[Bug stdio/26211] New: printf integer overflow calculating allocation size

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[Bug stdio/26211] New: printf integer overflow calculating allocation size

Sourceware - glibc-bugs mailing list
https://sourceware.org/bugzilla/show_bug.cgi?id=26211

            Bug ID: 26211
           Summary: printf integer overflow calculating allocation size
           Product: glibc
           Version: 2.32
            Status: NEW
          Severity: normal
          Priority: P2
         Component: stdio
          Assignee: unassigned at sourceware dot org
          Reporter: jsm28 at gcc dot gnu.org
  Target Milestone: ---

The following test segfaults on a 64-bit system (haven't tested 32-bit).  (Any
precision less than 0x7fffffe0 is OK; 0x7fffffe0 through 0x7fffffff result in
the crash.)

#include <stdio.h>
#include <stdlib.h>

int
main (void)
{
  FILE *fp;
  if ((fp = fopen ("/dev/null", "w")) == NULL)
    exit (1);
  fprintf (fp, "%2$.*1$a", 0x7fffffff, 1e200);
}

The problem is the code in printf_positional:

      /* Maybe the buffer is too small.  */
      if (MAX (prec, width) + EXTSIZ > WORK_BUFFER_SIZE)
        {
          if (__libc_use_alloca ((MAX (prec, width) + EXTSIZ)
                                 * sizeof (CHAR_T)))
            workend = ((CHAR_T *) alloca ((MAX (prec, width) + EXTSIZ)
                                          * sizeof (CHAR_T))
                       + (MAX (prec, width) + EXTSIZ));
          else
            {
              workstart = (CHAR_T *) malloc ((MAX (prec, width) + EXTSIZ)
                                             * sizeof (CHAR_T));
              if (workstart == NULL)
                {
                  done = -1;
                  goto all_done;
                }
              workend = workstart + (MAX (prec, width) + EXTSIZ);
            }
        }

The code in the implementation that doesn't use '$' argument numbers has a
check

        if (__glibc_unlikely (width >= INT_MAX / sizeof (CHAR_T) - EXTSIZ))
          {
            __set_errno (EOVERFLOW);
            done = -1;
            goto all_done;
          }

(and likewise for precision). The implementation for the case with '$' argument
numbers is missing that check. So MAX (prec, width) + EXTSIZ overflows.

--
You are receiving this mail because:
You are on the CC list for the bug.
Reply | Threaded
Open this post in threaded view
|

[Bug stdio/26211] printf integer overflow calculating allocation size

Sourceware - glibc-bugs mailing list
https://sourceware.org/bugzilla/show_bug.cgi?id=26211

--- Comment #1 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Joseph Myers <[hidden email]>:

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=6caddd34bd7ffb5ac4f36c8e036eee100c2cc535

commit 6caddd34bd7ffb5ac4f36c8e036eee100c2cc535
Author: Joseph Myers <[hidden email]>
Date:   Tue Jul 7 14:54:12 2020 +0000

    Remove most vfprintf width/precision-dependent allocations (bug 14231, bug
26211).

    The vfprintf implementation (used for all printf-family functions)
    contains complicated logic to allocate internal buffers of a size
    depending on the width and precision used for a format, using either
    malloc or alloca depending on that size, and with consequent checks
    for size overflow and allocation failure.

    As noted in bug 26211, the version of that logic used when '$' plus
    argument number formats are in use is missing the overflow checks,
    which can result in segfaults (quite possibly exploitable, I didn't
    try to work that out) when the width or precision is in the range
    0x7fffffe0 through 0x7fffffff (maybe smaller values as well in the
    wprintf case on 32-bit systems, when the multiplication by sizeof
    (CHAR_T) can overflow).

    All that complicated logic in fact appears to be useless.  As far as I
    can tell, there has been no need (outside the floating-point printf
    code, which does its own allocations) for allocations depending on
    width or precision since commit
    3e95f6602b226e0de06aaff686dc47b282d7cc16 ("Remove limitation on size
    of precision for integers", Sun Sep 12 21:23:32 1999 +0000).  Thus,
    this patch removes that logic completely, thereby fixing both problems
    with excessive allocations for large width and precision for
    non-floating-point formats, and the problem with missing overflow
    checks with such allocations.  Note that this does have the
    consequence that width and precision up to INT_MAX are now allowed
    where previously INT_MAX / sizeof (CHAR_T) - EXTSIZ or more would have
    been rejected, so could potentially expose any other overflows where
    the value would previously have been rejected by those removed checks.

