singlechair's profile - activity

Edited what I've already tried above to include other distros

Dell replaced heatsink and video card but problem persists.

Hi Everyone,

When I try to run computationally intensive applications (e.g. matlab, finite element simulations), I'm getting frequent notifications that a problem occurred. The frequency of the notifications increases with the size of the computation. On my older computer I was running these applications using RHEL 5 without a problem...

Looking at the details using the Problem Reporting GUI, I see that it is the kernel package (name: kernel-core). The last several lines of dmesg state that the CPU core temperature is above threshold, CPU clock throttled; then that CPU package temperature/speed is back to normal; and mce reports [hardware error]. Interestingly, the first notification appears pretty much at the start of the computation (seems strange that temperature at the CPU core would rise that fast). Computer is also relatively not that hot to touch.

Here are some details on the computer: M4800 Dell; Fedora 23 workstation 64bit; 32GB RAM, NVIDIA Quadro K2100M (using NVIDIA driver 358.16); processor is an Intel i7-4940mx @3.1 GHz. Note, when I run:

$ cat /proc/cpuinfo

It lists the correct cpu but shows cpu at 3300.000 MHz...

When running [watch sensors] the cpu temp is at about 45C when the computer is idle. When the computations are running, it's reading between 75C-96C.

I have tried the following without a successful resolution (problem still occurs).

• Updated to latest version of the BIOS
• Contacted Dell - ran diagnostics to stress system - no problems detected (and no thermal events recorded in the BIOS)
• Turbo ON and with it OFF in the BIOS and the OS
• With c-states OFF in BIOS (which should not matter considering OS runs the c-states regardless of BIOS setting)
• Thermald running; TLP running; Both Thermald and TLP running
• Changed x86_energy_perf_policy from normal to performance
• Changed NVIDIA setting from adaptive to performance
• Tried both the current kernel (4.3.3-300) and the previous running kernel (4.2.8-300)
• Tried at least one distribution from the major families: Debian 8.2, CentOS 7, StressLinux, Gecko Linux 3.16.2, Salix XFCE 14.1
• Sent computer back to Dell for hardware testing: they changed heatsink and videocard and called it a day

Computer is otherwise running great. I like Fedora and would like to solve problem without having to switch to Windows (booo). Have you seen this problem before? Is there a common solution that doesn't totally reduce the performance of these computationally intensive applications when solving? It seems to me that its either the CPU or a poor configuration (cpu, heatsink, fan) design in this notebook.

Any suggestions to solve this problem will be greatly appreciated. Thanks in advance!

A bit more info in your post (e.g. installed graphics driver) may help the community better diagnose your problem. You may want to start by booting into run level 3 and doing a

$ sudo dnf update

I read somewhere (don't remember where) that this is now the default (ie: /tmp type tmpfs). I confirmed this to be the case on my system running: $ mount | grep -i tmp

Good news! Before shutting down yesterday, I did my daily sudo dnf update. This morning, when I started the system, I noticed a new kernel (kernel-4.2.7-300.fc23.x86_64). I checked the 'Command Line' with dmesg and all still looked the same. I then checked the status of cryptsetup and finally the discard flag appeared. Ran fstrim and got a successful TRIM of encrypted ssd (all but the swap partition). Thanks to all that offered help. Moofed's answer is finally working with my system.

Thanks for your comment. It must be something with my specific setup. Still trying to figure it out. I have not added the discard option in /etc/fstab as described by Moofed and recommended against by others. @vitormp and @Florian, are you using discard in your /etc/fstab? Thanks!

Note: some of the power options you may be looking for can be set in /etc/systemd/logind.conf

I would also suggest you add some codecs. Follow configuration directions to enable RPM Fusion:

http://rpmfusion.org/Configuration

Install multimedia codecs:

sudo dnf install gstreamer1-plugins-* ffmpeg

Or if you want just some of the essential ones:

sudo dnf install gstreamer1-plugins-base gstreamer1-plugins-good gstreamer1-plugins1-good-extras gstreamer1-plugins-ugly gstreamer1-plugins-bad-free gstreamer1-plugins-bad-free gstreamer1-plugins-bad-freeworld gstreamer1-plugins-bad-free-extras ffmpeg

Follow configuration directions to enable RPM Fusion:

http://rpmfusion.org/Configuration

Install multimedia codecs:

sudo dnf install gstreamer1-plugins-* ffmpeg

Or if you want just some of the eesential ones:

sudo dnf install gstreamer1-plugins-base gstreamer1-plugins-good gstreamer1-plugins1-good-extras gstreamer1-plugins-ugly gstreamer1-plugins-bad-free gstreamer1-plugins-bad-free gstreamer1-plugins-bad-freeworld gstreamer1-plugins-bad-free-extras ffmpeg

Any other troubleshooting suggestions? Although I knew it should not matter, I also tried stopping the fstrim service and timer and lowering SELinux policy to permissive while updating grubby. No luck. Still only /boot on the ssd is trimmed.

By partition do you mean hard drive? Not sure why you would want to install only one partition.

to check SELinux status:

$ /usr/sbin/getenforce

or

$ /usr/sbin/sestatus

To temporarily change SELinux policy (use 1 at the end for enforcing and 0 for permissive). E.g.:

$ sudo /usr/sbin/setenforce 1

To changed read/write permission, I often use the chown command

To check SELinux status:

$ sudo /usr/sbin/getenforce

To temporarily change SELinux policy (use 1 at the end for enforcing and 0 for permissive). E.g.:

$ sudo /usr/sbin/setenforce 1

To changed read/write permission, I often use the chown command

I'm currently manually entering the encryption key at boot. If that matters...

