Re: dpt_i2o driver

['Christoph Hellwig' <hch@infradead.org>]

On Tue, Sep 02, 2003 at 01:26:53PM -0400, Salyzyn, Mark wrote:
> dpt_i2o holds on to the device structures in the `hostdata' areas of the
> device in order to speed up TID (I2O Target Identifier) lookup. This code
> may be in jeopardy as a result:
> 
>  (struct adpt_device*)(cmd->device->hostdata) = pDev;
> . . .
>  pDev->pScsi_dev = cmd->device;

That is bad indeed and I still need to audit all drivers for issues
like that.  Either we kill this cache (I need to take a look at the
actual code). OR you need to grab a reference using scsi_device_get
when you store it and release it in your ->slave_destroy method that's
called when the device is marked deleted.

> The pScsi_dev references can disappear, a hotplug reaction to clear this
> needs to be registered to free up the double linked references made like
> above (this is a plea for help to understand hotplug ;-/ ). Once this double
> linked reference is cleared, the busy check report ENODEV which is an
> `adequate' response indicating no longer busy.

Why do you need an ioctl then?  Just wait for the sysfs file to
disappear..

-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
On Tue, Sep 02, 2003 at 01:26:53PM -0400, Salyzyn, Mark wrote:
> dpt_i2o holds on to the device structures in the `hostdata' areas of the
> device in order to speed up TID (I2O Target Identifier) lookup. This code
> may be in jeopardy as a result:
> 
>  (struct adpt_device*)(cmd->device->hostdata) = pDev;
> . . .
>  pDev->pScsi_dev = cmd->device;

That is bad indeed and I still need to audit all drivers for issues
like that.  Either we kill this cache (I need to take a look at the
actual code). OR you need to grab a reference using scsi_device_get
when you store it and release it in your ->slave_destroy method that's
called when the device is marked deleted.

> The pScsi_dev references can disappear, a hotplug reaction to clear this
> needs to be registered to free up the double linked references made like
> above (this is a plea for help to understand hotplug ;-/ ). Once this double
> linked reference is cleared, the busy check report ENODEV which is an
> `adequate' response indicating no longer busy.

Why do you need an ioctl then?  Just wait for the sysfs file to
disappear..