Re: [PATCH storage v11 2/6] lvm: saferemove: zero out volumes range by range

["Michael Köppl" <m.koeppl@proxmox.com>]

On Thu Jul 23, 2026 at 3:56 PM CEST, Lukas Sichert wrote:

[snip]

> +                } elsif ($zeroout_variant eq 'syswrite') {
> +
> +                    # allow retrying once if syswrite writes zero bytes
> +                    $written = syswrite($fh, $zeroes, $stepsize, 0);
> +                    if (!defined($written)) {
> +                        die "syswrite failed: $!\n";
> +                    } elsif ($written == 0) {
> +                        warn "syswrite wrote 0 bytes, retrying";
> +                    }
> +
> +                    while ($written < $stepsize) {
> +                        my $remaining = $stepsize - $written;
> +                        my $retried_write = syswrite($fh, $zeroes, $remaining, $written);
> +                        if (!defined($retried_write)) {
> +                            die "syswrite failed: $!\n";
> +                        } elsif ($retried_write == 0) {
> +                            die "syswrite failed: wrote 0 bytes";
> +                        }
> +                        $written += $retried_write;
> +                    }

How do we ensure here that all data is actually zeroed out on disk as
well? Do we flush to disk in some way? Because if not, this might run,
then immediately discard and not all data might have been zeroed out.
Perhaps fsync? I think in Perl there'd be $fh->sync. Or am I missing
something here and this is already guaranteed to happen?

> +
> +                }
> +                $written_total += $written;
> +
> +                my $curr_time = clock_gettime(CLOCK_MONOTONIC);
> +                if (($curr_time - $lastprint) >= 3) {
> +                    my $percent_finished = 100 * $written_total / $size;

[snip]