[PATCH] arm64: restore get_current() optimisation

From
Mark Rutland <mark.rutland@arm.com>
Date
2017-03-03 19:48:47
Reply to
Jon Hunter, 2017-03-03
On Fri, Mar 03, 2017 at 03:32:22PM +0000, Jon Hunter wrote:
Hi Mark,

On 02/03/17 16:46, Mark Rutland wrote:
quoted
Perhaps a caller is deriving that from an uninitialised variable? The
code and stack layout changes as a result of my patch could easily
tickle that.
You are correct, the reg passed to regcache_flat_read() is indeed dodgy
and the above BUG() did trip it up.

Commit e411b0b5eb9b ('ASoC: dapm: Support second register for DAPM
control updates') introduced a 2nd register set into the
snd_soc_dapm_update struct and if the 'has_second_set' is true then it
tries to access this register. One of our out-of-tree audio patches was
not initialising the 'has_second_set' and hence it was trying to write a
bogus register (need to get this upstream!).
Phew. That's get_current() off the hook!
I should have caught this, but the bisect threw me off the scent! I
think that with reverting this patch we were just getting lucky and the
problem could have still occurred. I went back and tried it again and it
still works when reverting this change, but its just luck.

Sorry for the noise and thanks for pointing me in the right direction!
No worries; it's always a pain to debug this sort of thing, especially
with a bisect seeming so reliable.

Thanks,
Mark.
Thread (14 messages) 14 messages, 5 authors, last reply 2017-03-03
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help