12:15 a.m.
Even though F19 ARM Beta bits are making their way to the mirrors now
there is still quite a bit yet to be done before GA. In addition to
beta release notes still needing to be written there are still bugs and
regressions to track down. Let's squash as many of these as possible in
the next couple weeks, and preferably spin test images early&often to
confirm the fixes.
General image issues:
o LVM console spam on first boot
o The initial-setup does not appear to run
o Restorecon runs on first boot- can we avoid this?
Versatile express:
o No u-boot - would be nice to have for virt integration
o No boot-vexpress script or pre-extracted kernels - was in F18.
o No graphics support - regression since F18 (Kernel 3.6)
TI Boards in general:
o Images, as distributed, need manual steps to be made bootable.
o An easy script to handle the manual steps is not yet available.
o Supplied vfat partition possibly badly formed.
o No graphics - regression since F18 (Kernel 3.6)
OMAP3 (Beagle boards):
o The supplied uboot does not load boot.scr.
o The supplied kernel fails to boot.
AM335x (Beagle bone boards):
o Missing kernel support - this is probably an F19 remix or F20 feature.
Tegra2 (Trimslice, specifically)
o SSD support is unreliable due to evident hardware bug.
o X, though functional, is not stable.
Anaconda installs:
o a-b-c needs to be run manually in %post
Notwithstanding the above, OMAP4 (Panda), Trimslice, Highbank and
Versatile Express are in decent shape, which is why the Beta is going
out now. Plus it gives a readily available set of bits for advancing
these defects. If you are aware of additional issues please followup
here- let's get them all identified then work out solutions. Thanks!
--
Brendan Conoboy / Red Hat, Inc. / blc(a)redhat.com
12:20 p.m.
On 05/25/2013 10:15 PM, Brendan Conoboy wrote:
General image issues:
o LVM console spam on first boot
o The initial-setup does not appear to run
Paul is looking into these. If anybody else is investigating either
please let him know.
o Restorecon runs on first boot- can we avoid this?
After a discussion with others it appears we could avoid this by running
restorecon in %post, or otherwise having the image generation tools run
restorecon for us. We would be doing a service to end users' sd write
cycles by doing so. Dennis?
Versatile express:
o No u-boot - would be nice to have for virt integration
o No boot-vexpress script or pre-extracted kernels - was in F18.
No update here. We are building a uboot for versatile express, but it
does not appear to function.
o No graphics support - regression since F18 (Kernel 3.6)
Kyle is looking into this and may have an update next week.
TI Boards in general:
o Images, as distributed, need manual steps to be made bootable.
o An easy script to handle the manual steps is not yet available.
o Supplied vfat partition possibly badly formed.
Some have suggested using fossjon's livemedia-modifier script. Can
somebody comment here?
o No graphics - regression since F18 (Kernel 3.6)
Jon is looking into this for Panda only. He should have an update later
today.
OMAP3 (Beagle boards):
o The supplied uboot does not load boot.scr.
This is for Dennis.
o The supplied kernel fails to boot.
Nobody is looking at this that I am aware of.
AM335x (Beagle bone boards):
o Missing kernel support - this is probably an F19 remix or F20 feature.
Let's forget about this for F19, the kernel just isn't ready.
Tegra2 (Trimslice, specifically)
o SSD support is unreliable due to evident hardware bug.
This might be addressable by "hdparm -a 32". Or, less likely, tweaking
/sys/block/sda/queue/read_ahead_kb. Would somebody with a trimslice-pro
try this out and confirm?
o X, though functional, is not stable.
Nobody is looking at this that I am aware of.
Anaconda installs:
o a-b-c needs to be run manually in %post
I'll look into this.
--
Brendan Conoboy / Red Hat, Inc. / blc(a)redhat.com
5:35 a.m.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
El Wed, 29 May 2013 10:20:55 -0700
Brendan Conoboy <blc(a)redhat.com> escribió:
On 05/25/2013 10:15 PM, Brendan Conoboy wrote:
> General image issues:
> o LVM console spam on first boot
> o The initial-setup does not appear to run
Paul is looking into these. If anybody else is investigating either
please let him know.
i know what happened and it should already be fixed in the kickstarts
> o Restorecon runs on first boot- can we avoid this?
After a discussion with others it appears we could avoid this by
running restorecon in %post, or otherwise having the image generation
tools run restorecon for us. We would be doing a service to end
users' sd write cycles by doing so. Dennis?
> Versatile express:
> o No u-boot - would be nice to have for virt integration
> o No boot-vexpress script or pre-extracted kernels - was in F18.
No update here. We are building a uboot for versatile express, but
it does not appear to function.
> o No graphics support - regression since F18 (Kernel 3.6)
Kyle is looking into this and may have an update next week.
> TI Boards in general:
> o Images, as distributed, need manual steps to be made bootable.
> o An easy script to handle the manual steps is not yet available.
> o Supplied vfat partition possibly badly formed.
Some have suggested using fossjon's livemedia-modifier script. Can
somebody comment here?
> o No graphics - regression since F18 (Kernel 3.6)
Jon is looking into this for Panda only. He should have an update
later today.
> OMAP3 (Beagle boards):
> o The supplied uboot does not load boot.scr.
This is for Dennis.
> o The supplied kernel fails to boot.
Nobody is looking at this that I am aware of.
> AM335x (Beagle bone boards):
> o Missing kernel support - this is probably an F19 remix or F20
> feature.
Let's forget about this for F19, the kernel just isn't ready.
> Tegra2 (Trimslice, specifically)
> o SSD support is unreliable due to evident hardware bug.
This might be addressable by "hdparm -a 32". Or, less likely,
tweaking /sys/block/sda/queue/read_ahead_kb. Would somebody with a
trimslice-pro try this out and confirm?
> o X, though functional, is not stable.
Nobody is looking at this that I am aware of.
> Anaconda installs:
> o a-b-c needs to be run manually in %post
I'll look into this.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
iEYEARECAAYFAlGnK24ACgkQkSxm47BaWfcJXACgpFfN+qfECKlF6Azbeu1WCYxA
DiUAn14yz6NumnKrY980IB4IX+Hzxcz2
=WHO4
-----END PGP SIGNATURE-----
8:54 a.m.
On Sun, May 26, 2013 at 6:15 AM, Brendan Conoboy <blc(a)redhat.com> wrote:
Even though F19 ARM Beta bits are making their way to the mirrors now
there
is still quite a bit yet to be done before GA. In addition to beta release
notes still needing to be written there are still bugs and regressions to
track down. Let's squash as many of these as possible in the next couple
weeks, and preferably spin test images early&often to confirm the fixes.
General image issues:
o LVM console spam on first boot
o The initial-setup does not appear to run
Not sure when the first rcX is out to test Dennis's fix.
o Restorecon runs on first boot- can we avoid this?
Versatile express:
o No u-boot - would be nice to have for virt integration
It would be nice, not a regression though, by "virt integration" do
you mean a single .img for use with virt-manager or something else?
o No boot-vexpress script or pre-extracted kernels - was in F18.
o No graphics support - regression since F18 (Kernel 3.6)
TI Boards in general:
o Images, as distributed, need manual steps to be made bootable.
o An easy script to handle the manual steps is not yet available.
o Supplied vfat partition possibly badly formed.
o No graphics - regression since F18 (Kernel 3.6)
OMAP3 (Beagle boards):
o The supplied uboot does not load boot.scr.
o The supplied kernel fails to boot.
This is a nice to have. 3.8.x works minus USB, in theory this should
now be fixed (might be a 3.10.x fix though).
AM335x (Beagle bone boards):
o Missing kernel support - this is probably an F19 remix or F20 feature.
I'm working on this as I get time and hoping to land something for
Final but agreed it's a nice to have.
Tegra2 (Trimslice, specifically)
o SSD support is unreliable due to evident hardware bug.
o X, though functional, is not stable.
Again I think both of those are nice to have. It's solid as a rock on
mine when used as a /home directory even if I thump it with builds.
Anaconda installs:
o a-b-c needs to be run manually in %post
Is that a major blocker and how else would it run? Presumably as part
of the kernel install / anaconda boot loader setup?
Notwithstanding the above, OMAP4 (Panda), Trimslice, Highbank and
Versatile
Express are in decent shape, which is why the Beta is going out now. Plus
it gives a readily available set of bits for advancing these defects. If
you are aware of additional issues please followup here- let's get them all
identified then work out solutions. Thanks!
Agreed HW side is in reasonable shape. I would also like to bring to
attention the ARMTracker blocker bugs and if people have some spare
cycles to go through the list and see what they can help in fixing.
I'm reviewing package failure lists and (re)fixing or filing some bugs
as I get the time and it would be good if some people with more
programming skills can help fix some.
Peter
11:07 a.m.
On 05/29/2013 01:20 PM, Brendan Conoboy wrote:
On 05/25/2013 10:15 PM, Brendan Conoboy wrote:
> o No graphics - regression since F18 (Kernel 3.6)
Jon is looking into this for Panda only. He should have an update later
today.
So I'm just using a blunt instrument one-liner to request_module the
tfp4 panel driver on module init. This is ugly, and needs testing, etc.
It'll never go upstream. We can perhaps get away with a synchronous
request_module call during module init because that happens in the
context of modprobe (kmod) and so we probably won't deadlock stuff, but
the right thing in general is probably to DECLARE_WORK a worker that can
be scheduled in the kernel at probe time based on what chips are
actually present, e.g. after parsing the devicetree for that data. This
is just stream-of-consciousness thinking here, I'll think more about it.
I'll be ordering some new Panda ES boards for myself and Peter, since
the EB3 and other engineering builds seem to generate a backtrace on
boot. It's not worth fixing that if production ES boards are working.
Jon.
3976
days inactive
3981
days old
4 comments
4 participants
participants (4)
-
Brendan Conoboy -
Dennis Gilmore -
Jon Masters -
Peter Robinson