[Freedreno] PSA: ifc6410 / upstream kernel / clock status stuck at 'off' issues

[Rob Clark]

Just FYI, newer ifc6410 boards which are shipping with android 4.4
bootloaders, seem to have an issue with upstream kernel.

If you are seeing problems like:

-------
[    2.472644] WARNING: CPU: 3 PID: 51 at
drivers/clk/qcom/clk-branch.c:97 clk_branch_wait+0xd0/0x120()
[    2.472659] mdp_clk status stuck at 'off'
[    2.472667] Modules linked in:
[    2.472688] CPU: 3 PID: 51 Comm: kworker/u8:1 Not tainted 3.18.0-rc1 #912
[    2.472719] Workqueue: deferwq deferred_probe_work_func
[    2.472784] [<c02160fc>] (unwind_backtrace) from [<c0211afc>]
(show_stack+0x10/0x14)
[    2.472824] [<c0211afc>] (show_stack) from [<c0909864>]
(dump_stack+0x88/0x98)
[    2.472859] [<c0909864>] (dump_stack) from [<c0249fa4>]
(warn_slowpath_common+0x6c/0x88)
[    2.472889] [<c0249fa4>] (warn_slowpath_common) from [<c0249ff0>]
(warn_slowpath_fmt+0x30/0x40)
[    2.472917] [<c0249ff0>] (warn_slowpath_fmt) from [<c0792964>]
(clk_branch_wait+0xd0/0x120)
[    2.472952] [<c0792964>] (clk_branch_wait) from [<c0787f78>]
(__clk_enable+0x5c/0x9c)
[    2.472975] [<c0787f78>] (__clk_enable) from [<c0788464>]
(clk_enable+0x18/0x2c)
[    2.473013] [<c0788464>] (clk_enable) from [<c054121c>]
(mdp4_enable+0xe4/0x110)
[    2.473048] [<c054121c>] (mdp4_enable) from [<c0541270>]
(mdp4_hw_init+0x28/0x344)
[    2.473075] [<c0541270>] (mdp4_hw_init) from [<c0547e60>]
(msm_load+0x1c8/0x34c)
[    2.473108] [<c0547e60>] (msm_load) from [<c052596c>]
(drm_dev_register+0xa8/0x104)
[    2.473136] [<c052596c>] (drm_dev_register) from [<c0527824>]
(drm_platform_init+0x44/0xdc)
[    2.473166] [<c0527824>] (drm_platform_init) from [<c0566538>]
(try_to_bring_up_master.part.3+0xc8/0x108)
[    2.473190] [<c0566538>] (try_to_bring_up_master.part.3) from
[<c0566620>] (component_master_add_with_match+0xa8/0x120)
-------

Then the solution for now is to downgrade to version 4.2 rpm fw:

   fastboot flash rpm rpm.mbn

with rpm.mbn from Bootloaders_4.2_880261_V1.0.zip on inforce techweb site.

I've no idea what the root cause is, or why we don't see this on the
downstream kernel (but I guess it is easier for someone with docs to
debug).  But sooner or later others will hit the same issue so I
figured I should send this known workaround to an archived mailing
list so someone googling for the error has some chance to find it ;-)

BR,
-R