Revert "evm: Fix memleak in init_desc"
authorXiu Jianfeng <xiujianfeng@huawei.com>
Fri, 27 May 2022 11:17:26 +0000 (19:17 +0800)
committerMimi Zohar <zohar@linux.ibm.com>
Wed, 15 Jun 2022 18:03:47 +0000 (14:03 -0400)
This reverts commit ccf11dbaa07b328fa469415c362d33459c140a37.

Commit ccf11dbaa07b ("evm: Fix memleak in init_desc") said there is
memleak in init_desc. That may be incorrect, as we can see, tmp_tfm is
saved in one of the two global variables hmac_tfm or evm_tfm[hash_algo],
then if init_desc is called next time, there is no need to alloc tfm
again, so in the error path of kmalloc desc or crypto_shash_init(desc),
It is not a problem without freeing tmp_tfm.

And also that commit did not reset the global variable to NULL after
freeing tmp_tfm and this makes *tfm a dangling pointer which may cause a
UAF issue.

Reported-by: Guozihua (Scott) <guozihua@huawei.com>
Signed-off-by: Xiu Jianfeng <xiujianfeng@huawei.com>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
security/integrity/evm/evm_crypto.c

index a733aff..708de96 100644 (file)
@@ -75,7 +75,7 @@ static struct shash_desc *init_desc(char type, uint8_t hash_algo)
 {
        long rc;
        const char *algo;
-       struct crypto_shash **tfm, *tmp_tfm = NULL;
+       struct crypto_shash **tfm, *tmp_tfm;
        struct shash_desc *desc;
 
        if (type == EVM_XATTR_HMAC) {
@@ -120,16 +120,13 @@ unlock:
 alloc:
        desc = kmalloc(sizeof(*desc) + crypto_shash_descsize(*tfm),
                        GFP_KERNEL);
-       if (!desc) {
-               crypto_free_shash(tmp_tfm);
+       if (!desc)
                return ERR_PTR(-ENOMEM);
-       }
 
        desc->tfm = *tfm;
 
        rc = crypto_shash_init(desc);
        if (rc) {
-               crypto_free_shash(tmp_tfm);
                kfree(desc);
                return ERR_PTR(rc);
        }