Re: [patch] scsi: revert "[SCSI] Get rid of scsi_cmnd->done"

[Andrew Morton]

On Tue, 08 Jan 2008 23:01:07 -0500 Valdis.Kletnieks@xxxxxx wrote:

> On Mon, 07 Jan 2008 16:19:30 MST, Matthew Wilcox said:
> > On Mon, Jan 07, 2008 at 06:04:25PM -0500, Valdis.Kletnieks@xxxxxx wrote:
> > > Theoretically, at least.  Sometimes, in the real world, other constraints
> > > enter into it...
> > 
> > So you're saying that you can't find reliable ways to reproduce problems
> > on demand?  Those are some of the lower quality bug reports, so I don't
> > think we're losing much by having you not report them.
> 
> I'm sure that *everybody* on this list would *love* to know how you find
> a reliable way to reproduce all the bugs that start off with "after X days of
> uptime".   But when you're chasing what might be a race condition with a
> very small timing hole, you may need an event to happen several million times
> before the accumulated chance of hitting it becomes appreciable.
> 

I must say that the number of bugs which actually go away when the user
stops using nvidia/fglrx/ndiswrapper/etc is a small minority.

And you can usually tell beforehand too: if the user reports bad_page
warnings or pte table scroggage or whatever and they're using nvidia I just
hit 'd'.  But people who think that removing the nvidia driver will
magically fix that khubd-got-stuck-in-D-state bug are urinating up an
incline.


Facts:

- lots of people use nvidia/etc

- most bugs they report aren't caused by nvidia/etc

- we need lots of testers

draw you own conclusions.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/