#7928 closed Bugs (fixed)
boost::interprocess::shared_memory_object creation cause infinite loop
Reported by: | Owned by: | Ion Gaztañaga | |
---|---|---|---|
Milestone: | To Be Determined | Component: | interprocess |
Version: | Boost 1.52.0 | Severity: | Showstopper |
Keywords: | Cc: |
Description
Call of boost::interprocess::shared_memory_object creation cause infinite loop in case there is: boost::interprocess::shared_memory_object shm ( boost::interprocess::open_or_create, "some-name", boost::interprocess::read_write, 256); I dig inside of code and it happen in file:
boost/interprocess/shared_memory_object.hpp
this function :
inline bool shared_memory_object::priv_open_or_create
(ipcdetail::create_enum_t type,
const char *filename, mode_t mode, const permissions &perm)
{ .....
in switch/case: .... case ipcdetail::DoOpenOrCreate:
{
oflag |= O_CREAT; We need a loop to change permissions correctly using fchmod, since with "O_CREAT only" shm_open we don't know if we've created or opened the file. while(1){
m_handle = shm_open(m_filename.c_str(), oflag, unix_perm); if(m_handle >= 0){
::fchmod(m_handle, unix_perm); break;
} else if(errno == EEXIST){
if((m_handle = shm_open(m_filename.c_str(), oflag, unix_perm)) >= 0 errno != ENOENT){ break;
}
}
}
}
... as you can see there is a while and this while will remain infinite in case of any other errno then: "EEXIST"
... there are 9 different errors possible in case of shm_open() call, so they should be handled, in my case it was EACCES what brick my testing.
I'm testing it on 1.50 but saw 1.52 have the same code .....
Reagards
Ladislav Sopko lsopko@…
Change History (7)
comment:1 by , 10 years ago
comment:7 by , 10 years ago
I'm afraid it won't be out until Boost 1.54. We have not release date yet for that version.
Thanks for the report. I'll change the loop so that retries are only done if creation exactly returns EEXIST and opening returns ENOENT.