vrtd: "Internal error" while loading/programming a bitstream

[arktt]

Description

A regression has been introduced on the dev branch that prevents loading bitstreams onto the V80 cards. When attempting to execute any bitstream, the application aborts with the error: Exception: Internal error in vrtd daemon or local libvrtd. This appears to be an issue in the bitstream loading path, as running v80-smi validate succeeds without any errors (see logs below).

Regression Details

• Failing State: dev branch at commit 4c79427 (post-merge of PR #126).
• Known Working State: dev_v1.0.0 branch at commit cb1ef56.

Details of an attempted run of the 00_axilite example are give below:

./build/00_axilite 55:00 ../../../../axilite_hw.vbin 
VRT Version: 1.0.0
[2026-06-26 16:27:12.973] [DEBUG] void vrt::Vrtbin::extract()                                                     : Extracting vrtbin: ../../../../axilite_hw.vbin
[2026-06-26 16:27:13.016] [DEBUG] void vrt::Vrtbin::copy(const string&, const string&)                            : Copying file <some-path>/.cache/SLASH/vrt/vrtbin_55_00/system_map.xml to <some-path>/.cache/SLASH/vrt/metadata_55_00/system_map.xml
[2026-06-26 16:27:13.016] [DEBUG] void vrt::Vrtbin::copy(const string&, const string&)                            : Copying file <some-path>/.cache/SLASH/vrt/vrtbin_55_00/report_utilization_axilite_hw.xml to <some-path>/.cache/SLASH/vrt/metadata_55_00/report_utilization.xml
[2026-06-26 16:27:13.018] [INFO ] void vrt::impl::Device::programDevice()                                         : Programming PDI via vrtd design writer <some-path>/.cache/SLASH/vrt/vrtbin_55_00/images/top_i_slash_slash_axilite_hw_inst_0_partial.pdi
Exception: Internal error in vrtd daemon or local libvrtd

and its accompanying vrtd logs:

vrtd logs

sudo journalctl -u vrtd -b
Jun 26 15:10:39 hostname systemd[1]: Starting vrtd daemon for SLASH...
Jun 26 15:10:39 hostname vrtd[962]: Starting vrtd...
Jun 26 15:10:39 hostname vrtd[962]: Configuration loaded successfully
Jun 26 15:10:39 hostname vrtd[962]: Discovered 1 device(s) matching /dev/slash_ctl*
Jun 26 15:10:39 hostname vrtd[962]: Stripping PF function .2 from kernel BDF 0000:55:00.2; device tracked as board 0000:55:00
Jun 26 15:10:39 hostname vrtd[962]: Matched QDMA device /dev/slash_qdma_ctl0 for ctldev /dev/slash_ctl0 (BDF 0000:55:00)
Jun 26 15:10:39 hostname vrtd[962]: Discovered 1 device(s)
Jun 26 15:10:39 hostname vrtd[962]: Listening on unix socket api
Jun 26 15:10:39 hostname systemd[1]: Started vrtd daemon for SLASH.
Jun 26 16:27:13 hostname vrtd[962]: Design write submitted dev=0 uid=1001 conn_id=4
Jun 26 16:27:13 hostname vrtd[962]: Design writer transfer starting
Jun 26 16:27:13 hostname vrtd[962]: Transferring design writer payload to device offset 0x102100000 (bytes=8601408 buffer_bytes=8601600 aperture_bytes=65536)
Jun 26 16:27:13 hostname vrtd[962]: Short design writer transfer: Input/output error
Jun 26 16:27:13 hostname vrtd[962]: Failed to transfer design writer payload
Jun 26 16:27:13 hostname vrtd[962]: Design writer transfer failed: Input/output error
Jun 26 16:27:13 hostname vrtd[962]: Design write failed (error=5) for uid=1001 conn_id=4

v80-smi validate

v80-smi validate -d 55:00
Resetting device 0000:55:00...
MM channel: auto (all buffers)
Testing HBM data integrity (8 regions)...
    8/8 OK
Testing HBM read bandwidth (8 threads)...
    Read: 6538.52 MB/s
Testing HBM write bandwidth (8 threads)...
    Write: 13703.49 MB/s
Testing HBM bidirectional bandwidth (16 threads)...
    Read: 6480.90 MB/s
    Write: 6480.90 MB/s
    Total: 12961.80 MB/s
Testing DDR data integrity (8 buffers)...
    8/8 OK
Testing DDR read bandwidth (8 threads)...
    Read: 21234.14 MB/s
Testing DDR write bandwidth (8 threads)...
    Write: 21291.12 MB/s
Testing DDR bidirectional bandwidth (16 threads)...
    Read: 9115.64 MB/s
    Write: 9115.64 MB/s
    Total: 18231.27 MB/s
Testing HBM+DDR read bandwidth (16 threads)...
    Read: 10097.39 MB/s
Testing HBM+DDR write bandwidth (16 threads)...
    Write: 16736.22 MB/s
Testing HBM+DDR bidirectional bandwidth (32 threads)...
    Read: 9137.02 MB/s
    Write: 9137.02 MB/s
    Total: 18274.03 MB/s

[taraxacum45e9a]

I'm facing the same problem.

[quetric]

Hi @arktt thanks for reporting this issue and @taraxacum45e9a for confirming.

To help investigate, could you clarify your test methodology? For each of the failing and passing commits, do you 1) build the slashkit package including shell, 2) build example 0 against this shell, and 3) program the vbin via the example 0 host executable?