[Bug libstdc++/94906] New: memory corruption in std::pmr::memory_buffer_resource

[Bug libstdc++/94906] New: memory corruption in std::pmr::memory_buffer_resource

[arnaud02 at users dot sourceforge.net]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

            Bug ID: 94906
           Summary: memory corruption in std::pmr::memory_buffer_resource
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libstdc++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: arnaud02 at users dot sourceforge.net
  Target Milestone: ---

>cat pmr1.cpp
#include <array>
#include <memory_resource>
#include <unordered_map>
int main() {
  const std::size_t max_size = 75'000'000;
  std::pmr::monotonic_buffer_resource res;
  std::pmr::unordered_map<int, std::array<char,32>> counter{&res};
  for (size_t i = 0; i < max_size; ++i) counter[i];
}
>g++ -Wall -std=c++17 -O2 -g pmr1.cpp && ./a.out
*** Error in `./a.out': free(): invalid pointer: 0x00002b06379d000f ***
======= Backtrace: =========
/lib64/libc.so.6(+0x81299)[0x2b03c8a6f299]
opt/gcc/gcc-930-rh7/lib64/libstdc++.so.6(_ZNSt3pmr25monotonic_buffer_resource18_M_release_buffersEv+0x48)[0x2b03c825fe78]
opt/gcc/gcc-930-rh7/lib64/libstdc++.so.6(_ZNSt3pmr25monotonic_buffer_resourceD2Ev+0x1e)[0x2b03c825febe]
./a.out[0x400d51]
/lib64/libc.so.6(__libc_start_main+0xf5)[0x2b03c8a10555]
./a.out[0x400eb4]


In monotonic_buffer_resource::_Chunk::release, "__size" is not calculated
properly when _Chunk::_M_size is 32 or more. Here is a candidate patch:

--- gcc-9.3.0/libstdc++-v3/src/c++17/memory_resource.cc 2020-03-12
11:07:24.000000000 +0000
+++ gcc-9.3.0/libstdc++-v3/src/c++17/memory_resource.cc.new     2020-05-01
18:31:24.367672036 +0100
@@ -228,7 +228,7 @@
          if (__ch->_M_canary != (__ch->_M_size | __ch->_M_align))
            return; // buffer overflow detected!

-         size_t __size = (1u << __ch->_M_size);
+         size_t __size = (1lu << __ch->_M_size);
          size_t __align = (1u << __ch->_M_align);
          void* __start = (char*)(__ch + 1) - __size;
          __r->deallocate(__start, __size, __align);

[Bug libstdc++/94906] memory corruption in std::pmr::memory_buffer_resource

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-05-01
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |redi at gcc dot gnu.org
           Keywords|                            |wrong-code
     Ever confirmed|0                           |1

[Bug libstdc++/94906] memory corruption in std::pmr::memory_buffer_resource

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Better would be:
size_t __size = (((size_t)1) << __ch->_M_size);

As long could be 32bits (windows)

[Bug libstdc++/94906] memory corruption in std::pmr::monotonic_buffer_resource

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jonathan Wakely <redi@gcc.gnu.org>:

https://gcc.gnu.org/g:bb27781b64162e1769df15e0c97e8d2145d2d10d

commit r11-53-gbb27781b64162e1769df15e0c97e8d2145d2d10d
Author: Jonathan Wakely <jwakely@redhat.com>
Date:   Mon May 4 21:13:28 2020 +0100

    libstdc++: Fix incorrect size calculation in PMR resource  (PR 94906)

    Calculating the size of a chunk being returned to the upstream allocator
    was done with a 32-bit type, so it wrapped if the chunk was 4GB or
    larger.

    I don't know how to test this without allocating 4GB, so there's no test
    in the testsuite. It has been tested manually with allocations sizes and
    alignments exceeding 4GB.

            PR libstdc++/94906
            * src/c++17/memory_resource.cc
            (monotonic_buffer_resource::_Chunk::release): Use size_t for shift
            operands.

[Bug libstdc++/94906] memory corruption in std::pmr::monotonic_buffer_resource

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |9.4

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Thanks for the report and identifying the source of the bug.

Fixed on master only for now, but backports will come soon.

[Bug libstdc++/94906] memory corruption in std::pmr::monotonic_buffer_resource

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Jonathan Wakely
<redi@gcc.gnu.org>:

https://gcc.gnu.org/g:dc103060c18656affaecfdd57faa4e0237dadcd3

commit r10-8136-gdc103060c18656affaecfdd57faa4e0237dadcd3
Author: Jonathan Wakely <jwakely@redhat.com>
Date:   Tue May 12 09:54:44 2020 +0100

    libstdc++: Fix incorrect size calculation in PMR resource  (PR 94906)

    Calculating the size of a chunk being returned to the upstream allocator
    was done with a 32-bit type, so it wrapped if the chunk was 4GB or
    larger.

    I don't know how to test this without allocating 4GB, so there's no test
    in the testsuite. It has been tested manually of course.

    Backport from mainline
    2020-05-04  Jonathan Wakely  <jwakely@redhat.com>

            PR libstdc++/94906
            * src/c++17/memory_resource.cc
            (monotonic_buffer_resource::_Chunk::release): Use size_t for shift
            operands.

[Bug libstdc++/94906] memory corruption in std::pmr::monotonic_buffer_resource

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-9 branch has been updated by Jonathan Wakely
<redi@gcc.gnu.org>:

https://gcc.gnu.org/g:45a6686e76bfcd48f7c72a23d0e15186f76b4bfc

commit r9-8588-g45a6686e76bfcd48f7c72a23d0e15186f76b4bfc
Author: Jonathan Wakely <jwakely@redhat.com>
Date:   Tue May 12 09:54:44 2020 +0100

    libstdc++: Fix incorrect size calculation in PMR resource  (PR 94906)

    Calculating the size of a chunk being returned to the upstream allocator
    was done with a 32-bit type, so it wrapped if the chunk was 4GB or
    larger.

    I don't know how to test this without allocating 4GB, so there's no test
    in the testsuite. It has been tested manually of course.

    Backport from mainline
    2020-05-04  Jonathan Wakely  <jwakely@redhat.com>

            PR libstdc++/94906
            * src/c++17/memory_resource.cc
            (monotonic_buffer_resource::_Chunk::release): Use size_t for shift
            operands.

[Bug libstdc++/94906] memory corruption in std::pmr::monotonic_buffer_resource

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94906

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #6 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Fixed for 9.4 and 10.2