aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorStefan Hajnoczi <stefanha@redhat.com>2019-10-25 14:22:36 +0200
committerStefan Hajnoczi <stefanha@redhat.com>2019-12-13 11:24:07 +0000
commit86d2a49b41832355ab50cf60cec0cd50680fc0e5 (patch)
treeda4911e9fee08ec5f78b7a1fdda6448ce490563a
parent5f2585772fff7ada93cb6691c51603451228a09c (diff)
iothread: document -object iothread on man page
Add -object iothread documentation to the man page, including references to the query-iothread QMP command and qom-set syntax for adjusting adaptive polling parameters at run-time. Reported-by: Zhenyu Ye <yezhenyu2@huawei.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Message-id: 20191025122236.29815-1-stefanha@redhat.com Message-Id: <20191025122236.29815-1-stefanha@redhat.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
-rw-r--r--qemu-options.hx38
1 files changed, 38 insertions, 0 deletions
diff --git a/qemu-options.hx b/qemu-options.hx
index 65c9473b73..68d1592ccc 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -4926,6 +4926,44 @@ access
CN=laptop.example.com,O=Example Home,L=London,ST=London,C=GB
@end example
+@item -object iothread,id=@var{id},poll-max-ns=@var{poll-max-ns},poll-grow=@var{poll-grow},poll-shrink=@var{poll-shrink}
+
+Creates a dedicated event loop thread that devices can be assigned to. This is
+known as an IOThread. By default device emulation happens in vCPU threads or
+the main event loop thread. This can become a scalability bottleneck.
+IOThreads allow device emulation and I/O to run on other host CPUs.
+
+The @option{id} parameter is a unique ID that will be used to reference this
+IOThread from @option{-device ...,iothread=@var{id}}. Multiple devices can be
+assigned to an IOThread. Note that not all devices support an
+@option{iothread} parameter.
+
+The @code{query-iothreads} QMP command lists IOThreads and reports their thread
+IDs so that the user can configure host CPU pinning/affinity.
+
+IOThreads use an adaptive polling algorithm to reduce event loop latency.
+Instead of entering a blocking system call to monitor file descriptors and then
+pay the cost of being woken up when an event occurs, the polling algorithm
+spins waiting for events for a short time. The algorithm's default parameters
+are suitable for many cases but can be adjusted based on knowledge of the
+workload and/or host device latency.
+
+The @option{poll-max-ns} parameter is the maximum number of nanoseconds to busy
+wait for events. Polling can be disabled by setting this value to 0.
+
+The @option{poll-grow} parameter is the multiplier used to increase the polling
+time when the algorithm detects it is missing events due to not polling long
+enough.
+
+The @option{poll-shrink} parameter is the divisor used to decrease the polling
+time when the algorithm detects it is spending too long polling without
+encountering events.
+
+The polling parameters can be modified at run-time using the @code{qom-set} command (where @code{iothread1} is the IOThread's @code{id}):
+
+@example
+(qemu) qom-set /objects/iothread1 poll-max-ns 100000
+@end example
@end table