On Thu, 2025-05-15 at 13:09 +0200, Paolo Bonzini wrote:
> However, EBADE handling dates back to 2011 (commit 63583cca745f, > "[SCSI] > Add detailed SCSI I/O errors", 2011-02-12) and yet the Windows tests > for > PR were failing before QEMU switched to SG_IO for reads and writes. > I > guess I have to try reverting that and retest, though. Thanks! This makes me realize that we could summarize the goal for future efforts (independent of the current patch set) roughly like this: "Emulate a SCSI disk on top of a (host) multipath device in a way that 1) failover works properly (like it would work for regular IO from the host itself), 2) Windows tests for PR (plus test case X, Y, ...) can be run successfully". Does this make sense? It implies that PR commands don't just need to be forwarded appropriately, we also need to pass meaningful error codes back to the guest. Martin