Skip to content

arm: raw_spinlock annotations

Arnd Bergmann linux-arm-kernel 1 version New

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest0 of 12 patches 0 repliesNew

Patches in v1 No applies recorded yet

None of the 12 patches the cover letter announces are in this archive.

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [patch 00/12] arm: raw_spinlock annotations
Quoted 15 lines · On Tuesday 19 October 2010 22:03:32 Thomas Gleixner wrote:
Quoted 13 lines · Russell King - ARM Linux
Quoted 9 lines · Arnd Bergmann
Quoted 3 lines · Uwe Kleine-K?nig

While cleaning up my repo I refound the patches and rebased them on top of today's Linus' tree and only needed to fix up the l2x0_lock patch as in the meantime a new usage hit mainline.

The patches all look harmless, but none of them has any information on why the particular locks need to be raw_spin_lock. Ideally a raw spinlock should be the absolute exception, and IMHO should have a comment in front of it why it is special.

Or at least explained in the patch description.

Quoted 8 lines
Quoted 4 lines

For instance, can someone explain why the lock for leds and gpio stuff on Footbridge needs to be converted? What is the original problem? More importantly, what is the criteria for using a raw spinlock instead of a normal spinlock?

raw_spinlock is still a spinlock when PREEMPT_RT is enabled, mere spinlocks become magically "sleeping" spinlocks (aka. PI aware rtmutexes)

I think we all understood that part of the series description.

Quoted 4 lines

Vs. the patches: IIRC, it was all fallout from running -rt, but that needs to be looked at case by case. Some of those are obvious as they are called deep down in atomic irq disabled code, but others might be just due to laziness reasons.

Not as obvious as you'd think. The explanation "called in atomic irq disabled code" makes sense, but I hadn't thought of that before -- I had expected all that code to use IRQ threads in case of PREEMPT_RT.

Requiring a comment there would probably eliminate the laziness issues, adding a comment that you can't be bothered to do the right fix won't help getting a patch merged, so ideally you'd end up with a better solution getting merged.

Arnd