RoE | Blog

Category: linux

  • Different compression formats for initramfs vs uncompressed

    I’ve been working on improving boot times lately. I was doing uncompressed images in dracut.conf on the assumption that loading is faster without involving decompression, but I never tested this theory until now.

     

    First I generated the images, got the final size, and the amount of time it took to create the images. This is handled by:

     

    clear && sudo -e ‘/etc/dracut.conf.d/99-custom.conf’ &&  sudo time dracut –force –regenerate-all && sudo time lsinitrd | grep ‘Image:’ 

    Next, I rebooted and immediately ran systemd-analyze

    In-short, my theory was right with cat (uncompressed) coming in first at 15.112s, but gzip, lzo, and lz4 were still within 15 seconds and are half the file-size!

    Testing done on a Dell Latitude 5591 (BIOS 1.32), openSUSE Tumbleweed (GNOME, 6.8.8-1-default), i5-8400H, SK hynix Gold P31 (1TB), UEFI mode, unencrypted with root on ext4.

    cat
    52M
    3.13user 1.35system 0:04.18elapsed 107%CPU
    0.25user 0.45system 0:00.41elapsed 169%CPU
    Startup finished in 4.491s (firmware) + 603ms (loader) + 452ms (kernel) + 3.874s (initrd) + 5.691s (userspace) = 15.112s
    graphical.target reached after 5.690s in userspace.

    gzip
    26M
    11.61user 1.28system 0:05.05elapsed 255%CPU
    2.18user 0.27system 0:02.08elapsed 118%CPU
    Startup finished in 4.490s (firmware) + 504ms (loader) + 621ms (kernel) + 4.266s (initrd) + 5.602s (userspace) = 15.485s
    graphical.target reached after 5.601s in userspace.

    bzip2
    26M
    6.82user 1.28system 0:07.79elapsed 104%CPU
    9.31user 0.41system 0:09.19elapsed 105%CPU
    Startup finished in 4.477s (firmware) + 503ms (loader) + 2.651s (kernel) + 4.274s (initrd) + 5.572s (userspace) = 17.479s
    graphical.target reached after 5.572s in userspace.

    lzma
    21M
    19.94user 1.32system 0:20.89elapsed 101%CPU
    5.89user 0.34system 0:05.74elapsed 108%CPU
    Startup finished in 4.513s (firmware) + 486ms (loader) + 2.369s (kernel) + 4.151s (initrd) + 5.626s (userspace) = 17.147s
    graphical.target reached after 5.626s in userspace.

    xz
    23M
    23.75user 1.47system 0:07.23elapsed 348%CPU
    4.82user 0.34system 0:04.65elapsed 110%CPU
    Startup finished in 4.592s (firmware) + 496ms (loader) + 1.450s (kernel) + 4.246s (initrd) + 5.904s (userspace) = 16.690s
    graphical.target reached after 5.904s in userspace.

    lzo
    26M
    11.58user 1.35system 0:05.06elapsed 255%CPU
    2.14user 0.31system 0:02.07elapsed 118%CPU
    Startup finished in 4.647s (firmware) + 510ms (loader) + 631ms (kernel) + 4.004s (initrd) + 5.861s (userspace) = 15.655s
    graphical.target reached after 5.861s in userspace.

    lz4
    26M
    11.60user 1.31system 0:05.06elapsed 255%CPU
    2.14user 0.30system 0:02.07elapsed 118%CPU
    Startup finished in 4.507s (firmware) + 507ms (loader) + 631ms (kernel) + 4.165s (initrd) + 5.891s (userspace) = 15.702s
    graphical.target reached after 5.891s in userspace.

    zstd
    25M
    8.24user 1.39system 0:05.94elapsed 162%CPU
    0.60user 0.45system 0:00.47elapsed 224%CPU
    Startup finished in 4.500s (firmware) + 501ms (loader) + 503ms (kernel) + 4.071s (initrd) + 7.000s (userspace) = 16.577s
    graphical.target reached after 7.000s in userspace.

  • Getting More Into Cryptography, Part 1

    I’ve been studying up on various methods for improving my home network security across devices, internet privacy, and encryption methods for a few months now (pretty casually though). After hearing about the awesome community-driven site privacytools.io, I decided to finally start trying to break some old habits.

    The first objective I have is to break free of Google services, at least primarily. For the most part, I rely heavily on Chrome Sync, Google Voice, Gmail, Google+, and YouTube.

    For Gmail, I figure I can simply just switch to another provider, and update email addresses accordingly. Two services I have in-mind are ProtonMail and Tutanota. ProtonMail seems to have a waiting list currently due to the a large volume of sign-ups lately, but Tutanota is instant-access. Both providers offer a free service, end-to-end encryption, and 1GB of storage. I don’t use email heavily for “important” matters, so 1GB should be more than enough (I’m only using about 0.6GB on Gmail currently). I want to give ProtonMail a go before deciding on which provider I want, but my short time with Tutanota was pleasant.

    Update: One advantage that Tutanota has over ProtonMail is that it is open-source. ProtonMail’s waiting list is also apparently somewhat random when it comes to wait times. With both of those in-mind, I’ll accept Tutanota as my main email provider.

