Message ID | 1516046953-633-1-git-send-email-rosenp@gmail.com |
---|---|
State | Rejected |
Delegated to: | Mathias Kresin |
Headers | show |
Series | [LEDE-DEV] mt7621: Add back CONFIG_SCHED_HRTICK to kernel config | expand |
15.01.2018 21:09, Rosen Penev: > Fixes FS #1242. > > The way I figured this out was by using diff on config-4.4 from 17.01 > and config-4.9 from trunk. First I removed CONFIG options. That did not work. > Then I started adding. This one seems to do the trick. > > The issue is that anything in /dev/sdX starts returning bad data when read. > PCIe or USB does not matter. I have not tested NVME since I lack the hardware. > /dev/mtdblockX may or may not be impacted. No idea. > > Signed-off-by: Rosen Penev <rosenp@gmail.com> > --- > target/linux/ramips/mt7621/config-4.9 | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/target/linux/ramips/mt7621/config-4.9 b/target/linux/ramips/mt7621/config-4.9 > index f9765ed..0ea6798 100644 > --- a/target/linux/ramips/mt7621/config-4.9 > +++ b/target/linux/ramips/mt7621/config-4.9 > @@ -233,6 +233,7 @@ CONFIG_RTC_CLASS=y > CONFIG_RTC_DRV_PCF8563=y > CONFIG_RTC_I2C_AND_SPI=y > CONFIG_RTC_MC146818_LIB=y > +CONFIG_SCHED_HRTICK=y > # CONFIG_SCHED_INFO is not set > CONFIG_SCHED_SMT=y > # CONFIG_SCSI_DMA is not set What ever the root cause of your issue is, it isn't CONFIG_SCHED_HRTICK since it is already enabled for mt7621: $ git grep CONFIG_SCHED_HRTICK target/linux/gemini/config-4.4:# CONFIG_SCHED_HRTICK is not set target/linux/generic/config-4.14:CONFIG_SCHED_HRTICK=y target/linux/generic/config-4.4:CONFIG_SCHED_HRTICK=y target/linux/generic/config-4.9:CONFIG_SCHED_HRTICK=y target/linux/mcs814x/config-3.18:# CONFIG_SCHED_HRTICK is not set target/linux/omap24xx/config-4.1:CONFIG_SCHED_HRTICK=y target/linux/ppc40x/config-3.18:CONFIG_SCHED_HRTICK=y target/linux/ppc44x/config-3.18:CONFIG_SCHED_HRTICK=y target/linux/rb532/config-4.9:# CONFIG_SCHED_HRTICK is not set For further patches, please explain in the commit message why the change fixes the issue and move the not commit message related text below the tear line. Mathias
It is defined in generic yes. From discussions on IRC, the issue may be related to something else. I will do further testing to see whether or not this is correct. Should take me another week or so... I still have no idea why in 17.01 CONIG_SCHED_HRTICK is defined in generic and in a bunch of architectures whereas it;'s only defined in trunk in generic. On Mon, Jan 15, 2018 at 1:20 PM, Mathias Kresin <dev@kresin.me> wrote: > 15.01.2018 21:09, Rosen Penev: >> >> Fixes FS #1242. >> >> The way I figured this out was by using diff on config-4.4 from 17.01 >> and config-4.9 from trunk. First I removed CONFIG options. That did not >> work. >> Then I started adding. This one seems to do the trick. >> >> The issue is that anything in /dev/sdX starts returning bad data when >> read. >> PCIe or USB does not matter. I have not tested NVME since I lack the >> hardware. >> /dev/mtdblockX may or may not be impacted. No idea. >> >> Signed-off-by: Rosen Penev <rosenp@gmail.com> >> --- >> target/linux/ramips/mt7621/config-4.9 | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/target/linux/ramips/mt7621/config-4.9 >> b/target/linux/ramips/mt7621/config-4.9 >> index f9765ed..0ea6798 100644 >> --- a/target/linux/ramips/mt7621/config-4.9 >> +++ b/target/linux/ramips/mt7621/config-4.9 >> @@ -233,6 +233,7 @@ CONFIG_RTC_CLASS=y >> CONFIG_RTC_DRV_PCF8563=y >> CONFIG_RTC_I2C_AND_SPI=y >> CONFIG_RTC_MC146818_LIB=y >> +CONFIG_SCHED_HRTICK=y >> # CONFIG_SCHED_INFO is not set >> CONFIG_SCHED_SMT=y >> # CONFIG_SCSI_DMA is not set > > > What ever the root cause of your issue is, it isn't CONFIG_SCHED_HRTICK > since it is already enabled for mt7621: > > $ git grep CONFIG_SCHED_HRTICK > target/linux/gemini/config-4.4:# CONFIG_SCHED_HRTICK is not set > target/linux/generic/config-4.14:CONFIG_SCHED_HRTICK=y > target/linux/generic/config-4.4:CONFIG_SCHED_HRTICK=y > target/linux/generic/config-4.9:CONFIG_SCHED_HRTICK=y > target/linux/mcs814x/config-3.18:# CONFIG_SCHED_HRTICK is not set > target/linux/omap24xx/config-4.1:CONFIG_SCHED_HRTICK=y > target/linux/ppc40x/config-3.18:CONFIG_SCHED_HRTICK=y > target/linux/ppc44x/config-3.18:CONFIG_SCHED_HRTICK=y > target/linux/rb532/config-4.9:# CONFIG_SCHED_HRTICK is not set > > For further patches, please explain in the commit message why the change > fixes the issue and move the not commit message related text below the tear > line. > > Mathias
15.01.2018 22:30, Rosen Penev: > It is defined in generic yes. From discussions on IRC, the issue may > be related to something else. I will do further testing to see whether > or not this is correct. Should take me another week or so... > > I still have no idea why in 17.01 CONIG_SCHED_HRTICK is defined in > generic and in a bunch of architectures whereas it;'s only defined in > trunk in generic. It is as simple as the 17.01 (sub)target kernel configs weren't updated at some point. A 'make kernel_menuconfig CONFIG_TARGET=subtarget' will drop the redundant kernel config symbol from the 17.01 (sub)target configs as well. Mathias
Small update on this: It seems that all of the removed CONFIG settings are set anyways. That is, there is no difference in the final kernel .config file. Which leads me to believe that this issue was fixed upstream somehow. I'll be doing further testing to verify. On Mon, Jan 15, 2018 at 1:37 PM, Mathias Kresin <dev@kresin.me> wrote: > 15.01.2018 22:30, Rosen Penev: >> >> It is defined in generic yes. From discussions on IRC, the issue may >> be related to something else. I will do further testing to see whether >> or not this is correct. Should take me another week or so... >> >> I still have no idea why in 17.01 CONIG_SCHED_HRTICK is defined in >> generic and in a bunch of architectures whereas it;'s only defined in >> trunk in generic. > > > It is as simple as the 17.01 (sub)target kernel configs weren't updated at > some point. A 'make kernel_menuconfig CONFIG_TARGET=subtarget' will drop the > redundant kernel config symbol from the 17.01 (sub)target configs as well. > > Mathias
diff --git a/target/linux/ramips/mt7621/config-4.9 b/target/linux/ramips/mt7621/config-4.9 index f9765ed..0ea6798 100644 --- a/target/linux/ramips/mt7621/config-4.9 +++ b/target/linux/ramips/mt7621/config-4.9 @@ -233,6 +233,7 @@ CONFIG_RTC_CLASS=y CONFIG_RTC_DRV_PCF8563=y CONFIG_RTC_I2C_AND_SPI=y CONFIG_RTC_MC146818_LIB=y +CONFIG_SCHED_HRTICK=y # CONFIG_SCHED_INFO is not set CONFIG_SCHED_SMT=y # CONFIG_SCSI_DMA is not set
Fixes FS #1242. The way I figured this out was by using diff on config-4.4 from 17.01 and config-4.9 from trunk. First I removed CONFIG options. That did not work. Then I started adding. This one seems to do the trick. The issue is that anything in /dev/sdX starts returning bad data when read. PCIe or USB does not matter. I have not tested NVME since I lack the hardware. /dev/mtdblockX may or may not be impacted. No idea. Signed-off-by: Rosen Penev <rosenp@gmail.com> --- target/linux/ramips/mt7621/config-4.9 | 1 + 1 file changed, 1 insertion(+)