Messages in this thread | | | Date | Wed, 25 Nov 2009 05:35:16 -0500 | From | "Youquan,Song" <> | Subject | Re: [PATCH]crypto: Fix algorithm and driver duplicate registered |
| |
> On Mon, Nov 23, 2009 at 12:47:24PM -0500, Youquan,Song wrote: > > > > The alg->cra_name and alg->cra_driver_name, from the description, > > cra_name is the algorithm name and cra_driver_name is the driver name. > > For each algorithm you may have an arbitrary number of drivers > implementing it. The driver name must be unique. However, > duplicates are allowed so that a new version of a driver may > be loaded while the older instances of it are still in use. > > > But when I read the code, I often get confuse about these two names. > > They often mix each other. > > > > Can you give me some instruction about them? or Can we make them more clear? Thanks. > > Well can you tell me what exactly confuses you?
Thanks. For example:
struct crypto_alg *crypto_larval_lookup(const char *name, u32 type, u32 mask)
The parameter "name", sometime it is "alg->cra_name" while sometime it become "alg->cra_driver_name". What's to lookup, algorithm or driver, depends on the context of parameter, So it need confirm no duplicate name exists between cra_driver_name and cra_name. Therefore, there are some werid checking needed, such as following:
if (!strcmp(q->cra_driver_name, alg->cra_name) || !strcmp(q->cra_name, alg->cra_driver_name))
I wonder, Can we define two functions: one lookup algorithm other lookup driver? It will be more clear.
Thanks.
| |