tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

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

Andrew Doran writes:
> On Thu, Nov 27, 2008 at 11:19:35AM -0600, Greg Oster wrote:
> > But I'm completely failing to see how the "raid1: IO failed after 5 retries
> ." 
> > can occur without an actual component failure...
> nosleep memory allocation?

That sort of 'out-of-memory' condition was my first reaction, but I
couldn't find anything yesterday... digging a bit deeper reveals there
is a spot where we're not willing to wait for the Disk Queue Data pool 
to return us an item.  The low-water mark is set to just to just 64 
elements, and the high water mark is at 256... those values should be 
enough for 10-43 disks to run at full capacity..  

This could also be triggered if getiobuf() returns NULL too...  Hmm..
(For those playing along at home, src/sys/dev/raidframe/
rf_diskqueue.c:rf_CreateDiskQueueData() is the place where the 
allocations could fail...)


Greg Oster

Home | Main Index | Thread Index | Old Index