    For a social network, I’m highly considering Friendica on a local, self-hosted machine. At quick glance, it seems all I would need is an Apache and MySQL server, and a general GUI setup process, which doesn’t seem too bad at all (I’ve set-up some WordPress and Joomla servers in the past), but I’ll have to figure out some more details on this.

    I mainly used Chrome Sync for passwords, and I never really got into using things like LastPass. Two services that seem really interesting are Master Password and SuperGenPass. The idea behind those two services is that you take a “master password”, add some details to it (like a website name), and you get a hash-generated password to use. The generated passwords should be the same across all devices, and you don’t need to rely on a sync service or password vault (so basically, as long as you remember your master password, you’re good to go).

    My browser of choice for a good while was Google Chrome, mainly for Chrome Sync, but also because it worked well with pretty much anything I threw at it. Firefox in-comparison has/had bad hardware-acceleration support and inconsistent video playback abilities on Linux. As I’m typing this however, I’m running Firefox 38 on Fedora 22, and my experience for the most part has been pretty good with minor issues. I may give IceWeasel a try in the future (due to some more questionable Firefox additions by Mozilla), but it being pretty behind Firefox in versions is somewhat concerning.

    So pretty much, I’m in a planning phase currently. Transitioning data to new services and trying to get used to them will take a bit of time, but I’m sure it’ll pay off in the end along with being a great learning experience.

    As for what I’ve done so far, I flashed my D-Link DSL-2750B modem from Verizon firmware to stock firmware. Verizon hasn’t updated the firmware on these modems since 2013, but D-Link does release updates occasionally for this device (the latest update being on 2015-03-25). Updates are nice and all, but generally speaking, I trust D-Link a little bit more than Verizon when it comes to modem firmware from a security-standpoint.

    I’ve also blocked outside access to and from my Seagate NAS. I only use the NAS locally, so this isn’t a problem at all, but according to a reviewer, Seagate “reserves the right” to upload files from the NAS without consent according to the license agreement. Not sure on the validity of this claim, but better to be safe than sorry I suppose. I want to upgrade to a FreeNAS solution at some point in the future.

    And finally, I’ve stopped using Windows as a primary operating system. I mainly used Windows because of gaming, but PC gaming is meaning less and less to me nowadays. I’ve been using Linux primarily for a bit now, and my distro of choice currently is Fedora.

  • Story of my Gateway VX1120 and Custom Refresh Rates on Windows and Linux

    A few months ago I picked up a Gateway VX1120 from a computer store in town for $5, not knowing if it would work or not. I was always a fan of CRT monitors, and seeing something of this size, I had to jump on it.

    Got it home, plugged it in, and found out that it worked fine. Instantly replaced the Acer S201HL I was using without regrets. Colors were more vibrant, and animations are much smoother (one of my favorite things to do when switching from LCD to CRT is to move a window across my desktop to observe the smoothness). Plus the resolutions available were extensive, so I was sure I could find the right resolution to use daily. The other CRT monitors I have in the house were limited to around 1024×768, or something slightly higher at a lower refresh rate, so they weren’t too ideal.

    After some time, I found that the resolution of 1280×960 suited perfectly as a default resolution. It’s not high enough to make text hard to see, and at the same time, it’s large enough for the desktop not to feel cramped. The highest stock refresh rate for this resolution was 85Hz, which was great and all, but I soon looked into a way to improve that.

    On Windows, I used a tool called Custom Resolution Utility (CRU). It allowed me to generate custom resolutions and refresh rates. I wanted to run at 120Hz just to be able to claim I had such a monitor, so that was the first refresh rate I tried. It didn’t work; and the monitor complained about the frequency being out-of-range.

    After a little bit of research, I found that my horizontal scan frequency could go as high as 121kHz. So going by that, I then had the idea to keep trying resolutions with CRU until my horizontal scan frequency was just under that limit. Lucky me, 119Hz seemed to have worked fine.

    So in Windows, the refresh rate wasn’t too hard to handle. Things got a bit more interesting when I went over to Linux however. I use an AMD Radeon HD 7850 graphics card (this specifically), so I have the choice of either using AMD’s proprietary driver (fglrx), or the open-source drivers. The refresh rate on my screen however seems to act differently depending on which driver I use though.

    If I recall correctly, I was able to use the CVT-generated 119Hz with fglrx without issue (it’s been a while since I used fglrx with this monitor), but this wasn’t the case with the open-source driver.

    Looking back at cvt, I noticed that my hsync value was 122.59 for 119Hz. Definitely over the 121kHZ limit, so then I tried going to 118Hz. It ended up at 121.45, which was still over the limit. I wasn’t sure if 121kHZ was a hard limit, so I added the refresh rate and tried it anyway. Both 119Hz and 118Hz failed to work; so I tried 117Hz. Got the hsync value of 120.31, which was safely under the limit, and worked fine.

    I’m unsure how CRU and cvt generate the numbers they do, but a comparison at some point would be a great idea, whenever I find the time to do so.

    So far, this monitor is still working great, and I have no plans of replacing it anytime soon, unless I go back to my standing desk setup.