On 27 June 2013 17:13, Catalin Marinas [off-list ref] wrote:

On Thu, Jun 27, 2013 at 02:11:27PM +0100, Ard Biesheuvel wrote:

quoted

OK, just one more question before I respin the next (hopefully final) version:
if a caller does sleep after calling kernel_neon_begin() (and thus
receives no warning if he runs a non-PREEMPT build with
CONFIG_DEBUG_ATOMIC_SLEEP disabled), he will most likely find the
NEON/VFP unit disabled after waking up (as we disable it on a context
switch), so any subsequent NEON instructions will trigger the undef
handler.

Can you check on the VFP context switch path whether kernel_neon_begin()
has been called and we are moving away from the task? You could even
store the LR in kernel_neon_begin() to give better error information.

I guess it should be quite doable to add the LR of the most recent
kernel_neon_begin() call to the vfpstate (and clear it in
kernel_neon_end), so we can check it when vfp_notifier() is called.
We'll still hit the vfp undef handler on swapping the task back in, so
it may be a bit redundant.

What I could do is use this LR field in vfpstate when hitting the
undef handler to distinguish between (a) kernel_neon_begin()
erroneously not having called at all and (b) kernel_neon_begin()
having been called before sleeping (and print a stacktrace in the
latter case)

-- 
Ard.