An improper locking bug(e.g., deadlock) on the lock ltc_ecc_fp_lock

[Ryan]

Hi, developers, thank you for your checking. It seems the lock ltc_ecc_fp_lock is not released correctly when cache_entry == NULL in the function ltc_ecc_fp_save_state?

libtomcrypt/src/math/fp/ltc_ecc_fp_mulmod.c

Lines 1412 to 1433 in 673f5ce

    LTC_MUTEX_LOCK(&ltc_ecc_fp_lock); 
    /* 
     * build the list; 
       Cache DEFINITIONS ::= 
       BEGIN 
        CacheDump ::= SEQUENCE { 
          numEntries    SHORTINTEGER, 
          maxEntries    SHORTINTEGER, 
          numLUT        SHORTINTEGER, 
          cache         SEQUENCE OF INTEGER 
        } 
       END 
     * 
     */ 
    /* 
     * The cache itself is a point (3 INTEGERS), 
     * the LUT as pairs of INTEGERS (2 * 1<<FP_LUT), 
     * and the mu INTEGER 
     */ 
    cache_entry = XCALLOC(FP_ENTRIES*(2*(1U<<FP_LUT)+4)+3, sizeof(ltc_asn1_list)); 
    if (cache_entry == NULL) 
       return CRYPT_MEM; 

[Ryan]

Hi, developers @sjaeckel @fperrad @karel-m

Any comments would be highly appreciated. Thank you.

Best Regards,

[Steffen Jaeckel]

btw. referring to OP-TEE/optee_os#4842 would have been nice :)

looks fine, I'll apply the patch you posted to optee

[Ryan]

@sjaeckel Thanks~

[Steffen Jaeckel]

Can you please check whether #626 correctly fixes this?

[Ryan]

It looks good to me.