Re: module: save load_info for livepatch modules

[Jessica Yu <jeyu@redhat.com>]

+++ Petr Mladek [11/11/15 15:31 +0100]:
>On Mon 2015-11-09 23:45:52, Jessica Yu wrote:
>> In livepatch modules, preserve section, symbol, string information from
>> the load_info struct in the module loader. This information is used to
>> patch modules that are not loaded in memory yet; specifically it is used
>> to resolve remaining symbols and write relocations when the target
>> module loads.
>>
>> Signed-off-by: Jessica Yu <jeyu@redhat.com>
>> ---
>>  include/linux/module.h  | 25 +++++++++++++++++++++++++
>>  kernel/livepatch/core.c | 17 +++++++++++++++++
>>  kernel/module.c         | 36 ++++++++++++++++++++++--------------
>>  3 files changed, 64 insertions(+), 14 deletions(-)
>>
>> diff --git a/include/linux/module.h b/include/linux/module.h
>> index 3a19c79..c8680b1 100644
>> --- a/include/linux/module.h
>> +++ b/include/linux/module.h
>[...]
>> @@ -635,6 +651,15 @@ static inline bool module_requested_async_probing(struct module *module)
>>  	return module && module->async_probe_requested;
>>  }
>>
>> +#ifdef CONFIG_LIVEPATCH
>> +extern void klp_prepare_patch_module(struct module *mod,
>> +				     struct load_info *info);
>> +extern int
>> +apply_relocate_add(Elf64_Shdr *sechdrs, const char *strtab,
>> +		   unsigned int symindex, unsigned int relsec,
>> +		   struct module *me);
>> +#endif
>
>This function is already declared in moduleloader.h.
>It is implemted only when CONFIG_MODULES_USE_ELF_RELA is defined.
>
>I guess that we want to include moduleloader.h in livepatch.
>
>> +
>>  #else /* !CONFIG_MODULES... */
>>
>>  /* Given an address, look for it in the exception tables. */
>> diff --git a/kernel/livepatch/core.c b/kernel/livepatch/core.c
>> index 6e53441..087a8c7 100644
>> --- a/kernel/livepatch/core.c
>> +++ b/kernel/livepatch/core.c
>> @@ -1001,6 +1001,23 @@ static struct notifier_block klp_module_nb = {
>>  	.priority = INT_MIN+1, /* called late but before ftrace notifier */
>>  };
>>
>> +/*
>> + * Save necessary information from info in order to be able to
>> + * patch modules that might be loaded later
>> + */
>> +void klp_prepare_patch_module(struct module *mod, struct load_info *info)
>> +{
>> +	Elf_Shdr *symsect;
>> +
>> +	symsect = info->sechdrs + info->index.sym;
>> +	/* update sh_addr to point to symtab */
>> +	symsect->sh_addr = (unsigned long)info->hdr + symsect->sh_offset;
>
>Is livepatch the only user of this value? By other words, is this safe?

I think it is safe to say yes. klp_prepare_patch_module() is only
called at the very end of load_module(), right before
do_init_module(). Normally, at that point, info->hdr will have already
been freed by free_copy() along with the elf section information
associated with it. But if we have a livepatch module, we don't free.
So we should be the very last user, and there should be nobody
utilizing the memory associated with the load_info struct anymore at
that point.

>> +	mod->info = kzalloc(sizeof(*info), GFP_KERNEL);
>> +	memcpy(mod->info, info, sizeof(*info));
>> +
>> +}
>
>It is strange that this funtion is defined in livepatch/core.c
>but declared in module.h. I would move the definition to
>module.c.

Right, I was trying to keep all the livepatch-related functions
together in livepatch/core.c. but I can move it to module.c if it
makes more sense/Rusty doesn't object to it :-)

>>  static int __init klp_init(void)
>>  {
>>  	int ret;
>> diff --git a/kernel/module.c b/kernel/module.c
>> index 8f051a1..8ae3ca5 100644
>> --- a/kernel/module.c
>> +++ b/kernel/module.c
>> @@ -318,20 +318,6 @@ int unregister_module_notifier(struct notifier_block *nb)
>>  }
>>  EXPORT_SYMBOL(unregister_module_notifier);
>>
>> -struct load_info {
>> -	Elf_Ehdr *hdr;
>> -	unsigned long len;
>> -	Elf_Shdr *sechdrs;
>> -	char *secstrings, *strtab;
>> -	unsigned long symoffs, stroffs;
>> -	struct _ddebug *debug;
>> -	unsigned int num_debug;
>> -	bool sig_ok;
>> -	struct {
>> -		unsigned int sym, str, mod, vers, info, pcpu;
>> -	} index;
>> -};
>> -
>>  /* We require a truly strong try_module_get(): 0 means failure due to
>>     ongoing or failed initialization etc. */
>>  static inline int strong_try_module_get(struct module *mod)
>> @@ -2137,6 +2123,11 @@ static int simplify_symbols(struct module *mod, const struct load_info *info)
>>  			       (long)sym[i].st_value);
>>  			break;
>>
>> +#ifdef CONFIG_LIVEPATCH
>> +		case SHN_LIVEPATCH:
>> +			break;
>> +#endif
>
>IMHO, even a kernel compiled without CONFIG_LIVEPATCH should handle livepatch
>modules with grace. It means to reject loading.

I think even right now, without considering this patchset, we don't
reject modules "gracefully" when we load a livepatch module without
CONFIG_LIVEPATCH. The module loader will complain and reject the
livepatch module, saying something like "Unknown symbol
klp_register_patch." This behavior is the same with or without
this patch series applied. If we want to add a bit more logic to
gracefully reject patch modules, perhaps that should be a different
patch altogether, as I think it is unrelated to the goal of this one :-)

>>  		case SHN_UNDEF:
>>  			ksym = resolve_symbol_wait(mod, info, name);
>>  			/* Ok if resolved.  */
>> @@ -2185,6 +2176,11 @@ static int apply_relocations(struct module *mod, const struct load_info *info)
>>  		if (!(info->sechdrs[infosec].sh_flags & SHF_ALLOC))
>>  			continue;
>>
>> +#ifdef CONFIG_LIVEPATCH
>> +		if (info->sechdrs[i].sh_flags & SHF_RELA_LIVEPATCH)
>> +			continue;
>> +#endif
>> +
>>  		if (info->sechdrs[i].sh_type == SHT_REL)
>>  			err = apply_relocate(info->sechdrs, info->strtab,
>>  					     info->index.sym, i, mod);
>> @@ -3530,8 +3526,20 @@ static int load_module(struct load_info *info, const char __user *uargs,
>>  	if (err < 0)
>>  		goto bug_cleanup;
>>
>> +#ifdef CONFIG_LIVEPATCH
>> +	/*
>> +	 * Save sechdrs, indices, and other data from info
>> +	 * in order to patch to-be-loaded modules.
>> +	 * Do not call free_copy() for livepatch modules.
>> +	 */
>> +	if (get_modinfo((struct load_info *)info, "livepatch"))
>> +		klp_prepare_patch_module(mod, info);
>> +	else
>> +		free_copy(info);
>> +#else
>
>I would move this #else one line above and get rid of the
>double free_copy(info); But it is a matter of taste.

Maybe I'm missing something, but I think we do need the double
free_copy(), because in the CONFIG_LIVEPATCH case, we still want to
call free_copy() for non-livepatch modules. And we want to avoid
calling free_copy() for livepatch modules (hence the extra else).

>>  	/* Get rid of temporary copy. */
>>  	free_copy(info);
>> +#endif
>

Thanks for the comments,
Jessica