Re: [Bugme-new] [Bug 8122] New: Scsi_Host's n_io_port forgot to be initialized?

Re: [Bugme-new] [Bug 8122] New: Scsi_Host's n_io_port forgot to be initialized?

[Andrew Morton]

On Sun, 4 Mar 2007 01:19:32 -0800 bugme-daemon@bugzilla.kernel.org wrote:

> http://bugzilla.kernel.org/show_bug.cgi?id=8122
> 
>            Summary: Scsi_Host's n_io_port forgot to be initialized?
>     Kernel Version: Linux 2.6.20
>             Status: NEW
>           Severity: normal
>              Owner: scsi_drivers-other@kernel-bugs.osdl.org
>          Submitter: shanlu@cs.uiuc.edu
> 
> 
> Most recent kernel where this bug did *NOT* occur: No
> Distribution: Linux-2.6.20
> Problem Description:
> It seems that usually structure Scsi_Host's io_port field and n_io_port field
> needs to be set together.
> However, in file drivers/scsi/dmx3191d.c,
> function dmx3191d_probe_one
> 71  static int __devinit dmx3191d_probe_one(struct pci_dev *pdev,
> 72              const struct pci_device_id *id)
> ...
> 88      shost = scsi_host_alloc(&dmx3191d_driver_template,
> 89                      sizeof(struct NCR5380_hostdata));
> 90      if (!shost)
> 91              goto out_release_region;
> 92      shost->io_port = io;
> 93      shost->irq = pdev->irq;
> ...
> 109     error = scsi_add_host(shost, &pdev->dev);
> 
> only io_port is set and n_io_port is uninitialized. Won't this cause some
> problem? I fear when some code tries to calculate the io_port region (from
> io_port to io_port+n_io_port-1 is regarded as the port region), there would be
> problem.
> 
> It seems that in all other similar scsi device's similar functions, like
> dc390_probe_one and inia100_probe_one, io_port and n_io_port are both set. This
> makes me think above dmx3191d.c's missing n_io_port at least a potential danger.
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.

Re: [Bugme-new] [Bug 8122] New: Scsi_Host's n_io_port forgot to be initialized?

[Christoph Hellwig]

On Mon, Mar 05, 2007 at 07:17:30PM -0800, Andrew Morton wrote:
> > Most recent kernel where this bug did *NOT* occur: No
> > Distribution: Linux-2.6.20
> > Problem Description:
> > It seems that usually structure Scsi_Host's io_port field and n_io_port field
> > needs to be set together.

io_port and n_io_port are legacy fields left over purely for driver use.
The scsi midlayer doesn't care, and if you want to "fix" things in this
area please just remove use of it from the offending driver completely
so might be able to get rid of these fields one day.