Re: [PATCH] btrfs-progs: mkfs: use device_discard_blocks() to replace prepare_discard_device()

[Roman Mamedov <rm@romanrm.net>]

On Thu, 8 Jan 2026 15:40:07 +1030
Qu Wenruo <wqu@suse.com> wrote:

> Well, this is a special handling for full device trim, where we don't do 
> any special probing on if the device support TRIM.
> But based on the first discard try to determine and output needed message.
> 
> Either we need a reliable way to do the probe (which is not that simple, 
> previously we had some try to use 0 length discard, but it's not 
> reliable), thus we changed to use the current way.
> 
> So it's not that easy to change without changing the timing of the 
> message or the behavior.

One idea would be to first call device_discard_blocks with "0, 1M" arguments
(or less than 1M, maybe even 4K), and if that succeeds then print the message
and call it again with the full device size.

But I am not certain if this will turn out simpler or more straightforward
than the current code.

-- 
With respect,
Roman