You are not logged in.
I was wondering whether the lsinitcpio "Estimated extraction time" would include or exclude disk i/o.
To test this, I made both speed (lz4 -l; 5.2MiB) and size (lzma -9; 2.5MiB) copies of the initramfs file and copied these files also onto a much slower USB2 stick (flash drive).
hdparm -t /dev/sdX1stick A: 132 MB/sec (USB-3)
stick B: 36 MB/sec (USB-2)
I noticed that lsinitcpio estimated extraction time results from both drives were near identical. Hmmm, what could that be? Ah, disk caching.
After purging the disk caches with
sync && echo 3 | tee /proc/sys/vm/drop_cachesin between each lsinitcpio, extraction time results are different:
stick A 5.2MiB initramfs.img: Estimated extraction time: 0.077s (cached: 0.028s)
stick B 5.2MiB initramfs.img: Estimated extraction time: 0.171s (cached: 0.029s)
stick A 2.4MiB initramfs.img: Estimated extraction time: 0.514s (cached: 0.497s)
stick B 2.4MiB initramfs.img: Estimated extraction time: 0.512s (cached: 0.497s)
Would lsinitcpio need some wording/warning that the estimates are cached results?
---
==> Image: /boot/initramfs-linux-tiny.img.lzma
==> Created with mkinitcpio 24
==> Kernel: 4.13.12-1-ARCH
==> Size: 2.46 MiB
==> Compressed with: lzma
-> Uncompressed size: 9.28 MiB (.265 ratio)
-> Estimated extraction time: 0.497s
==> Included modules:
atkbd [explicit] hid scsi_mod usb-storage
btrfs [explicit] hid-generic [explicit] sd_mod [explicit] usbcore
crc32c-intel [explicit] i8042 [explicit] serio usbhid [explicit]
crc32c_generic libcrc32c serio_raw [explicit] xhci-hcd
ehci-hcd libps2 uas [explicit] xhci-pci [explicit]
ehci-pci [explicit] raid6_pq usb-common xor
==> Included binaries:
blkid btrfsck dosfsck fsck.btrfs kmod switch_root
btrfs busybox fsck fsck.vfat mount
==> Hook run order:
without-udevLast edited by probackup-nl (2017-11-23 01:43:42)
Offline
I suppose the idea is to give estimate of initramfs extraction time during system boot-up, which should always be non-cached reads. Therefore lsinitcpio reporting estimate based on cached reads would seem like uninteded behavior.
In any case, I don't think the estimate should differ between subsequent runs.
I think this is something that should be reported as a bug.
Offline
https://bugs.archlinux.org/task/56417
Last edited by probackup-nl (2017-11-25 15:24:25)
Offline