Re: Subject: RFC: Read repair for md RAID1 after mirror read failures

On 7/17/26 12:55 PM, Roger Heflin wrote:
> On Fri, Jul 17, 2026 at 1:42 AM G.W. Kant - Hunenet B.V.
> <[email protected]> wrote:
>>
>> On 7/15/26 8:04 PM, Abd-Alrhman Masalkhi wrote:
>>> On Wed, Jul 15, 2026 at 19:41 +0200, Abd-Alrhman Masalkhi wrote:
>>>> Hi Dion,
>>>>
>>>
>>> I'm not entirely sure about the media refresh part on modern drives...,
>>> I'll have to look into that first.
>>>
>>
>> Maybe, we could ask the smartmontools people about this. They should
>> know this. But I have never read about a refresh action in their
>> documentation.
>>
>> Best regards,
>>
>> Dion
> 
> I am not sure why you believe the smartmontools people would have ANY
> idea about the refresh action.  Smartmontools simply exposes the
> internal drive counters that the drive manufacturer exposes and/or
> documents.
> 
> The refresh action is 100% entirely controlled by the drive's firmware
> and has nothing to do with the smartmontools, and the smart counters
> may not show any of the internal ECC forced rewrites.
> 
> As far as I can tell in general the drive does what it wants to
> (refresh or not, relocation bad sectors or not) and there is NO way
> that I have ever found to force the firmware do do something else.

When you refer to "the refresh action", are you referring to a 
documented firmware mechanism that periodically rewrites successfully 
readable sectors, or are you using the term more generally for any 
internal media maintenance the drive may perform?

I have not yet found documentation describing such a refresh mechanism 
for commodity HDDs. That is why I mentioned the smartmontools 
developers: not because I expect smartmontools to implement or control 
such a mechanism, but because they are familiar with the capabilities 
that modern drives expose through SMART and related standards.

I'm trying to understand what is publicly known about this.

Best regards,

Dion Kant