On Tue, Aug 18, 2015 at 6:38 PM, Andrew Morton
[off-list ref] wrote:

On Tue, 18 Aug 2015 16:06:00 -0400 Dan Streetman [off-list ref] wrote:

quoted

Add ZPOOL_MAX_TYPE_NAME define, and change zpool_driver *type field to
type[ZPOOL_MAX_TYPE_NAME].  Remove redundant type field from struct zpool
and use zpool->driver->type instead.

The define will be used by zswap for its zpool param type name length.

Patchset is fugly.  All this putzing around with fixed-length strings,
worrying about overflow and is-it-null-terminated-or-isnt-it.  Shudder.

It's much better to use variable-length strings everywhere.  We're not
operating in contexts which can't use kmalloc, we're not
performance-intensive and these strings aren't being written to
fixed-size fields on disk or anything.  Why do we need any fixed-length
strings?

IOW, why not just replace that alloca with a kstrdup()?

for the zpool drivers (zbud and zsmalloc), the type is actually just
statically assigned, e.g. .type = "zbud", so you're right the *type is
better than type[].  I'll update it.

quoted

--- a/include/linux/zpool.h
+++ b/include/linux/zpool.h

...

@@ -79,7 +77,7 @@ static struct zpool_driver *zpool_get_driver(char *type)

      spin_lock(&drivers_lock);
      list_for_each_entry(driver, &drivers_head, list) {
-             if (!strcmp(driver->type, type)) {
+             if (!strncmp(driver->type, type, ZPOOL_MAX_TYPE_NAME)) {

Why strncmp?  Please tell me these strings are always null-terminated.

Yep, you're right.  The driver->type always is, and the type param is
passed in from sysfs, which we can rely on to be null-terminated.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>