On Tue, 02 Oct 2012 21:35:05 +0100 John Robinson <john.robinson@xxxxxxxxxxxxxxxx> wrote: > On 02/10/2012 21:29, NeilBrown wrote: > > On Tue, 02 Oct 2012 20:51:31 +0100 John Robinson > > <john.robinson@xxxxxxxxxxxxxxxx> wrote: > > > >> On 02/10/2012 16:56, Brian Candler wrote: > >>> On Tue, Oct 02, 2012 at 03:56:42PM +1000, NeilBrown wrote: > >>>> Lesson: always do a read test as well as a write test! > >>> > >>> And there was me thinking that drives read back data after writing it - > >>> clearly they do not :-) > >> > >> Would it be worth adding a re-read onto the end of the usual read > >> failure reconstruct-write cycle? > > > > What? Read twice? > > > > We already read after the write when attempting to fix a failure. > > I didn't realise you already did that. > > > I suspect it doesn't do much good though as the data is in cache in the drive. > > No, maybe not. There isn't an option you can add to ask for an uncached > read? But I dare say you'd have thought of that already if there was :-) Maybe setting REQ_FUA would work. Or maybe it would cause crashes and random data corruption. I suspect most people think of REQ_FUA as being associated with WRITEs. Maybe I'm to cynical, but this would be extremely hard to test, and I don't feel up to the code review that would be required to provide sufficient confidence :-( NeilBrown > > Cheers, > > John. > > -- > To unsubscribe from this list: send the line "unsubscribe linux-raid" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html
Attachment:
signature.asc
Description: PGP signature
