On Tue, Jul 14, 2020 at 02:24:18PM +0300, Boaz Harrosh wrote:
> Matthew Hi
>
> First I want to thank you for the great xarray tool. I use it heavily with great joy & ease
>
> However I have encountered a bug in my code which I did not expect, as follows:
>
> I need code in the very hot-path that is looping on the xarray in an unusual way.
> What I need is to scan a range from x-x+l but I need to break on first "hole" ie.
> first entry that was not __xa_store() to. So I am using this loop:
> rcu_read_lock();
>
> for (xae = xas_load(&xas); xae; xae = xas_next(&xas)) {
> ...
> }
>
> Every thing works fine and I usually get a NULL from xas_next() (or xas_load())
> on first hole, And the loop exits.
>
> But in the case that I have entered a *single* xa_store() *at index 0*, but then try
> to GET a range 0-Y I get these unexpected results:
> xas_next() will return the same entry repeatedly
I thought this was fixed in commit 91abab83839aa2eba073e4a63c729832fdb27ea1
Do you have that commit in your tree?
