Re: panic: softdep_deallocate_dependencies: unrecovered I/O error (but no I/O error)

[Greg Oster <oster%cs.usask.ca@localhost>]

Brian Buhrow writes:
>       Hello.  I've seen that message on my 4.0 based system with a raid5 set
> with no underlying disk drive error messages.  What I think  it indicates
> is that the raid driver ran into a resource shortage, no buffers for
> example, before it even passed the i/o requests down to the disk layer.  In
> my case, it didn't cause a panic, and the machine is continuing to run some
> months later.

Hmmmm... This should be in a PR somewhere then... (RAIDframe 
shouldn't be returning EIO unless there really is an IO error on the 
underlying components..)

Later...

Greg Oster

> On Nov 25, 12:10pm, Edgar =?iso-8859-1?B?RnXf?= wrote:
> } Subject: Re: panic: softdep_deallocate_dependencies: unrecovered I/O error
> } > No other messages?
> } No, the lines immediately before the "raid1: IO failed after 5 retries." ar
> e
> } "wsdisplay0: screen n added (...)" from the last boot.
> } 
> } > Note that our softdep implementation has no capability to recover from
> } > I/O errors whatsoever, so it just panics.
> } Is this any different in the FreeBSD implementation?
> } 
> } However, it looks like what I had was not a real I/O error.
> } Would it be possible that the softdep code tried to write a block outside
> } the partition, and that failed without a further error message from the
> } disk driver?
> } 
> >-- End of excerpt from Edgar =?iso-8859-1?B?RnXf?=
> 

Later...

Greg Oster