My boot is on the ssd and I'm booting using the correct kernel (4.2.6-301.fc23.x86_64). As I mention in my preliminary checks above, my drive does support TRIM. Returns:

Data Set Management TRIM supported (limit 1 block)

Yes. Output looks like:

BOOT_IMAGE=/vmlinuz-4.2.6-301.fc23.x86_64 root=/dev/mapper/fedora-root ro rd.lvm.lv=fedora/root rd.luks.uuid=luks-e35...db0 rd.lvm.lv=fedora/swap rhgb quiet rd.luks.options=discard

Not sure what the ro is doing in there but yes it shows up.

Did exactly as you suggested (all the steps again). Only difference now is that the discard flag that was showing up for hdd is gone. Ssd still doesn't see discard flag in crypttab. As before, when I run sudo fstrim --verbose --all; only the /boot is trimmed...

@Florian: Is your root directory encrypted? My root directory is encrypted.

@Florian: Thanks for your help. I don't have discard in fstab. Interestingly, looking at the status of cryptsetup for my ssd shows no discard flags but my hdd does show the flag (since discard is at the end of both lines in my crypttab). I understand the hdd will not TRIM and I did confirm that I did not mix up the hdd with the sdd drives. According to: https://wiki.archlinux.org/index.php/... For the root filesystems, TRIM support for SSD needs adding the right kernel parameter to the bootloaler configuration? Do we know what that is for Fedora?

Yes. Just tried it again and no discard flag appears when checking the status of cryptsetup and only the /boot partition is trimmed. I also tried $ sudo dracut -f -I /etc/crypttab with a reboot after editing crypttab.

Here is also what the linux command line looks in my /etc/default/grub:

GRUB_CMDLINE_LINUX="rd.lvm.lv=fedora/root rd.luks.uuid=luks-e35...db0 rd.luks.options=discard rd.lvm.lv=fedora/swap rhgb quiet"

Anything look strange here?

Yes. Thank you. I'll make the edit.

Fedora now automatically mounts /tmp to tmpfs

I have not been able to get the discard flag to show when I: sudo cryptsetup status luks-e35...db0

My crypttab file only includes the two drives:

luks-b88...c40 UUID=b88...c40 none discard

luks-e35...db0 UUID=e35...db0 none discard

Looks different than example in man crypttab (first column is not the name of the mount points eg: swap). Could this somehow be the problem? Computer boots up fine.

Thank you for your response. I'm still missing something. Still does not work for encrypted mounts. Does not see discard... As a note, from your directions concerning grub, I did not get the rd.luks.options=discard at boot in grub configuration.

I added discard in /etc/crypttab -tried adding it to both drives and just to the ssd: luks-b88...c40 UUID=b88...c40 none discard luks-e35...db0 UUID=e35...db0 none discard

Edited /etc/default/grub. Now looks like: "rd.lvm.lv=fedora/root rd.luks.uuid=luks-e35...db0 rd.luks.options=discard rd.lvm.lv=fedora/swap rhgb quiet

fstrim works only for /boot

Hi Everyone, This is my first post. I've been using Linux on and off for several years now but am a newbie to Fedora.

I performed a fresh Fedora 23 install and partitioned using LVM and ext4. I used the default anaconda encryption during install. The computer has both an SSD (sda) and an HDD (sdb).

After reading different posts all over the internet offering advice on how to optimize an SSD, I'm starting to get more confused than clear on what needs to be done to correctly fstrim my encrypted ssd, especially with recent advances and using Fedora 23. I can really use your expert help.

• Preliminary checks:

$ sudo blockdev --getalignoff /dev/sda I get a 0 (which as I understand confirms alignment)

$ su -c 'hdparm -I /dev/sda' | grep TRIM I get that TRIM is supported

My goal is to to use the fstrim.service to TRIM weekly and verify that it works. From my understanding (this may be wrong), before this will work correctly, I need to enable TRIM on all the filesystem layers (LVM, ext4, and crypt). To set this up, I edited as follows:

/etc/lvm/lvm.conf set issue_discards=1

/etc/crypttab

Tried three different ways to allow discard (none seem to work) by adding the following at the end of the lines in crypttab:

1. ...none allow-discards
2. ...none discard
3. ...none luks,discard

Then trying both:

• $ sudo dracut -f

and

• $ sudo dracut -f -I /etc/crypttab

Then after reboot:

$ sudo cryptsetup status luks-e34...11 (for the sda) I get no flags showing discard enabled

• I also read that I'm suppose to edit /etc/default/grub... I tried:

1. rd.luks.allow-discards=e43...11

2. rd.luks.options=discard

3. both (1) and (2) together

followed each time with:

$ sudo grub2-mkconfig -o /boot/grub2/grub.cfg; reboot

Starting fstrim:

$ sudo systemctl enable fstrim.timer;

$ sudo systemctl start fstrim.service;

$ sudo systemctl status fstrim.service;

Gives: fstrim.service - Discard unused blocks Loaded: loaded (/usr/lib/systemd/system/fstrim.service; static; vendor preset: disabled) Active: inactive (dead)

• Testing using:

$ sudo systemctl start fstrim

Gives nothing

$ sudo fstrim -all

Gives: fstrim: failed to parse length: 'l'

$ sudo fstrim -v /home

Gives: fstrim: /home: the discard operation is not supported

$ sudo lsblk -D

Does seem to show the appropriate DISC-GRAN (512B) and DISC-MAX (2G) for the ssd but shows DISC-ZERO to be 0. From my understanding, this is a sign that the command does not propagate.

Basically, when trying to test fstrim I get operation not supported. So, as you can see, I'm in need of help. Your feedback on how to correctly setup fstrim to TRIM weekly for an encrypted SSD in Fedora 23 will be greatly appreciated. Many thanks in advance.