https://bugzilla.kernel.org/show_bug.cgi?id=218982
--- Comment #20 from Vasant Hegde (vasant.he...@amd.com) --- Hi Tony, (In reply to Tony Houghton from comment #18) > (In reply to Vasant Hegde from comment #13) > > Thanks for the dmesg Tony. > > Thank you for investigating and helping; sorry about the delay in replying. > > > Only other noticeable difference is X2APIC support. Looks like x2apic is > not > > enabled in BIOS. But driver will use MSI interrupt and it should work fine. > > (I am referring to iommu_init_irq() code path). > > Is that likely to be significant? I Googled x2apic, and apparently it > doesn't do much besides provide support for more than 256 CPU cores. Again, > I can't find a BIOS option for this, but I did eventually find one to > disable IOMMU (see below). Its not significant. Even with APIC mode it should work fine. > > > By any chance do you see any message on console once you resume the system? > > No, the display output stays off, and so does networking, so I can't see > anything. All that happens is the fans and LEDs etc come on, but the system > is completely unresponsive. My keyboard lights up too, but the caps lock LED > doesn't toggle when I press the caps lock key. > > > Just to double check, can you try out below change? This is queued for > > v6.13 merge window. Basically this patch will skip copy_device_table() code > > path in resume path (This is to rule out the resume code taking wrong path > > in copy_device_table() function). > ... > > commit 3f6eeada6930056c38f20964120ac402cf0030b9 > > I've added that patch, and it doesn't make a difference that I can tell. Thanks for testing. > > I see that de111f6b4f6a3010020825d22a068f416bc29c95, which causes the issue > for me, is also very simple. AIUI, de111f6b4f6a removes the calls to > iommu_enable_gt from early_enable_iommu and early_enable_iommus and calls > iommu_enable_gt from iommu_init_pci instead. Then the patch from > <https://lore.kernel.org/linux-iommu/20240621101533.20216-1-vasant.hegde@amd. > com/> restored iommu_enable_gt to early_enable_iommu but didn't affect the > other two locations. So that makes me think calling iommu_enable_gt from > iommu_init_pci is what's upsetting my board. Right. That's commit 150bdf5f8d. I don't think iommu_init_pci() path gets called in resume path. Only resume callback handler gets called. -Vasant -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug. _______________________________________________ acpi-bugzilla mailing list acpi-bugzilla@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/acpi-bugzilla
