dm-integrity

dm-integrity

[Gandalf Corvotempesta]

I'm testing dm-integrity.
Simple question: when corrupted data are found, repair is done
immediately or on next scrub?

This is what I have:

[ 6727.395808] md: data-check of RAID array md0
[ 6727.528589] device-mapper: integrity: Checksum failed at sector 0xe228
[ 6727.938689] md: md0: data-check done.
[ 6749.125075] md: data-check of RAID array md0
[ 6749.664269] md: md0: data-check done.

if repair is done immediatly, would be possible to add a single log
line saying that ?
something like:
[ 6727.528589] md: md0: Repaired data at sector 0xe228

Re: dm-integrity

[Song Liu]

+ dm-devel

On Sun, Jan 12, 2020 at 5:43 AM Gandalf Corvotempesta
<gandalf.corvotempesta@gmail.com> wrote:
>
> I'm testing dm-integrity.
> Simple question: when corrupted data are found, repair is done
> immediately or on next scrub?
>
> This is what I have:
>
> [ 6727.395808] md: data-check of RAID array md0
> [ 6727.528589] device-mapper: integrity: Checksum failed at sector 0xe228
> [ 6727.938689] md: md0: data-check done.
> [ 6749.125075] md: data-check of RAID array md0
> [ 6749.664269] md: md0: data-check done.
>
> if repair is done immediatly, would be possible to add a single log
> line saying that ?
> something like:
> [ 6727.528589] md: md0: Repaired data at sector 0xe228

I guess this belongs to dm-integrity instead of md?

Thanks,
Song

Re: dm-integrity

[Piergiorgio Sartor]

On Mon, Jan 13, 2020 at 10:11:00AM -0800, Song Liu wrote:
> + dm-devel
> 
> On Sun, Jan 12, 2020 at 5:43 AM Gandalf Corvotempesta
> <gandalf.corvotempesta@gmail.com> wrote:
> >
> > I'm testing dm-integrity.
> > Simple question: when corrupted data are found, repair is done
> > immediately or on next scrub?
> >
> > This is what I have:
> >
> > [ 6727.395808] md: data-check of RAID array md0
> > [ 6727.528589] device-mapper: integrity: Checksum failed at sector 0xe228
> > [ 6727.938689] md: md0: data-check done.
> > [ 6749.125075] md: data-check of RAID array md0
> > [ 6749.664269] md: md0: data-check done.
> >
> > if repair is done immediatly, would be possible to add a single log
> > line saying that ?
> > something like:
> > [ 6727.528589] md: md0: Repaired data at sector 0xe228
> 
> I guess this belongs to dm-integrity instead of md?

Eh, well, no.
He is asking about "md" in case the underlying
layer, dm-integrity in this case, returns an error.

This could be the case also if the HDD returns
a read error which the RAID will correct and,
if I get it right, rewrite.

But I guess "md" already returns where the
correction happened, isn't like that?
I recall seeing in the logs something about
it, but it was some time ago...

bye,

-- 

piergiorgio

Re: dm-integrity

[Song Liu]

On Mon, Jan 13, 2020 at 10:17 AM Piergiorgio Sartor
<piergiorgio.sartor@nexgo.de> wrote:
>
> On Mon, Jan 13, 2020 at 10:11:00AM -0800, Song Liu wrote:
> > + dm-devel
> >
> > On Sun, Jan 12, 2020 at 5:43 AM Gandalf Corvotempesta
> > <gandalf.corvotempesta@gmail.com> wrote:
> > >
> > > I'm testing dm-integrity.
> > > Simple question: when corrupted data are found, repair is done
> > > immediately or on next scrub?
> > >
> > > This is what I have:
> > >
> > > [ 6727.395808] md: data-check of RAID array md0
> > > [ 6727.528589] device-mapper: integrity: Checksum failed at sector 0xe228
> > > [ 6727.938689] md: md0: data-check done.
> > > [ 6749.125075] md: data-check of RAID array md0
> > > [ 6749.664269] md: md0: data-check done.
> > >
> > > if repair is done immediatly, would be possible to add a single log
> > > line saying that ?
> > > something like:
> > > [ 6727.528589] md: md0: Repaired data at sector 0xe228
> >
> > I guess this belongs to dm-integrity instead of md?
>
> Eh, well, no.
> He is asking about "md" in case the underlying
> layer, dm-integrity in this case, returns an error.
>
> This could be the case also if the HDD returns
> a read error which the RAID will correct and,
> if I get it right, rewrite.

I see. Thanks for the clarification.

>
> But I guess "md" already returns where the
> correction happened, isn't like that?
> I recall seeing in the logs something about
> it, but it was some time ago...

Right now, md_done_sync() doesn't really print any message. I think this is
easy to add. However, md check/recovery is at block granularity, so we
probably cannot print exact which sector got fixed.

Thanks,
Song

Re: dm-integrity

[Gandalf Corvotempesta]

Il giorno lun 13 gen 2020 alle ore 19:58 Song Liu <song@kernel.org> ha scritto:
> Right now, md_done_sync() doesn't really print any message. I think this is
> easy to add. However, md check/recovery is at block granularity, so we
> probably cannot print exact which sector got fixed.

Well, having the exact sector fixed is not mandatory, but really useful.
Better than nothing would be md logging that has fixed "something" and
it's location (a setor, a block, ....)
9 times out of 10 the dm-integrity log would be just above the md log,
so a match between dm-integrity sector and md block would be easy,
just read the line below or above.

I think that md should be a little bit more verbose on certain
actions. (in example, saying what was fixed, the last time of scrub
and it's result, as per my previous thread and so on)