Alveo U200 shows only Xilinx JTAG/UART (903f) , no 5004/5005 devices

[xuanzhengyang]

I’m trying to bring up an Alveo U200 on a Dell PowerEdge R750 running AlmaLinux 9.6 (kernel 5.14.0-570.39.1.el9_6.x86_64). XRT 2025.1 (2.19.194) installs fine and the xocl/xclmgmt modules load, but the PCIe bus only ever shows the JTAG/UART function (Device ID 903f). The expected accelerator functions (Device IDs 5004/5005) never appear, so XRT can’t detect the card.

Environment

• Server: Dell PowerEdge R750, BIOS 1.14.1, both CPU sockets populated, PSU 1400 W
• OS: AlmaLinux 9.6 (GLIBC 2.34)
• Kernel: 5.14.0-570.39.1.el9_6.x86_64
• XRT: 2.19.194 (branch 2025.1), installed from RPM, /opt/xilinx/xrt/bin/xbmgmt --version matches

Key commands and outputs

1. sudo lspci -vd 10ee:

   b1:00.0 Serial controller: Xilinx Corporation Device 903f (prog-if 01 [16450])
           Subsystem: Xilinx Corporation Device 0007
           Flags: fast devsel, NUMA node 1
           Memory at dbc10000 (32-bit, non-prefetchable) [size=4K]
           Memory at dbc00000 (64-bit, non-prefetchable) [size=64K]
           Capabilities: [40] Power Management version 3
           Capabilities: [48] MSI: Enable- Count=1/1 Maskable- 64bit+
           Capabilities: [70] Express Endpoint, MSI 00
           Capabilities: [100] Advanced Error Reporting
           Capabilities: [1c0] Secondary PCI Express
   

2. sudo lspci -vvv -s b1:00.0

   b1:00.0 Serial controller: Xilinx Corporation Device 903f (prog-if 01 [16450])
           Subsystem: Xilinx Corporation Device 0007
           Control: I/O+ Mem+ BusMaster-
           Status: Cap+ … INTx-
           NUMA node: 1
           Region 0: Memory at dbc10000 (32-bit, non-prefetchable) [size=4K]
           Region 2: Memory at dbc00000 (64-bit, non-prefetchable) [size=64K]
           …
           LnkSta: Speed 8GT/s (ok), Width x8 (downgraded)
           …
   

3. sudo /opt/xilinx/xrt/bin/xbmgmt examine

   System Configuration
     OS Name              : Linux
     Release              : 5.14.0-570.39.1.el9_6.x86_64
     Machine              : x86_64
     CPU Cores            : 64
     Memory               : 514906 MB
     Distribution         : AlmaLinux 9.6 (Sage Margay)
     GLIBC                : 2.34
     Model                : PowerEdge R750
     BIOS Vendor          : Dell Inc.
     BIOS Version         : 1.14.1
   
   XRT
     Version              : 2.19.194
     Branch               : 2025.1
     Hash                 : 7d8151e6ee73c6ec2e99501a58c9c2eca6cc68ce
     Hash Date            : 2025-05-18 11:45:44
     xocl                 : 2.19.194, 7d8151e6ee73c6ec2e99501a58c9c2eca6cc68ce
     xclmgmt              : 2.19.194, 7d8151e6ee73c6ec2e99501a58c9c2eca6cc68ce
   
   Device(s) Present
     0 devices found
   

4. sudo lsmod | grep -E "xocl|xclmgmt"

   xocl     2609152  0
   xclmgmt  1417216  0
   

Current symptoms

• lspci never shows Device ID 5004/5005, only 903f.
• BusMaster remains disabled.
• xbmgmt examine reports “0 devices found”; no /dev/xclmgmt* nodes.
• No driver is bound under /sys/bus/pci/devices/0000:b1:00.0/driver.

Troubleshooting done

• Installed SC/CMC/base/validate RPMs for U200 (xilinx-u200-gen3x16-xdma-*).
• Rebuilt DKMS modules (sudo dkms remove/install xrt/2.19.194).
• Verified xocl/xclmgmt load (modprobe).
• Cold booted (shutdown, power cords removed for several minutes, reapply power).
• Reseated U200 card, reattached 8‑pin AUX power (checked firmly seated).
• No USB cable attached to the card.
• Tried xbmgmt program --base --device <BDF> but it fails because the card isn’t detected beyond 903f.

Suspicions / Questions

• Could AUX power still be an issue despite reseating? Does the R750 need a specific power profile for Alveo cards?
• Is the card stuck in golden/safe mode that only exposes the JTAG function, requiring JTAG reprogramming?
• Are there BIOS or PCIe settings (slot bifurcation, hot-plug, etc.) that could prevent BusMaster from enabling?
• Any additional diagnostics recommended (e.g., lspci -xxxx, forcing golden image reload, different XRT version, different OS)?

Has anyone seen an Alveo (U200 or otherwise) that only enumerates as Device 903f with BusMaster disabled? Any pointers on recovering it to the normal 5004/5005 functions would be greatly appreciated.

Thanks!