On 4 August 2016 at 06:09, Hillf Danton <hillf.zj@xxxxxxxxxxxxxxx> wrote: >> >> If something we call in the fail_dump() code path tries to acquire a >> resource that might fail (due to fault injection), then we should not >> try to recurse back into the fault injection code. >> >> I've seen this happen with the console semaphore in the upcoming >> semaphore trylock fault injection code. >> >> Signed-off-by: Vegard Nossum <vegard.nossum@xxxxxxxxxx> >> --- >> lib/fault-inject.c | 34 ++++++++++++++++++++++++++++------ >> 1 file changed, 28 insertions(+), 6 deletions(-) >> >> diff --git a/lib/fault-inject.c b/lib/fault-inject.c >> index 6a823a5..adba7c9 100644 >> --- a/lib/fault-inject.c >> +++ b/lib/fault-inject.c >> @@ -100,6 +100,33 @@ static inline bool fail_stacktrace(struct fault_attr *attr) >> >> #endif /* CONFIG_FAULT_INJECTION_STACKTRACE_FILTER */ >> >> +static DEFINE_PER_CPU(int, fault_active); >> + >> +static bool __fail(struct fault_attr *attr) >> +{ >> + bool ret = false; >> + >> + /* >> + * Prevent recursive fault injection (this could happen if for >> + * example printing the fault would itself run some code that >> + * could fail) >> + */ >> + preempt_disable(); >> + if (unlikely(__this_cpu_inc_return(fault_active) != 1)) >> + goto out; >> + >> + ret = true; >> + fail_dump(attr); >> + >> + if (atomic_read(&attr->times) != -1) >> + atomic_dec_not_zero(&attr->times); >> + >> +out: >> + __this_cpu_dec(fault_active); >> + preempt_enable(); > > Well schedule entry point is add in paths like > rt_mutex_trylock > __alloc_pages_nodemask > and please add one or two sentences in log > message for it. I'm sorry, but I don't really get what you are saying or what you want me to add. Are you saying that because I'm adding a fail_dump() call to mutex_trylock() that we can now end up calling schedule() from a weird context? This patch is just to prevent __fail() from looping on itself, I don't see what the connection is to rt_mutex_trylock(), __alloc_pages_nodemask(), or schedule(). Could you please clarify? Thanks, Vegard -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx"> email@xxxxxxxxx </a>