<KanjiMonster>
russell--: so it looks like the commit comment is correct, and the old macros cause a 0x22 -> 0x22 -> 0x23 -> 0x23. So I wonder if the fix made something work to be effective which isn't actually supported, and now is causing the hang on warm boot
<KanjiMonster>
russell--: my next step would be to do printk debugging and add a pr_info("foo\n"); before any changed lines by the commit, which should tell you (hopefully) which lines it doesn't like
<KanjiMonster>
assuming these are only called on boot, and not during regular runtime ...
<KanjiMonster>
alternative would be to revert them one by one to the _old again; a re-build should be quick then
<KanjiMonster>
you can just edit the file in build_dir/target-*/linux-*/linux-*, a make target/linux/install should then pick up the changes and build a new image
<KanjiMonster>
russell--: so the three suspect places are the one for /* Enable cxMMX extensions (GX1 Datasheet 54) */, and the two changes in geode_configure()
<russell-->
in geode_configure() "fixes" it, doesn't hang anymore
<KanjiMonster>
now you could try to use 0x80 instead of 0x88 to just enable the #SUSP pin
<russell-->
trying ...
<russell-->
hung on the second warm boot
<KanjiMonster>
interesting. So next test would be 0x08, just the suspend on halt feature
guidosarducci has joined #openwrt-devel
guidosarducci_ has quit [Remote host closed the connection]
<russell-->
survived the first warm boot (the reboot after flashing), warm booting again and it seemed to survive that as well, one more for good measure ...
<russell-->
and it survived that as well with:
<russell-->
/* Suspend on halt power saving and enable #SUSP pin */
Mangix has quit [Read error: Connection reset by peer]
Mangix has joined #openwrt-devel
noltari_ has joined #openwrt-devel
noltari has quit [Ping timeout: 480 seconds]
mrnuke has quit [Ping timeout: 480 seconds]
<Habbie>
can i tell .config "i just want to build for arm926ej-s, don't care about the target"?
<schmars[m]>
i dont think so - i think i've done that in the past and the defconfig call afterwards just wiped it, if i hadn't selected target+subtarget+profile
<schmars[m]>
(selected, as in echo'd into .config)
<Habbie>
yeah, i remember failing at that too
<Habbie>
i found the right target/subtarget combo now, but it's a bit of a puzzle sometimes
<schmars[m]>
in my tooling i save the output of 'scripts/dump-target-info.pl architectures' and make use of it a lot
<schmars[m]>
if you're just saying that translating architecture -> target+subtarget is a bit of a hassle every time
<Habbie>
oh yes, i was going to try to remember the tool
<Habbie>
let me see
<Habbie>
arm_arm926ej-s at91/sam9x mxs/generic
<Habbie>
right
<Habbie>
same result as my poor git grep
<Habbie>
but great if i ever need to do this automatically
<Habbie>
yes, that's perfect
<Habbie>
pity that in 'make menuconfig' it doesn't say "mxs" anywhere
<Habbie>
i'm doing something worse, i want to (ab)use openwrt's build system to build binaries (not packages) for a system to which openwrt hasn't even been ported