CAMBRIA and more than one atheros card

[John Hay]

On Mon, Jul 14, 2014 at 01:02:03PM -0700, Adrian Chadd wrote:
> On 14 July 2014 13:00, John Hay <jhay at meraka.org.za> wrote:
> > On Mon, Jul 14, 2014 at 12:55:19PM -0700, Adrian Chadd wrote:
> >> Right. What's the output of hw.ath ?
> >
> > tst-11-arm:~ # sysctl hw.ath
> > hw.ath.bstuck: 4
> > hw.ath.txbuf_mgmt: 32
> > hw.ath.txbuf: 200
> > hw.ath.rxbuf: 40
> > hw.ath.anical: 100
> > hw.ath.resetcal: 1200
> > hw.ath.shortcal: 100
> > hw.ath.longcal: 30
> > hw.ath.debug: 0
> > hw.ath.hal.debug: 0
> >
> > All 3 interfaces are in adhoc mode if it makes a difference.
> 
> Ah, you're not even using it in 11n mode. Ok.
> 
> Well, that's 72 mbufs allocated, and up to 200 more being allocated
> during transmit. Then there's the handful of pages being used for DMA
> descriptors. So yeah, over time that may become a problem.
> 
> Try doubling the number of bounce buffers available and see what happens.

The first boot looks good, all 3 aths came up. Bounce buffers just
after login:

tst-11-arm:~ # sysctl hw.busdma
hw.busdma.total_bpages: 128
hw.busdma.zone0.total_bpages: 128
hw.busdma.zone0.free_bpages: 128
hw.busdma.zone0.reserved_bpages: 0
hw.busdma.zone0.active_bpages: 0
hw.busdma.zone0.total_bounced: 0
hw.busdma.zone0.total_deferred: 0
hw.busdma.zone0.lowaddr: 0x4000fff
hw.busdma.zone0.alignment: 4096

It stayed like that for about 6-7 minutes, then looked like this:

tst-11-arm:~ # sysctl hw.busdma
hw.busdma.total_bpages: 128
hw.busdma.zone0.total_bpages: 128
hw.busdma.zone0.free_bpages: 126
hw.busdma.zone0.reserved_bpages: 0
hw.busdma.zone0.active_bpages: 2
hw.busdma.zone0.total_bounced: 208
hw.busdma.zone0.total_deferred: 0
hw.busdma.zone0.lowaddr: 0x4000fff
hw.busdma.zone0.alignment: 4096

After 13 minutes:

tst-11-arm:~ # sysctl hw.busdma
hw.busdma.total_bpages: 128
hw.busdma.zone0.total_bpages: 128
hw.busdma.zone0.free_bpages: 124
hw.busdma.zone0.reserved_bpages: 0
hw.busdma.zone0.active_bpages: 4
hw.busdma.zone0.total_bounced: 667
hw.busdma.zone0.total_deferred: 0
hw.busdma.zone0.lowaddr: 0x4000fff
hw.busdma.zone0.alignment: 4096

After 26 minutes:

tst-11-arm:~ # sysctl hw.busdma
hw.busdma.total_bpages: 128
hw.busdma.zone0.total_bpages: 128
hw.busdma.zone0.free_bpages: 121
hw.busdma.zone0.reserved_bpages: 0
hw.busdma.zone0.active_bpages: 7
hw.busdma.zone0.total_bounced: 2433
hw.busdma.zone0.total_deferred: 0
hw.busdma.zone0.lowaddr: 0x4000fff
hw.busdma.zone0.alignment: 4096

Great! At last it seems like we are making progress. I'll leave it
till the morning and then do a few reboots and then some traffic.

One question, should the active bounce pages keep growing or stabilize
somewhere? How do I know if I have to few or if there is a leak?

Thanks Adrian for your help.

John
-- 
John Hay -- jhay at meraka.csir.co.za / jhay at meraka.org.za