Re: [smartmontools-support] Several questions regarding smartctl

[Bruce A. <ba...@gr...>]

Hi Alexander,

See comments below.

> First of all, I'd like to thank all of the people involved with 
> smartmontools. The work you're doing is invaluable.

Thanks!

> I'm writing a graphical interface for smartctl in gtkmm, the basic 
> functionality is almost complete and there are just a few things left to 
> solve.

Will this graphical interface be usable under all the OSes that 
smartmontools supports?  If so, I'm interested to learn some more about 
it.

> I was wondering if you could shred some light on a couple of
> questions I have which I couldn't find any answers for elsewhere.
>
>
> 1. Is the state of --offlineauto=value preserved across resets?

I'm not sure.  This was part of the ATA-3 and SFF-8035i specs. I suggest 
you look at those to find out.


> 2. Regarding --saveauto=on: I'm don't quite understand what exactly it
>    does. Does it enable remembering of device attributes (--attributes)?
>    Or just the worst values? Does it mean that if it's disabled, no
>    attributes will be updated ever?

Again, I'm not sure.  I suggest you check the documentation above to see 
if there are any hints.

> Is there any gain in disabling it? That is, is there any need to provide
> an option for it at all? Right now I just enable it automatically whenever the user
> chooses to enable SMART, so is it ok?

I think that's fine.


> 3. Can the --all option be harmful in any way? That is, can a drive hang
>    or suffer some kind of damage with --all, but not with --info? The
>    reason I ask is that I have to decide - do I retrieve the complete
>    information when scanning for drives, or just the basic one (--info)
>    if it's unsafe and do --all only when the user requests it.

In the six years since smartmontools was released, I have never heard of 
--all causing trouble.

> 4. This is related to question number 3: Are there any harmful
>    operations smartctl can perform (apart from captive tests, which
>    AFAIK may hang an OS)? For example, can a selftest hang a drive until
>    reset, or just plain damage it (by incorrect reallocating, etc...)?
>    What about informational options - --capabilities or --health?

As far as I know, captive self-tests are the only 'dangerous' operation.

> I need to provide warnings whenever the user chooses an option which
> may be potentially harmful.
>
> 5. What does "self-test routine recommended polling time" mean? Does it
>    mean that I shouldn't execute "-l selftest" before that time passes?
>    (It's doubtful, but that's what polling usually means). My drive's
>    "extended" one is 227min, which is about the time the whole extended
>    selftest needs to complete.

This time is the approximate time that a drive will take to run a 
self-test, when otherwise idle.  If the drive is also doing read/write 
operations, then the self-test will take longer.

> 6. Can a sector be reallocated during reads? (I need to provide a correct
> description for reallocations).

Yes, sometimes, if the read is successful.  For example of the drive has 
to retry a read of a sector 100 times and finally this succeeds, the drive 
may determine that the sector is bad and then reallocate it.

> I found conflicting information on this: some say that reallocations are 
> performed only during writes, while others say that if a hard drive has 
> trouble reading a sector (say, 9 times of 10), it marks it "unstable" 
> and may reallocate it whenever it manages to read it.

This is drive- and manufacturer-dependent.

> I know that I asked A LOT of questions, and I apologize if some of them
> are too obvious or involve my incorrect understanding of the problems.

To polish your knowledge, I suggest that you get a copy of the ATA specs 
on-line (for example the ATA-8 draft spec) and read the SMART section.

Cheers,
 	Bruce