diff options
author | Paul Mundt <lethal@linux-sh.org> | 2011-05-31 15:23:20 +0900 |
---|---|---|
committer | Paul Mundt <lethal@linux-sh.org> | 2011-05-31 15:23:20 +0900 |
commit | d4905ce38c73964b868037e49a5945e1cf47a7f2 (patch) | |
tree | f7b41c1720dabb7b14a92ff0c47aeff1d6e4d70f /drivers/clocksource/sh_cmt.c | |
parent | db7eba292e913390fa881272bfbc3da0a5380513 (diff) | |
download | op-kernel-dev-d4905ce38c73964b868037e49a5945e1cf47a7f2.zip op-kernel-dev-d4905ce38c73964b868037e49a5945e1cf47a7f2.tar.gz |
Revert "clocksource: sh_tmu: Runtime PM support"
This reverts commit 1b842e91fea9447eff5eb687e28ad61c02f5033e.
There is a fundamental ordering race between the early and late probe
paths and the runtime PM tie-in that results in __pm_runtime_resume()
attempting to take a lock that hasn't been initialized yet (which by
proxy also suggests that pm_runtime_init() hasn't yet been run on the
device either, making the entire thing unsafe) -- resulting in instant
death on SMP or on UP with spinlock debugging enabled:
sh_tmu.0: used for clock events
sh_tmu.0: used for periodic clock events
BUG: spinlock trylock failure on UP on CPU#0, swapper/0
lock: 804db198, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
...
Revert it for now until the ordering issues can be resolved, or we can get
some more help from the runtime PM framework to make this possible.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Diffstat (limited to 'drivers/clocksource/sh_cmt.c')
0 files changed, 0 insertions, 0 deletions