Web lists-archives.com

Re: all files moved to lost+found




On Mon, 1 Oct 2018 23:18:13 -0700 David Christensen said:
> On 10/1/18 8:40 PM, Abdullah Ramazanoğlu wrote:

> > I'm not sure how an on-disk cache problem could definitively be caught
> > without power cycling. What if on-disk controller is ignoring all cache
> > related commands? (cache bypass, cache invalidate, cache flush)  
> 
> I must depend upon Seagate to get their firmware and their diagnostics 
> right.  If their firmware were so bad as to ignore all cache related 
> commands, we'd probably hear about it.

I am not questioning the quality of the diagnostic tool, but there can
be situations where it is technically impossible to detect a failure. For
instance, a destructive cache test can write some data to disk, flush the
cache, invalidate the cache, and then reread the data just written. But if both
cache flush and cache invalidate don't work, then this scheme will produce a
false positive result.

* Write test data to disk --> OK

* Flush on-disk cache --> Controller returns "OK" but doesn't do it.

* Invalidate the cache --> Controller again returns "OK", without doing it.
  Now, diag tool thinks the data is recorded on media and the cache is clear.
  In fact, it is not recorded and is solely on cache.

* Reread the data --> Success. Controller reads it from cache, while the diag
  tool thinks it is coming from the media.

In such a case, there is no way the diag tool can definitively test on-disk
cache.

Only sure way is power-cycling, as far as I can see.

> Then again, as the OP's computer seems to be a newer model, some 
> hardware could be unsupported and/or the supporting device drivers could 
> be incomplete and/or buggy.  I know Intel provides reference FOSS 
> drivers for their stuff; does AMD?  Without a shopping list of the key 
> chips in the OP's computer, there is no way to check support status.

Agreed.

Regards
-- 
Abdullah Ramazanoğlu