* Bert Karwatzki <spasswolf@xxxxxx> [241016 06:29]:
> Am Montag, dem 14.10.2024 um 10:46 +0100 schrieb Lorenzo Stoakes:
> > On Mon, Oct 14, 2024 at 12:35:59AM +0200, Bert Karwatzki wrote:
> > > I created a program which can trigger the bug on newer kernel (after the
> > > "Avoid zeroing vma tree in mmap_region()" patch and before the fix).
> > > My original goal was to trigger the bug on older kernels,
> > > but that does not work, yet.
> > >
> > > Bert Karwatzki
> >
> > Thanks, that's great!
> >
> > For older kernels the problem should still be present, the fundamental
> > thing that changed from the point of view of this bug is that merge won't
> > contribute to the number of VMAs being overwritten at once.
> >
> > To trigger prior to commit f8d112a4e657 ("mm/mmap: avoid zeroing vma tree
> > in mmap_region()") you would need to create a situation where the _clear_
> > triggers the bug, i.e. you must consistute all the VMAs that are being
> > overwritten by the store from existing VMAs you are overwriting with a
> > MAP_FIXED.
> >
> > So some tweaks should get you there...
> > >
>
> I don't think triggering the bug on a clear works, because a write of a %NULL
> that will cause a node to end with a %NULL becomes a spanning write into the
> next node:
I think there is room for improvement on this algorithm and have been
debating changing how this works.
>
> /*
> * mas_is_span_wr() - Check if the write needs to be treated as a write that
> * spans the node.
> * @mas: The maple state
> * @piv: The pivot value being written
> * @type: The maple node type
> * @entry: The data to write
> *
> * Spanning writes are writes that start in one node and end in another OR if
> * the write of a %NULL will cause the node to end with a %NULL.
> *
> * Return: True if this is a spanning write, false otherwise.
> */
> static bool mas_is_span_wr(struct ma_wr_state *wr_mas)
> {
>
>
> I think the could would trigger in this situation
>
> Node_0
> /
> /
> Node_1
> / \
> / \
> Node_2 Node_3
>
> but only if Node_3 contained only two ranges, one empty range and one normal
> range, and if the mmap into empty range of Node_3 would merge with the last
> range of Node_2 and the last range of Node_3. But I think the rebalancing of the
> tree will make it very hard if not impossible to create such a node.
I can confirm it is impossible to create a node with two entries that is
not a root node (if the root is a leaf node then there isn't a lower
bound to the entry count, if it isn't a leaf then the lower limit is 2
like other b-trees).
