Date: Thu, 11 Dec 2025 08:35:53 +0000 From: bugzilla-noreply@freebsd.org To: virtualization@FreeBSD.org Subject: [Bug 290920] FreeBSD 15 Regression: BHyve PCI Passthrough Broken on 15-STABLE (No GSI/Interrupt Remapping Failure) Message-ID: <bug-290920-27103-VfTFlZcxYP@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-290920-27103@https.bugs.freebsd.org/bugzilla/> References: <bug-290920-27103@https.bugs.freebsd.org/bugzilla/>
index | next in thread | previous in thread | raw e-mail
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=290920 --- Comment #23 from dan@sunsaturn.com --- (In reply to Bojan Novković from comment #20) I appreciate you sending the patch. I tested it by first reverting my local vmmapi.c to the unpatched original, then rebuilding both libvmmapi and bhyve to ensure a fresh link: Test Procedure: cd /usr/src/lib/libvmmapi make clean cp vmmapi.c.old vmmapi.c # Revert to unpatched original patch -p1 -d /usr/src < /root/freebsd-usb2.patch # Apply your patch make && make install cd /usr/src/usr.sbin/bhyve make clean make && make install I have completed testing of your proposed patch for libvmmapi.c (using GPA as offset). The testing confirms that the ordering of memory segment mapping remains the critical factor for my environment, and the proposed GPA-based offset patch introduces a new issue. TEST RESULTS SUMMARY: 1. Your Patch (GPA Offset applied): - VM Boots: Yes - Passthrough Works: No - Guest Error Observed: PCI INT A: no GSI (interrupt routing failure) 2. Original Patch (High -> Low Ordering re-applied for verification): - VM Boots: Yes - Passthrough Works: Yes (Full functionality restored for GPU and USB) CONCLUSION: The original patch that forces the High Memory segment to be mapped before the Low Memory segment successfully resolves the passthrough issue on my updated 15.0 system. The failure of the GPA-based patch to correctly set up the PCI interrupt routing (PCI INT A: no GSI) suggests that the explicit High -> Low ordering of the map updates is required by the VMM or IOMMU for proper initialization of passthrough devices, overriding any need for the GPA offset correction in this configuration. I will continue using the ordering patch for now. -- You are receiving this mail because: You are the assignee for the bug.help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-290920-27103-VfTFlZcxYP>