if (rq->kobj.parent)
> > +		error = bsg_register_queue(rq, kobject_name(rq->kobj.parent));
> > +	else
> > +		error = bsg_register_queue(rq, kobject_name(&sdev->sdev_gendev.kobj));
> > +	if (error) {
> > +		sdev_printk(KERN_INFO, sdev, "Failed to register bsg queue\n");
> > +		goto out;
> 
> Needs more cleanup here?

No ... this bit's magic and clever.  Once you've set up the devices and
done a get_device, cleanup is simply doing a put_device because it's all
done in the release routine.

> We might just ignore the error here since it's not fatal not to create
> a bsg device, I guess.
> 
> I updated the patch against the latest code (which has just be merged
> to Linus's tree).

Thanks,

James


-
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/