[Qemu-devel] Fragile MAX_OP_PER_INSTR

[Jan Kiszka]

Hi,

we ran into a weird SEGFAULT of QEMU which turned out to be a corruption
of tcg_ctx due to writing beyond gen_opc_buf. The reason for this was a
too small MAX_OP_PER_INSTR, given a particular toolchain (here the one
of SLES 10 for x86-32). So far it runs fine with

#define MAX_OP_PER_INSTR 64

Does anyone have an idea how to resolve the related comment in exec-all.h?

/* XXX: make safe guess about sizes */

Or should we just increase the safety margin?

Jan

-- 
Siemens AG, Corporate Technology, CT SE 2
Corporate Competence Center Embedded Linux