<div style="font-family: Arial, sans-serif; font-size: 14px;"><span>I can confirm this bug on independent hardware, and I have some additional</span><div><span>analysis suggesting it is systematic rather than configuration-specific.</span></div><div><br></div><div><span>Reproduction</span></div><div><span>------------</span></div><div><span>Same failure, different machine:</span></div><div><br></div><div><span> nvidia-open-kernel-dkms 550.163.01-2</span></div><div><span> dkms 3.2.2-1~deb13u1</span></div><div><span> gcc Debian 14.2.0-19 (14.2.0)</span></div><div><span> Debian GNU/Linux 13 (trixie)</span></div><div><span> GPU: NVIDIA GA106M [GeForce RTX 3060 Mobile / Max-Q] [10de:2520]</span></div><div><span> (hybrid/Optimus alongside Intel Alder Lake-P Iris Xe)</span></div><div><br></div><div><span>Build against 6.12.100+deb13-amd64 fails at the same location:</span></div><div><br></div><div><span> nvidia/nv-pci.c:237:9: error: too few arguments to function</span></div><div><span> 'pci_resize_resource'</span></div><div><br></div><div><span>Building the same source against 6.12.96+deb13-amd64 succeeds.</span></div><div><br></div><div><span>Root cause</span></div><div><span>----------</span></div><div><span>The prototype gained a fourth parameter between the two kernels.</span></div><div><br></div><div><span>6.12.96, /usr/src/linux-headers-6.12.96+deb13-common/include/linux/pci.h:1421</span></div><div><br></div><div><span> int __must_check pci_resize_resource(struct pci_dev *dev, int i, int size);</span></div><div><br></div><div><span>6.12.100, /usr/src/linux-headers-6.12.100+deb13-common/include/linux/pci.h:1421</span></div><div><br></div><div><span> int __must_check pci_resize_resource(struct pci_dev *dev, int i, int size,</span></div><div><span> int exclude_bars);</span></div><div><br></div><div><span>Why this is not configuration-dependent</span></div><div><span>---------------------------------------</span></div><div><span>This is the part I think is worth adding to the report: the driver has no</span></div><div><span>compatibility shim for this symbol at all.</span></div><div><br></div><div><span>1. pci_resize_resource appears exactly once in the unpacked source tree:</span></div><div><br></div><div><span> $ grep -rn pci_resize_resource . --include='*.c' --include='*.h'</span></div><div><span> ./nvidia/nv-pci.c:237: r = pci_resize_resource(pci_dev, NV_GPU_BAR1, requested_size);</span></div><div><br></div><div><span>2. conftest.sh does not probe it:</span></div><div><br></div><div><span> $ grep -c pci_resize_resource conftest.sh</span></div><div><span> 0</span></div><div><br></div><div><span> ...while it does probe many other PCI symbols (pci_bus_address,</span></div><div><span> pci_channel_state, pci_enable_msix_range, pci_get_domain_bus_and_slot,</span></div><div><span> pci_driver_has_driver_managed_dma, and others), which is the mechanism</span></div><div><span> the driver normally uses to absorb this kind of API change.</span></div><div><br></div><div><span>3. The call site is not inside any preprocessor conditional. The nearest</span></div><div><span> #endif above it closes an unrelated</span></div><div><span> NV_IS_EXPORT_SYMBOL_PRESENT_pci_find_host_bridge block.</span></div><div><br></div><div><span>So the call is unconditional and there is no code path that can adapt to the</span></div><div><span>new signature. Any build of 550.163.01 against a kernel carrying the</span></div><div><span>4-argument prototype will fail identically, regardless of GPU model, Optimus</span></div><div><span>versus discrete-only, or local configuration.</span></div><div><br></div><div><span>Scope across suites</span></div><div><span>-------------------</span></div><div><span>No currently packaged revision fixes this:</span></div><div><br></div><div><span> bookworm 535.261.03-1</span></div><div><span> trixie 550.163.01-2</span></div><div><span> trixie-backports 550.163.01-4~bpo13+1</span></div><div><span> sid 550.163.01-4</span></div><div><span> experimental 555.58.02-2</span></div><div><br></div><div><span>trixie-backports is worth calling out explicitly, since it is the natural</span></div><div><span>first thing an affected user will reach for. It is the same upstream</span></div><div><span>550.163.01, and the Debian revisions since -2 do not touch this code path:</span></div><div><span>per the changelog, -3 backported the NV_DMA_BUF_HAS_DYNAMIC_ATTACHMENT</span></div><div><span>changes to fix the open kernel module build for Linux 6.16, and -4 was a</span></div><div><span>sync with src:nvidia-graphics-drivers. Installing the backport rebuilds the</span></div><div><span>same source and fails the same way.</span></div><div><br></div><div><span>Impact</span></div><div><span>------</span></div><div><span>The DKMS failure makes the linux-image postinst exit non-zero, which leaves</span></div><div><span>the kernel packages unconfigured and blocks dpkg:</span></div><div><br></div><div><span> iF linux-image-6.12.100+deb13-amd64</span></div><div><span> iU linux-image-amd64</span></div><div><span> iU linux-headers-6.12.100+deb13-amd64</span></div><div><span> iU linux-headers-amd64</span></div><div><br></div><div><span>Users who reboot into 6.12.100 before noticing will come up with no NVIDIA</span></div><div><span>driver. On this machine the other DKMS modules (acpi-call, kbdlight, xtrx)</span></div><div><span>all built successfully against 6.12.100; nvidia-current-open was the only</span></div><div><span>failure.</span></div><div><br></div><div><span>I have not checked whether nvidia-kernel-dkms (non-free) shares this call</span></div><div><span>site and is affected in the same way. It is not installed here, so I cannot</span></div><div><span>confirm either way.</span></div><div><br></div><div><span>Workaround</span></div><div><span>----------</span></div><div><span>Staying on 6.12.96+deb13-amd64, with linux-image-amd64 and</span></div><span>linux-headers-amd64 held, until a fixed driver is available.</span><br></div>
<div style="font-family: Arial, sans-serif; font-size: 14px;" class="protonmail_signature_block protonmail_signature_block-empty">
<div class="protonmail_signature_block-user protonmail_signature_block-empty"></div>
<div class="protonmail_signature_block-proton protonmail_signature_block-empty">
</div>
</div>