Reading allegedly bad ZFS blocks

[Zaphod Beeblebrox]

I have a system I suspect of not being entirely sane.  It has accused some
files in my ZFS RAIDZ1 array of being corrupt --- ie: not enough information
to recover them.

However, I suspect strongly that the data may be OK and simply the system
itself (reading those blocks) was bad.

Now... I attach the ZFS array to a new system ... and the new system refuses
to check those files again --- they are "known" bad when I first import the
filesystem.

I would like the ability to read the underlying data again --- possibly
re-check the checksum ... or just get the whole file.

Current ZFS behaviour is to return EIO for any of the blocks it doesn't wish
to read.  Someone in the archives mentioned patching ZFS to ignore it's
checksums.  Strikes me this should be sysctl.

Could someone post that patch?