Re: Verbosity during boot

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Thu, Feb 27, 2025 at 08:38:27PM +0000, Andrew Cooper wrote:
> I've raised this during review before, but:
> 
> > (XEN) [    1.209230] AMD-Vi: IOMMU Extended Features:
> > (XEN) [    1.213998] - Peripheral Page Service Request
> > (XEN) [    1.218849] - x2APIC
> > (XEN) [    1.221536] - NX bit
> > (XEN) [    1.224221] - Invalidate All Command
> > (XEN) [    1.228297] - Guest APIC
> > (XEN) [    1.236062] - Performance Counters
> > (XEN) [    1.244692] - Host Address Translation Size: 0x2
> > (XEN) [    1.254547] - Guest Address Translation Size: 0
> > (XEN) [    1.264313] - Guest CR3 Root Table Level: 0x1
> > (XEN) [    1.273925] - Maximum PASID: 0xf
> > (XEN) [    1.282338] - SMI Filter Register: 0x1
> > (XEN) [    1.291241] - SMI Filter Register Count: 0x2
> > (XEN) [    1.300607] - Guest Virtual APIC Modes: 0
> > (XEN) [    1.309655] - Dual PPR Log: 0x2
> > (XEN) [    1.317801] - Dual Event Log: 0x2
> > (XEN) [    1.326078] - Secure ATS
> > (XEN) [    1.333490] - User / Supervisor Page Protection
> > (XEN) [    1.342892] - Device Table Segmentation: 0x3
> > (XEN) [    1.351981] - PPR Log Overflow Early Warning
> > (XEN) [    1.361040] - PPR Automatic Response
> > (XEN) [    1.369341] - Memory Access Routing and Control: 0x1
> > (XEN) [    1.379012] - Block StopMark Message
> > (XEN) [    1.387273] - Performance Optimization
> > (XEN) [    1.395637] - MSI Capability MMIO Access
> > (XEN) [    1.404138] - Guest I/O Protection
> > (XEN) [    1.412042] - Host Access
> > (XEN) [    1.419105] - Enhanced PPR Handling
> > (XEN) [    1.427008] - Attribute Forward
> > (XEN) [    1.434494] - Host Dirty
> > (XEN) [    1.441308] - Virtualized IOMMU
> > (XEN) [    1.448699] - VMGuard I/O Support
> > (XEN) [    1.456345] - VM Table Size: 0x2
> > (XEN) [    1.491312] AMD-Vi: IOMMU 0 Enabled.
> > (XEN) [    1.499087] AMD-Vi: IOMMU 1 Enabled.
> > (XEN) [    1.506835] AMD-Vi: IOMMU 2 Enabled.
> > (XEN) [    1.514554] AMD-Vi: IOMMU 3 Enabled.
> > (XEN) [    1.522452] I/O virtualisation enabled
> 
> Lots of that information is not actually useful, not even for
> developers.  What's worse is that this is a release build of Xen and it
> still takes 0.3s to print the feature list alone.

VT-d is kind of similar, but not that verbose in the list of features.
We should probably adjust there too.

I would be fine with doing (didn't test this at all):

diff --git a/xen/drivers/passthrough/amd/iommu_detect.c 
b/xen/drivers/passthrough/amd/iommu_detect.c
index cede44e6518f..6bb5d5db9ac7 100644
--- a/xen/drivers/passthrough/amd/iommu_detect.c
+++ b/xen/drivers/passthrough/amd/iommu_detect.c
@@ -72,6 +72,9 @@ void __init get_iommu_features(struct amd_iommu *iommu)
             amd_iommu_max_paging_mode = 4 + iommu->features.flds.hats;
     }
 
+    if ( !iommu_verbose )
+        return;
+
     /* Don't log the same set of features over and over. */
     first = list_first_entry(&amd_iommu_head, struct amd_iommu, list);
     if ( iommu != first && iommu->features.raw == first->features.raw )