2072633 - jpake_Round1/Round2/Final in jpakesftk.c ignore PORT_NewArena() failure

Status: RESOLVED FIXED

(NSS :: Libraries, defect)

Description

[Tigran Sogomonian]

Steps to reproduce:

Found by static analysis and code review of lib/softoken/jpakesftk.c (current master).

All three J-PAKE functions handle a PORT_NewArena() failure like this:

arena = PORT_NewArena(NSS_SOFTOKEN_DEFAULT_CHUNKSIZE);
if (arena == NULL)
    crv = CKR_HOST_MEMORY;          /* no return */

crv = sftk_MultipleAttribute2SecItem(arena, ...);   /* crv overwritten */

Actual results:

CKR_HOST_MEMORY is overwritten and the NULL arena is passed on:

• sftk_MultipleAttribute2SecItem() falls back to heap allocations for session objects (as does DSA_NewRandom() for x1 in round 1). These buffers, including secret values (x1, x2, x2s), are leaked without being zeroed.
• The next freebl call (JPAKE_Sign, JPAKE_Verify or JPAKE_Round2) rejects the NULL arena with SEC_ERROR_INVALID_ARGS, so the caller gets CKR_MECHANISM_PARAM_INVALID instead of CKR_HOST_MEMORY.
• In debug builds, round 1 hits PORT_Assert(arena != NULL) in jpake_Sign().

Expected results:

Each function should return CKR_HOST_MEMORY immediately when PORT_NewArena() fails, as the rest of lib/softoken does. Nothing is allocated before that point, so an early return is safe:

if (arena == NULL)
    return CKR_HOST_MEMORY;

Comment 1

[Tigran Sogomonian]

Attachment: Bug 2072633 - return CKR_HOST_MEMORY when PORT_NewArena fails in jpakesftk.c. r?#nss-reviewers

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/projects/nss/rev/ae4848691f24
return CKR_HOST_MEMORY when PORT_NewArena fails in jpakesftk.c. r=nss-reviewers,keeler