Linux-NVME Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Hannes Reinecke" <hare@kernel.org>
Cc: linux-nvme@lists.infradead.org
Subject: Re: [PATCH v4 5/9] nvmet-auth: parse dhchap key from configfs attribute
Date: Fri, 20 Nov 2026 11:28:54 +0000	[thread overview]
Message-ID: <sashiko-outbox-225043@kernel.org> (raw)
In-Reply-To: <20261120-dhchap-keyring-v4-v4-5-5f8c7d9926ee@kernel.org>

Thank you for your contribution! Sashiko AI review found 5 potential
issues to consider:
- [High] nvmet-auth: missing error abort and invalid status in
  nvmet_setup_auth()
- [High] nvmet-auth: missing initialization of ctrl->shash_id in
  nvmet_setup_auth()
- [High] nvmet-auth: missing length checks in nvme_dhchap_psk_preparse()
- [Medium] nvmet-auth: memory leak of keydata in
  nvmet_host_dhchap_ctrl_key_store()
- [Medium] nvmet-auth: empty writes rejected in configfs key store
  callbacks
--

commit 4bc851b3a915dfe8fe3d7c3db8fe8c3fa58c01ca
Author: Hannes Reinecke <hare@kernel.org>

nvmet-auth: parse dhchap key from configfs attribute

This commit converts the nvmet configfs DH-HMAC-CHAP key handling to store
host and controller keys as struct key pointers rather than raw secret
strings.

> diff --git a/drivers/nvme/target/auth.c b/drivers/nvme/target/auth.c
> index 26457ce4acc2a..33b45b4b202d4 100644
> --- a/drivers/nvme/target/auth.c
> +++ b/drivers/nvme/target/auth.c
> @@ -19,52 +19,63 @@

[ ... ]

>  int nvmet_auth_set_key(struct nvmet_host *host, const char *secret,
>  		       bool set_ctrl)
>  {
> -	unsigned char key_hash;
> -	char *dhchap_secret;
> -	int offset = 0;
> +	struct key *key;
> +	size_t len;
>  
>  	if (!strlen(secret)) {
>  		down_write(&nvmet_config_sem);
> -		if (set_ctrl) {
> -			kfree(host->dhchap_ctrl_secret);
> -			host->dhchap_ctrl_secret = NULL;
> -			host->dhchap_ctrl_key_hash = 0;
> -			host->dhchap_ctrl_key_offset = 0;
> -		} else {
> -			kfree(host->dhchap_secret);
> -			host->dhchap_secret = NULL;
> -			host->dhchap_key_hash = 0;
> -			host->dhchap_key_offset = 0;
> -		}
> +		nvmet_auth_revoke_key(host, set_ctrl);
>  		up_write(&nvmet_config_sem);
>  		return 0;
>  	}
> -	if (sscanf(secret, "DHHC-1:%hhd:%n", &key_hash, &offset) != 1 ||
> -	    !offset)
> -		return -EINVAL;
> -	if (key_hash > 3) {
> -		pr_warn("Invalid DH-HMAC-CHAP hash id %d\n",
> -			 key_hash);
> -		return -EINVAL;
> -	}
> -	dhchap_secret = kstrdup(secret, GFP_KERNEL);
> -	if (!dhchap_secret)
> -		return -ENOMEM;
> +	len = strcspn(secret, "\n");
> +	key = nvme_auth_extract_key(NULL, secret, len);
> +	if (IS_ERR(key)) {
> +		pr_debug("%s: invalid key specification\n", __func__);
> +		return PTR_ERR(key);
> +	}

[Severity: High]
Can passing unvalidated configfs strings into nvme_auth_extract_key()
underflow datalen and ret in nvme_dhchap_psk_preparse()?

nvme_auth_extract_key() calls nvme_dhchap_psk_create() -> key_create() ->
nvme_dhchap_psk_preparse():

drivers/nvme/common/keyring.c:nvme_dhchap_psk_preparse() {
	...
	if (sscanf(prep->data, "DHHC-%01hhu:%02hhu:%*s",
		   &version, &hmac) != 2) {
		...
		return -EINVAL;
	}

	/* skip header and final ':' character */
	datalen -= 11;
	...
	ret = base64_decode(prep->data + 10, datalen, upayload->data,
			    true, BASE64_STD);
	if (ret < 0) {
		...
		return ret;
	}
	ret -= 4;
	crc = ~crc32(~0, upayload->data, ret);
	...
}

Because %02hhu matches 1 or 2 digits and %*s is not counted in sscanf()'s
return value, a short string such as "DHHC-1:1:" (9 bytes) returns 2 with
datalen < 11, so datalen -= 11 underflows size_t datalen and passes a huge
length to base64_decode().  Similarly, a short payload that decodes to 0..3
bytes (such as "DHHC-1:01:AA==:") makes ret -= 4 negative, which is then
converted to size_t in crc32() and reads past upayload->data.

