[Bug report] hash_name() may cross page boundary and trigger

[Xie Yuanbin]

Hi, Linus Torvalds and Will Deacon!

We have some discussion and solutions on other threads, and it seems
that there are somthing missing on this discussion thread. Therefore,
I think it is necessary to synchronize some information here.

1. There is a test case that can consistently reproduce the bug, which
might be helpful for us to do the test. The test case is located after
the '---' maker line in the following patch:
Link: https://lore.kernel.org/20251126101952.174467-1-xieyuanbin1@huawei.com

2. Al Viro give a suggest on 2025-11-26 19:26:
Link: https://lore.kernel.org/20251126192640.GD3538@ZenIV

This patch is similar to one I submitted long time ago, which was
intended fix another bug: missing branch predictor mitigation:
Link: https://lore.kernel.org/20250925025744.6807-1-xieyuanbin1@huawei.com

My patch was not accepted, Sebastian's patch:
Link: https://lore.kernel.org/20251110145555.2555055-2-bigeasy@linutronix.de
fixed this bug, but Sebastian's patch has not yet been merged into the
linux-next branch, so this bug still exists in the current linux-next
branch.

I hope there is a simple solution to fix both bugs, so I submitted this
patch on 2025-11-27 14:49:
Link: https://lore.kernel.org/20251127140109.191657-1-xieyuanbin1@huawei.com
This patch is based on the linux-next branch, therefore it does not
contain Sebastian's patch.

3. On 2025-11-28 17:06, Linus Torvalds provided a solution similar to
Al Viro's suggestion and my patch:
Link: https://lore.kernel.org/CAHk-=wh+cFLLi2x6u61pvL07phSyHPVBTo9Lac2uuqK4eRG_=w@mail.gmail.com

Currently, all solutions have been tested that can fix this one bug.
I still hold the view that perhaps there is a simpler way to fix another
bug at the same time, because the solutions of these two bugs are very
similar.

Thanks very much!

Xie Yuanbin