Re: [PATCH] btrfs: raid56: Use correct stolen pages to calculate P/Q

[Andrei Borzenkov <arvidjaar@gmail.com>]

28.11.2016 06:37, Christoph Anton Mitterer пишет:
> On Sat, 2016-11-26 at 14:12 +0100, Goffredo Baroncelli wrote:
>> I cant agree. If the filesystem is mounted read-only this behavior
>> may be correct; bur in others cases I don't see any reason to not
>> correct wrong data even in the read case. If your ram is unreliable
>> you have big problem anyway.
> 
> I'd agree with that - more or less.
> 
> If the memory is broken, then even without repairing (on read) a
> filesystem that is written to will likely be further corrupted.
> 
> 
> I think for safety it's best to repair as early as possible (and thus
> on read when a damage is detected), as further  blocks/devices may fail
> till eventually a scrub(with repair) would be run manually.
> 
> However, there may some workloads under which such auto-repair is
> undesirable as it may cost performance and safety may be less important
> than that.
> 
> Thus I think, there should be a mount-option that let users control
> whether repair should happen on normal reads or not... and this should
> IMO be independent of whether the fs was mounted ro or rw.

If you allow any write to filesystem before resuming from hibernation
you risk corrupted filesystem. I strongly believe that "ro" must be
really read-only - having separate option to control it will require
update of every tool that generates initramfs and even then you cannot
avoid using older initramfs with newer kernel.

> I'd say the default should go for data safety (i.e. repair as soon as
> possible).
> 

Safety means exactly opposite for me - do not modify data if explicitly
requested not to do it. Having filesystem in half-read-only state will
be too error prone.