Do not free a pointer from another pool #1161
Conversation
Signed-off-by: Lukasz Dorau <lukasz.dorau@intel.com>
|
|
||
| umf_result_t umfPoolFree(umf_memory_pool_handle_t hPool, void *ptr) { | ||
| UMF_CHECK((hPool != NULL), UMF_RESULT_ERROR_INVALID_ARGUMENT); | ||
| umf_memory_pool_handle_t poolTracker = umfMemoryTrackerGetPool(ptr); |
There was a problem hiding this comment.
with this call, we would search the tracker 2 times during the free?
There was a problem hiding this comment.
AFAIK in case of Proxy Pool and Disjoint Pool - yes
There was a problem hiding this comment.
so this could significantly hurt the performance. Could you run the benchmarks and ask @lplewa for interpretation of results?
There was a problem hiding this comment.
Is this problem specific to the disjoint and proxy pools? Does the issue exist with the scalable or with the jemalloc pools?
There was a problem hiding this comment.
I think it is theoretically possible for every pool.
There was a problem hiding this comment.
As @bratpiorka mentioned it might impact performance.
so it is obvious that if an incorrect pool handle is passed by the client it is an ill-formed client's code. The question is do we want to catch such issues (and probably pay for the performance impact) or we should just say it is "undefined behavior" if an incorrect pool handle is passed as a parameter?
There was a problem hiding this comment.
I vote for UB in such case.
|
Passing an incorrect pool handle to |
perhaps should we update the docs? |
4 participants
Description
Passing an incorrect pool handle to umfPoolFree() is an Undefined Behavior.
Checklist