[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: fstrim and Luks / dm-raid

On Sunday, October 06, 2019 04:32:36 AM Pascal Hambourg wrote:
> Le 05/10/2019 à 21:12, Reco a écrit :
> >>> The way I heard it, to trigger the corruption one should issue TRIM
> >>> asynchronously *and* utilize NCQ for it. fstrim is synchronous.
> >> 
> >> Asynchronous and synchronous to what ?
> > 
> > To SSD's I/O queue.
> Can you explain what it means or provide any pointers ?

Without looking it up, I'll try to explain synchronous and asynchronous in my 

For things to be synchronous means that they are done in a (prescribed or 
foreordained) sequence.  LIke standing in a bank queue with one teller, 
customers are taken in order and the business of one customer is finished 
before the next is started.

When things are asynchronous, they don't (necessarily) follow such a sequence, 
a new customer may be served before the previous is finished, and any customer 
in the queue may be served in any sequence, before or after the business of 
some previous customer is finished.

I guess, to be synchronous to the SSD's I/O queue means that the requests for 
service by the SSD are done in a defined sequence, like FIFO (First In, First 
Out -- service in the order of arrival at the queue -- no jumping the line),

Reply to: