[PATCHv2 00/27] ARM: OMAP2+: clock code migration to drivers/clk/ti

[Tony Lindgren]

* Nishanth Menon <nm at ti.com> [150521 00:03]:
> On 05/21/2015 01:38 AM, Tero Kristo wrote:
> > On 05/21/2015 01:40 AM, Paul Walmsley wrote:
> >> On Tue, 19 May 2015, Tero Kristo wrote:
> >>
> >>> Any news on this? As noted previously, I am not able to reproduce the
> >>> issue
> >>> you are seeing currently, can you give DEBUG_LL a shot?
> >>
> >> Yeah I just bisected it, it was caused by this:
> >>
> >> commit cc4a5fe972ad7834e8662b49b3a5fdb597e9e15e
> >> Author: Felipe Balbi <balbi at ti.com>
> >> Date:   Fri Jan 30 11:18:56 2015 -0600
> >>
> >>      arm: config: omap2plus_defconfig: switch over to LZMA compression
> >>
> >>      LZMA compression makes about 33% smaller zImage
> >>      with just a slight extra decompression time.
> >>
> >>      Before this patch, zImage built with o2+_dc
> >>      is 4.5MiB and after it's about 3.3MiB.
> >>
> >>      Suggested-by: David Cohen <david.a.cohen at linux.intel.com>
> >>      Signed-off-by: Felipe Balbi <balbi at ti.com>
> >>      Signed-off-by: Tony Lindgren <tony at atomide.com>
> >>
> >>
> >> and the timeouts on the testbed being set to fail if a kernel takes
> >> longer
> >> than five seconds to start.  Seems that the part about a "slight extra
> >> decompression time" probably only applies to relatively recent chips.
> >>
> >>
> >> - Paul
> >>
> > 
> > Oh, so this explains why I was thinking it took very long time to boot
> > the recent kernels also. The boot lag is clearly noticeable without any
> > measurement. I wonder if we should probably revert this patch.
> 
> we already have issues with zImage size bloating up and running headlong
> into dtb - esp on platforms like N900. Felipe spend quiet some time
> getting things into a manageable size here -> loosing 33% sounds like
> bad idea to me :(

If it affects people using the slower 34xx systems we should probably
revert it though. Or make the uncompress somehow faster :)

Regards,

Tony