Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- [19:30] <freemangordon> Hi there. I am running Ubuntu 12.04 (armHF+thumb2)on Nokia N900 (Cortex-A8 r1p2) - the one with broken thumb2. AFAIK there are 2 erratas to be workarounded to have working thumb - 430973 and 657417. The first is workarounded in kernel (with a little help from u-boot) and the other - by linker. Can anyone tell me if something else is needed(I know it is #beagle, was pointed here by dm8tbr @ #maemo). BTW Ubuntu runs just fine, no SIGILL crashes at all.
- [19:51] <dm8tbr> freemangordon: aholler recently saw the thumb2 problem. don't know though if and which errata he had addressed.
- [19:52] <freemangordon> dm8tbr, thanks, but I will wait for a couple of hours to see if I will get an answer. Will repeat my question if there is no. Don't want to be pushy, noone's on a hurry after all
- [19:54] <dm8tbr> I don't remember who was involved in the discussion back then, but I think koen, mru and probably rcn-ee
- [19:55] <aholler> dm8tbr: I just didn't know if it's a toolchain or hw-bug. I've only remembered that I couldn't use thumb2 to build a system.
- [19:55] <mru> thumb2 on r1pX cortex-a8 is finicky
- [19:55] <mru> there are multiple hardware bugs
- [19:55] <dm8tbr> yes, and my recollection was that there was no way to work around all of them at the same time
- [19:55] <dm8tbr> at least not reliably
- [19:56] <koen> dm8tbr: does disabling t2 count as a workaround?
- [19:56] <mru> something like that
- [19:56] <freemangordon> mru, I know that, but I think i found the way to workaround all of them, just need someone more knowledgeable than me to confirm/not
- [19:56] <dm8tbr> koen: freemangordon claims he's found a way
- [19:56] <mru> and the secure monitor makes it hard/impossible to apply all the workarounds required
- [19:56] <freemangordon> mru, there is an SCM call on n900 to play with ACR
- [19:57] <freemangordon> SMC*
- [19:57] <mru> right, n900 uses hs chips and might have its own secure monitor
- [19:57] <freemangordon> mru: so I can set IBE bit which is non-accessible uotside the secure world
- [19:57] <mru> so supposing you have write access to all the required registers, it's still hard to get it right
- [19:58] <mru> and even if you do, there will be a performance hit
- [19:58] <mru> just stick to arm code
- [19:58] <mru> much simpler
- [19:58] <freemangordon> mru: may give a short description on my findings before a definite conclusion. I am working on that for about 6 months or so.
- [19:59] <mru> 6 months that could have been better spent imo
- [19:59] <mru> if you need thumb2, get an r3 chip instead
- [20:00] <mru> also, you don't actually need thumb2
- [20:00] <freemangordon> mru: Nokia N900. No way to touch the HW
- [20:00] <mru> so stick with arm code
- [20:00] <freemangordon> and Ubuntu for OMAP3 if thumb2 comiled
- [20:00] <freemangordon> *is
- [20:00] <mru> so?
- [20:00] <mru> nobody really runs that on beagle classic
- [20:00] <denix> _av500_: indeed again!
- [20:00] <mru> and xm doesn't have the bug
- [20:00] <freemangordon> no option to run it on n900 without thumb workarounds
- [20:01] <mru> then run something else
- [20:01] <freemangordon> yeah
- [20:03] <freemangordon> mru: ok, are you aware of anything beside erratas 430973 and 657417 which prevents thumb2 on thumb-broken omap cores. have in mind that we never hit 687067 as 430973 workaround does not use BTB invalidate by MVA instructions?
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement