Re: [KVM-autotest] virtio_console test v3.0
On (Wed) Sep 15 2010 [17:41:28], Lukas Doktor wrote: New version of virtio_console test is more similar to upstream C virtio_console test so it will be easier to maintain. Also we moved the scripts/console_switch.py to more sensible scripts/virtio_guest.py There are still some virtio_console bugs which restricts some usage (mostly with console). We have created workarounds which enables us to run the tests. They are marked as FIXME comments. I'm guessing you're seeing those bugs because you try open/close/etc. calls on the hvc devices. That's not how they're supposed to be used. As I mentioned previously, you should spawn a terminal on hvc and use it for console IO. I'll take a look at the rest in a bit. Amit -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [KVM-autotest] virtio_console test
On (Fri) Aug 20 2010 [16:12:51], Lukáš Doktor wrote: Dne 20.8.2010 15:40, Lukas Doktor napsal(a): Hi, This patch adds new test for virtio_console. It supports booth, serialport and console, virtio_console types and it contains three tests: 1) smoke 2) loopback 3) perf This is great, thanks for the tests. I was working with Lucas at the KVM Forum to get my virtio-console tests integrated upstream. I have a few micro tests that test correctness of various bits in the virtio_console code: http://fedorapeople.org/gitweb?p=amitshah/public_git/test-virtserial.git It would be great to sync up with Lucas and add those tests to autotest as well. Eventually, I'd also like to adapt the C code to python so that it integrates with KVM-autotest. Just a note about kernels: serialport works great but console have big issues. Use kernels = 2.6.35 for testing. Can you tell me what the issues are? I test console io in the testsuite mentioned above and it's been passing fine. Amit -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [KVM-autotest] virtio_console test
Hi Amit, Dne 23.8.2010 09:53, Amit Shah napsal(a): On (Fri) Aug 20 2010 [16:12:51], Lukáš Doktor wrote: Dne 20.8.2010 15:40, Lukas Doktor napsal(a): Hi, This patch adds new test for virtio_console. It supports booth, serialport and console, virtio_console types and it contains three tests: 1) smoke 2) loopback 3) perf This is great, thanks for the tests. I was working with Lucas at the KVM Forum to get my virtio-console tests integrated upstream. I have a few micro tests that test correctness of various bits in the virtio_console code: http://fedorapeople.org/gitweb?p=amitshah/public_git/test-virtserial.git yes, I'm aware of your tests and the fedora page about virtio-console. I took an inspiration from them ;-) (thanks) It would be great to sync up with Lucas and add those tests to autotest as well. I went through the code trying to find tests missing in our autotest virtio_console.py test. Correct me if I'm wrong, but the missing tests are: * variations on opening/writing/reading host/guest closed/open consoles and check the right handling * guest caching (if necessarily) Eventually, I'd also like to adapt the C code to python so that it integrates with KVM-autotest. Yes, either you or I can do that. I would just need the test-requirement-list. Just a note about kernels: serialport works great but console have big issues. Use kernels= 2.6.35 for testing. Can you tell me what the issues are? I test console io in the testsuite mentioned above and it's been passing fine. Sometimes on F13 there were Oops while booting guests with virtconsoles. Also the sysfs informations about virtconsoles were sometimes mismatched. With vanilla kernel 2.6.35 it worked better with occasional Oops while booting the guest. My colleague was working on that part so when he's back from holiday he can provide more information. With virtserialport it always worked correctly. Amit Lukáš -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [KVM-autotest] virtio_console test
On (Mon) Aug 23 2010 [15:20:17], Lukáš Doktor wrote: Hi Amit, Dne 23.8.2010 09:53, Amit Shah napsal(a): On (Fri) Aug 20 2010 [16:12:51], Lukáš Doktor wrote: Dne 20.8.2010 15:40, Lukas Doktor napsal(a): Hi, This patch adds new test for virtio_console. It supports booth, serialport and console, virtio_console types and it contains three tests: 1) smoke 2) loopback 3) perf This is great, thanks for the tests. I was working with Lucas at the KVM Forum to get my virtio-console tests integrated upstream. I have a few micro tests that test correctness of various bits in the virtio_console code: http://fedorapeople.org/gitweb?p=amitshah/public_git/test-virtserial.git yes, I'm aware of your tests and the fedora page about virtio-console. I took an inspiration from them ;-) (thanks) It would be great to sync up with Lucas and add those tests to autotest as well. I went through the code trying to find tests missing in our autotest virtio_console.py test. Correct me if I'm wrong, but the missing tests are: * variations on opening/writing/reading host/guest closed/open consoles and check the right handling Mostly checking for the open/read/write/close semantics, poll(), a new test that I added today for checking of blocking calls (read) in one thread while doing write in another. I actually want to add a lot more (and even convert existing ones) to threads, since currently I can't guarantee that the guest commands are processed before any subsequent host commands (this means threading in the host, obviously). * guest caching (if necessarily) Yes, guest caching