Hi Martin !
That is strange - I didn't expect that, especially as the previous code
was wrong with respect to state tracking.
Can you check whether the addresses do not have the DEtACHED flag?
You could try the dtrace script from the PR - it shows a little bit what
is going on.
There was also some mii fixing just before my patch, but I assume you
testes with reverting just
this commit.
Sorry for the issue - I have yet to find an explaination for that behavior.
Maybe comparing the dtrace outputs for both varainst can shed a light on
what happens.
Frank
On 08/27/20 10:20, Martin Husemann wrote:
On Mon, Aug 24, 2020 at 12:46:04PM +0000, Frank Kardel wrote:
Module Name: src
Committed By: kardel
Date: Mon Aug 24 12:46:04 UTC 2020
Modified Files:
src/sys/dev/mii: mii_physubr.c
Log Message:
Keep the change check invariant intact. The previous code could miss
status updates by picking up a new status different from the tested
status. This left addresses in the DETACHED state although the
link status is already UP again.
addresses PR/kern 55538
To generate a diff of this commit:
cvs rdiff -u -r1.92 -r1.93 src/sys/dev/mii/mii_physubr.c
Hi Frank,
this change breaks the network on my macppc, with r1.93 it only seems to be
able to send packets, but never receives answers (ARP does not complete,
but other hosts see the ARP requests).
gem0 at pci2 dev 15 function 0: Apple Computer GMAC Ethernet (rev. 0x01)
gem0: interrupting at irq 41
brgphy0 at gem0 phy 0: BCM5411 1000BASE-T media interface, rev. 1
brgphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT,
1000baseT-FDX, auto
gem0: Ethernet address 00:03:93:71:ff:cc, 10KB RX fifo, 4KB TX fifo
It is connected to a gige switch:
media: Ethernet autoselect (1000baseT full-duplex,master)
status: active
(which looks the same in the non-working kernel).
Any ideas how to debug?
Martin