Messages in this thread | | | Date | Wed, 1 Sep 2004 14:54:28 -0400 (EDT) | From | Bill Davidsen <> | Subject | Re: Driver retries disk errors. |
| |
On Wed, 1 Sep 2004, Alan Cox wrote:
> On Mer, 2004-09-01 at 16:18, Bill Davidsen wrote: > > If would probably be good to retry "read what you were asked, nothing > > more" on error, to avoid passing back errors caused by readahead. I > > suspect this would avoid some issues reading data off CD as well, where > > one software can read clean and another ends with a short image and error. > > Sure but as I understand the block layer currently (and I may be missing > something in the 2.6 code) I can't do that from a driver. > Sorry, that was unclear. I was speaking of a general approach rather than what would be done in the driver. Clearly that's best done at a higher level. Drivers should not be making policy decisions of that type, but I don't think it's good to return a read error caused by data the program didn't request (ie. readahead).
Unless S.M.A.R.T is lying, that happens so seldom on disk that the overhead of a retry doesn't matter. And on CD it makes things work where currently they fail.
-- bill davidsen <davidsen@tmr.com> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |