Lest anyone think this is resolved, I just experienced it on buster, running kernel 4.19.0-16-amd64 and 5.2+dfsg-9~bpo10+1 from buster-backports.
In my case, though, I've used qemu-nbd without difficulty for raw files on the order of 40G before without difficulty - but this time, with a VDI of actual size ~19G and...let's call it "virtual" size 100GB, I wrote around a GB of data to it and then suddenly [2005064.948700] block nbd0: Connection timed out [2005064.951474] block nbd0: shutting down sockets [2005064.951479] print_req_error: I/O error, dev nbd0, sector 39028592 [2005064.954230] block nbd0: Connection timed out [2005064.956982] print_req_error: I/O error, dev nbd0, sector 39029104 [2005064.958271] block nbd0: Connection timed out [2005064.959345] print_req_error: I/O error, dev nbd0, sector 39029360 [2005064.960527] block nbd0: Connection timed out [2005064.961608] print_req_error: I/O error, dev nbd0, sector 39029616 [2005064.962677] block nbd0: Connection timed out [2005064.963726] print_req_error: I/O error, dev nbd0, sector 39029872 (I presume if I hadn't noticed and had waited long enough I too would see "task blocked for xyz seconds") The fact that the original report on this particular bug was using a VDI as well makes me suspect it might be a problem with handling VDIs - I'm going to try converting it and report back... - Rich
