It seems the battstat applet likes to poll /proc/apm (in APM mode) rather
regularly: once per second I get:

1129538207.808299 poll([{fd=17, events=POLLIN}, {fd=3, events=POLLIN}, {fd=26,
events=POLLIN|POLLPRI}, {fd=28, events=POLLIN|POLLPRI}, {fd=29,
events=POLLIN|POLLPRI}, {fd=30, events=POLLIN|POLLPRI}, {fd=33,
events=POLLIN|POLLPRI}, {fd=31, events=POLLIN|POLLPRI}, {fd=34,
events=POLLIN|POLLPRI}, {fd=35, events=POLLIN|POLLPRI}], 10, 717) = 0
1129538208.533235 gettimeofday({1129538208, 535807}, NULL) = 0
1129538208.535994 open("/proc/apm", O_RDONLY) = 36
1129538208.536172 fstat64(36, {st_mode=S_IFREG|0444, st_size=0, ...}) = 0
1129538208.536276 mmap2(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40029000
1129538208.536321 read(36, "1.16ac 1.2 0x03 0x01 0x00 0x01 1"..., 1024) = 41
1129538208.553297 close(36)             = 0
1129538208.553462 munmap(0x40029000, 4096) = 0
1129538208.553521 ioctl(3, FIONREAD, [0]) = 0
1129538208.553562 gettimeofday({1129538208, 553569}, NULL) = 0

It often comes onto my 'top' - chewing a percentage of CPU.

It's really not clear to me that this polling is necessary; in an ideal world
surely we'd get a kernel generated HAL event ? but - worse - hald is also doing
a chunk of work here:

1129538316.102195 gettimeofday({1129538316, 102278}, NULL) = 0
1129538316.104001 open("//proc/apm", O_RDONLY|O_LARGEFILE) = 3
1129538316.111954 fstat64(3, {st_mode=S_IFREG|0444, st_size=0, ...}) = 0
1129538316.120840 mmap2(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7f47000
1129538316.123954 read(3, "1.16ac 1.2 0x03 0x01 0x00 0x01 1"..., 1024) = 41
1129538316.136679 close(3)              = 0
1129538316.142844 munmap(0xb7f47000, 4096) = 0
1129538316.143961 open("//proc/apm", O_RDONLY|O_LARGEFILE) = 3
1129538316.151995 fstat64(3, {st_mode=S_IFREG|0444, st_size=0, ...}) = 0
1129538316.159956 mmap2(NULL, 4096, PROT_READ|PROT_WRITE,
MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7f47000
1129538316.167379 read(3, "1.16ac 1.2 0x03 0x01 0x00 0x01 1"..., 1024) = 41
1129538316.171957 close(3)              = 0
1129538316.179958 munmap(0xb7f47000, 4096) = 0
1129538316.188001 gettimeofday({1129538316, 188484}, NULL) = 0
1129538316.194678 poll([{fd=5, events=POLLIN}, {fd=10, events=POLLIN}, {fd=7,
events=POLLIN}, {fd=9, events=POLLIN|POLLPRI}, {fd=8, events=POLLIN}, {fd=12,
events=POLLIN}], 6, 1914) = 0

Of course - hald appears to like to read /proc/apm -twice- every two seconds
~thus presumably giving higher latency without reduced cost ;-> - all looks a
little bit odd to me ...