[Bug 296201] D50067 ix(4)/ixgbe: Intel E610-XT2 (8086:57b0) hangs in attach_pre (ACI) on amd64; works under Linux ixgbe

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=296201

--- Comment #3 from Giovanni Manzoni <[email protected]> ---
(In reply to Krzysztof Galazka from comment #2)

I swtiched to paid pfSense Plus 26.03 and there, the NIC work as expected. I
also tried with RELEASE 15.1 as your request, thanks for your help.


# pciconf -lv | grep -A1 57b0
ix0@pci0:1:0:0 class 0x02000 rev=0x00 hdr=0x00 vendor=0x8086 device=0x57b0
subvendor=0x8086 subdevice=0x0002 vendor = 'Intel Corporation'
ix1@pci0:1:0:1 class 0x02000 rev=0x00 hdr=0x00 vendor=0x8086 device=0x57b0
subvendor=0x8086 subdevice=0x0002 vendor = 'Intel Corporation'



form pfSense 26.03

[26.03-RELEASE][[email protected]]/root: pciconf -lv | grep -A1 57b0
ix0@pci0:1:0:0: class=0x020000 rev=0x00 hdr=0x00 vendor=0x8086 device=0x57b0
subvendor=0x8086 subdevice=0x0002
    vendor     = 'Intel Corporation'
--
ix1@pci0:1:0:1: class=0x020000 rev=0x00 hdr=0x00 vendor=0x8086 device=0x57b0
subvendor=0x8086 subdevice=0x0002
    vendor     = 'Intel Corporation'



But that doesn't solve the problem. The issue was never compiling a driver—it
was making one that was KBI-compatible with the pfSense 2.8.1 kernel, which is
based on FreeBSD 15.0-CURRENT@bf06074106cf, not 15.1

What Claude.ai said:  The real problem: A kernel module must be KBI-compatible
with the running kernel. The pfSense 2.8.1 kernel is based on FreeBSD
15.0-CURRENT @ bf06074106cf, plus pfSense patches, renumbered to 1500068. To
build an if_ix.ko that loads without the kobj_class_compile1 crash, you need
exactly those source files—and the pfSense release sources (RELENG_2_8_1) are
not publicly available. The E610 driver that works, on the other hand, runs on
15.1, a different base.
So the constraint wasn’t “you can’t compile”; it was: you can’t compile a
KBI-correct module for pfSense 2.8.1 using the E610 source code, because:

you don’t have the source code that matches 2.8.1 (it’s not public), and
the source code that includes the E610 driver (15.1) has a different KBI →
causing a crash on load.

Translated with DeepL.com (free version)

-- 
You are receiving this mail because:
You are the assignee for the bug.