From 297b41d61536ae3016f46e99c5fd6448100b81cb Mon Sep 17 00:00:00 2001
From: Bo Chen <bchen@crusoe.ai>
Date: Thu, 20 Feb 2025 09:37:12 -0800
Subject: [PATCH] pci, vmm: vfio: Report device path on host with DMA map/unmap
 errors

In addition to the BDF information on the guest, reporting the device
path of the VFIO device on the host is more useful when it comes to
debug DMA map/unmap errors particularly ones caused by failing hardware
on the host.

Signed-off-by: Bo Chen <bchen@crusoe.ai>
---
 pci/src/vfio.rs           | 20 +++++++++++++-------
 vmm/src/device_manager.rs |  1 +
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/pci/src/vfio.rs b/pci/src/vfio.rs
index e9d768a9a..41481ad5d 100644
--- a/pci/src/vfio.rs
+++ b/pci/src/vfio.rs
@@ -7,6 +7,7 @@ use std::any::Any;
 use std::collections::{BTreeMap, HashMap};
 use std::io;
 use std::os::unix::io::AsRawFd;
+use std::path::PathBuf;
 use std::ptr::null_mut;
 use std::sync::{Arc, Barrier, Mutex};
 
@@ -48,10 +49,10 @@ pub(crate) const VFIO_COMMON_ID: &str = "vfio_common";
 pub enum VfioPciError {
     #[error("Failed to create user memory region: {0}")]
     CreateUserMemoryRegion(#[source] HypervisorVmError),
-    #[error("Failed to DMA map: {0} for device {1}")]
-    DmaMap(#[source] vfio_ioctls::VfioError, PciBdf),
-    #[error("Failed to DMA unmap: {0} for device {1}")]
-    DmaUnmap(#[source] vfio_ioctls::VfioError, PciBdf),
+    #[error("Failed to DMA map: {0} for device {1} (guest BDF: {2})")]
+    DmaMap(#[source] vfio_ioctls::VfioError, PathBuf, PciBdf),
+    #[error("Failed to DMA unmap: {0} for device {1} (guest BDF: {2})")]
+    DmaUnmap(#[source] vfio_ioctls::VfioError, PathBuf, PciBdf),
     #[error("Failed to enable INTx: {0}")]
     EnableIntx(#[source] VfioError),
     #[error("Failed to enable MSI: {0}")]
@@ -1411,6 +1412,7 @@ pub struct VfioPciDevice {
     iommu_attached: bool,
     memory_slot_allocator: MemorySlotAllocator,
     bdf: PciBdf,
+    device_path: PathBuf,
 }
 
 impl VfioPciDevice {
@@ -1428,6 +1430,7 @@ impl VfioPciDevice {
         memory_slot_allocator: MemorySlotAllocator,
         snapshot: Option<Snapshot>,
         x_nv_gpudirect_clique: Option<u8>,
+        device_path: PathBuf,
     ) -> Result<Self, VfioPciError> {
         let device = Arc::new(device);
         device.reset();
@@ -1453,6 +1456,7 @@ impl VfioPciDevice {
             iommu_attached,
             memory_slot_allocator,
             bdf,
+            device_path: device_path.clone(),
         };
 
         Ok(vfio_pci_device)
@@ -1658,7 +1662,9 @@ impl VfioPciDevice {
                                 user_memory_region.size,
                                 user_memory_region.host_addr,
                             )
-                            .map_err(|e| VfioPciError::DmaMap(e, self.bdf))?;
+                            .map_err(|e| {
+                                VfioPciError::DmaMap(e, self.device_path.clone(), self.bdf)
+                            })?;
                     }
                 }
             }
@@ -1719,7 +1725,7 @@ impl VfioPciDevice {
         if !self.iommu_attached {
             self.container
                 .vfio_dma_map(iova, size, user_addr)
-                .map_err(|e| VfioPciError::DmaMap(e, self.bdf))?;
+                .map_err(|e| VfioPciError::DmaMap(e, self.device_path.clone(), self.bdf))?;
         }
 
         Ok(())
@@ -1729,7 +1735,7 @@ impl VfioPciDevice {
         if !self.iommu_attached {
             self.container
                 .vfio_dma_unmap(iova, size)
-                .map_err(|e| VfioPciError::DmaUnmap(e, self.bdf))?;
+                .map_err(|e| VfioPciError::DmaUnmap(e, self.device_path.clone(), self.bdf))?;
         }
 
         Ok(())
diff --git a/vmm/src/device_manager.rs b/vmm/src/device_manager.rs
index 6cb0a3d4a..aaf3e5eb0 100644
--- a/vmm/src/device_manager.rs
+++ b/vmm/src/device_manager.rs
@@ -3431,6 +3431,7 @@ impl DeviceManager {
             memory_manager.lock().unwrap().memory_slot_allocator(),
             vm_migration::snapshot_from_id(self.snapshot.as_ref(), vfio_name.as_str()),
             device_cfg.x_nv_gpudirect_clique,
+            device_cfg.path.clone(),
         )
         .map_err(DeviceManagerError::VfioPciCreate)?;