From mboxrd@z Thu Jan 1 00:00:00 1970 From: Aaron Lu Subject: Re: [PATCH v6 1/7] scsi: sr: support runtime pm for ODD Date: Tue, 11 Sep 2012 17:24:13 +0800 Message-ID: <20120911092410.GA1704@mint-spring.sh.intel.com> References: <4663685.bsC7gzLV6E@linux-lqwf.site> <20120911064429.GA2196@mint-spring.sh.intel.com> <3578472.dQIepChrCX@linux-lqwf.site> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mga01.intel.com ([192.55.52.88]:63939 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755928Ab2IKJYQ (ORCPT ); Tue, 11 Sep 2012 05:24:16 -0400 Content-Disposition: inline In-Reply-To: <3578472.dQIepChrCX@linux-lqwf.site> Sender: linux-acpi-owner@vger.kernel.org List-Id: linux-acpi@vger.kernel.org To: Oliver Neukum Cc: Alan Stern , Aaron Lu , James Bottomley , Jeff Garzik , linux-scsi@vger.kernel.org, linux-ide@vger.kernel.org, linux-pm@vger.kernel.org, linux-acpi@vger.kernel.org On Tue, Sep 11, 2012 at 10:55:44AM +0200, Oliver Neukum wrote: > On Tuesday 11 September 2012 14:44:30 Aaron Lu wrote: > > On Mon, Sep 10, 2012 at 12:45:51PM +0200, Oliver Neukum wrote: > > > On Monday 10 September 2012 17:16:22 Aaron Lu wrote: > > > > > > > +static int sr_resume(struct device *dev) > > > > +{ > > > > + struct scsi_cd *cd; > > > > + struct scsi_sense_hdr sshdr; > > > > + > > > > + cd = dev_get_drvdata(dev); > > > > + > > > > + if (!cd->device->powered_off) > > > > + return 0; > > > > + > > > > + /* get the disk ready */ > > > > + scsi_test_unit_ready(cd->device, SR_TIMEOUT, MAX_RETRIES, &sshdr); > > > > + > > > > + /* if user wakes up the ODD, eject the tray */ > > > > + if (cd->device->need_eject) { > > > > + cd->device->need_eject = 0; > > > > + if (!(cd->cdi.mask & CDC_CLOSE_TRAY)) > > > > + sr_tray_move(&cd->cdi, 1); > > > > > > 1. Even if the door is locked? > > > > This piece of code of sr_resume is intended for ZPODD devices, for > > normal devices it will simply return as the powered_off flag will never > > be set for them. > > > > And for ZPODD devices, the door shouldn't be in locked state here since > > for it to enter power off state, "no medium inside" has to be satisfied > > This is true only if the device is autosuspended. But sr_resume() > will also be called if the whole system was suspended. You mean when system resumes, right? > What happens if the user presses the door button on the drive > while the system was suspended? The drive does not have the capability to wake up the system from S3, so nothing happens. > > > and when there is no medium inside, we do not lock its door. > > > > > 2. sr_tray_move allocates memory with GFP_KERNEL. This smells of a deadlock. > > > > Sorry, don't get this. Can you please elaborate? > > Suppose you need to wake up two devices for the same reason, a ZPODD and a hard > drive. The ZPODD is woken first and sr_resume() requests memory with GFP_KERNEL. > The VM layer decides to page out to the hard drive to be woken up -> deadlock. While, I don't understand what stops the hard disk from being resumed, the hard drive's runtime resume does not depend on the ODD's, and I don't think there should be an order enforced on them. Thanks, Aaron