Jan Kiszka wrote:
Hi again,
here comes the first update of the new latency tracer.
arch/i386/kernel/entry.S | 27 +++
arch/i386/kernel/ipipe-root.c | 4
include/asm-i386/system.h | 70 +++++++++
include/linux/ipipe_trace.h | 3
kernel/ipipe/Kconfig | 18 ++
kernel/ipipe/tracer.c | 247 +++++++++++++++++++++++-----------
6 files changed, 288 insertions(+), 81 deletions(-)
The list of changes:
o save ipipe_trace_freeze() (and only this function!) from NMI context
o CONFIG_IPIPE_TRACE_IRQSOFF instrumentation traces the longest
hard-IRQs-off phase of the system (regarding limitations see my
earlier posting today)
o fixed subtle trace mess-up
o added some "unlikely" to the tracer's hotpath
o added ipipe_trace_{max|frozen}_reset() to the API
To-do
o fix leaking IRQs-off trace
o documentation...
Philippe, I think this should already be merged into Adeos although
TRACE_IRQSOFF is not perfect yet.
Ok, it's queued for review. I'm busy right now integrating Dmitry's
shared IRQ infrastructure into all supported archs and revalidating Xeno
with the shortened interrupt path; once that's done and a first round of
patches is rolled out, a second one will follow for x86 with the tracer
fixes.
--
Philippe.