    I believe this completely fixes bugs 14231 and 26211.

    Excessive allocations are still possible in the floating-point case
    (bug 21127), as are other integer or buffer overflows (see bug 26201).
    This does not address the cases where a precision larger than INT_MAX
    (embedded in the format string) would be meaningful without printf's
    return value overflowing (when it's used with a string format, or %g
    without the '#' flag, so the actual output will be much smaller), as
    mentioned in bug 17829 comment 8; using size_t internally for
    precision to handle that case would be complicated by struct
    printf_info being a public ABI.  Nor does it address the matter of an
    INT_MIN width being negated (bug 17829 comment 7; the same logic
    appears a second time in the file as well, in the form of multiplying
    by -1).  There may be other sources of memory allocations with malloc
    in printf functions as well (bug 24988, bug 16060).  From inspection,
    I think there are also integer overflows in two copies of "if ((width
    -= len) < 0)" logic (where width is int, len is size_t and a very long
    string could result in spurious padding being output on a 32-bit
    system before printf overflows the count of output characters).

    Tested for x86-64 and x86.

--
You are receiving this mail because:
You are on the CC list for the bug.
Reply | Threaded
Open this post in threaded view
|

[Bug stdio/26211] printf integer overflow calculating allocation size

Sourceware - glibc-bugs mailing list
In reply to this post by Sourceware - glibc-bugs mailing list
https://sourceware.org/bugzilla/show_bug.cgi?id=26211

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |2.32
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #2 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
Fixed for 2.32.

commit 6caddd34bd7ffb5ac4f36c8e036eee100c2cc535
Author: Joseph Myers <[hidden email]>
Date:   Tue Jul 7 14:54:12 2020 +0000

    Remove most vfprintf width/precision-dependent allocations (bug 14231, bug
26211).

    The vfprintf implementation (used for all printf-family functions)
    contains complicated logic to allocate internal buffers of a size
    depending on the width and precision used for a format, using either
    malloc or alloca depending on that size, and with consequent checks
    for size overflow and allocation failure.

    As noted in bug 26211, the version of that logic used when '$' plus
    argument number formats are in use is missing the overflow checks,
    which can result in segfaults (quite possibly exploitable, I didn't
    try to work that out) when the width or precision is in the range
    0x7fffffe0 through 0x7fffffff (maybe smaller values as well in the
    wprintf case on 32-bit systems, when the multiplication by sizeof
    (CHAR_T) can overflow).

    All that complicated logic in fact appears to be useless.  As far as I
    can tell, there has been no need (outside the floating-point printf
    code, which does its own allocations) for allocations depending on
    width or precision since commit
    3e95f6602b226e0de06aaff686dc47b282d7cc16 ("Remove limitation on size
    of precision for integers", Sun Sep 12 21:23:32 1999 +0000).  Thus,
    this patch removes that logic completely, thereby fixing both problems
    with excessive allocations for large width and precision for
    non-floating-point formats, and the problem with missing overflow
    checks with such allocations.  Note that this does have the
    consequence that width and precision up to INT_MAX are now allowed
    where previously INT_MAX / sizeof (CHAR_T) - EXTSIZ or more would have
    been rejected, so could potentially expose any other overflows where
    the value would previously have been rejected by those removed checks.

    I believe this completely fixes bugs 14231 and 26211.

    Excessive allocations are still possible in the floating-point case
    (bug 21127), as are other integer or buffer overflows (see bug 26201).
    This does not address the cases where a precision larger than INT_MAX
    (embedded in the format string) would be meaningful without printf's
    return value overflowing (when it's used with a string format, or %g
    without the '#' flag, so the actual output will be much smaller), as
    mentioned in bug 17829 comment 8; using size_t internally for
    precision to handle that case would be complicated by struct
    printf_info being a public ABI.  Nor does it address the matter of an
    INT_MIN width being negated (bug 17829 comment 7; the same logic
    appears a second time in the file as well, in the form of multiplying
    by -1).  There may be other sources of memory allocations with malloc
    in printf functions as well (bug 24988, bug 16060).  From inspection,
    I think there are also integer overflows in two copies of "if ((width
    -= len) < 0)" logic (where width is int, len is size_t and a very long
    string could result in spurious padding being output on a 32-bit
    system before printf overflows the count of output characters).

    Tested for x86-64 and x86.

--
You are receiving this mail because:
You are on the CC list for the bug.