Re: [PATCH v9 00/22] Automatic Disk Unlock with TPM2

[Gary Lin]

On Mon, Jun 24, 2024 at 07:28:14PM +0200, Daniel Kiper wrote:
> On Thu, Mar 07, 2024 at 04:59:05PM +0800, Gary Lin via Grub-devel wrote:
> > On Thu, Feb 08, 2024 at 08:58:43PM +0100, Daniel Kiper wrote:
> > > Hey,
> > >
> > --8<--
> > >
> > > And I have attached the Coverity report. All issues reported there have
> > > to be fixed. If you cannot fix an issue you have to explain why you
> > > cannot do that and what is potential impact on the code stability,
> > > security, etc.
> > >
> > I have went through all the coverity issues. There are 6 issues in the
> > TPM2 stack and the utility:
> >
> > CID 435775, CID 435771, CID 435770, CID 435769, CID 435767, CID 435761
> >
> > Those will be fixed in the next version.
> >
> > The 9 issues are from libtasn1 and gnulib.
> >
> > There are two memory corruption issue: CID 435762 and CID 435766, and
> > I've filed the issues in libtasn1 upstream.
> >
> > - CID 435762
> >   https://gitlab.com/gnutls/libtasn1/-/issues/49
> >   This is a valid case. However, the only exploitable function,
> >   _asn1_insert_tag_der(), is disabled in grub2 patch, so the severity is
> >   low. I have a quick fix but upstream would like to fix it in another
> >   way.
> > - CID 435766
> >   https://gitlab.com/gnutls/libtasn1/-/issues/50
> >   IMO, this is false positive, but I need libtasn1 upstream to confirm
> >   that.
> >
> > Then, the remaining 7 Integer handling issues are involved with the macros
> > from gnulib, and I believe those are false positive.
> >
> > The following are my analyses:
> >
> > ________________________________________________________________________________________________________
> > *** CID 435774:  Integer handling issues  (CONSTANT_EXPRESSION_RESULT)
> > /grub-core/lib/libtasn1/lib/decoding.c: 481 in asn1_get_object_id_der()
> > 475            */
> > 476           if (leading != 0 && der[len_len + k] == 0x80)
> > 477         return ASN1_DER_ERROR;
> > 478           leading = 0;
> > 479
> > 480           /* check for wrap around */
> > >>>     CID 435774:  Integer handling issues  (CONSTANT_EXPRESSION_RESULT)
> > >>>     "val < ((((1 ? 0 : val) - 1 < 0) ? ~((((1 ? 0 : val) + 1 << 62UL /* 
> > >>> sizeof (+val) * 8 - 2 */) - 1) * 2 + 1) : ((1 ? 0 : val) + 0)) >> 7)" 
> > >>> is always false regardless of the values of its operands. This occurs 
> > >>> as the second operand of "?:".
> > 481           if (INT_LEFT_SHIFT_OVERFLOW (val, 7))
> > 482         return ASN1_DER_ERROR;
> > 483
> > 484           val = val << 7;
> > 485           val |= der[len_len + k] & 0x7F;
> > 486
> >
> > Here are the related macros:
> >
> > #define EXPR_SIGNED(e) (_GL_INT_NEGATE_CONVERT (e, 1) < 0)
> >
> > #define _GL_INT_CONVERT(e, v) ((1 ? 0 : (e)) + (v))
> >
> > #define _GL_SIGNED_INT_MAXIMUM(e)                                       \
> >   (((_GL_INT_CONVERT (e, 1) << (TYPE_WIDTH (+ (e)) - 2)) - 1) * 2 + 1)
> >
> > #define _GL_INT_MINIMUM(e)                                              \
> >   (EXPR_SIGNED (e)                                                      \
> >    ? ~ _GL_SIGNED_INT_MAXIMUM (e)                                       \
> >    : _GL_INT_CONVERT (e, 0))
> >
> > #define INT_LEFT_SHIFT_OVERFLOW(a, b) \
> >   INT_LEFT_SHIFT_RANGE_OVERFLOW (a, b, \
> >                                  _GL_INT_MINIMUM (a), _GL_INT_MAXIMUM (a))
> >
> > #define INT_LEFT_SHIFT_RANGE_OVERFLOW(a, b, min, max)   \
> >   ((a) < 0                                              \
> >    ? (a) < (min) >> (b)                                 \
> >    : (max) >> (b) < (a))
> >
> > The statement in question is expanded "(a) < (min) >> (b)" of
> > INT_LEFT_SHIFT_RANGE_OVERFLOW. Since 'val' is 'uint64_t', '(a) < 0' is 
> > always
> > false, so the result of that statement doen't matter.
> 
> Something is missing and/or requires clarification/expansion here.
> AFAICT at least your description completely ignores "(max) >> (b) < (a)"
> expression result.
> 
Because Coverity only complained "(a) < (min) >> (b)".
"(max) >> (b) < (a)" does the right thing to check whether 'a' is
overflowed after the left shift.

