On Sun, 2007-12-16 at 09:05 -0800, Greg KH wrote: > On Sun, Dec 16, 2007 at 05:37:59PM +0100, Mattias Nissler wrote: > > This makes debugfs use its own file_operations for the value accessor files > > created by debugfs_create_XXX. Having that, we can also have proper versions > > for signed integers. > > Why not tweak the "SIMPLE_ATTRIBUTE" code to support this instead? That > way debugfs and all other filesystems could also use these attributes?
I expected that question ;-) Actually, I had a version that did this. But it ended up being ugly, cause SIMPLE_ATTRIBUTEs expect to access the values via get/set functions using u64. Here is what I did and why I didn't like it: * I made the set/get function interface more generic (as it is in the debugfs patch), i.e. getting passed the buf, so they have to decode/encode whatever they like into the buf instead of working with u64 * This means either all code using SIMPLE_ATTRIBUTES must be changed, or I need to add another wrapper macro that produces suitable get/set functions from the u64 versions it gets passed. * Changing all SIMPLE_ATTRIBUTE users is not what I want, cause moving the csnprintf()/strtoull() calls out of the SIMPLE_ATTRIBUTE code means it becomes less simple to use it, right? * The wrapper is actually possible, but it suffers from cases where somebody passes NULL for a get or set function (which is actually expected, compare the code in libfs.c) It is possible to hack around this, but it's ugly. Then I thought, well, SIMPLE_ATTRIBUTEs are made to work with a set/get function pair. But debugfs expects a pointer to an actual variable anyway, so just bypass the SIMPLE_ATTRIBUTE code and make debugfs use it's own file_operations (most of them shamelessly stolen from the SIMPLE_ATTRIBUTE code, I admit). Mattias -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
