On Sun, Feb 08, 2026 at 12:18:40PM +0100, Matthieu Herrb wrote:
> On Sun, Feb 08, 2026 at 10:40:53AM +0100, Matthieu Herrb wrote:
> > Hi,
> >
> > while upgrading my X40 from 7.8 to -current, I found that the kernel
> > from current snapshots panic on boot. here's an AI powered transcript
> > of the screen shot (I did proof read it, it looks like it is correct):
> >
> > 0x1014), c1(@1 halt()), PSS
> > acpipwrres0 at acpi0: PUBS, resource for USB0, USB1, USB7
> > acpitz0 at acpi0
> > acpitz0: critical temperature is 95 degC
> > panic: pci_make_tag: bad request
> > Stopped at db_enter+0x4: popl %ebp
> > TID PID UID PRFLAGS PFLAGS CPU COMMAND
> > * 0 0 0 0x10000 0x200 0 swapper
> > db_enter(d1264a7c,d0f10565,1,2,d1264a80) at db_enter+0x4
> > panic(d0d1a60b,0,d1264b10,d0949fb3,0) at panic+0x76
> > pci_make_tag(0,c7,84,d0a0) at pci_make_tag+0x93
> > aml_rdpciaddr(d3dd9704,d1264b20) at aml_rdpciaddr+0x173
> > acpi_found(d3ddc744,d3d94800,d0a0d9b0,0,d3ddc744) at acpi_found+0xb7
> > aml_walknodes(d3ddc744,0,d0a0d9b0,d3d94800,d0a0d9b0) at aml_walknodes+0x65
> > aml_walknodes(d3ddc784,0,d0a0d9b0,d3d94800,d0a0d9b0) at aml_walknodes+0x39
> > aml_walknodes(d3dd9704,0,d0a0d9b0,d3d94800,d0a0d9b0) at aml_walknodes+0x39
> > aml_walknodes(d3db8584,0,d0a0d9b0,d3d94800,d0a0d9b0) at aml_walknodes+0x39
> > aml_walknodes(d3d9c504,0,d0a0d9b0,d3d94800,0) at aml_walknodes+0x39
> > aml_walknodes(d103c23c,0,d0a0d9b0,d3d94800) at aml_walknodes+0x39
> > acpi_attach_common(d3d94800,f6f90) at acpi_attach_common+0x663
> > acpi_attach(d3d9c000,d3d94800,d1264dd8) at acpi_attach+0x2c
> > config_attach(d3d9c000,d0f200b4,d1264dd8,d07e40c0) at config_attach+0x188
> > https://www.openbsd.org/ddb.html describes the minimum info required in bug
> > reports. Insufficient info makes it difficult to find and fix bugs.
> > ddb>
> >
>
> Reverting the following commit makes the machine boot fine.
>
> commit 0c90e2b526b26d9f50dde7f7712cec107cda3326
> Author: jcs <[email protected]>
> Date: Sun Nov 23 19:56:24 2025 +0000
>
> Make aml_rdpciaddr work better through PCI bridges
>
> When looking up the address of a PCI device from which to read for
> an opregion, walk from the root inward and each time a PCI bridge is
> encountered, read its bus info register and note its bus id.
>
> Previously we were walking outward from the device and only changing
> buses if there was a _BBN node which is not always the case.
>
> This should fix booting of late Intel Macs which have AML that tries
> to read the PCI class of the SSD after powering up the PCI port in
> its _PS0 method, waiting in a loop for it to read 1 (Mass Storage).
> Since we were previously reading the PCI class of the bridge instead
> and always getting 6, this loop would have to timeout which took
> many minutes. Now we are properly reading from the device itself
> which is on a different bus, and the PCI class reads correctly.
>
> feedback from miod
> ok kettenis
the initial paths don't have _ADR and aren't pci devices
aml_nodename on the nodes that take the path in the diff shows
\\ and \\_SB_
for example:
path[0] = \\_SB_.PCI0.IDE0
path[1] = \\_SB_.PCI0
path[2] = \\_SB_
path[3] = \\
where n starts at 3
Index: sys/dev/acpi/dsdt.c
===================================================================
RCS file: /cvs/src/sys/dev/acpi/dsdt.c,v
diff -u -p -r1.278 dsdt.c
--- sys/dev/acpi/dsdt.c 19 Jan 2026 20:37:46 -0000 1.278
+++ sys/dev/acpi/dsdt.c 9 Feb 2026 01:06:55 -0000
@@ -2387,7 +2387,8 @@ aml_rdpciaddr(struct aml_node *pcidev, u
addr->dev = 0;
addr->fun = 0;
}
- }
+ } else
+ continue;
if (n == 0)
break;