> > ________________________________________________________________________________________________________
> > *** CID 435773:  Integer handling issues  (NO_EFFECT)
> > /grub-core/lib/libtasn1/lib/decoding.c: 439 in asn1_get_object_id_der()
> > 433         return ASN1_DER_ERROR;
> > 434
> > 435       val0 = 0;
> > 436
> > 437       for (k = 0; k < len; k++)
> > 438         {
> > >>>     CID 435773:  Integer handling issues  (NO_EFFECT)
> > >>>     This less-than-zero comparison of an unsigned value is never true. 
> > >>> "(1 ? 0UL : val0) - 1UL < 0UL".
> > 439           if (INT_LEFT_SHIFT_OVERFLOW (val0, 7))
> > 440         return ASN1_DER_ERROR;
> > 441
> > 442           val0 <<= 7;
> > 443           val0 |= der[len_len + k] & 0x7F;
> > 444           if (!(der[len_len + k] & 0x80))
> >
> > Here are the related macros:
> >
> > #define _GL_INT_NEGATE_CONVERT(e, v) ((1 ? 0 : (e)) - (v))
> > #define EXPR_SIGNED(e) (_GL_INT_NEGATE_CONVERT (e, 1) < 0)
> >
> > The statement in question is the expanded 'EXPR_SIGNED'. Since that macro is
> > to test whether the variable is signed or not. For 'uint64_t val0', it's
> > expected to be always false.
> >
> > ________________________________________________________________________________________________________
> > *** CID 435772:  Integer handling issues  (NO_EFFECT)
> > /grub-core/lib/libtasn1/lib/decoding.c: 204 in asn1_get_tag_der()
> > 198           /* Long form */
> > 199           punt = 1;
> > 200           ris = 0;
> > 201           while (punt < der_len && der[punt] & 128)
> > 202         {
> > 203
> > >>>     CID 435772:  Integer handling issues  (NO_EFFECT)
> > >>>     This less-than-zero comparison of an unsigned value is never true. 
> > >>> "(1 ? 0U : ((1 ? 0U : ris) + 128U)) - 1U < 0U".
> > 204           if (INT_MULTIPLY_OVERFLOW (ris, 128))
> > 205             return ASN1_DER_ERROR;
> > 206           ris *= 128;
> > 207
> > 208           if (INT_ADD_OVERFLOW (ris, ((unsigned) (der[punt] & 0x7F))))
> > 209             return ASN1_DER_ERROR;
> >
> > Here are the related macros:
> >
> > #define _GL_INT_CONVERT(e, v) ((1 ? 0 : (e)) + (v))
> > #define _GL_INT_NEGATE_CONVERT(e, v) ((1 ? 0 : (e)) - (v))
> > #define EXPR_SIGNED(e) (_GL_INT_NEGATE_CONVERT (e, 1) < 0)
> >
> > The statement in question is the expanded 'EXPR_SIGNED(_GL_INT_CONVERT 
> > (ris, 128))'.
> > Since 'ris' is 'unsigned int', it's expected that 'EXPR_SIGNED' always
> > returns false.
> 
> At least INT_MULTIPLY_OVERFLOW definitions is missing.
> 
I got to admit that lots of details were skipped since the expansion of
INT_MULTIPLY_OVERFLOW is really lengthy.

I'll add the details back in my v18 patchset cover letter.

Gary Lin

> Please double check other analysis too. If you provide correct analysis
> for all non-fixable CIDs I will mark them as ignored. It would be nice
> if updated analysis is a part of latest version of the patch set.