isn't needed in the current code (they're kept around from the previous code where we had such a thing). Eventually, I'd also like to adapt the C code to python so that it integrates with KVM-autotest. Yes, either you or I can do that. I would just need the test-requirement-list. I'd much prefer someone else did it, as I'm not familiar with python (yet). I'll be willing to help you or whoever picks it up. Just a note about kernels: serialport works great but console have big issues. Use kernels= 2.6.35 for testing. Can you tell me what the issues are? I test console io in the testsuite mentioned above and it's been passing fine. Sometimes on F13 there were Oops while booting guests with virtconsoles. Also the sysfs informations about virtconsoles were sometimes mismatched. With vanilla kernel 2.6.35 it worked better with occasional Oops while booting the guest. I see. Can you give me any logs / oops messages / traces? 2.6.35 is the first kernel version to officially support the new features and is complete barring a few fixes which I haven't pushed out yet. There's a chance that the F13 kernel doesn't have some fixes, but I did backport all of them AFAIR. My colleague was working on that part so when he's back from holiday he can provide more information. OK, let me know if you get more info. Thanks, Amit -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[KVM-autotest] virtio_console test
Hi, This patch adds new test for virtio_console. It supports booth, serialport and console, virtio_console types and it contains three tests: 1) smoke 2) loopback 3) perf Before any tests are executed it starts the machine with required number of virtio_consoles. Than it allows user to run all three tests. Using the parameters user can control which tests are executed and what setting is used. All tests supports multiple run using ';' separated list of settings. Most of the settings are optional only. The mandatory ones are written in CAPITALS. ad1) virtio_console_smoke format: $VIRTIO_CONSOLE_TYPE:$custom_data It creates a loopback via $VIRTIO_CONSOLE_TYPE console and sends $custom_data. If the received data match the original test pass ad2) virtio_console_loopback format: $source_console_t...@$buffer_length:$destination_console1_t...@$buffer_length:...:$destination_consolex_t...@$buffer_length:$loopback_buffer_length Creates loopback between the $SOURCE_CONSOLE_TYPE console and all following $DESTINATION_CONSOLEn_TYPE consoles. Than it sends data by $buffer_length to the source port. The loopback resends the data by $loopback_buffer_length to all destination consoles. The test listens on the destination consoles and controls the received data. NOTE: in the debug mode you can see the send/received data's buffers in every second during the test. ad3) virtio_console_perf format: $virtio_console_t...@$buffer_size:$test_duration First it sends the prepared data in a loop over $VIRTIO_CONSOLE_TYPE console from host to guest. Guest only reads all the data and throw them away. This part runs $test_duration seconds. Second it does the same from guest to host. For booth runs it provides information of minimum/median/maximum throughput and guest/host average loads. Best regards, Lukas Doktor -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [KVM-autotest] virtio_console test
Dne 20.8.2010 15:40, Lukas Doktor napsal(a): Hi, This patch adds new test for virtio_console. It supports booth, serialport and console, virtio_console types and it contains three tests: 1) smoke 2) loopback 3) perf Before any tests are executed it starts the machine with required number of virtio_consoles. Than it allows user to run all three tests. Using the parameters user can control which tests are executed and what setting is used. All tests supports multiple run using ';' separated list of settings. Most of the settings are optional only. The mandatory ones are written in CAPITALS. ad1) virtio_console_smoke format: $VIRTIO_CONSOLE_TYPE:$custom_data It creates a loopback via $VIRTIO_CONSOLE_TYPE console and sends $custom_data. If the received data match the original test pass ad2) virtio_console_loopback format: $source_console_t...@$buffer_length:$destination_console1_t...@$buffer_length:...:$destination_consolex_t...@$buffer_length:$loopback_buffer_length Creates loopback between the $SOURCE_CONSOLE_TYPE console and all following $DESTINATION_CONSOLEn_TYPE consoles. Than it sends data by $buffer_length to the source port. The loopback resends the data by $loopback_buffer_length to all destination consoles. The test listens on the destination consoles and controls the received data. NOTE: in the debug mode you can see the send/received data's buffers in every second during the test. ad3) virtio_console_perf format: $virtio_console_t...@$buffer_size:$test_duration First it sends the prepared data in a loop over $VIRTIO_CONSOLE_TYPE console from host to guest. Guest only reads all the data and throw them away. This part runs $test_duration seconds. Second it does the same from guest to host. For booth runs it provides information of minimum/median/maximum throughput and guest/host average loads. Best regards, Lukas Doktor -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Just a note about kernels: serialport works great but console have big issues. Use kernels = 2.6.35 for testing. -- To unsubscribe from this list: send the line unsubscribe kvm in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html