[PATCH] systemd: disable problematic GCC 5.2 optimizations

[PATCH] systemd: disable problematic GCC 5.2 optimizations

[Jonathan Liu]

This fixes systemd failing to start on Raspberry Pi 2 if it is compiled
with GCC 5.2.

It would try to start "Journal Service" and "udev Kernel Device Manager"
but fail repeatedly.

[YOCTO #8291]

Signed-off-by: Jonathan Liu <net147@gmail.com>
---
 meta/recipes-core/systemd/systemd_225.bb | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/meta/recipes-core/systemd/systemd_225.bb b/meta/recipes-core/systemd/systemd_225.bb
index f7d4c7d..4a19ff4 100644
--- a/meta/recipes-core/systemd/systemd_225.bb
+++ b/meta/recipes-core/systemd/systemd_225.bb
@@ -123,6 +123,9 @@ EXTRA_OECONF = " --with-rootprefix=${rootprefix} \
 # uclibc does not have NSS
 EXTRA_OECONF_append_libc-uclibc = " --disable-myhostname "
 
+# disable problematic GCC 5.2 optimizations [YOCTO #8291]
+FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
+
 do_configure_prepend() {
 	export NM="${HOST_PREFIX}gcc-nm"
 	export AR="${HOST_PREFIX}gcc-ar"
-- 
2.5.0

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 436 bytes --]

On 21 September 2015 at 12:23, Jonathan Liu <net147@gmail.com> wrote:

> +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
>

Are these bugs with the gcc optimisations or the systemd source code, and
if its gcc's fault are the problematic optimisations specific to a
particular architecture (such as ARM)?  Just wondering if we can use
overrides to only change the optimisations where they're a problem.

Ross

[-- Attachment #2: Type: text/html, Size: 855 bytes --]

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Jonathan Liu]

[-- Attachment #1: Type: text/plain, Size: 836 bytes --]

On 21/09/2015 10:23 PM, Burton, Ross wrote:
>
> On 21 September 2015 at 12:23, Jonathan Liu <net147@gmail.com 
> <mailto:net147@gmail.com>> wrote:
>
>     +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
>
>
> Are these bugs with the gcc optimisations or the systemd source code, 
> and if its gcc's fault are the problematic optimisations specific to a 
> particular architecture (such as ARM)?  Just wondering if we can use 
> overrides to only change the optimisations where they're a problem.
>
> Ross
I am not sure how to isolate if it's an issue in the systemd source code 
but the issue is not present when using GCC 4.9.
Haven't checked if this is occuring on other architectures. Any feedback 
whether this occurs on other platforms (Raspberry Pi 2 is ARMv7) is welcome.

Regards,
Jonathan

[-- Attachment #2: Type: text/html, Size: 1893 bytes --]

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Adrian Freihofer]

[-- Attachment #1: Type: text/plain, Size: 1220 bytes --]

Hi,

On Beaglebone-black Systemd 215 does not boot if compiled with -O2.
Compiling with -O1 works.

Regards,
Adrian

2015-09-21 14:31 GMT+02:00 Jonathan Liu <net147@gmail.com>:

> On 21/09/2015 10:23 PM, Burton, Ross wrote:
>
>
> On 21 September 2015 at 12:23, Jonathan Liu <net147@gmail.com> wrote:
>
>> +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
>>
>
> Are these bugs with the gcc optimisations or the systemd source code, and
> if its gcc's fault are the problematic optimisations specific to a
> particular architecture (such as ARM)?  Just wondering if we can use
> overrides to only change the optimisations where they're a problem.
>
> Ross
>
> I am not sure how to isolate if it's an issue in the systemd source code
> but the issue is not present when using GCC 4.9.
> Haven't checked if this is occuring on other architectures. Any feedback
> whether this occurs on other platforms (Raspberry Pi 2 is ARMv7) is welcome.
>
> Regards,
> Jonathan
>
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>
>

[-- Attachment #2: Type: text/html, Size: 2542 bytes --]

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Jonathan Liu]

[-- Attachment #1: Type: text/plain, Size: 1317 bytes --]

Hi Adrian,

On 21/09/2015 10:45 PM, Adrian Freihofer wrote:
> On Beaglebone-black Systemd 215 does not boot if compiled with -O2. 
> Compiling with -O1 works.
> 2015-09-21 14:31 GMT+02:00 Jonathan Liu <net147@gmail.com 
> <mailto:net147@gmail.com>>:
>
>     On 21/09/2015 10:23 PM, Burton, Ross wrote:
>>
>>     On 21 September 2015 at 12:23, Jonathan Liu <net147@gmail.com
>>     <mailto:net147@gmail.com>> wrote:
>>
>>         +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
>>
>>
>>     Are these bugs with the gcc optimisations or the systemd source
>>     code, and if its gcc's fault are the problematic optimisations
>>     specific to a particular architecture (such as ARM)?  Just
>>     wondering if we can use overrides to only change the
>>     optimisations where they're a problem.
>>
>>     Ross
>     I am not sure how to isolate if it's an issue in the systemd
>     source code but the issue is not present when using GCC 4.9.
>     Haven't checked if this is occuring on other architectures. Any
>     feedback whether this occurs on other platforms (Raspberry Pi 2 is
>     ARMv7) is welcome.
>
>     Regards,
>     Jonathan
>

Can you try the patch instead of changing -O2 to -O1 to see if it fixes 
booting with systemd?

Regards,
Jonathan

[-- Attachment #2: Type: text/html, Size: 3559 bytes --]

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Adrian Freihofer]

Hi Jonathan

Your patch is fine on beaglebone black. At least it is possible to
login. Thank you.

Regards,
Adrian

On Mon, 2015-09-21 at 22:49 +1000, Jonathan Liu wrote:
> Hi Adrian,
> 
> On 21/09/2015 10:45 PM, Adrian Freihofer wrote:
> > On Beaglebone-black Systemd 215 does not boot if compiled with -O2.
> > Compiling with -O1 works.
> > 2015-09-21 14:31 GMT+02:00 Jonathan Liu <net147@gmail.com>:
> > > On 21/09/2015 10:23 PM, Burton, Ross wrote:
> > > > 
> > > > On 21 September 2015 at 12:23, Jonathan Liu <net147@gmail.com>
> > > > wrote:
> > > > > +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule
> > > > > -insns2"
> > > > > 
> > > > Are these bugs with the gcc optimisations or the systemd source
> > > > code, and if its gcc's fault are the problematic optimisations
> > > > specific to a particular architecture (such as ARM)?  Just
> > > > wondering if we can use overrides to only change the
> > > > optimisations where they're a problem.
> > > > 
> > > > Ross
> > > I am not sure how to isolate if it's an issue in the systemd
> > > source code but the issue is not present when using GCC 4.9.
> > > Haven't checked if this is occuring on other architectures. Any
> > > feedback whether this occurs on other platforms (Raspberry Pi 2
> > > is ARMv7) is welcome.
> > > 
> > > Regards,
> > > Jonathan
> > > 
> Can you try the patch instead of changing -O2 to -O1 to see if it
> fixes booting with systemd?
> 
> Regards,
> Jonathan

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1348 bytes --]

Please make it arm specific
On Sep 21, 2015 4:22 AM, "Jonathan Liu" <net147@gmail.com> wrote:

> This fixes systemd failing to start on Raspberry Pi 2 if it is compiled
> with GCC 5.2.
>
> It would try to start "Journal Service" and "udev Kernel Device Manager"
> but fail repeatedly.
>
> [YOCTO #8291]
>
> Signed-off-by: Jonathan Liu <net147@gmail.com>
> ---
>  meta/recipes-core/systemd/systemd_225.bb | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/meta/recipes-core/systemd/systemd_225.bb
> b/meta/recipes-core/systemd/systemd_225.bb
> index f7d4c7d..4a19ff4 100644
> --- a/meta/recipes-core/systemd/systemd_225.bb
> +++ b/meta/recipes-core/systemd/systemd_225.bb
> @@ -123,6 +123,9 @@ EXTRA_OECONF = " --with-rootprefix=${rootprefix} \
>  # uclibc does not have NSS
>  EXTRA_OECONF_append_libc-uclibc = " --disable-myhostname "
>
> +# disable problematic GCC 5.2 optimizations [YOCTO #8291]
> +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
> +
>  do_configure_prepend() {
>         export NM="${HOST_PREFIX}gcc-nm"
>         export AR="${HOST_PREFIX}gcc-ar"
> --
> 2.5.0
>
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>

[-- Attachment #2: Type: text/html, Size: 2298 bytes --]

Re: [PATCH] systemd: disable problematic GCC 5.2 optimizations

[Jonathan Liu]

[-- Attachment #1: Type: text/plain, Size: 1865 bytes --]

Hi Khem,

On 22/09/2015 3:05 AM, Khem Raj wrote:
>
> Please make it arm specific
>
I have sent V2 that only makes the change for arm.
Interestly, this issue doesn't appear for qemuarm (ARMv5) when running 
under qemu.

I am wondering if this issue affects platforms other than ARM on real 
boards depending on TUNE_FEATURES and TARGET_FPU.
> On Sep 21, 2015 4:22 AM, "Jonathan Liu" <net147@gmail.com 
> <mailto:net147@gmail.com>> wrote:
>
>     This fixes systemd failing to start on Raspberry Pi 2 if it is
>     compiled
>     with GCC 5.2.
>
>     It would try to start "Journal Service" and "udev Kernel Device
>     Manager"
>     but fail repeatedly.
>
>     [YOCTO #8291]
>
>     Signed-off-by: Jonathan Liu <net147@gmail.com
>     <mailto:net147@gmail.com>>
>     ---
>      meta/recipes-core/systemd/systemd_225.bb <http://systemd_225.bb>
>     | 3 +++
>      1 file changed, 3 insertions(+)
>
>     diff --git a/meta/recipes-core/systemd/systemd_225.bb
>     <http://systemd_225.bb> b/meta/recipes-core/systemd/systemd_225.bb
>     <http://systemd_225.bb>
>     index f7d4c7d..4a19ff4 100644
>     --- a/meta/recipes-core/systemd/systemd_225.bb <http://systemd_225.bb>
>     +++ b/meta/recipes-core/systemd/systemd_225.bb <http://systemd_225.bb>
>     @@ -123,6 +123,9 @@ EXTRA_OECONF = " --with-rootprefix=${rootprefix} \
>      # uclibc does not have NSS
>      EXTRA_OECONF_append_libc-uclibc = " --disable-myhostname "
>
>     +# disable problematic GCC 5.2 optimizations [YOCTO #8291]
>     +FULL_OPTIMIZATION += "-fno-schedule-insns -fno-schedule-insns2"
>     +
>      do_configure_prepend() {
>             export NM="${HOST_PREFIX}gcc-nm"
>             export AR="${HOST_PREFIX}gcc-ar"
>     --
>     2.5.0<http://lists.openembedded.org/mailman/listinfo/openembedded-core>
>
Regards,
Jonathan

[-- Attachment #2: Type: text/html, Size: 3590 bytes --]