Re: [RFC] PATCH to fix rescan_partitions to return errors properly - take 2

[Suzuki K P <suzuki@in.ibm.com>]

Erik Mouw wrote:
> On Thu, Oct 05, 2006 at 01:32:34PM -0700, Suzuki Kp wrote:
> 
>>Btw, do you think it is a good idea to let the other partition checkers 
>>run, even if one of them has failed ?
> 
> 
> Yes, just let them run. Partition information doesn't need to be on the
> very first sector of the drive. If the first sector is bad and the
> partition table for your funky XYZ partition table format lives on the
> tenth sector, then a checker that checks the first sector would fail
> and prevent your checker from running.
> 
> OTOH: having ten partition checkers check the same bad first sector
> doesn't really speed up the partion check process (for that reason we
> disable partition checking for drives we get for recovery). A way to
> solve that would be to keep a list of bad sectors: if the first checker
> finds a bad sector, it notes it down in the list so the next checker
> wouldn't have to try to read that particular sector. Maybe that's too
> much work to do in kernel and we'd better move the partition checking
> to userland.
> 
> 
>>Right now, the check_partition runs the partition checkers in a 
>>sequential manner, until it finds a success or an error.
> 
> 
> I think it's best not to change the current behaviour and let all
> partition checkers run, even if one of them failed due to device
> errors. I wouldn't mind if the behaviour changed like you propose,
> though.
> 
At present, the partition checkers doesn't run, if one of the preceeding 
checker has reported an error ! *But*, some of the checkers doesn't 
report the I/O error which they came across! So, this may let others 
run. Thats not we want, right. We would like them to return I/O errors, 
and and the check_partition should let other partition checkers continue.

Comments ?

Thanks,

Suzuki
> 
> Erik
>