#! /usr/bin/env stap

# This script tries to identify contended user-space locks by hooking
# into the futex system call.

/* Bits indicating sharing in futex_key.both.offset */
global FUT_OFF_INODE = 1 /* We set bit 0 if key has a reference on inode */
global FUT_OFF_MMSHARED = 2 /* We set bit 1 if key has a reference on mm */

global FUTEX_WAIT = 0 /*, FUTEX_WAKE = 1 */
global FUTEX_PRIVATE_FLAG = 128 /* linux 2.6.22+ */
global FUTEX_CLOCK_REALTIME = 256 /* linux 2.6.29+ */

global wait_keys # per-thread key strings for waiting futexes

global lock_waits # long-lived stats on (tid,lock) blockage elapsed time
global process_names # long-lived pid-to-execname mapping

@define KEY %( @cast(key, "union futex_key") %)
probe kernel.function("futex_wait_setup").return {
  if ($return != 0) next

  key = @entry(&$q->key)

  offset = @KEY->both->offset
  masked_offset = offset & ~(FUT_OFF_INODE | FUT_OFF_MMSHARED)

  if (offset & FUT_OFF_INODE) {
    address = @KEY->shared->pgoff * mem_page_size() + masked_offset
    path = inode_path(@KEY->shared->inode)
    wait_keys[tid()] = sprintf("INODE:%s+%p", path, address)
  } else if (offset & FUT_OFF_MMSHARED) {
    address = @KEY->private->address + masked_offset
    wait_keys[tid()] = sprintf("MMSHARED:%p", address)
  }
}

global entry_times%

probe syscall.futex {
  if ((op & ~(FUTEX_PRIVATE_FLAG|FUTEX_CLOCK_REALTIME)) != FUTEX_WAIT) next
  entry_times[tid()] = gettimeofday_us()
}

probe syscall.futex.return {  
  if (!(tid() in wait_keys)) next
  key = wait_keys[tid()]
  delete wait_keys[tid()]

  if (!(tid() in entry_times)) next

  elapsed = gettimeofday_us() - entry_times[tid()]
  lock_waits[pid(), key] <<< elapsed
  delete entry_times[tid()]
  if (!(pid() in process_names))
    process_names[pid()] = execname()
}

probe end {
  foreach ([pid+, lock] in lock_waits) 
    printf ("%s[%d] lock %s contended %d times, %d avg us\n",
            process_names[pid], pid, lock, @count(lock_waits[pid,lock]),
            @avg(lock_waits[pid,lock]))
}