**** BEGIN LOGGING AT Wed Nov 09 03:00:00 2016 Nov 09 17:14:56 Crofton|work: you on? figured out the whole broken SDHCI thing Nov 09 17:42:43 nrossi, yeah aaround Nov 09 17:42:47 what happened Nov 09 17:44:30 Crofton|work: The commit that i mentioned that the bisect on was the trigger. It causes DEBUG_RODATA to default to y. Nov 09 17:44:48 OK Nov 09 17:45:00 and that triggers the issue? Nov 09 17:45:07 but the actual bug is caused by DEBUG_RODATA padding out the .text section Nov 09 17:45:23 I'm glad smart people look at stuff like this Nov 09 17:45:49 which causes .head.text to be small enough that it fits within the first 512KB... Nov 09 17:46:01 which you can't dma into Nov 09 17:46:03 ? Nov 09 17:46:06 yep Nov 09 17:46:37 but that was already setup as reserved, it just didn't take into account that there could be a hole between .head.text and .text Nov 09 17:46:45 https://github.com/nathanrossi/linux/commit/f8a0fd1a13e13346d929edcf1fe4f83e5dc58eb2 Nov 09 17:47:07 OMG, I can't help myself Nov 09 17:47:12 that is the fix i have Nov 09 17:47:26 nrossi, will make zynq great again! Nov 09 17:48:26 well i learnt much about how the linker scripts work, and how memblock works :) Nov 09 17:48:39 yep Nov 09 17:48:56 and found an awful heisenbug Nov 09 17:49:34 yep, definitely a wild goose chase :) Nov 09 17:49:53 well you did end up finding an actual problem Nov 09 17:51:13 For linux-yocto 4.8/etc i will just disable CONFIG_DEBUG_RODATA, since that will remove the padding. And you can backport it easy :) Nov 09 17:52:03 ok, so key for now is make sure that never gets enabled? Nov 09 17:52:33 yep, if you are using fragments, "# CONFIG_DEBUG_RODATA is not set" is the magic line Nov 09 17:52:53 rgr **** ENDING LOGGING AT Thu Nov 10 03:00:00 2016