Hi Andreas. On Mon, 8 Sep 2014 11:11:14 +0200, Andreas Fritiofson wrote: > On Sun, Sep 7, 2014 at 10:29 PM, Jens Bauer <[email protected]> wrote: >> >> [...] NOP that is HardFaulting. [...] >> >> What I find peculiar, is that ... reading memory from OpenOCD seems >> to be inconsistent. > > Just a thought... I have no experience with the LPC controllers but I > assume that they, like most, need a configurable number of > wait-states on flash access, depending on clock speed, Vdd range and > so on. Is it configured correctly?
Yes, correct. > Maybe try to add an extra wait-state to see if it makes a difference. This is a really good suggestion - I'll try that! > Perhaps the oddly behaving addresses belong to flash cells that are on the > limit, > and temperature increase pushes it over after a while... It could explain why reading bytes from the addresses may behave differently from reading halfwords / words too! As far as I understand, a halfword / word will be read multiple times when reading bytes only. So that's very likely. > If the target crashes on it's own, I think we can rule out OpenOCD, for now. I think that the crash isn't the most interesting thing; it's the strange behaviour of reading odd/even addresses gives different results. But as you say ... it may be because of those flash wait-states that OpenOCD reads the content incorrectly. The current wait-state setting might be fine for reading byte positions, but may fail on 16 / 32 bit reads (or the other way round). > I have no idea how the disassemble routine handles odd/even > addresses. I guess the correct address to pass is the even one, > without the T-bit set. OK. This too could be due to the wait-states. I believe odd addresses would use the byte-reader, rather than a complex 32-bit read with shifting; because it logically, shouldn't attempt to touch bytes outside the requested range. I'll try checking that the wait-states are set up correctly (and inserting more wait-states to see how it helps). Thank you for the suggestion. =) Love Jens ------------------------------------------------------------------------------ Want excitement? Manually upgrade your production database. When you want reliability, choose Perforce Perforce version control. Predictably reliable. http://pubads.g.doubleclick.net/gampad/clk?id=157508191&iu=/4140/ostg.clktrk _______________________________________________ OpenOCD-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/openocd-devel
