Обсуждение: [sqlsmith] Infinite recursion in bitshift

Hi,

sqlsmith just found another crasher:

    select bit '1' >> (-2^31)::int;

This is due to an integer overflow in bitshiftright()/bitshiftleft()
leading to them recursively calling each other.  Patch attached.

regards,
Andreas
>From cfdc425f75da268e1c2af08f936c59f34b69e577 Mon Sep 17 00:00:00 2001
From: Andreas Seltenreich <seltenreich@gmx.de>
Date: Fri, 14 Oct 2016 20:52:52 +0200
Subject: [PATCH] Fix possible infinite recursion on bitshift.

bitshiftright() and bitshiftleft() would recursively call each other
infinitely when the user passed a MIN_INT for the shift amount due to
an integer overflow.  This patch reduces a negative shift amount to
-VARBITMAXLEN to avoid overflow.
---
 src/backend/utils/adt/varbit.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/src/backend/utils/adt/varbit.c b/src/backend/utils/adt/varbit.c
index 75e6a46..d8ecfb6 100644
--- a/src/backend/utils/adt/varbit.c
+++ b/src/backend/utils/adt/varbit.c
@@ -1387,9 +1387,15 @@ bitshiftleft(PG_FUNCTION_ARGS)

     /* Negative shift is a shift to the right */
     if (shft < 0)
+    {
+        /* Protect against overflow */
+        if (shft < -VARBITMAXLEN)
+            shft = -VARBITMAXLEN;
+
         PG_RETURN_DATUM(DirectFunctionCall2(bitshiftright,
                                             VarBitPGetDatum(arg),
                                             Int32GetDatum(-shft)));
+    }

     result = (VarBit *) palloc(VARSIZE(arg));
     SET_VARSIZE(result, VARSIZE(arg));
@@ -1447,9 +1453,15 @@ bitshiftright(PG_FUNCTION_ARGS)

     /* Negative shift is a shift to the left */
     if (shft < 0)
+    {
+        /* Protect against overflow */
+        if (shft < -VARBITMAXLEN)
+            shft = -VARBITMAXLEN;
+
         PG_RETURN_DATUM(DirectFunctionCall2(bitshiftleft,
                                             VarBitPGetDatum(arg),
                                             Int32GetDatum(-shft)));
+    }

     result = (VarBit *) palloc(VARSIZE(arg));
     SET_VARSIZE(result, VARSIZE(arg));
--
2.9.3

Andreas Seltenreich <seltenreich@gmx.de> writes:
> sqlsmith just found another crasher:
>     select bit '1' >> (-2^31)::int;

Nice catch :-)

> This is due to an integer overflow in bitshiftright()/bitshiftleft()
> leading to them recursively calling each other.  Patch attached.

Seems sane, though I wonder if it'd be better to use -INT_MAX rather
than -VARBITMAXLEN.
        regards, tom lane

Tom Lane writes:

>> This is due to an integer overflow in bitshiftright()/bitshiftleft()
>> leading to them recursively calling each other.  Patch attached.
>
> Seems sane, though I wonder if it'd be better to use -INT_MAX rather
> than -VARBITMAXLEN.

I am undecided between those two.  -INT_MAX might be a more precise fix
for the problem, but the extra distance to the danger zone was kind of
soothing :-).

regards,
Andreas

Andreas Seltenreich <seltenreich@gmx.de> writes:
> Tom Lane writes:
>> Seems sane, though I wonder if it'd be better to use -INT_MAX rather
>> than -VARBITMAXLEN.

> I am undecided between those two.  -INT_MAX might be a more precise fix
> for the problem, but the extra distance to the danger zone was kind of
> soothing :-).

Yeah, might as well use the tighter limit.

Poking around in varbit.c, I noticed some other places that were assuming
that a typmod couldn't exceed VARBITMAXLEN.  anybit_typmodin() enforces
that, but there are places where a user can shove in an arbitrary integer,
eg

regression=# select "bit"(42, 2147483647);
ERROR:  invalid memory alloc request size 18446744073441116169

I fixed those too and pushed it.  Thanks for the report!
        regards, tom lane

On Fri, Oct 14, 2016 at 3:31 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Andreas Seltenreich <seltenreich@gmx.de> writes:
>> Tom Lane writes:
>>> Seems sane, though I wonder if it'd be better to use -INT_MAX rather
>>> than -VARBITMAXLEN.
>
>> I am undecided between those two.  -INT_MAX might be a more precise fix
>> for the problem, but the extra distance to the danger zone was kind of
>> soothing :-).
>
> Yeah, might as well use the tighter limit.
>
> Poking around in varbit.c, I noticed some other places that were assuming
> that a typmod couldn't exceed VARBITMAXLEN.  anybit_typmodin() enforces
> that, but there are places where a user can shove in an arbitrary integer,
> eg
>
> regression=# select "bit"(42, 2147483647);
> ERROR:  invalid memory alloc request size 18446744073441116169
>
> I fixed those too and pushed it.  Thanks for the report!

Curious -- are there real world scenarios where this would happen?

merlin

Merlin Moncure <mmoncure@gmail.com> writes:
> On Fri, Oct 14, 2016 at 3:31 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Poking around in varbit.c, I noticed some other places that were assuming
>> that a typmod couldn't exceed VARBITMAXLEN.

> Curious -- are there real world scenarios where this would happen?

I think you'd have to be intentionally trying to break it.  The largest
varbit typmod you're allowed to declare normally is only ~ 80 million.
        regards, tom lane