Re: magic sysrq keys functionality

On Mon, 26 Jul 2004, Brian Fundakowski Feldman wrote:
> On Mon, Jul 26, 2004 at 11:49:55AM -0600, Scott Long wrote:
> > On Mon, 26 Jul 2004, Brian Fundakowski Feldman wrote:
> > B> On Sun, Jul 25, 2004 at 09:23:36PM +0000, bugghy wrote:
> > > > Yeah but it sometimes "freezes" (no reboot) ... and I'd rather umount my
> > > > filesystems before rebooting.
> > >
> > > SoftUpdates guarantess that your file systems will not get corrupt.
> > >
> >
> > This isn't entirely correct.  Softupdates guarantees that you won't get
> > corruption due to metadata pointing to invalid or stale data blocks.
> > That's not the same as guaranteeing that there won't be any corruption.
> > Write caching on the drive combined with an in-opportune power loss or
> > other failure can easily leave you with corrupt or incomplete metadata
> > and/or data blocks.  A panic while metadata is being committed to disk can
> > also leave the metadata highly inconsistent and prone to corruption.
> > This isn't to say the SU is bad or that other strategies are necessarily
> > better, just that there are definite risks.
>
> If you just want to generalize it, you can say that "SoftUpdates
> guarantees that your file systems will not get corrupt due to just
> software errors."  I don't particularly think not having UPS is a
> good idea, but those can fail, and even so the ordering is such
> that a truncated inode won't result in a truly corrupt filesystem,
> and the inode doesn't get written until its contents are written
> out.
>
> Also, hw.ata.wc really shouldn't default to 1.
>

GAH!  No, please don't start this war again!  The last time that we tried
turning this off in a release (4.1 IIRC), were were plagqued by months of
earthquakes, plagues, and deaths of first-born youngsters.  I 100% agree
that write caching in ATA is not compatible with data integrety, but the
ATA marketting machine has convinced us that cached+untagged speed is
better than uncached+tagged safety.  C'est la vie, or so they say here.

Scott