Re: [Qemu-devel] [PATCH 2/3 V3] s390: implement pci instructions

[Markus Armbruster]

Frank Blaschka <address@hidden> writes:

> On Tue, Jan 20, 2015 at 01:56:09PM +0100, Markus Armbruster wrote:
>> Markus Armbruster <address@hidden> writes:
>> 
>> > Cornelia Huck <address@hidden> writes:
>> >
>> >> On Tue, 20 Jan 2015 10:45:41 +0100
>> >> Markus Armbruster <address@hidden> wrote:
>> >>
>> >>> This patch makes Coverity unhappy:
>> >>> 
>> >>> *** CID 1264326:  Unintended sign extension  (SIGN_EXTENSION)
>> >>> /hw/s390x/s390-pci-inst.c: 787 in stpcifc_service_call()
>> >>> 781         stq_p(&fib.pal, pbdev->pal);
>> >>> 782         stq_p(&fib.iota, pbdev->g_iota);
>> >>> 783         stq_p(&fib.aibv, pbdev->routes.adapter.ind_addr);
>> >>> 784         stq_p(&fib.aisb, pbdev->routes.adapter.summary_addr);
>> >>> 785         stq_p(&fib.fmb_addr, pbdev->fmb_addr);
>> >>> 786     
>> >>> >>>     CID 1264326:  Unintended sign extension  (SIGN_EXTENSION)
>> >>> >>>     Suspicious implicit sign extension: "pbdev->isc" with type
>> >>> >>> "unsigned char" (8 bits, unsigned) is promoted in "(pbdev->isc <<
>> >>> >>> 28) | (pbdev->noi << 16)" to type "int" (32 bits, signed), then
>> >>> >>> sign-extended to type "unsigned long" (64 bits, unsigned).  If
>> >>> >>> "(pbdev->isc << 28) | (pbdev->noi << 16)" is greater than
>> >>> >>> 0x7FFFFFFF, the upper bits of the result will all be 1.
>> >>> 787         data = (pbdev->isc << 28) | (pbdev->noi << 16) |
>> >>> 788 (pbdev->routes.adapter.ind_offset << 8) | (pbdev->sum << 7) |
>> >>> 789                pbdev->routes.adapter.summary_offset;
>> >>> 790         stw_p(&fib.data, data);
>> >>> 791     
>> >>> 792         if (pbdev->fh >> ENABLE_BIT_OFFSET) {
>> >>
>> >> There's a fix for this (and the memory leak):
>> >>
>> >> http://marc.info/?l=qemu-devel&m=142124886620078&w=2
>> >>
>> >> The patch is sitting in my queue, will send with the next pile of s390x
>> >> updates.
>> >
>> > I can't see how
>> >
>> > @@ -787,7 +787,7 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, 
>> > uint64_t fiba)
>> >      data = (pbdev->isc << 28) | (pbdev->noi << 16) |
>> >             (pbdev->routes.adapter.ind_offset << 8) | (pbdev->sum << 7) |
>> >             pbdev->routes.adapter.summary_offset;
>> > -    stw_p(&fib.data, data);
>> > +    stl_p(&fib.data, data);
>> >  
>> >      if (pbdev->fh >> ENABLE_BIT_OFFSET) {
>> >          fib.fc |= 0x80;
>> >
>> > fixes the implicit sign extension within the assignment preceding it.
>> > Let me explain it again real slow:
>> >
>> > 1. pbdev->isc gets promoted from uint8_t to int as operand of binary <<
>> >    (usual arithmetic conversions ISO/IEC 9899:1999 6.3.1.8)
>> >
>> > 2. The int result is shifted left 28 bits.  This can set the MSB.
>> >
>> > 3. Likewise: pbdev->noi gets promoted from uint64_t to int, and shifted
>> >    left 16 bits.
> uint16_t to int

Yes, that's what I meant :)

>> >
>> > 4. The two shift results stay int and get ored.
>> >
>> > 5. pbdev->routes.adapter.ind_offset stays uint64_t, and is shifted left
>> >    8 bits.
>> >
>> > 6. The next or's left operand is the int result of 4 and the right
>> >    operant is the uint64_t result of 5.  Therefore, the left operand is
>> >    *sign-extended* from int to uint64_t.  This copies bit#7 of
>> >    pbdev->isc to bits#31..63.  Whoops.
>> 
>> I neglected to say: we don't currently use the upper 32 bits, and as
>> long as we do that, the sign extension is harmless.  I'd recommend to
>> avoid it all the same, for robustness, and to hush up Coverity.
>>
>
> Hi Markus,
>
> thx for your explanation. I did not see a problem since ISC is not bigger
> than 0x7 so MSB is never set. But the time I wrote the code I was not aware of
> ind_offset is uint64_t since zpci defines only a 6 bit field for this value.

Okay.

> How can I avoid the sign extension and make Coverity happy?

Casting pbdev->routes.adapter.ind_offset to uint32_t should do.  Then,
all operands of | are either int (promoted from narrower unsigned type)
or uint32_t (type cast).  Conversion from int to uint32_t won't
sign-extend as long as int is at least 32 bits.  Surely the case for
anything that can run QEMU.

>> > Regarding the leak, I prefer my patch, because it avoids the free on
>> > error.  But you're the maintainer.
>
> This is fine for me as well ...
>
> Thx,
>
> Frank
>>