https://bugs.kde.org/show_bug.cgi?id=411047
--- Comment #6 from Kevin Funk <kf...@kde.org> ---
Just did some some quick investigation, it's indeed from loading the album art.
Plasmashell output (trimmed):
```
[12:10:03.102] kio_http(21273)/kf5.kio.kio_http: unknown:
QUrl("http://o.scdn.co/image/ab67616d00001e0295bbdbc2ef2d339f081fbd13";)
...
[12:10:03.147] kio_http(21273)/kf5.kio.kio_http: unknown: ============ Sending
Header:
[12:10:03.148] kio_http(21273)/kf5.kio.kio_http: unknown: "GET
/image/ab67616d00001e0295bbdbc2ef2d339f081fbd13 HTTP/1.1"
[12:10:03.148] kio_http(21273)/kf5.kio.kio_http: unknown: "Host: o.scdn.co"
[12:10:03.148] kio_http(21273)/kf5.kio.kio_http: unknown: "Connection:
keep-alive"
...
[12:10:04.721] kio_http(21273)/kf5.kio.kio_http: unknown: keepAlive = true
[12:10:04.721] kio_http(21273)/kf5.kio.kio_http: unknown:
[12:10:04.722] kio_http(21273)/kf5.kio.kio_http: unknown: keep alive ( 60 )
...
```
=> So the reason for the http.so process staying around is the keep-alive
connection
I think that makes sense it general, but the http.so shouldn't be using that
much CPU when it in fact does nothing...
Attached the debugger to the http.so process and found this:
```
(gdb) bt
#0 0x00007f081befad26 in __GI_ppoll (fds=0x7fffae5af410, nfds=1,
timeout=<optimized out>, sigmask=0x0) at ../sysdeps/unix/sysv/linux/ppoll.c:39
#1 0x00007f081c6e5906 in qt_ppoll (fds=0x7fffae5af410, nfds=1,
timeout_ts=0x7fffae5af340) at
/home/kfunk/devel/src/qt5.12/qtbase/src/corelib/kernel/qcore_unix.cpp:132
#2 0x00007f081c6e59a1 in qt_safe_poll (fds=0x7fffae5af410, nfds=1,
timeout_ts=0x7fffae5af3a0) at
/home/kfunk/devel/src/qt5.12/qtbase/src/corelib/kernel/qcore_unix.cpp:162
#3 0x00007f081888c825 in qt_poll_msecs (fds=0x7fffae5af410, nfds=1, timeout=1)
at
../../include/QtCore/5.12.6/QtCore/private/../../../../../../../../src/qt5.12/qtbase/src/corelib/kernel/qcore_unix_p.h:381
#4 0x00007f081888f33d in QNativeSocketEnginePrivate::nativeSelect
(this=0x8bd700, timeout=1, checkRead=true, checkWrite=false,
selectForRead=0x7fffae5af4de, selectForWrite=0x7fffae5af4df) at
/home/kfunk/devel/src/qt5.12/qtbase/src/network/socket/qnativesocketengine_unix.cpp:1432
#5 0x00007f081888bd1f in QNativeSocketEngine::waitForReadOrWrite
(this=0x9220f0, readyToRead=0x7fffae5af4de, readyToWrite=0x7fffae5af4df,
checkRead=true, checkWrite=false, msecs=1, timedOut=0x0) at
/home/kfunk/devel/src/qt5.12/qtbase/src/network/socket/qnativesocketengine.cpp:1120
#6 0x00007f081887474c in QAbstractSocket::waitForReadyRead (this=0x9070b0,
msecs=1) at
/home/kfunk/devel/src/qt5.12/qtbase/src/network/socket/qabstractsocket.cpp:2257
#7 0x00007f0818aafbc8 in KIO::ConnectionBackend::waitForIncomingTask
(this=0x8fd580, ms=1) at
/home/kfunk/devel/src/kf5/kio/src/core/connectionbackend.cpp:205
#8 0x00007f0818ae9ae2 in KIO::SlaveBase::dispatchLoop (this=0x7fffae5af5f0) at
/home/kfunk/devel/src/kf5/kio/src/core/slavebase.cpp:328
#9 0x00007f0816a7f0cb in kdemain (argc=<optimized out>, argv=<optimized out>)
at /home/kfunk/devel/src/kf5/kio/src/ioslaves/http/http.cpp:115
#10 0x000000000020a49f in launch (argc=4, _name=0x84dad8
"/home/kfunk/devel/install/kf5/lib/x86_64-linux-gnu/plugins/kf5/kio/http.so",
args=<optimized out>, cwd=0x0, envc=0, envs=<optimized out>,
reset_env=<optimized out>, tty=0x0, avoid_loops=<optimized out>,
startup_id_str=0x2041a6 "0") at /home/kfunk/devel/src/k
f5/kinit/src/kdeinit/kinit.cpp:704
#11 0x000000000020abe5 in handle_launcher_request (sock=7, who=0x84dad0 "\004")
at /home/kfunk/devel/src/kf5/kinit/src/kdeinit/kinit.cpp:1142
#12 0x0000000000208ccb in handle_requests (waitForPid=<optimized out>) at
/home/kfunk/devel/src/kf5/kinit/src/kdeinit/kinit.cpp:1335
#13 0x00000000002084e1 in main (argc=1, argv=0x7fffae5aff48) at
/home/kfunk/devel/src/kf5/kinit/src/kdeinit/kinit.cpp:1773
```
We're apparently excessively calling QAbstractSocket::waitForReadyRead(msecs=1)
in a loop which causes high CPU usage. Adding a few breakpoints and continuing
running the process seems to confirm this:
```
Thread 1 "http.so" hit Breakpoint 2, 0x00007f081887474c in
QAbstractSocket::waitForReadyRead (this=0x9070b0, msecs=1) at
/home/kfunk/devel/src/qt5.12/qtbase/src/network/socket/qabstractsocket.cpp:2257
(gdb) c
Continuing.
Thread 1 "http.so" hit Breakpoint 1, 0x00007f0818ae9ae2 in
KIO::SlaveBase::dispatchLoop (this=0x7fffae5af5f0) at
/home/kfunk/devel/src/kf5/kio/src/core/slavebase.cpp:328
(gdb) c
Continuing.
(and repeating)
```
So the problem seems to be deeply in KIO Core.
Can someone more familiar with the framework shed some light on this and ensure
that the implementation of KIO::SlaveBase::dispatchLoop() is sane?
--
You are receiving this mail because:
You are watching all bug changes.
