Re: [PATCH] sunrpc: Improve exception handling in krb5_etm_checksum()

[Markus Elfring]

…
>> Thus use another label.
…
>> ---
>>  net/sunrpc/auth_gss/gss_krb5_crypto.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
…
> As has undoubtedly been pointed out in other forums, calling kfree()
> with a NULL argument is perfectly valid.

The function call “kfree(NULL)” is not really useful for error/exception handling
while it is tolerated at various source code places.


>                                          Since this small GFP_KERNEL
> allocation almost never fails, it's unlikely this change is going to
> make any difference except for readability.

I became curious if development interests can grow for the usage of
an additional label.
https://wiki.sei.cmu.edu/confluence/display/c/MEM12-C.+Consider+using+a+goto+chain+when+leaving+a+function+on+error+when+using+and+releasing+resources


> Now if we want to clean up the error flows in here to look more
> idiomatic, how about this:
…
> +++ b/net/sunrpc/auth_gss/gss_krb5_crypto.c
…
> @@ -970,8 +970,9 @@ u32 krb5_etm_checksum(struct crypto_sync_skcipher *cipher,
>
>  out_free_ahash:
>  	ahash_request_free(req);
> -out_free_mem:
> +out_free_iv:
>  	kfree(iv);
> +out_free_cksumdata:
>  	kfree_sensitive(checksumdata);
…

I find it nice that you show another possible adjustment of corresponding identifiers.

Regards,
Markus