https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=166724
Artem Viklenko <ar...@viklenko.net> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ar...@viklenko.net
--- Comment #41 from Artem Viklenko <ar...@viklenko.net> ---
My story.. also had this issue.
Two my home routers run 11.3-RELEASE-p5 amd64 and have realtek nics like this
one:
re0 pnpinfo vendor=0x10ec device=0x8168 subvendor=0x1458
subdevice=0xe000 class=0x020000 at slot=0 function=0 dbsf=pci0:1:0:0
handle=\_SB_.PCI0.RP01.PXSX
Interrupt request lines:
258
pcib1 I/O port window:
0xe000-0xe0ff
pcib1 memory window:
0xd0700000-0xd0703fff
0xd0704000-0xd0704fff
I switched to Realtek's driver. But still have wtchdog timeouts. After some
Googling I found duscussion about issues with jumbo buffers. After cheking this
idea I found confirmation - after some time (depending on traffic rate/amount)
memory became fragmented and requests to 9k buffers fails.
Now I use 1.95 driver from vendor but with very-very dirty hack. I've replaced
Jumbo_Frame_9k with value 3072. So now re driver use only 4k buffers.
I'm ok with MTU of 1500 (this change limits max MTU). But now it is stable and
no watchdog timeouts.
And no more failures on buffers:
artem@gate$ vmstat -z | grep buf
mbuf_packet: 256, 2362080, 2, 1265, 9732282, 0, 0
mbuf: 256, 2362080, 514, 1789,17068586803, 0, 0
mbuf_cluster: 2048, 369076, 1267, 21, 531078, 0, 0
mbuf_jumbo_page: 4096, 184537, 513, 263,7618134369, 0, 0
mbuf_jumbo_9k: 9216, 54677, 0, 0, 0, 0, 0
mbuf_jumbo_16k: 16384, 30756, 0, 0, 0, 0, 0
I know it is stupid trick but at least it works. :)
Hope it can help.
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
freebsd-net@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-net
To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
