There are two issues with ratelimiting as far a I can see: 1. Messages may be lost even if their amount fit burst limit; 2. "suppressed" message may not be printed if there is no call to ___ratelimit() after interval ends. I address (1) issue in the second patch. While the (2) issue will require adding timers to print "suppressed" message and care if ratelimit is on stack and no more exists. Which looks too much at this point. Dmitry Safonov (2): random: Omit double-printing ratelimit messages lib/ratelimit: Lockless ratelimiting drivers/char/random.c | 22 +++++++-------- drivers/gpu/drm/i915/i915_perf.c | 4 +-- include/linux/ratelimit.h | 34 ++++++++++++++--------- lib/ratelimit.c | 59 +++++++++++++++++++--------------------- 4 files changed, 61 insertions(+), 58 deletions(-) -- 2.13.6 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
