Re: [Crash-utility] [PATCH 0/2] The new API about ikconfig.

----- Original Message ----- Hello Toshi, I have a few issues/questions with your patch: 1. This should be impossible, because even if preload_extensions() is used, that function is called long after the init-time call to read_in_kernel_config(IKCFG_INIT) is done: + if (command == IKCFG_INIT && kt->ikconfig_setup) { + /* fastpath */ + if (get_kernel_config("DEBUG_BUGVERBOSE", NULL) == IKCONFIG_N) + kt->flags |= BUGVERBOSE_OFF; + + if (get_kernel_config("NR_CPUS", &val) == IKCONFIG_STR) { + kt->kernel_NR_CPUS = atoi(val); + if (CRASHDEBUG(1)) + error(INFO, "CONFIG_NR_CPUS: %d\n", + kt->kernel_NR_CPUS); + } + if (get_kernel_config("PGTABLE_4", NULL) == IKCONFIG_Y) { + machdep->flags |= VM_4_LEVEL; + if (CRASHDEBUG(1)) + error(INFO, "CONFIG_PGTABLE_4\n"); + } + if (get_kernel_config("HZ", &val) == IKCONFIG_STR) { + machdep->hz = atoi(val); + if (CRASHDEBUG(1)) + error(INFO, "CONFIG_HZ: %d\n", machdep->hz); + } + + goto out1; + } 2. I didn't test this, but I think this will be a problem because it will not return back to the _init() function in the extension module: if ((sp = symbol_search("kernel_config_data")) == NULL) { - if (command == IKCFG_READ) + if (command == IKCFG_READ || command == IKCFG_SETUP) error(FATAL, "kernel_config_data does not exist in this kernel\n"); return; It should do this: If IKCONFIG_INIT: return quietly with no error message If IKCONFIG_READ: error(FATAL, ... If IKCONFIG_SETUP: error(WARNING, ... 2. Can you explain the need for the IKCONFIG_SORT_INTERVAL stuff? 3. When adding items to any global structure (such as to the kernel_table), I always add them to the end of the structure so that any pre-existing extension modules that use something in the unmodified structure will still work even if additional entries are added to the end. 4. Just a preference on my part, but I would allow get_kernel_config() to accept both "CONFIG_XXXX" and "XXX" as a "name" argument, and then just strip off "CONFIG_" if it's there. Dave