[Intel-gfx] [PATCH v5 1/3] drm: Use XArray instead of IDR for minors

[Matthew Wilcox]

On Sun, Nov 06, 2022 at 04:51:39PM +0200, Oded Gabbay wrote:
> I tried executing the following code in a dummy driver I wrote:

You don't need to write a dummy driver; you can write test-cases
entirely in userspace.  Just add them to lib/test_xarray.c and then
make -C tools/testing/radix-tree/

> static DEFINE_XARRAY_ALLOC(xa_dummy);
> void check_xa(void *pdev)
> {
>   void *entry;
>   int ret, index;
> 
>   ret = xa_alloc(&xa_dummy, &index, NULL, XA_LIMIT(0, 63), GFP_NOWAIT);
>   if (ret < 0)
>       return ret;
> 
>   entry = xa_cmpxchg(&xa_dummy, index, NULL, pdev, GFP_KERNEL);
>   if (xa_is_err(entry))
>        return;
> 
>   xa_lock(&xa_dummy);
>   xa_dev = xa_load(&xa_dummy, index);
>   xa_unlock(&xa_dummy);
> }
> 
> And to my surprise xa_dev is always NULL, when it should be pdev.
> I believe that because we first allocate the entry with NULL, it is
> considered a "zero" entry in the XA.
> And when we replace it, this attribute doesn't change so when we do
> xa_load, the xa code thinks the entry is a "zero" entry and returns
> NULL.

There's no "attribute" to mark a zero entry.  It's just a zero entry.
The problem is that you're cmpxchg'ing against NULL, and it's not NULL,
it's the ZERO entry.  This is even documented in the test code:

        /* cmpxchg sees a reserved entry as ZERO */
        XA_BUG_ON(xa, xa_reserve(xa, 12345678, GFP_KERNEL) != 0);
        XA_BUG_ON(xa, xa_cmpxchg(xa, 12345678, XA_ZERO_ENTRY,
                                xa_mk_value(12345678), GFP_NOWAIT) != NULL);

I'm not quite sure why you're using xa_cmpxchg() here anyway; if you
allocated it, it's yours and you can just xa_store() to it.