On Mon, 2014-12-15 at 11:12 +0800, Charles Chiou wrote:
>
> On 12/10/2014 05:02 PM, Oliver Neukum wrote:
> > On Wed, 2014-12-10 at 09:38 +0800, Charles Chiou wrote:
> >> From 91868d4afe10533b8a4496075109e411100217bb Mon Sep 17 00:00:00 2001
> >> From: Charles Chiou <charles.chiou@xxxxxxxxxxxxxx>
> >> Date: Fri, 7 Nov 2014 10:15:18 +0800
> >> Subject: [PATCH 4/4] scsi:stex.c Add S3/S4 support
> >>
> >> Add S3/S4 support, add .suspend and .resume function in pci_driver.
> >>
> >> Pegasus need 30~40 seconds to boot up. We don't want to OS wait
> >> in .resume function. Create a thread to handle device boot up.
> >>
> >
> >> +static int stex_resume(struct pci_dev *pdev)
> >> +{
> >> + struct st_hba *hba = pci_get_drvdata(pdev);
> >> + struct hba_handshake_workstruct *hswork;
> >> + int sts;
> >> +
> >> + hba->mu_status = MU_STATE_STARTING;
> >> + hswork = kzalloc(sizeof(struct hba_handshake_workstruct), GFP_KERNEL);
> >
> > The system is coming back from sleep. You cannot swap or page out
> > as disks may still be asleep. GFP_KERNEL is automatically changed
> > to GFP_NOIO. It would be nice to outright use GFP_NOIO.
> >
> >> + INIT_WORK(&hswork->handshake_work, resume_handshake);
> >
> > Memory allocations can fail.
> > I suggest you allocate the memory in suspend(). There you can just
> > return -ENOMEM in the error case.
> >
> >
> Hi Oliver, sorry for the late reply.
>
> Good point, could we move kzalloc function from suspend to probe and
> return -ENOMEM when allocation fail? We can avoid to allocate memory
> again and again in suspend/resume cycles.
Yes, that would work.
Regards
Oliver
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
