On 7/21/21 9:08 AM, Hannes Reinecke wrote:

On 7/20/21 10:27 PM, Vladislav Bolkhovitin wrote:

quoted

On 7/16/21 2:04 PM, Hannes Reinecke wrote:

[...]

quoted

+struct nvmet_dhchap_hash_map {
+	int id;
+	int hash_len;
+	const char hmac[15];
+	const char digest[15];
+} hash_map[] = {
+	{.id = NVME_AUTH_DHCHAP_HASH_SHA256,
+	 .hash_len = 32,
+	 .hmac = "hmac(sha256)", .digest = "sha256" },
+	{.id = NVME_AUTH_DHCHAP_HASH_SHA384,
+	 .hash_len = 48,
+	 .hmac = "hmac(sha384)", .digest = "sha384" },
+	{.id = NVME_AUTH_DHCHAP_HASH_SHA512,
+	 .hash_len = 64,
+	 .hmac = "hmac(sha512)", .digest = "sha512" },
+};

"hmac()" is always here, so why not to just auto-generate hmac(sha512)
from sha512?

... all part of the learning curve ...
If that's true then of course I can auto-generate the hmac name.

As far as I understand, this naming convention is as stable as any other
interface in the kernel.

Vlad