Re: Improving Linux laptop battery life: Testers Wanted
Hi Hans, do you have any F27 kernel to test SATA Link Power Management (LPM)? Best regards ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
On Mon, Oct 23, 2017, 3:13 PM Hans de Goedewrote: > Hi, > > On 21-10-17 20:50, Kevin Fenzi wrote: > > On 09/14/2017 04:22 AM, Hans de Goede wrote: > >> Hi All, > >> > >> My next project for Red Hat is to work on improving Linux laptop battery > >> life. > >> Part of the (hopefully) low hanging fruit here is using kernel tunables > to > >> enable more runtime powermanagement. My first target here is SATA Link > >> Power > >> Management (LPM) which, as Matthew Garrett blogged about 2 years ago: > >> https://mjg59.dreamwidth.org/34868.html > >> can lead to a significant improvement in battery life. > >> > >> There is only one small problem, there have been some reports that some > >> disks/SSDs don't play well with Linux' min_power LPM policy and that > this > >> may lead to system crashes and even data corruption. > >> > >> As such I've written a new LPM policy, which matches the > power-management > >> defaults from the Intel RST Windows drivers. Since it mimicks Windows, > >> this new policy will hopefully not hit any SSD firmware bugs like > min_power > >> sometimes does. > >> > >> So now I'm looking for people with a laptop with a SATA SSD or HDD to > help > >> me test this to make sure this won't cause any issues when we enable > this > >> by default for F28, for more details and test instructions see: > >> > >> https://hansdegoede.livejournal.com/18412.html > > > > Hey Hans. I had marked this to try and do to help gather info, but > > things have been busy and I haven't ever gotten to it. > > > > I thought before I did I might ask if you already have the info you were > > looking for since it's been over a month now. > > > > How have reports been? Is it perhaps time to enable this in rawhide? > > I've had successful test reports from 10 people, of which 1 person > was actually seeing data corruption on his ssd with min_power and not > with the new med_power_with_dipm policy my patches add, which is good > news really :) > > But 10 testers is not that much, so if you still feel like testing this, > then some more testing would definitely be welcome. > I'm curious what the manifestation of corruption is, and whether any on the 10 were using Btrfs? Chris Murphy > ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi, On 21-10-17 20:50, Kevin Fenzi wrote: On 09/14/2017 04:22 AM, Hans de Goede wrote: Hi All, My next project for Red Hat is to work on improving Linux laptop battery life. Part of the (hopefully) low hanging fruit here is using kernel tunables to enable more runtime powermanagement. My first target here is SATA Link Power Management (LPM) which, as Matthew Garrett blogged about 2 years ago: https://mjg59.dreamwidth.org/34868.html can lead to a significant improvement in battery life. There is only one small problem, there have been some reports that some disks/SSDs don't play well with Linux' min_power LPM policy and that this may lead to system crashes and even data corruption. As such I've written a new LPM policy, which matches the power-management defaults from the Intel RST Windows drivers. Since it mimicks Windows, this new policy will hopefully not hit any SSD firmware bugs like min_power sometimes does. So now I'm looking for people with a laptop with a SATA SSD or HDD to help me test this to make sure this won't cause any issues when we enable this by default for F28, for more details and test instructions see: https://hansdegoede.livejournal.com/18412.html Hey Hans. I had marked this to try and do to help gather info, but things have been busy and I haven't ever gotten to it. I thought before I did I might ask if you already have the info you were looking for since it's been over a month now. How have reports been? Is it perhaps time to enable this in rawhide? I've had successful test reports from 10 people, of which 1 person was actually seeing data corruption on his ssd with min_power and not with the new med_power_with_dipm policy my patches add, which is good news really :) But 10 testers is not that much, so if you still feel like testing this, then some more testing would definitely be welcome. Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
On 09/14/2017 04:22 AM, Hans de Goede wrote: > Hi All, > > My next project for Red Hat is to work on improving Linux laptop battery > life. > Part of the (hopefully) low hanging fruit here is using kernel tunables to > enable more runtime powermanagement. My first target here is SATA Link > Power > Management (LPM) which, as Matthew Garrett blogged about 2 years ago: > https://mjg59.dreamwidth.org/34868.html > can lead to a significant improvement in battery life. > > There is only one small problem, there have been some reports that some > disks/SSDs don't play well with Linux' min_power LPM policy and that this > may lead to system crashes and even data corruption. > > As such I've written a new LPM policy, which matches the power-management > defaults from the Intel RST Windows drivers. Since it mimicks Windows, > this new policy will hopefully not hit any SSD firmware bugs like min_power > sometimes does. > > So now I'm looking for people with a laptop with a SATA SSD or HDD to help > me test this to make sure this won't cause any issues when we enable this > by default for F28, for more details and test instructions see: > > https://hansdegoede.livejournal.com/18412.html Hey Hans. I had marked this to try and do to help gather info, but things have been busy and I haven't ever gotten to it. I thought before I did I might ask if you already have the info you were looking for since it's been over a month now. How have reports been? Is it perhaps time to enable this in rawhide? kevin signature.asc Description: OpenPGP digital signature ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
On Wed, Sep 27, 2017 at 11:54:29PM +0200, Germano Massullo wrote: > > > It might be that systemd does not like that rc.local is owned > > by "user" rather then "root" > Unfortunately > # chown root:root /etc/rc.d/rc.local > did not solve the problem, instead manually running the script as root, > changed the > /sys/class/scsi_host/host0/link_power_management_policy > state to > med_power_with_dipm (No idea why rc.local is not working for you.) A simpler option might be to use add a tmpfiles.d entry: # /etc/tmpfiles.d/sata.conf w /sys/class/scsi_host/host*/link_power_management_policy - - - - med_power_with_dipm Zbyszek ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
> It might be that systemd does not like that rc.local is owned > by "user" rather then "root" Unfortunately # chown root:root /etc/rc.d/rc.local did not solve the problem, instead manually running the script as root, changed the /sys/class/scsi_host/host0/link_power_management_policy state to med_power_with_dipm In powertop I noticed a saving of ~1 Watt. I will send you complete test results as you requested in https://hansdegoede.livejournal.com/18412.html For people who wants to stress the disk: # dnf install phoronix-test-suite $ phoronix-test-suite install disk $ phoronix-test-suite run disk I will run disk suite tests many times per day for 2 weeks ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi, On 26-09-17 16:59, Germano Massullo wrote: This message may be useful to others having my similar problem. On a Thinkpad X220, after having booted the custom kernel, running [root@machine]# cat /sys/class/scsi_host/host0/link_power_management_policy I get max_performance instead of med_power_with_dipm Here some machine details [root@machine]# cat /etc/rc.d/rc.local #!/bin/sh for i in /sys/class/scsi_host/host*/link_power_management_policy; do echo med_power_with_dipm > $i done [root@machine]# ls -latr /etc/rc.d/rc.local -rwxrwxr-x. 1 user user 119 13 set 14.53 /etc/rc.d/rc.local It might be that systemd does not like that rc.local is owned by "user" rather then "root" ? Have you tried manually executing rc.local as root: /etc/rc.d/rc.local And / or have you tried to do the echo manually as root ? Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
This message may be useful to others having my similar problem. On a Thinkpad X220, after having booted the custom kernel, running [root@machine]# cat /sys/class/scsi_host/host0/link_power_management_policy I get max_performance instead of med_power_with_dipm Here some machine details [root@machine]# cat /etc/rc.d/rc.local #!/bin/sh for i in /sys/class/scsi_host/host*/link_power_management_policy; do echo med_power_with_dipm > $i done [root@machine]# ls -latr /etc/rc.d/rc.local -rwxrwxr-x. 1 user user 119 13 set 14.53 /etc/rc.d/rc.local [root@machine]# cat /proc/cpuinfo | grep "model name" model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz [root@machine]# cat /sys/class/scsi_device/*/device/model Samsung SSD 850 (it is a Samsung 850 PRO) ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
This message may be useful to others having my similar problem. On a Thinkpad X220, after having booted the custom kernel, running [root@machine]# cat /sys/class/scsi_host/host0/link_power_management_policy I get max_performance instead of med_power_with_dipm Here some machine details [root@machine]# cat /etc/rc.d/rc.local #!/bin/sh for i in /sys/class/scsi_host/host*/link_power_management_policy; do echo med_power_with_dipm > $i done [root@machine]# ls -latr /etc/rc.d/rc.local -rwxrwxr-x. 1 user user 119 13 set 14.53 /etc/rc.d/rc.local [root@machine]# cat /proc/cpuinfo | grep "model name" model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz model name : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz [root@machine]# cat /sys/class/scsi_device/*/device/model Samsung SSD 850 (it is a Samsung 850 PRO) ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi, Germano Massullo wrote: > Thank you for the extensive explanation! > > Ok, at the end of each day I will give a look to "dmesg", just to be > sure that everything went fine. > I will also check for firmware updates before starting the tests. > > A few questions/statements: > 1) is it okay to run the tests with the power cord plugged? Or do I need > to use battery in order to trigger such kind of power saving features? Running the tests with the power cord plugged in is fine, but to get info on how much power this actually saves you need to be running on battery since most laptops can only measure power-consumption while running on battery. After getting the initial idle and idle power consumption with the new med_power_with_dipm policy activated you can run with the cord plugged in. > 2) In your guide you specified "[...] on a clean Fedora". I actually > cannot run a new Fedora installation on my SSDs, but I haven't messed up > with any kind of strange powersaving features, so it should be fine to > use my regular Fedora installation for the tests. Do you confirm? Yes that is fine. > To be > sure, I will attach to test results e-mail, also: > - /etc/fstab > - /proc/cmdline > - # lshw > 3) Since in these days I am not using much the Thinkpads, I think I will > run Phoronix suite in order to simulate machine activity > Among all Phoronix test suite tests, If you think that there are some > that best fit your case, please let me know. To read a list of disk > oriented benchmarks, you can do: > # dnf install phoronix-test-suite > $ phoronix-test-suite list-available-tests | grep Disk Sounds good. Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Thank you for the extensive explanation! > TL;DR: I do not expect you to see any silent data corruption if > anything goes wrong, you should know. Ok, at the end of each day I will give a look to "dmesg", just to be sure that everything went fine. I will also check for firmware updates before starting the tests. A few questions/statements: 1) is it okay to run the tests with the power cord plugged? Or do I need to use battery in order to trigger such kind of power saving features? 2) In your guide you specified "[...] on a clean Fedora". I actually cannot run a new Fedora installation on my SSDs, but I haven't messed up with any kind of strange powersaving features, so it should be fine to use my regular Fedora installation for the tests. Do you confirm? To be sure, I will attach to test results e-mail, also: - /etc/fstab - /proc/cmdline - # lshw 3) Since in these days I am not using much the Thinkpads, I think I will run Phoronix suite in order to simulate machine activity Among all Phoronix test suite tests, If you think that there are some that best fit your case, please let me know. To read a list of disk oriented benchmarks, you can do: # dnf install phoronix-test-suite $ phoronix-test-suite list-available-tests | grep Disk Best regards ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi, On 17-09-17 17:26, Germano Massullo wrote: Il 17/09/2017 15:39, Hans de Goede ha scritto: Hi, On 14-09-17 14:44, Germano Massullo wrote: Hi Hans, I can provide ~10 different Thinkpad models to test your code. Looking forward to start! That is great, thank you. I'm looking forward to hear back from you with the results. Regards, Hans Just a question before I start with tests on some machines. Concerning the very little probability that some data corruption could happen: which kind of data corruption would be? Bit flips, etc. Yes I do have backups, but I wanted to ask that just to understand if I better to resume data from backups even if I had no apparent crashes during tests. There are 2 sorts of problems which have been reported in the past: 1) The disk stops responding, so the system is effectively dead until you reboot, with a journaling filesystem this should not be a problem really. This is the most frequent problem (still not very frequent). 2) Some SSD firmwares seem to mess up their internal bookkeeping of which part of the flash holds what when putting the SSD in a low power state at just the wrong time. This typically lead to all sort of filesystem errors, etc. and requires a full re-install. This is a very rare occurrence, but is the reason why I've been asking for wide testing as this one is nasty. TL;DR: I do not expect you to see any silent data corruption if anything goes wrong, you should know. Note hitting the race for problem 2. may take some time, which is why I've asked people to test with the patched kernel and the med_power_with_dipm policy for at least 2 weeks. Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Il 17/09/2017 15:39, Hans de Goede ha scritto: > Hi, > > On 14-09-17 14:44, Germano Massullo wrote: >> Hi Hans, I can provide ~10 different Thinkpad models to test your code. >> Looking forward to start! > > That is great, thank you. > > I'm looking forward to hear back from you with the results. > > Regards, > > Hans > Just a question before I start with tests on some machines. Concerning the very little probability that some data corruption could happen: which kind of data corruption would be? Bit flips, etc. Yes I do have backups, but I wanted to ask that just to understand if I better to resume data from backups even if I had no apparent crashes during tests. Have a nice day ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi, On 14-09-17 14:44, Germano Massullo wrote: Hi Hans, I can provide ~10 different Thinkpad models to test your code. Looking forward to start! That is great, thank you. I'm looking forward to hear back from you with the results. Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Re: Improving Linux laptop battery life: Testers Wanted
Hi Hans, I can provide ~10 different Thinkpad models to test your code. Looking forward to start! Have a nice day ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Improving Linux laptop battery life: Testers Wanted
Hi All, My next project for Red Hat is to work on improving Linux laptop battery life. Part of the (hopefully) low hanging fruit here is using kernel tunables to enable more runtime powermanagement. My first target here is SATA Link Power Management (LPM) which, as Matthew Garrett blogged about 2 years ago: https://mjg59.dreamwidth.org/34868.html can lead to a significant improvement in battery life. There is only one small problem, there have been some reports that some disks/SSDs don't play well with Linux' min_power LPM policy and that this may lead to system crashes and even data corruption. As such I've written a new LPM policy, which matches the power-management defaults from the Intel RST Windows drivers. Since it mimicks Windows, this new policy will hopefully not hit any SSD firmware bugs like min_power sometimes does. So now I'm looking for people with a laptop with a SATA SSD or HDD to help me test this to make sure this won't cause any issues when we enable this by default for F28, for more details and test instructions see: https://hansdegoede.livejournal.com/18412.html Regards, Hans ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org