Hi all,
I have a Realtek physical net interface in a laptop that I used as a
home server:
re0 at pci2 dev 0 function 1 "Realtek 8168" rev 0x12: RTL8411B
(0x5c80), msi, address 08:97:98:6e:16:eb
rgephy0 at re0 phy 7: RTL8251 PHY, rev. 0
I have been plagued by by the issue:
re0: watchdog timeout
which sometimes is recovered from and other time does not.
Digging around a bit on Google it seems that an often suggestion is
just to replace the network card, but in my case it is integrated into
the laptop so that is not an option.
I have no idea what I am doing but started playing around with the
kernel sys/dev/ic/re.c code and made a change that makes the watchdog
timeout go away, even after 100's of Gigs of transfers.
Index: sys/dev/ic/re.c
===================================================================
RCS file: /cvs/src/sys/dev/ic/re.c,v
retrieving revision 1.211
diff -u -p -r1.211 re.c
--- sys/dev/ic/re.c 17 May 2021 11:59:53 -0000 1.211
+++ sys/dev/ic/re.c 12 Apr 2022 23:45:05 -0000
@@ -1021,6 +1021,7 @@ re_attach(struct rl_softc *sc, const cha
case RL_HWREV_8168C:
case RL_HWREV_8168C_SPIN2:
case RL_HWREV_8168CP:
+ case RL_HWREV_8411B:
break;
default:
ifp->if_capabilities |= IFCAP_CSUM_IPv4;
Some questions for those that know what they are doing:
1. Should the RL_HWREV_8411B be added to this exception list? Are
there consequences that I don't understand?
2. If so, then how could I test it in order to help?
3. If the removal of HW checksumming causes issues with the stability
of the network device does that imply that checksumming itself needs to
be investigated? My thought is that if the hardware checksum is
disabled then the software backup should be stable enough to handle
doing so...??
Thanks,
Justin
