[RFC PATCH] vfs: Fix might sleep in load_unaligned_zeropad() with rcu read lock held

[Xie Yuanbin]

On, Thu, 27 Nov 2025 08:20:57 +0100, Sebastian Andrzej Siewior wrote:
> This all should be covered by the series here
> 	https://lore.kernel.org/all/20251110145555.2555055-1-bigeasy@linutronix.de/

Yes, I know it.

> or do I miss something.

We had some discussions about this bug:
Link: https://lore.kernel.org/lkml/20251126090505.3057219-1-wozizhi@huaweicloud.com/

The discussions:
Link: https://lore.kernel.org/CAHk-=wh1Wfwt9OFB4AfBbjyeu4JVZuSWQ4A8OoT3W6x9btddfw@mail.gmail.com
Link: https://lore.kernel.org/20251126192640.GD3538@ZenIV
Link: https://lore.kernel.org/aSeNtFxD1WRjFaiR@shell.armlinux.org.uk

According to the discussion, in do_page_fault(), when addr >= TASK_SIZE,
we should not try to acquire the mm read lock or find vma. Instead, we
should directly call __do_kernel_fault() or __do_user_fault(). Your
submission just moved harden_branch_predictor() forward. I think we can
have more discussions about the patches to fix the missing spectre.

I am trying to write a new patch, I hope it will better handle these two
bugs and be compatible with PREEMPT_RT scenarios.

> Sebastian

Thanks!

Xie Yuanbin