Re: [PATCH V3] scsi: core: put LLD module refcnt after SCSI device is released

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Thu, Oct 07, 2021 at 03:12:52PM +0800, Ming Lei wrote:
> On Tue, Oct 05, 2021 at 03:35:40PM +0200, Greg Kroah-Hartman wrote:
> > On Thu, Sep 30, 2021 at 08:44:15PM +0800, Ming Lei wrote:
> > > SCSI host release is triggered when SCSI device is freed, and we have to
> > > make sure that LLD module won't be unloaded before SCSI host instance is
> > > released because shost->hostt is required in host release handler.
> > > 
> > > So make sure to put LLD module refcnt after SCSI device is released.
> > 
> > What is a "LLD"?
> 
> Lower level driver, which is used often as one scsi term.
> 
> > 
> > > Fix one kernel panic of 'BUG: unable to handle page fault for address'
> > > reported by Changhui and Yi.
> > > 
> > > Reported-by: Changhui Zhong <czhong@redhat.com>
> > > Reported-by: Yi Zhang <yi.zhang@redhat.com>
> > > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> > > Signed-off-by: Ming Lei <ming.lei@redhat.com>
> > > ---
> > >  drivers/scsi/scsi.c        |  4 +++-
> > >  drivers/scsi/scsi_sysfs.c  | 12 ++++++++++++
> > >  include/scsi/scsi_device.h |  1 +
> > >  3 files changed, 16 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
> > > index b241f9e3885c..291ecc33b1fe 100644
> > > --- a/drivers/scsi/scsi.c
> > > +++ b/drivers/scsi/scsi.c
> > > @@ -553,8 +553,10 @@ EXPORT_SYMBOL(scsi_device_get);
> > >   */
> > >  void scsi_device_put(struct scsi_device *sdev)
> > >  {
> > > -	module_put(sdev->host->hostt->module);
> > > +	struct module *mod = sdev->host->hostt->module;
> > > +
> > >  	put_device(&sdev->sdev_gendev);
> > > +	module_put(mod);
> > >  }
> > >  EXPORT_SYMBOL(scsi_device_put);
> > >  
> > > diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
> > > index 86793259e541..9ada26814011 100644
> > > --- a/drivers/scsi/scsi_sysfs.c
> > > +++ b/drivers/scsi/scsi_sysfs.c
> > > @@ -449,9 +449,16 @@ static void scsi_device_dev_release_usercontext(struct work_struct *work)
> > >  	struct scsi_vpd *vpd_pg80 = NULL, *vpd_pg83 = NULL;
> > >  	struct scsi_vpd *vpd_pg0 = NULL, *vpd_pg89 = NULL;
> > >  	unsigned long flags;
> > > +	struct module *mod;
> > > +	bool put_mod = false;
> > >  
> > >  	sdev = container_of(work, struct scsi_device, ew.work);
> > >  
> > > +	if (sdev->put_lld_mod_ref) {
> > 
> > Why do you need this flag at all?
> > 
> > Shouldn't you just always grab/release the module?  Why would you not
> > want to?
> 
> try_module_get() may fail in scsi_device_dev_release() in case that
> unloading is started, then we don't need to put it in
> scsi_device_dev_release_usercontext(), so this flag is required.

If grabing the module reference failed then you CAN NOT call into it so
you have a problem here that you seem to be papering over.  Just set the
module pointer to NULL and then you don't need to test it when calling
module_put, right?

thanks,

greg k-h