summary refs log tree commit diff stats
path: root/results/classifier/zero-shot/108/other/1772086
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/zero-shot/108/other/1772086')
-rw-r--r--results/classifier/zero-shot/108/other/177208648
1 files changed, 48 insertions, 0 deletions
diff --git a/results/classifier/zero-shot/108/other/1772086 b/results/classifier/zero-shot/108/other/1772086
new file mode 100644
index 000000000..a6a7b0bf6
--- /dev/null
+++ b/results/classifier/zero-shot/108/other/1772086
@@ -0,0 +1,48 @@
+semantic: 0.766
+other: 0.657
+device: 0.626
+PID: 0.554
+graphic: 0.549
+files: 0.546
+network: 0.489
+performance: 0.468
+vnc: 0.429
+socket: 0.409
+permissions: 0.385
+debug: 0.383
+boot: 0.294
+KVM: 0.257
+
+malformed serial data being sent from guest
+
+When sending data through serial from guest each time 0x0A byte is sent 0x0D is sent before it. For example, when sending {0x29, 0x0A} on the other end I receive {0x29, 0x0D, 0x0A}.
+
+Something somewhere in the stack is converting LF to CRLF. This could be something inside your guest, or in QEMU, or in the host; to find out where we need more detail.
+
+Can you describe your setup, including:
+ * complete QEMU command line
+ * how you're sending data inside the guest
+ * how you're reading it on the host end
+please?
+
+
+I am unable to provide complete QEMU command line as I'm using virt-manager to deal with configuration. I can say that two serial ports are linked with physical ones through the /dev/ttyS* files.
+The guests I tested it with are Windows 98 and Windows XP. For the testing I connected one port to another. I could confirm through a kernel level serial monitor that I was indeed sending just \n but on the second port I received \r\n. I also received \r\n when the port was read by the host.
+Host is Ubuntu Xenial. 
+
+After doing a bit of research I think line 142 in file chardev/char-serial.c is problematic. https://github.com/qemu/qemu/blob/master/chardev/char-serial.c#L142
+
+It enables output processing, which is something unwanted here. With a simple test program I found out that by default, besides OPOST, ONLCR flag is set in c_oflag. I guess fix would be removing OPOST flag, which would disable any output processing, or setting c_oflag to 0 just to be sure.
+
+Seems like the problems isn't really new and might be at least 6 years old if not more. https://robert.penz.name/550/mapping-a-serial-device-to-a-kvm-guest-may-lead-to-communication-problems/
+
+It's even older than 6 years, see:
+https://lists.nongnu.org/archive/html/qemu-devel/2006-06/msg00196.html
+and:
+https://bugs.launchpad.net/qemu/+bug/1407813
+https://bugs.launchpad.net/qemu/+bug/1715296
+
+
+Patch has now been committed here:
+https://git.qemu.org/?p=qemu.git;a=commitdiff;h=12fb0ac0575df83cec72ec
+