On Tue, Mar 16, 2010 at 12:35 PM, Hans Verkuil <hverkuil@xxxxxxxxx> wrote: > That is my opinion, yes. However, there is one case where this is actually > useful. Take for example the function videobuf_to_dma in > videobuf-dma-sg.c. This is called by drivers and it makes sense that that > function should double-check that the videobuf_buffer is associated with > the dma_sg memtype. > > But calling this 'magic' is a poor choice of name. There is nothing magic > about it, in this case it is just an identifier of the memtype. And there > may be better ways to do this check anyway. It's funny, because when I saw the word "magic", I knew *exactly* what it did. The notion of putting a magic value at the top of a structure is not that uncommon (although you don't see it in Linux much). I've seen it done many times over the years (all using the term "magic" which is why I knew what it did). The cases where it is really helpful is when you have lots of (void *) pointers in your buffer management code, as it lets you detect cases where a function gets passed the wrong buffer type (which wouldn't fail a compile time due to the void * pointer). And by using different magic values for different structure types, in many cases it will give you a hint where the problem is (because you would now know *what* type of structure got passed into the function). That said, these sorts of cases usually aren't usually intended to catch *random* memory corruption as they are to help isolate problems in buffer handling code. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- 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
