I'm debugging a custom Bay Trail-I board design that uses either the E3815 or E3805 Bay Trail-I SKUs (both D0-step). I'm using a coreboot BIOS based on the bayleybay_fsp mainboard to run DOS and Linux test utilities. Boards with the E3815 processor boot, but with the E3805 it hangs in FspInit. The main difference with the E3805 SKU is the lack of built-in iGFX.
Running a debug version of the FSP, "BAYTRAIL_FSP_GOLD_003_DEBUG_29-SEP-2014", it stops while querying the graphics. I've attached the coreboot log with Fsp debug messages for both the E3805 (hang) and E3815 (boot) for comparison.
I've tried setting the FSP graphics parameters to a logical value for "no graphics" (see logs). I've tried both M0130679901 and M0130679903 microcode with no difference. I'm using TXE binary "BYT-I_DUAL_BOOT_TXE_KIT_GOLD_RELEASE_126.96.36.1999".
Has anyone else tried the E3805 with the GOLD3 FSP? I suspect the FSP doesn't properly handle the lack of iGFX at the moment.
Hello Scott Radcliffe.
When you talk about a debug version of FSP, does it mean that you have modified the file?
Please use the https://downloadcenter.intel.com/download/24496 Intel® FSP for Intel® Atom™ Processor E3800 Series Gold3: Linux* Release , and be sure to follow the steps mentioned in section
4.0 Example Boot Loader from http://www.intel.com/content/dam/www/public/us/en/documents/guides/atom-e3800-crb-guide.pdf Intel® Atom™ Processor E3800 Product Family Customer Reference Board Platform Guide.
I used the Gold3 FSP you mention initially, which stops during FspInit() with the E3805. The POST code at this point is 0x2F (port 80h), 0x01 (port 81h). From the Bayley Bay CRB documentation, this indicates that memory initialization is complete. Using the same firmware with the E3815 has no issues.
I then tried a debug build of the FSP. This came from Intel (I think it was released on IBP). This FSP acts the same way, but provides debug logging through the serial console. The last debug message from the FSP before stopping is "Polling Render Force Wake Acknowledge Bit". If you look at the attached file "E3805 Processor coreboot log.txt", everything from the line "============= PEIM FSP (VLYVIEW0 0x00000303) =============" and after is generated by the debug FSP.
I'm not sure if you got an answer on this one. The E3805 is the only version of this processor that does not have graphics, so the graphics portion will not load. The closest alternative processor with graphics would be the E3815.
We're not trying to use the graphics, which is why we want to use the E3805. The problem is we hang just after memory initialization, while in the Intel FSP binary blob, when using the E3805. My best guess is the Intel FSP doesn't work with the E3805. I'm wondering if anyone has tested the E3805 with the Intel FSP?
Checking with our engineering staff, but sure we used FSP Gold 3 for the minnowboard OSP (coreboot based). We should have a Bay Trail OSP for testing available soon (also FSP Gold 3-based) on our website. www.se-eng.com, but I don't know that we ported specifically to the Atom E3805. I'll have a sales engineer contact you, if you wish.
I thought I posted this answer. Can you access this? http://www.intel.com/content/www/us/en/intelligent-systems/intel-firmware-support-package/intel-fsp-overview.html http://www.intel.com/content/www/us/en/intelligent-systems/intel-firmware-support-package/intel-fsp-overview.html My AE suggested this site for all E38XX processors.
That page is public and is where I grabbed the first FSP I tried (Linux version of E3800 GOLD3). I was recently enabled to ask questions about the FSP through Intel's premier support portal. I'm not sure whether I can share the response here, but I will say I'm waiting for an update to the FSP.
Hello Scott, et al,
Just wondering if there is any public resolution to this problem? We're having what looks to be the same issue. We have a few development boards, some with 3825's, and some with 3805's populated (the boards are identical except for the populated processors). We can boot the 3825s with no problem, but the 3805s spin forever after a console message saying memory init is complete. We're waiting for access to the restricted support site, and have just ordered a Dediprog SF100...trying to figure out what to try once we're able to reprogram the flash.