Repository navigation
BUG: PyArray_BufferConverter is unsafe #15450
Description
This function:
- Calls
PyObject_GetBuffer - Extracts a data pointer
- Calls
PyBuffer_Release - Returns the pointer to the caller
PyBuffer_Release calls PyBufferProcs.bf_releasebuffer(PyObject *exporter, Py_buffer *view), which according to the docs may "free all memory associated with view", and gives no requirement that view remains alive as long as exporter.
So in principle a type that allocates a buffer for itself on the fly in bf_getbuffer and deletes it when the last uses releases it will cause a use-after-free in numpy.
I don't know if any implementers of the buffer protocol actually do this, but my reading of it is that they are permitted to.
Activity
seberg commentedon Jan 27, 2020
This is strange. I think I would like if the buffer interface specified that while the buffer struct fields (such as strides, etc.) get free'd by
PyBuffer_Releasethe actual memory pointed to must be owned by the original object (this would solve our problem with the very annoying_dealloc_cached_buffer_info). I think this is the intention and true, but...Now this function returns a
PyArray_Chunkand not a buffer, so if you make the clarification that I say above, it is actually completely fine.eric-wieser commentedon Jan 27, 2020
Your interpretation sounds more reasonable - perhaps a pull request against python/cpython would be wise, to update the docs to tell people not to implement what I described above.
seberg commentedon Jan 27, 2020
Was going to open, but then went to lunch first: https://bugs.python.org/issue39471 I may look into proposing actually changes to the text. If we can agree on my interpretation, we can clean up very ugly and very slow code in the buffer protocol implementation (it currently slows down scalar math by 20+%). (EDIT: But I suppose only after Python fixes their ArgParse code :()
eric-wieser commentedon Jan 27, 2020
That adds some clarity, thanks - I hadn't realized that python was using the same reading as
bf_releasebufferas I was.seberg commentedon Jan 27, 2020
Damn, I had overread that before (from the PEP):
Which reads like it is valid to give the buffer a new chunk of memory :(.
Although: looking at the implementation of
array(python one), all it does is use it to ensure that it disables resizing if a buffer to it exists.