The lifetime_ of the zone will be lost after remount zenfs

[ywang-wnlo]

zenfs/fs/zbd_zenfs.cc

Lines 51 to 64 in b04ca0c

	Zone::Zone(ZonedBlockDevice *zbd, ZonedBlockDeviceBackend *zbd_be,
	std::unique_ptr<ZoneList> &zones, unsigned int idx)
	: zbd_(zbd),
	zbd_be_(zbd_be),
	busy_(false),
	start_(zbd_be->ZoneStart(zones, idx)),
	max_capacity_(zbd_be->ZoneMaxCapacity(zones, idx)),
	wp_(zbd_be->ZoneWp(zones, idx)) {
	lifetime_ = Env::WLTH_NOT_SET;
	used_capacity_ = 0;
	capacity_ = 0;
	if (zbd_be->ZoneIsWritable(zones, idx))
	capacity_ = max_capacity_ - (wp_ - start_);
	}

The zone is configured with a lifetime_ of Env::WLTH_NOT_SET during initialization, but if the zone already has data(such as remount), it should be configured as the maximum value of the lifetime_ of the files in it.

zenfs/fs/zbd_zenfs.cc

Lines 393 to 413 in b04ca0c

	#define LIFETIME_DIFF_NOT_GOOD (100)
	#define LIFETIME_DIFF_COULD_BE_WORSE (50)
	unsigned int GetLifeTimeDiff(Env::WriteLifeTimeHint zone_lifetime,
	Env::WriteLifeTimeHint file_lifetime) {
	assert(file_lifetime <= Env::WLTH_EXTREME);
	if ((file_lifetime == Env::WLTH_NOT_SET) ||
	(file_lifetime == Env::WLTH_NONE)) {
	if (file_lifetime == zone_lifetime) {
	return 0;
	} else {
	return LIFETIME_DIFF_NOT_GOOD;
	}
	}
	if (zone_lifetime > file_lifetime) return zone_lifetime - file_lifetime;
	if (zone_lifetime == file_lifetime) return LIFETIME_DIFF_COULD_BE_WORSE;
	return LIFETIME_DIFF_NOT_GOOD;
	}

Due to the fact that the lifetime_ of zones is related to the zone allocation algorithm, which will always return LIFETIME_DIFF_NOT_GOOD as the zone_lifetime is Env::WLTH_NOT_SET

[ywang-wnlo]

I tried to fix the problem in this commit

I will submit a pull request if needed

[yhr]

@ywang-wnlo , this looks like a good fix, please submit a PR. Pardon the delay in my response :)