DWC OTG TX path optimisation for 11-current
Svatopluk Kraus
onwahe at gmail.com
Tue Aug 18 11:22:29 UTC 2015
On Sat, Aug 15, 2015 at 2:33 PM, Hans Petter Selasky <hps at selasky.org> wrote:
> On 08/09/15 05:09, Andreas Schwarz wrote:
>>
>> On 31.07.15, Hans Petter Selasky wrote:
>>
>>> The attached patch will add some more debug prints so that I know what
>>> is going on in the problematic function. You'll need to C&P me the
>>> resulting console prints.
>>
>>
>> I've applied the patch to r286117 and started to build the world in a
>> loop until the problem occur. See the output below. After the problem
>> was there (last lines), no further debug output was generated by your
>> patch.
>>
>
> Hi,
>
> Can you re-test the latest 11-current after:
>
> https://svnweb.freebsd.org/changeset/base/286802
>
I have re-tested it. I noticed that system behaved a bit differently.
Almost from the beginning of buildworld, disk was very busy and
interrupt rate was not so stable - it was like when the trigger is
pulled. However, it did not fix the problem.
After buildword was finished, I have swap off, network down, system
99% idle, only the following processes exist
root at rpi2:~ # ps -ax
PID TT STAT TIME COMMAND
0 - DLs 0:00.87 [kernel]
1 - ILs 0:06.51 /sbin/init --
2 - RL 0:41.99 [rand_harvestq]
3 - DL 0:00.00 [crypto]
4 - DL 0:00.00 [crypto returns]
5 - DL 0:28.66 [cam]
6 - IL 0:00.00 [VCHIQ-0]
7 - IL 0:00.00 [VCHIQr-0]
8 - IL 0:00.00 [VCHIQs-0]
9 - DL 0:00.00 [sctp_iterator]
10 - RL 4774:38.93 [idle]
11 - WL 27:37.49 [intr]
12 - DL 0:35.24 [geom]
13 - DL 29:50.89 [usb]
14 - DL 0:00.01 [mmcsd0: mmc/sd card]
15 - IL 0:00.00 [VCHIQka-0]
16 - IL 0:00.00 [bcm2835_audio_worke]
17 - DL 0:46.60 [pagedaemon]
18 - DL 0:03.80 [vmdaemon]
19 - DL 0:00.00 [pagezero]
20 - DL 0:07.70 [bufdaemon]
21 - DL 2:17.08 [syncer]
22 - RL 0:02.01 [vnlru]
369 - Is 0:00.26 /sbin/devd
683 u0 IWs 0:00.00 login [pam] (login)
684 u0 R 0:00.40 -csh (csh)
22887 u0 R+ 0:00.01 ps -ax
22873 v0 Is+ 0:00.01 /usr/libexec/getty Pc ttyv0
interrupt's rate is following
root at rpi2:~ # vmstat -i -w1
interrupt total rate
irq1: mbox0 27 0
irq2: vchiq0 7 0
irq17: bcm283x_dwco 567896005 10
irq26: bcm_dma0 962 0
irq65: uart0 7942705 40
irq70: sdhci_bcm0 496 0
irq73: generic_time 19187335 22
irq76: ipi 845121 9
Total 595872658 6487
irq1: mbox0 0 0
irq2: vchiq0 0 0
irq17: bcm283x_dwco 641 606
irq26: bcm_dma0 0 0
irq65: uart0 525 496
irq70: sdhci_bcm0 0 0
irq73: generic_time 18 17
irq76: ipi 2 2
Total 1186 1121
irq1: mbox0 0 0
irq2: vchiq0 0 0
irq17: bcm283x_dwco 277 262
irq26: bcm_dma0 0 0
irq65: uart0 476 450
irq70: sdhci_bcm0 0 0
irq73: generic_time 13 12
irq76: ipi 3 3
Total 769 727
irq1: mbox0 0 0
irq2: vchiq0 0 0
irq17: bcm283x_dwco 653 641
irq26: bcm_dma0 0 0
irq65: uart0 475 466
irq70: sdhci_bcm0 0 0
irq73: generic_time 11 11
irq76: ipi 1 1
Total 1140 1119
irq1: mbox0 0 0
irq2: vchiq0 0 0
irq17: bcm283x_dwco 434 405
irq26: bcm_dma0 0 0
irq65: uart0 474 442
irq70: sdhci_bcm0 0 0
irq73: generic_time 12 11
irq76: ipi 3 3
Total 923 861
and still the system response time is terrible.
Do I understand it well that you are not able to trigger the problem?
Unfortunatelly, I cannot get any reasonable results from my
investigation except some mirror fixes which did not fix the problem.
Thus I need to work out how to investigate it better and it takes some
time.
Svata
> --HPS
> _______________________________________________
> freebsd-arm at freebsd.org mailing list
> https://lists.freebsd.org/mailman/listinfo/freebsd-arm
>
> To unsubscribe, send any mail to "freebsd-arm-unsubscribe at freebsd.org"
More information about the freebsd-arm
mailing list