[ ... ]

> @@ -186,44 +195,74 @@ u8 nvmet_setup_auth(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq, bool reset)
>  		goto out_unlock;
>  	}
>  
> -	if (!host->dhchap_secret) {
> +	if (!host->dhchap_key) {
>  		pr_debug("No authentication provided\n");
>  		goto out_unlock;
>  	}
>  
> -	host_key = nvme_auth_extract_key(NULL, host->dhchap_secret,
> -					 strlen(host->dhchap_secret));
> -	if (IS_ERR(host_key)) {
> -		ret = NVME_AUTH_DHCHAP_FAILURE_NOT_USABLE;
> +	key = key_get(host->dhchap_key);
> +	if (!key) {
> +		pr_warn("%s: host key %08x not found\n",
> +			__func__, key_serial(host->dhchap_key));
>  		goto out_unlock;
>  	}
> -	if (host->dhchap_ctrl_secret) {
> -		ctrl_key = nvme_auth_extract_key(NULL, host->dhchap_ctrl_secret,
> -						 strlen(host->dhchap_ctrl_secret));
> -		if (IS_ERR(ctrl_key)) {
> -			key_put(host_key);
> -			ret = NVME_AUTH_DHCHAP_FAILURE_NOT_USABLE;
> -			goto out_unlock;
> +	down_read(&key->sem);
> +	ret = key_validate(key);
> +	if (!ret) {
> +		if (ctrl->host_key) {
> +			pr_debug("%s: drop host key %08x\n",
> +				 __func__, key_serial(ctrl->host_key));
> +			key_put(ctrl->host_key);
>  		}
> +		ctrl->host_key = key;
> +	}
> +	key_id = key_serial(key);
> +	up_read(&key->sem);
> +	if (ret) {
> +		pr_warn("host key %08x invalidated\n", key_id);
> +		key_put(key);
> +		key = ERR_PTR(-EKEYREVOKED);
>  	}

[Severity: High]
When key_validate(key) fails on host->dhchap_key, doesn't execution fall
through instead of setting ret = NVME_AUTH_DHCHAP_FAILURE_NOT_USABLE and
jumping to out_unlock?

If host->dhchap_ctrl_key is present and valid, the second key_validate()
overwrites ret with 0, so nvmet_setup_auth() returns 0 while leaving
ctrl->host_key NULL on initial connect, bypassing authentication because
nvmet_has_auth() and nvmet_check_auth_status() treat a NULL ctrl->host_key
as no authentication required.  If host->dhchap_ctrl_key is absent or also
invalid, nvmet_setup_auth() returns the negative errno from key_validate()
truncated to u8 instead of NVME_AUTH_DHCHAP_FAILURE_NOT_USABLE, and if
key_get() returns NULL it jumps to out_unlock with ret == 0.

> -	key_put(ctrl->host_key);
> -	ctrl->host_key = host_key;
> -	key_put(ctrl->ctrl_key);
> -	ctrl->ctrl_key = ctrl_key;
> -	ctrl->shash_id = host->dhchap_hash_id;
>  #ifdef CONFIG_NVME_TARGET_AUTH_DEBUG
> -	host_hash = nvme_dhchap_psk_hash(ctrl->host_key);
> -	pr_debug("%s: using hash %s key %u\n", __func__,
> -		 ctrl_hash > 0 ?
> -		 nvme_auth_hmac_name(ctrl_hash) : "none",
> -		 key_serial(ctrl->host_key));
> -	if (ctrl->ctrl_key) {
> -		ctrl_hash = nvme_dhchap_psk_hash(ctrl->ctrl_key);
> -		pr_debug("%s: using ctrl hash %s key %u\n", __func__,
> -			 ctrl_hash > 0 ?
> -			 nvme_auth_hmac_name(ctrl_hash) : "none",
> -			 key_serial(ctrl->ctrl_key));
> +	pr_debug("%s: using dhchap hash %s key %08x\n", __func__,
> +		 nvme_auth_hmac_name(ctrl->shash_id), key_id);
> +#endif

[Severity: High]
Does dropping ctrl->shash_id = host->dhchap_hash_id leave ctrl->shash_id at
0?

nvmet_auth_negotiate() matches the host's proposed HMAC IDs against
ctrl->shash_id, so leaving it at 0 ignores the host's configured dhchap_hash
attribute and always falls back to fallback_hash_id (and passes 0 to
nvme_auth_hmac_name() in the debug prints here).

