Network stopped working on haiku nightly hrev50967

[addos]

I did a bisection of tests with the nightlies to track down which nightly I last had working internet on. The last build was hrev50779. After that, it looks like dhcp assigns an ip to the interface ipro1000, but I can't ping anything and websites don't load in webpositive. I can't connect to irc in vision. Need assistance tracking down what changed.

[diver]

So the last working build was hrev50779? What was the next you tried which was broken?

[addos]

hrev50779 was last build that worked hrev50967 was first nightly that networking stopped working.

[diver]

That's 188 commits: ​http://cgit.haiku-os.org/haiku/log/?qt=range&q=hrev50967+%5ehrev50779&ofs=250

[addos]

Those were the only two builds I could download from the haiku nightly site. Are there any intermediate builds available somewhere I could try, to try to narrow it down further?

​https://download.haiku-os.org/nightly-images/x86_gcc2_hybrid/

[humdinger]

Thanks to package management, you can easily ​downgrade to any hrev.

[addos]

So using the pkgman system, it looks like the build where it last worked was 50812 and it broke in 50813, assuming I did everything right.

[korli]

these changes were added in hrev50813:

​http://cgit.haiku-os.org/haiku/commit/?id=e32c64c113053d4a3e4cfdb50a8ce64b8c1a60eb

​http://cgit.haiku-os.org/haiku/commit/?id=6e02aa9d1ce55ac600f75ac0ba895b2bdfd2d23b

​http://cgit.haiku-os.org/haiku/commit/?id=dfe2eac4c6555119c3590f90c51dd1636a483699

​http://cgit.haiku-os.org/haiku/commit/?id=80d8697c0ec83093b2552c90a9c947189245f51f

Please add your syslog.

[addos]

​http://cgit.haiku-os.org/haiku/commit/?id=e32c64c113053d4a3e4cfdb50a8ce64b8c1a60eb looks suspicious since it impacts the network card I am using

[diver]

Odd, that commit simply added new PCI IDs and didn't change anything.

[korli]

addos, the syslog is incomplete.

[korli]

Replying to diver:

Odd, that commit simply added new PCI IDs and didn't change anything.

At least the part with driver.c should be reverted as it affects the old driver beos_ipro1000.

Seeing the typos in driver.c, it doesn't compile anyway (";" instead of ":").

[diver]

Ah, so the old ipro1000 is used for these cards now? Nice catch!

[korli]

grep -R E1000_DEV_ID_82541GI_LF
dev/e1000/if_em.c:	{ 0x8086, E1000_DEV_ID_82541GI_LF,	PCI_ANY_ID, PCI_ANY_ID, 0},
dev/e1000/e1000_hw.h:#define E1000_DEV_ID_82541GI_LF			0x107C
dev/e1000/if_lem.c:	{ 0x8086, E1000_DEV_ID_82541GI_LF,	PCI_ANY_ID, PCI_ANY_ID, 0},
dev/e1000/e1000_api.c:	case E1000_DEV_ID_82541GI_LF:

Your device ID is now declared for both driver if_em.c and if_lem.c. To compare with FreeBSD: ​http://fxr.watson.org/fxr/ident?v=FREEBSD9;i=E1000_DEV_ID_82541GI_LF

Defined as a preprocessor macro in:

    dev/e1000/e1000_hw.h, line 75 

Referenced (in 3 files total) in:

    dev/e1000/e1000_hw.h, line 75
    dev/e1000/if_lem.c, line 115
    dev/e1000/e1000_api.c, line 207 

Obviously the addition of IDs is the problem.

[korli]

So let's try what's common:

egrep  "(E1000_DEV_ID_[^,]+)," if_em.c  | sed -e "s/^.*8086, //g" | sed -e "s/,.*//g" | sort > em.ids
egrep  "(E1000_DEV_ID_[^,]+)," if_lem.c  | sed -e "s/^.*8086, //g" | sed -e "s/,.*//g" | sort > lem.ids
comm -12 em.ids lem.ids

Result:

E1000_DEV_ID_82540EM
E1000_DEV_ID_82540EM_LOM
E1000_DEV_ID_82540EP
E1000_DEV_ID_82540EP_LOM
E1000_DEV_ID_82540EP_LP
E1000_DEV_ID_82541EI
E1000_DEV_ID_82541EI_MOBILE
E1000_DEV_ID_82541ER
E1000_DEV_ID_82541ER_LOM
E1000_DEV_ID_82541GI
E1000_DEV_ID_82541GI_LF
E1000_DEV_ID_82541GI_MOBILE
E1000_DEV_ID_82542
E1000_DEV_ID_82543GC_COPPER
E1000_DEV_ID_82543GC_FIBER
E1000_DEV_ID_82544EI_COPPER
E1000_DEV_ID_82544EI_FIBER
E1000_DEV_ID_82544GC_COPPER
E1000_DEV_ID_82544GC_LOM
E1000_DEV_ID_82545EM_COPPER
E1000_DEV_ID_82545EM_FIBER
E1000_DEV_ID_82545GM_COPPER
E1000_DEV_ID_82545GM_FIBER
E1000_DEV_ID_82545GM_SERDES
E1000_DEV_ID_82546EB_COPPER
E1000_DEV_ID_82546EB_FIBER
E1000_DEV_ID_82546EB_QUAD_COPPER
E1000_DEV_ID_82546GB_COPPER
E1000_DEV_ID_82546GB_FIBER
E1000_DEV_ID_82546GB_PCIE
E1000_DEV_ID_82546GB_QUAD_COPPER
E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3
E1000_DEV_ID_82546GB_SERDES
E1000_DEV_ID_82547EI
E1000_DEV_ID_82547EI_MOBILE
E1000_DEV_ID_82547GI

[korli]

Clean up done in hrev51417.

[addos]

Hey, wanted to confirm that I booted the latest nightly and networking is working as expected.