Hi, First of all, please do not recycle old threads to send your messages. Please use the "New" button of your e-mail client, and not the "Reply" button when you want to start a new thread. Thanks. Le Fri, 10 Mar 2006 22:22:51 +0530, Vamsi <krishna.vamsi@xxxxxxxxx> a écrit : > gng by the Kernel Module initialisation code ( > kernel/module.c-sys_init_module() function ), it is trying flush > i-cache ranges and i really failed to understand why it is doing it... > > if (mod->module_init) > flush_icache_range((unsigned long)mod->module_init, > (unsigned long)mod->module_init > + mod->init_size); > flush_icache_range((unsigned long)mod->module_core, > (unsigned long)mod->module_core + > mod->core_size); This happens right after module loading, and before execution of any code of the module. From my understand, this flush of Instruction Cache is needed, because this cache might contain spurious cache lines, not coherent with what we just loaded into memory. In order to make sure that we're going to execute the code we have just loaded, we need to flush the cache for the range of adresses where we loaded the code. BTW, on i386, flush_icache_range() is a no-op, because the processor already takes care of this kind of problem. See http://lxr.linux.no/source/include/asm-i386/cacheflush.h#L15. But for m68k, some stuff is needed, see http://lxr.linux.no/source/arch/m68k/mm/memory.c#L426. This is related to the cache coherency problem, see http://en.wikipedia.org/wiki/Cache_coherency. (Please note that I'm not a specialist of these caches, barriers and other architecture-dependant subtelties, so do not hesitate to correct these informations if needed.) Sincerly, Thomas -- Thomas Petazzoni - thomas.petazzoni@xxxxxxxx http://{thomas,sos,kos}.enix.org - http://www.toulibre.org http://www.{livret,agenda}dulibre.org -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
