FreeBSD 6.5-prerelease and if_re - patches needed?

Pyun YongHyeon pyunyh at gmail.com
Sun Nov 2 16:49:20 PST 2008


On Sun, Nov 02, 2008 at 12:29:06PM +0100, Torfinn Ingolfsen wrote:
 > On Sun, 02 Nov 2008 14:09:15 +0900
 > Pyun YongHyeon <pyunyh at gmail.com> wrote:
 > 
 > > http://people.freebsd.org/~yongari/re/6.x/README
 > > 
 > > Hope this helps.
 > 
 > Yes it does, thanks!
 > 
 > On boot, trherer is a noticable delay (tens of seconds) after printing
 > these lines:
 > re0: <RealTek 8168/8168B/8168C/8168CP/8111B/8111C/8111CP PCIe Gigabit Ethernet> port 0xee00-0xeeff mem 0xfdfff000-0xfdffffff irq 19 at device 0.0 on pci2
 > re0: turning off MSI enable bit.
 > re0: Chip rev. 0x38000000
 > re0: MAC rev. 0x00000000
 > 
 > The original didn't have that delay.

I've changed to have re(4) wait the completion of DMAable memory
allocation during bus_dma cleanups. The delay you've seen may be
related with that change. Previously it just failed to load the
driver if there is no available memory at the time of driver
loading. However I guess that delay wouldn't happen if the driver
is statically linked into kernel.
Did you use kernel module?

In theory PCIe variants of RealTek controllers would work with DAC
so I could alleviate memory allocation restrictions imposed by
bus_dma by allowing 64bits DMA addressing. Since I don't have PCIe
based RealTek controllers and no datasheets are available for
PCIe based controllers it's somewaht difficult to chage current
allocation restrictions.

 > Otrher than that it works much better. Details:
 > root at kg-vm# uname -a
 > FreeBSD kg-vm.kg4.no 6.4-PRERELEASE FreeBSD 6.4-PRERELEASE #3: Sun Nov  2 10:44:32 CET 2008     root at kg-vm.kg4.no:/usr/obj/usr/src/sys/SMP  amd64
 > root at kg-vm# pciconf -lv | grep re0 -A 4
 > re0 at pci2:0:0:	class=0x020000 card=0x81aa1043 chip=0x816810ec rev=0x01 hdr=0x00
 >     vendor     = 'Realtek Semiconductor'
 >     device     = 'RTL8168/8111 PCI-E Gigabit Ethernet NIC'
 >     class      = network
 >     subclass   = ethernet
 > 
 > Note: I haven't testet if_rl, so I don't know how this patch affects that.

rl(4) has a single change to build with updated if_rlreg.h and I
don't think that would affect any stability of rl(4).

 > I'll get back with a note on stability sometime next week (after I have done losts of data transfers to and from this box).

Ok.

-- 
Regards,
Pyun YongHyeon


More information about the freebsd-stable mailing list