[Boost-users] Managed shared memory size restrictions

Subject: [Boost-users] Managed shared memory size restrictions
From: Veranika Liaukevich (veronika.levkevich_at_[hidden])
Date: 2015-05-14 08:39:20

Hello everybody,

Is there any restriction to the size of managed shared memory regions?
Should it be in multiples of the page size? I did not manage to find any
relevant information in the documentation, but managed_shared_memory
creation and managed_shared_memory::grow sometimes fail with small
memory sizes specified.

Examples:
1) When I create a managed_shared_memory segment with size 40, the
creation fails with interprocess_exception::library_error.

2) I created managed_shared_memory with initial size 1000, then
increased the size by 24 bytes, I got this:

/usr/include/boost/interprocess/mem_algo/rbtree_best_fit.hpp:514: void
boost::interprocess::rbtree_best_fit<MutexFamily, VoidMutex,
MemAlignment>::grow(boost::interprocess::rbtree_best_fit<MutexFamily,
VoidMutex, MemAlignment>::size_type) [with MutexFamily =
boost::interprocess::mutex_family; VoidPointer =
boost::interprocess::offset_ptr<void>; long unsigned int MemAlignment =
0ul; boost::interprocess::rbtree_best_fit<MutexFamily, VoidMutex,
MemAlignment>::size_type = long unsigned int]: Assertion
`new_block->m_size >= BlockCtrlUnits' failed.
Aborted (core dumped)

I use boost 1.55.0 on Ubuntu 14.04 64bit.

Code used for testing:

#include <boost/interprocess/managed_shared_memory.hpp>
const char *shmname = "MyBoostShmTest";
// Usage: ./progname [init-memory-size] [grow-memory-size]

int main(int argc, char **argv) {
    using namespace boost::interprocess;

    struct shm_remove {
        shm_remove() { shared_memory_object::remove(shmname); }
        ~shm_remove(){ shared_memory_object::remove(shmname); }
    } remover;

    size_t size = atoi(argv[1]);
    size_t grow_size = atoi(argv[2]);

    {
        managed_shared_memory segment(create_only, shmname, size);
    }
    {
        managed_shared_memory::grow(shmname, grow_size);
    }

    return 0;
}

Thanks,
Veranika