Package: less
Version: 590-2.1~deb12u2
Severity:|important|
Since Bookworm I observed this several times, a less +F process constantly in
run mode
needing a lot of CPU processing time when there are no more newer data in log
file buffer:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 92642 0.0 0.0 37748 6364 ? S May15 0:00 sudo
su root 92643 0.0 0.0 37748 1044 ? Ss May15 0:00
\_ sudo su root 92644 0.0 0.0 20528 4328 ? S May15
0:00 \_ su root 92645 0.0 0.0 7320 4248 ? S
May15 0:00 \_ bash root 92649 85.3 0.0 5972 2360
? R May15 830:48 \_ less +F bkclientd_debug.log
In that case of zero reads less normally polls every 2 ms if I recall
correctly but here we see no waits (clock_nanosleep()) on zero reads, it
polls every 40 µs: 09:45:34.782289 read(3, "", 1) = 0 <0.000004>
09:45:34.782304 read(3, "", 1) = 0 <0.000013> 09:45:34.782344 read(3,
"", 1) = 0 <0.000014> 09:45:34.782374 read(3, "", 1) = 0 <0.000013>
09:45:34.782404 read(3, "", 1) = 0 <0.000013> 09:45:34.782442 read(3,
"", 1) = 0 <0.000015> 09:45:34.782482 read(3, "", 1) = 0 <0.000004>
09:45:34.782523 read(3, "", 1) = 0 <0.000014> 09:45:34.782562 read(3,
"", 1) = 0 <0.000014> 09:45:34.782589 read(3, "", 1) = 0 <0.000003>
09:45:34.782627 read(3, "", 1) = 0 <0.000014> 09:45:34.782667 read(3,
"", 1) = 0 <0.000004> 09:45:34.782707 read(3, "", 1) = 0 <0.000015>
09:45:34.782747 read(3, "", 1) = 0 <0.000003> 09:45:34.782819 read(3,
"", 1) = 0 <0.000011> 09:45:34.782853 read(3, "", 1) = 0 <0.000003>
09:45:34.782882 read(3, "", 1) = 0 <0.000004> 09:45:34.782924 read(3,
"", 1) = 0 <0.000004> 09:45:34.782964 read(3, "", 1) = 0 <0.000005>
09:45:34.782991 read(3, "", 1) = 0 <0.000014> 09:45:34.783032 read(3,
"", 1) = 0 <0.000014> 09:45:34.783071 read(3, "", 1) = 0 <0.000004>
09:45:34.783100 read(3, "", 1) = 0 <0.000014> 09:45:34.783139 read(3,
"", 1) = 0 <0.000014> 09:45:34.783178 read(3, "", 1) = 0 <0.000019>
09:45:34.783224 read(3, "", 1) = 0 <0.000014> 09:45:34.783264 read(3,
"", 1) = 0 <0.000014> 09:45:34.783303 read(3, "", 1) = 0 <0.000014>
09:45:34.783342 read(3, "", 1) = 0 <0.000015> 09:45:34.783385 read(3,
"", 1) = 0 <0.000015> 09:45:34.783425 read(3, "", 1) = 0 <0.000014> An
intact less +F has these nano_sleep() - poll() sequences between the
read() which we don't have anymore for less delivered with Bookworm:
12:20:19.583389 read(4, "", 4672) = 0 <0.000005> 12:20:19.583406
clock_nanosleep(CLOCK_REALTIME, 0, {tv_sec=0, tv_nsec=2000000}, NULL) =
0 <0.002061> 12:20:19.585484 poll([{fd=3, events=POLLIN}], 1, 0) = 0
(Timeout) <0.000006> 12:20:19.585507 poll([{fd=4,
events=POLLERR|POLLHUP}], 1, 0) = 0 (Timeout) <0.000005> 12:20:19.585527
read(4, "", 4672) = 0 <0.000005> 12:20:19.585545
clock_nanosleep(CLOCK_REALTIME, 0, {tv_sec=0, tv_nsec=2000000}, NULL) =
0 <0.002093> 12:20:19.587658 poll([{fd=3, events=POLLIN}], 1, 0) = 0
(Timeout) <0.000008> 12:20:19.587684 poll([{fd=4,
events=POLLERR|POLLHUP}], 1, 0) = 0 (Timeout) <0.000006> 12:20:19.587707
read(4, "", 4672) = 0 <0.000007> 12:20:19.587729
clock_nanosleep(CLOCK_REALTIME, 0, {tv_sec=0, tv_nsec=2000000}, NULL) =
0 <0.002073> 12:20:19.589843 poll([{fd=3, events=POLLIN}], 1, 0) = 0
(Timeout) <0.000010> 12:20:19.589898 poll([{fd=4,
events=POLLERR|POLLHUP}], 1, 0) = 0 (Timeout) <0.000010> 12:20:19.589929
read(4, "", 4672) = 0 <0.000007> 12:20:19.589963
clock_nanosleep(CLOCK_REALTIME, 0, {tv_sec=0, tv_nsec=2000000}, NULL) =
0 <0.002102> 12:20:19.592101 poll([{fd=3, events=POLLIN}], 1, 0) = 0
(Timeout) <0.000005> 12:20:19.592127 poll([{fd=4,
events=POLLERR|POLLHUP}], 1, 0) = 0 (Timeout) <0.000005> 12:20:19.592146
read(4, "", 4672) = 0 <0.000005> 12:20:19.592162
clock_nanosleep(CLOCK_REALTIME, 0, {tv_sec=0, tv_nsec=2000000}, NULL) =
0 <0.002099> 12:20:19.594299 poll([{fd=3, events=POLLIN}], 1, 0) = 0
(Timeout) <0.000009>
