On Thu, 16 Aug 2012, Hans de Goede wrote:
> Ah right, these are bus_driver operations. That explains some things, so I've
> done some more research asking myself: "Why does generic_ide_suspend(), which
> is a *bus* op, call dev_get_drvdata?", the answer to that seems to be that
> the ide subsystem is abusing (IMHO) drvdata to store per device bus_driver
> data. Which I believe is not how drvdata is intended to be used.
>
> With that said, the above knowledge has allowed me to write an (ugly) fix for
> the regression Miklos is seeing. Miklos can you give the attached patch a
> try please?
>
> > It clearly should check if drive is not NULL before using that pointer.
>
> I assume you mean drive*r*, yes I agree that generic_ide_remove should
> check for that. So who is going to write a patch for that?
The existing code could certainly be improved. Your patch does:
> + /*
> + * device_register() will have cleared drvdata on
> + * device_attach failure, but we use drvdata to store per
> + * device bus info, rather then for driver info, so restore
> it.
> + */
> + dev_set_drvdata(dev, drive);
But at this point, dev is defined by:
struct device *dev = &drive->gendev;
So why bother setting anything? It seems to me that
generic_ide_suspend() and generic_ide_resume() could easily replace
ide_drive_t *drive = dev_get_drvdata(dev);
with
ide_drive_t *drive = dev_to_ide_drive(dev);
where dev_to_ide_drive is defined as "container_of(dev, ide_drive_t,
gendev)" (if this isn't defined already).
Alan Stern
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
