Hi A recent report http://thread.gmane.org/gmane.linux.drivers.video-input-infrastructure/47594 has revealed the following asymmetry in how videobuf2 functions: as is also documented in videobuf2-core.h, the user's struct vb2_ops::buf_init() method in the MMAP case is called after allocating the respective buffer, which happens at REQBUFS time, in the USERPTR case it is called after acquiring a new buffer at QBUF time. If the allocation in MMAP case fails, the respective buffer simply doesn't get created. However, if acquiring a new USERPTR buffer at QBUF time fails, the buffer object remains on the queue, but the user-provided .buf_init() method is not called for it. When the queue is destroyed, the user's .buf_cleanup() method is called on an uninitialised buffer. This is exactly the reason for the BUG() in the above referenced report. Therefore my question: is this videobuf2-core behaviour really correct and we should be prepared in .buf_cleanup() to process uninitialised buffers, or should the videobuf2-core be adjusted? Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer http://www.open-technology.de/ -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
