brintos

brintos / linux-shallow public Read only

0
0
Text · 2.3 KiB · bdd2082 Raw
53 lines · plain
1=======================2IRQ-flags state tracing3=======================4 5:Author: started by Ingo Molnar <mingo@redhat.com>6 7The "irq-flags tracing" feature "traces" hardirq and softirq state, in8that it gives interested subsystems an opportunity to be notified of9every hardirqs-off/hardirqs-on, softirqs-off/softirqs-on event that10happens in the kernel.11 12CONFIG_TRACE_IRQFLAGS_SUPPORT is needed for CONFIG_PROVE_SPIN_LOCKING13and CONFIG_PROVE_RW_LOCKING to be offered by the generic lock debugging14code. Otherwise only CONFIG_PROVE_MUTEX_LOCKING and15CONFIG_PROVE_RWSEM_LOCKING will be offered on an architecture - these16are locking APIs that are not used in IRQ context. (the one exception17for rwsems is worked around)18 19Architecture support for this is certainly not in the "trivial"20category, because lots of lowlevel assembly code deal with irq-flags21state changes. But an architecture can be irq-flags-tracing enabled in a22rather straightforward and risk-free manner.23 24Architectures that want to support this need to do a couple of25code-organizational changes first:26 27- add and enable TRACE_IRQFLAGS_SUPPORT in their arch level Kconfig file28 29and then a couple of functional changes are needed as well to implement30irq-flags-tracing support:31 32- in lowlevel entry code add (build-conditional) calls to the33  trace_hardirqs_off()/trace_hardirqs_on() functions. The lock validator34  closely guards whether the 'real' irq-flags matches the 'virtual'35  irq-flags state, and complains loudly (and turns itself off) if the36  two do not match. Usually most of the time for arch support for37  irq-flags-tracing is spent in this state: look at the lockdep38  complaint, try to figure out the assembly code we did not cover yet,39  fix and repeat. Once the system has booted up and works without a40  lockdep complaint in the irq-flags-tracing functions arch support is41  complete.42- if the architecture has non-maskable interrupts then those need to be43  excluded from the irq-tracing [and lock validation] mechanism via44  lockdep_off()/lockdep_on().45 46In general there is no risk from having an incomplete irq-flags-tracing47implementation in an architecture: lockdep will detect that and will48turn itself off. I.e. the lock validator will still be reliable. There49should be no crashes due to irq-tracing bugs. (except if the assembly50changes break other code by modifying conditions or registers that51shouldn't be)52 53