[coreboot] [regression] Increased romstage boot time on ASRock E350M1 (AMD Family 14h)

Mon Jan 11 17:11:30 CET 2016

On Mon, Jan 11, 2016 at 7:33 AM, Aaron Durbin <adurbin at google.com> wrote:
> On Sun, Jan 10, 2016 at 4:40 AM, Paul Menzel
> <paulepanter at users.sourceforge.net> wrote:
>> Dear coreboot folks,
>>
>>
>> on the ASRock E350M1, I lately noticed that the SeaBIOS banner takes
>> longer to appear. And looking at the logs board status [1], the time
>> stamps stored in CBMEM confirm this.
>>
>
> What were your typical times like?
>
>> ```
>> $ grep 1st asrock/e350m1/4.2-*/*/coreboot_timestamps.txt
>> asrock/e350m1/4.2-33-g42444f6/2015-10-30T17:54:28Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,199
>> asrock/e350m1/4.2-36-g0ace013/2015-10-31T13:22:12Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,416
>> asrock/e350m1/4.2-37-gab35575/2015-10-31T18:23:47Z/coreboot_timestamps.txt:   0:1st timestamp                                     367,904
>> asrock/e350m1/4.2-41-g3c47e8a/2015-10-31T20:39:02Z/coreboot_timestamps.txt:   0:1st timestamp                                     367,829
>> asrock/e350m1/4.2-42-g0746452/2015-10-31T21:11:04Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,081
>> asrock/e350m1/4.2-43-g160ad6a/2015-10-31T21:12:09Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,290
>> asrock/e350m1/4.2-44-gbabb2e6/2015-10-31T21:14:48Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,023
>> asrock/e350m1/4.2-53-gf6dc544/2015-11-01T13:27:02Z/coreboot_timestamps.txt:   0:1st timestamp                                     368,470
>> asrock/e350m1/4.2-58-g65eec4d/2015-11-02T15:41:33Z/coreboot_timestamps.txt:   0:1st timestamp                                     679,462
>> asrock/e350m1/4.2-628-g62c0276/2015-12-29T17:17:01Z/coreboot_timestamps.txt:   0:1st timestamp                                     1,528,198
>> asrock/e350m1/4.2-701-gb95a074/2016-01-08T01:44:15Z/coreboot_timestamps.txt:   0:1st timestamp                                     1,298,841
>> asrock/e350m1/4.2-702-gfecc24a/2016-01-08T16:21:59Z/coreboot_timestamps.txt:   0:1st timestamp                                     1,289,489
>> asrock/e350m1/4.2-703-g8846382/2016-01-09T21:18:59Z/coreboot_timestamps.txt:   0:1st timestamp                                     1,289,756
>> ```
>>
>> Unfortunately, there was a time, where I had forgotten to select this
>> option, so I am still bisecting this.
>>
>> I thought, it might have been fixed with the commit 4.2-630-g65e33c0
>> below [1], but it’s not.
>>
>> ```
>> commit 65e33c08a9a88c52baaadaf515b9591856115a77
>> Author: Nico Huber <nico.huber at secunet.com>
>> Date:   Mon Dec 28 20:17:13 2015 +0100
>>
>>     x86: Align CBFS on top of ROM
>>
>>     Since the introduction of the new (interim?) master header, coreboot
>>     searches the whole ROM for CBFS entries. Fix that by aligning it on top
>>     of the ROM.
>>
>>     Change-Id: I080cd4b746169a36462a49baff5e114b1f6f224a
>> […]
>> ```
>>
>> Do you know, which commit the commit message referred to?
>>
>> Looking more into it, Nico’s commit was reverted in commit 4.2-673-
>> g12c55ed [3] and the logic reworked.
>>
>> ```
>> commit 12c55eda11453ed1e7a24e218338831f67cd5de6
>> Author: Aaron Durbin <adurbin at chromium.org>
>> Date:   Mon Jan 4 13:57:07 2016 -0600
>>
>>     Revert "x86: Align CBFS on top of ROM"
>>
>>     This reverts commit 65e33c08a9a88c52baaadaf515b9591856115a77.
>>     This was the wrong logic to fix the master header.
>>
>>     Change-Id: I4688034831f09ac69abfd0660c76112deabd62ec
>> […]
>> ```
>>
>> If you have any suggestions, please tell me. Otherwise, I’d continue
>> trying to bisect this.
>>
>> And unfortunately, I am unable to provide romstage messages, as I still
>> haven’t got the serial header for the board.
>>
>> So if somebody else, Stefan, Martin, Kevin, could provide that, that
>> would be awesome.
>
> Did you rebuild cbfstool that contains this patch?
> https://review.coreboot.org/12825 It was the one I said fixed the
> logic when you asked in https://review.coreboot.org/#/c/12824/.
> Additionally, there's also the comments I left in
> https://review.coreboot.org/12810 that explained how that patch was
> incorrect.
>
> Please provide a hexdump snippet (hexdump -C image.rom) that shows the
>  "ORBC" section. That's the master header, and we can analyze if you
> did indeed build w/ an updated cbfstool as well as determine if there
> are other issues.
>

I just looked at the 703-* log:

CBFS: 'Master Header Locator' located CBFS at [100:3fffc0)

That corresponds w/ your config:
CONFIG_CBFS_SIZE=0x400000
CONFIG_ROM_SIZE=0x400000

So that's all correct. It's not cbfs as far as I can tell (seems
unlikely). I suspect that your timestamps are just more correct. Where
does your first timestamp original from? Is that in ramstage?

>>
>>
>> Thanks,
>>
>> Paul
>>
>>
>> PS: I’ll try to create a ticket for this issue in the bug tracker [4]
>> this evening.
>>
>>
>> [1] https://review.coreboot.org/gitweb?p=board-status.git;a=summary
>> [2] https://review.coreboot.org/12810
>> [3] https://review.coreboot.org/12824
>> [4] https://ticket.coreboot.org/
>> --
>> coreboot mailing list: coreboot at coreboot.org
>> http://www.coreboot.org/mailman/listinfo/coreboot