Re: Call for testers: Atheros AR8121(L1E)/AR8113/AR8114(L2E) ethernet

Pyun YongHyeon <pyunyh_at_gmail.com> writes:
> On Tue, Nov 04, 2008 at 10:36:21PM +0100, Jan Henrik Sylvester wrote:
>  > Pyun YongHyeon wrote:
>  > >On Mon, Nov 03, 2008 at 07:49:34PM +0100, Jan Henrik Sylvester wrote:
>  > > > >Pyun YongHyeon <pyunyh at gmail.com> writes:
>  > > > >> http://people.freebsd.org/~yongari/ale/if_ale.c
>  > > > 
>  > > > I got an Eee PC 1000H, too, with 7.1-BETA2 and just used scp to copy 
>  > > > with 11.5MB/s -- much better than wlan for big files. Great!
>  > > > 
>  > >
>  > >Thanks for testing!
>  > 
>  > I was happy too early. Now I keep getting these:
>  > ale0: DMA read error! -- resetting
>  > ale0: could not disable Tx/Rx MAC(0x00000008)!
>
> It looks like Tx queue processing is still in progress.
> Would you show me hardware MAC statistics?
> ("sysctl dev.ale.0.stats" will show you detailed information.)
> Also ale(4) related dmesg output might be useful.
>
>  > ale0: interrupt moderation is 100 us.
>  > 
>  > The speed of copying a file has dropped to <50kB/s. Stopping and 
>  > restarting the interface does not help, even if I kldunload the driver 
>  > in the meantime. I have yet to reboot the machine.
>  > 
>  > I do not know what is different, though. I left the computer alone for 
>  > an hour before using the network...
>  > 
>
> Hmm, this is the first severe report that is related to stability
> of driver. Personally I've nerver met this kind of error during
> writing of ale(4). ATM I have no idea why DMA read error happens on
> your box though. As you guess I don't have access to datasheet.
>
> Is there any reliable way to reproduce it?

I tried to reproduce under heavy net load but failed. I created a tmpfs
500M disk and fetched big files (i.e. openoffice.org3 distro as of 300M)
to /tmp (at 11-12MB/s rate). No traces of failure. Here is some info:
-----
# ifconfig ale0
ale0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
	options=3199<RXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,TSO4,WOL_MCAST,WOL_MAGIC>
	ether 00:22:15:XX:XX:XX
	inet 192.168.233.80 netmask 0xffffffc0 broadcast 192.168.233.127
	media: Ethernet autoselect (100baseTX <full-duplex>)
	status: active
# sysctl dev.ale.0.stats
dev.ale.0.stats.rx.good_frames: 779733
dev.ale.0.stats.rx.good_bcast_frames: 232
dev.ale.0.stats.rx.good_mcast_frames: 0
dev.ale.0.stats.rx.pause_frames: 0
dev.ale.0.stats.rx.control_frames: 0
dev.ale.0.stats.rx.crc_errs: 0
dev.ale.0.stats.rx.len_errs: 0
dev.ale.0.stats.rx.good_octets: 1115319790
dev.ale.0.stats.rx.good_bcast_octets: 20376
dev.ale.0.stats.rx.good_mcast_octets: 0
dev.ale.0.stats.rx.runts: 0
dev.ale.0.stats.rx.fragments: 0
dev.ale.0.stats.rx.frames_64: 370
dev.ale.0.stats.rx.frames_65_127: 159
dev.ale.0.stats.rx.frames_128_255: 108
dev.ale.0.stats.rx.frames_256_511: 124
dev.ale.0.stats.rx.frames_512_1023: 58
dev.ale.0.stats.rx.frames_1024_1518: 779088
dev.ale.0.stats.rx.frames_1519_max: 0
dev.ale.0.stats.rx.trunc_errs: 0
dev.ale.0.stats.rx.fifo_oflows: 0
dev.ale.0.stats.rx.desc_oflows: 0
dev.ale.0.stats.rx.align_errs: 0
dev.ale.0.stats.rx.filtered: 174
dev.ale.0.stats.tx.good_frames: 516845
dev.ale.0.stats.tx.good_bcast_frames: 4
dev.ale.0.stats.tx.good_mcast_frames: 0
dev.ale.0.stats.tx.pause_frames: 0
dev.ale.0.stats.tx.control_frames: 0
dev.ale.0.stats.tx.excess_defers: 0
dev.ale.0.stats.tx.defers: 0
dev.ale.0.stats.tx.good_octets: 34113213
dev.ale.0.stats.tx.good_bcast_octets: 0
dev.ale.0.stats.tx.good_mcast_octets: 0
dev.ale.0.stats.tx.frames_64: 5
dev.ale.0.stats.tx.frames_65_127: 516840
dev.ale.0.stats.tx.frames_128_255: 0
dev.ale.0.stats.tx.frames_256_511: 0
dev.ale.0.stats.tx.frames_512_1023: 0
dev.ale.0.stats.tx.frames_1024_1518: 0
dev.ale.0.stats.tx.frames_1519_max: 0
dev.ale.0.stats.tx.single_colls: 0
dev.ale.0.stats.tx.multi_colls: 0
dev.ale.0.stats.tx.late_colls: 0
dev.ale.0.stats.tx.excess_colls: 0
dev.ale.0.stats.tx.abort: 0
dev.ale.0.stats.tx.underruns: 0
dev.ale.0.stats.tx.desc_underruns: 0
dev.ale.0.stats.tx.len_errs: 0
dev.ale.0.stats.tx.trunc_errs: 240
-----

> Also I've slightly changed code which may help to stop MAC in case
> of critical condition. Would you get the latest if_ale.c at the
> following URL and try again?
> http://people.freebsd.org/~yongari/ale/if_ale.c

I didn't use this file version though.

>  > > > This will probably not go into 7.1-RELEASE, will it?
>  > >
>  > >It was too late to go into 7.1-RELEASE, I guess.
>  > 
>  > It seems to be too early, anyhow.
>  > 
>
> That's one of reason why I need more wider testing. :-)
>
>  > Do you want me to try anything specific? (I hope it does not involve 
>  > recompiling the kernel on this slow Eee PC.)
>  > 
>
> Since you're using kernel module, just unload ale(4) and
> rebuilding ale(4) is enough. For example,
> #cd /path/to/ale_directory/
> #fetch http://people.freebsd.org/~yongari/ale/if_ale.c
> #ifconfig ale0 down
> #kldunload if_ale
> #make && make install
> #kldload if_ale
> #ifconfig ale0 inet x.x.x.x netmask y.y.y.y
> or
> #dhclinet ale0 if you use DHCP.

WBR
-- 
Boris Samorodov (bsam)
Research Engineer, http://www.ipt.ru Telephone & Internet SP
FreeBSD committer, http://www.FreeBSD.org The Power To Serve