On 09/06/2013 08:15 AM, Peter Maydell wrote: > On 6 September 2013 16:12, Richard Henderson <r...@twiddle.net> wrote: >> On 09/05/2013 09:53 PM, Stefan Weil wrote: >>> After lots of SIGSEGV, the program indeed finishes successfully, >>> so my report was wrong - SIGSEGV is not a fatal signal for sparc64. >>> That's interesting - thank you for this information. >> >> It's not just sparc64. >> >> That's changed page detection, for determining when to invalidate >> translation blocks. For self-modifying code, and more. > > Yeah, but in practice sparc64 does this a huge number of > times on simple binaries in a way that other linux-user > guests I looked at just don't. Presumably it's just that > gcc has for some reason put writable data next to code > for that target.
It's the (old?) PLT scheme for sparc -- it modifies code on linking. If you use LD_BIND_NOW=1 it can help speed things up when debugging. r~
