Ethernet – Intel e1000e Still Broken with 5.4-rc6
Just tested Kernel 5.4-rc6 on a system with Intel e1000e, and this is still broken, despite being reported when -rc1 was released.. http://lkml.iu.edu/hypermail/linux/kernel/1910.3/07624.html
Kernel release notes for 5.4-rc6 are here: http://lkml.iu.edu/hypermail/linux/kernel/1911.0/01280.html
The problem, is that the Ethernet connection repeatedly activates/deactivates, making the connection useless:
(From KDE/Plasma)……………
Connection 'eth0' deactivated.
Connection 'eth0' activated.
Connection 'eth0' deactivated.
Connection 'eth0' activated.
Connection 'eth0' activated.
…… repeated ….
The fix is to revert the patch here: https://patchwork.ozlabs.org/patch/1131961/
Robert Gadsdon. November 4th, 2019.
Then kernel sources (even the last one 5.4-rc6) still contain an old version (3.2.6) of the e1000e driver.
Compiling the current version (3.6.0) of the module from intel sources solved our problems here. (Dell Optiplex 7460 AIO, Ethernet Intel 15bb rev 10)
https://downloadcenter.intel.com/download/15817
Thanks for the info..
| ^~~~~~~~~~~~~~~~~~
I tried this with kernel 5.4-rc6, but the compile fails:
# make
make[1]: Entering directory ‘/usr/src/linux-5.4-rc6’
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/netdev.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/ethtool.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/ich8lan.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/mac.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/nvm.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/manage.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/phy.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/82571.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/80003es2lan.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/param.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/ptp.o
CC [M] /home/rgadsdon/kernel/e1000e-3.6.0/src/kcompat.o
In file included from /home/rgadsdon/kernel/e1000e-3.6.0/src/e1000.h:14,
from /home/rgadsdon/kernel/e1000e-3.6.0/src/ich8lan.c:41:
/home/rgadsdon/kernel/e1000e-3.6.0/src/kcompat.h:2799:10: fatal error: linux/pci-aspm.h: No such file or directory
2799 | #include
compilation terminated.
– and it seems that pci-aspm.h no longer exists, in the 5.4-rc tree..
RG.
Yes and i am also having problems with this new e1000e-3.6.0 driver on linux-5.4.2 with the missing linux/pci-aspm.h in the kernel source.
like this:
cd /home/Dave/Downloads/e1000e-3.6.0/src
bash-5.0# make install
make[1]: Entering directory ‘/mnt/disk/Downloads/k/linux-5.4.2’
CC [M] /mnt/disk/Downloads/e1000e-3.6.0/src/netdev.o
In file included from /mnt/disk/Downloads/e1000e-3.6.0/src/e1000.h:14,
from /mnt/disk/Downloads/e1000e-3.6.0/src/netdev.c:30:
/mnt/disk/Downloads/e1000e-3.6.0/src/kcompat.h:2799:10: fatal error: linux/pci-aspm.h: No such file or directory
2799 | #include
| ^~~~~~~~~~~~~~~~~~
compilation terminated.
make[2]: *** [scripts/Makefile.build:266: /mnt/disk/Downloads/e1000e-3.6.0/src/netdev.o] Error 1
make[1]: *** [Makefile:1652: /mnt/disk/Downloads/e1000e-3.6.0/src] Error 2
make[1]: Leaving directory ‘/mnt/disk/Downloads/k/linux-5.4.2’
make: *** [Makefile:73: default] Error 2
A bit of a mess to be sure. My system is an asus rog strix z390-e gaming running slackware-current multilib 64 bit. I get a pregnant 2 minute pause on boot up then
my system runs ok. It was only when i tried to update the kernel that i obtained this hang on boot and then tried to update of the e1000e driver to see if it would help my system as my system seems to hang around the time my nic initializes. I also have an msi apache pro laptop running the same
software that does not have the same problem (Also doesn’t have an e1000e driver needed)
I am sure there is a work around (eventually) so it is not too big of a deal i just don’t like the long boot up time.
Perhaps next i will tell intel. It is their driver after all.
just comment the include as the defs within it seems to have been moved to pci.h (as far as i read elsewhere); compilation worked for me that way and connection started again to be stable!
I’ve got the same issue with 5.4.0, but for me it’s not stable on a Lenovo ThinkPad P52 using driver version 3.6.0
Same problem here as well on 5.4.0, using the kernel.org version of e1000e; 3.2.6-k. I’ve not tried to compile the upstream 3.6.0 yet.