Em seg., 1 de fev. de 2021 às 12:09, Jan Beulich escreveu: > > On 01.02.2021 15:46, Claudemir Todo Bom wrote: > > Tested first without the debug patch and with following parameters: > > And this test was all three of the non-debugging patches? Yes, all three patches. > > xen: dom0_mem=1024M,max:2048M dom0_max_vcpus=4 dom0_vcpus_pin=true smt=true > > kernel: loglevel=3 > > > > same behaviour as before... black screen right after the xen messages. > > > > adding earlyprintk=xen to the kernel command line is sufficient to > > make it boot, I can imagine this can be happening because Xen is not > > releasing console to the kernel at that moment. > > If the answer to the above question is "yes", then I start > suspecting this to be a different problem. I'm not sure I > see a way to debug this without having access to any output > (i.e. neither video nor serial). Without "earlyprintk=xen" > and instead with "vga=keep watchdog" on the Xen command > line, is there anything helpful (without or if need be with > the debugging patch in place)? with "vga=text-80x25,keep watchdog" and without the earlyprintk, system booted. I'm attaching the "xl dmesg" and "dmesg" outputs here. > > > The system worked well (with earlyprintk=xen), tested with the "yes > > stress test" mentioned before on a guest and on dom0. > > > > Then, I installed the debug patch and booted it again, it also needed > > the earlyprintk=xen parameter on the kernel command line. I've also > > added console_timestamps=boot to the xen command line in order to get > > the time of the messages. > > > > I'm attaching the outputs of "xl dmesg" and "dmesg" on this message. > > > > Think it is almost done! WIll wait for the next round of tests! > > As per above, not sure if there's going to be one. Thanks > for your patient testing! I can live with the "earlyprintk=xen" or any other solution that makes it boot. I'm pretty sure this problem will appear in tests of other people with a setup that helps debug deeper. Thank you for your work! Best regards, Claudemir