[Opinion] [Bug 15324] Random hard freezes with ATAPI drive on an SATA controller

[Ben Collins]

On Mon, 2005-12-12 at 14:36 +0200, Sandis Neilands wrote:
> Hi! 
> 
> There was an issue with pmount (it was impossible to access floppy
> drives from gnome) which was solved by putting pmount from dapper into
> backports. I don't know if it's possible to do the same with kernel
> though. 
> 
> However this is major flaw in the official update policy. From the
> user's perspective if the key hardware is not working even after
> installing official updates, the distro is broken and it's time to
> switch - again. Of course, home users and hobbyist could use dapper,
> but what about enterprises, schools and other organizations, where
> stability AND working hardware is crucial? As for testing and new bugs
> - I'm 100% sure that people will give you more feedback using updated
> breezy, rather than using fedora core, suse, mandrake or debian.
> 
> IMO hardware-not-working bugs should be treated as important as
> security bugs.

The problem is that most times these bugs aren't fixed by a trivial
patch. Quite often, "fixing" the bug was done by upstream kernel, at the
expense of major rewrites to the subsystem in question (acpi and ata are
good examples of this). One good reason we wouldn't backport such a
change is that in a lot of cases, we have no idea what fixed it (just
that the bug was reported fixed in linux 2.6.x, and so we accept that it
is).

The upstream kernel also has a policy of not retaining compatibility
code. So it's not like we can grab the ata subsystem from 2.6.15, and
drop it into our 2.6.12 kernel.

Even if we can identify the patch(es), and yes, they have fixed your
problem, this code is so new, we don't know if it breaks things for
someone else.

Bottom line is generally, we only consider regressions to be critical
for hardware failures (works in hoary, not in breezy). Even then, after
the dist has been released, there's only slight changes that things will
get fixed, assuming it is possible.

I am hoping that this improves post-dapper. I went through the kernel
bug list, and am busy trying to get it manageable. Once it is at a point
where maintaining the bugs isn't such a chore, then hopefully we can
revisit our policy on these sorts of things.

-- 
   Ben Collins <ben.collins at ubuntu.com>
   Developer
   Ubuntu Linux