> diff --git a/drivers/nvme/target/configfs.c b/drivers/nvme/target/configfs.c
> index 27c50fa533e70..880cf5151588a 100644
> --- a/drivers/nvme/target/configfs.c
> +++ b/drivers/nvme/target/configfs.c

[ ... ]

> @@ -2251,9 +2266,19 @@ static ssize_t nvmet_host_dhchap_key_store(struct config_item *item,
>  		const char *page, size_t count)
>  {
>  	struct nvmet_host *host = to_host(item);
> +	u8 *keydata;
> +	size_t len;
>  	int ret;
>  
> -	ret = nvmet_auth_set_key(host, page, false);
> +	len = strcspn(page, "\n");
> +	if (!len)
> +		return -EINVAL;
> +	keydata = kstrndup(page, len, GFP_KERNEL);
> +	if (!keydata)
> +		return -ENOMEM;
> +
> +	ret = nvmet_auth_set_key(host, keydata, false);
> +	kfree_sensitive(keydata);

[Severity: Medium]
Does returning -EINVAL when !len in nvmet_host_dhchap_key_store() and
nvmet_host_dhchap_ctrl_key_store() prevent clearing a configured key via
configfs?  An empty or newline-only write is rejected before reaching the
!strlen(secret) revocation branch in nvmet_auth_set_key().

[ ... ]

> @@ -2284,9 +2323,18 @@ static ssize_t nvmet_host_dhchap_ctrl_key_store(struct config_item *item,
>  		const char *page, size_t count)
>  {
>  	struct nvmet_host *host = to_host(item);
> +	u8 *keydata;
> +	size_t len;
>  	int ret;
>  
> -	ret = nvmet_auth_set_key(host, page, true);
> +	len = strcspn(page, "\n");
> +	if (!len)
> +		return -EINVAL;
> +	keydata = kstrndup(page, len, GFP_KERNEL);
> +	if (!keydata)
> +		return -ENOMEM;
> +
> +	ret = nvmet_auth_set_key(host, keydata, true);
>  	/*
>  	 * Re-authentication is a soft state, so keep the
>  	 * current authentication valid until the host

[Severity: Medium]
Does nvmet_host_dhchap_ctrl_key_store() leak keydata?  Unlike
nvmet_host_dhchap_key_store(), it does not call kfree_sensitive(keydata)
after nvmet_auth_set_key().

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261120-dhchap-keyring-v4-v4-0-5f8c7d9926ee@kernel.org?part=5


  reply	other threads:[~2026-11-20 11:28 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20 11:13 [PATCH v4 0/9] nvme-auth: switch to use the kernel keyring Hannes Reinecke
2026-11-20 11:14 ` [PATCH v4 1/9] nvme-auth: modify nvme_auth_transform_key() to return status Hannes Reinecke
2026-11-20 11:19   ` sashiko-bot
2026-11-20 11:42     ` Hannes Reinecke
2026-11-20 11:14 ` [PATCH v4 2/9] nvme-keyring: add 'dhchap' key type Hannes Reinecke
2026-11-20 11:28   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 3/9] nvme-auth: switch to use 'struct key' Hannes Reinecke
2026-11-20 11:31   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 4/9] nvme: parse dhchap keys during option parsing Hannes Reinecke
2026-11-20 11:32   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 5/9] nvmet-auth: parse dhchap key from configfs attribute Hannes Reinecke
2026-11-20 11:28   ` sashiko-bot [this message]
2026-11-20 11:14 ` [PATCH v4 6/9] nvme: allow to pass in key description as dhchap secret Hannes Reinecke
2026-11-20 11:31   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 7/9] nvme-auth: wait for authentication to finish when changing keys Hannes Reinecke
2026-11-20 11:26   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 8/9] nvme-fabrics: allow to pass in keyring by name Hannes Reinecke
2026-11-20 11:37   ` sashiko-bot
2026-11-20 11:14 ` [PATCH v4 9/9] nvmet: add configfs attribute 'dhchap_keyring' Hannes Reinecke
2026-11-20 11:39   ` sashiko-bot

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-225043@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=hare@kernel.org \
    --cc=linux-nvme@lists.infradead.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox