So far, we've let the arm64 kernel start its meaningful time stamping of the kernel log pretty late, which is caused by sched_clock() being initialised rather late compared to other architectures.
Pavel Tatashin proposed[1] to move the initialisation of sched_clock much earlier, which I had objections to. The reason for initialising sched_clock late is that a number of systems have broken counters, and we need to apply all kind of terrifying workarounds to avoid time going backward on the affected platforms. Being able to identify the right workaround comes pretty late in the kernel boot, and providing an unreliable sched_clock, even for a short period of time, isn't an appealing prospect. To address this, I'm proposing that we allow an architecture to chose to (1) divorce time stamping and sched_clock during the early phase of booting, and (2) inherit the time stamping clock as the new epoch the first time a sched_sched clock gets registered. (1) would allow arm64 to provide a time stamping clock, however unreliable it might be, while (2) would allow sched_clock to provide time stamps that are continuous with the time-stamping clock. The last patch in the series adds the necessary logic to arm64, allowing the (potentially unreliable) time stamping of early kernel messages. Tested on a bunch of arm64 systems, both bare-metal and in VMs. Boot tested on a x86 guest. [1] https://lore.kernel.org/patchwork/patch/1015110/ Marc Zyngier (3): printk: Allow architecture-specific timestamping function sched/clock: Allow sched_clock to inherit timestamp_clock epoch arm64: Allow early time stamping arch/arm64/Kconfig | 3 +++ arch/arm64/kernel/setup.c | 44 +++++++++++++++++++++++++++++++++++++ include/linux/sched/clock.h | 13 +++++++++++ kernel/printk/printk.c | 4 ++-- kernel/time/sched_clock.c | 10 +++++++++ 5 files changed, 72 insertions(+), 2 deletions(-) -- 2.20.1
