Re: walk_pgd_range BUG: unable to handle page fault

[Tytus Rogalewski <tytanick@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1545 bytes --]

Hi,

Well it seemed "like mem leak" but honestly it was my bad assumption i
guess.

And does it correlate with io_uring somehow ?
Because i have no such issues when using threads mode in proxmox with
network drive that is on fuse.
Could be that triggered perhaps when someone was writing hard to his disk
and that would cause other fuse drives to hung for few secs and that would
trigger that BUG ?
And using threads somehow fixes that issue ? Does that make any sense ? I
am not familiar with low level memory management and io_uring too.



--

tel. 790 202 300

*Tytus Rogalewski*

Dolina Krzemowa 6A

83-010 Jagatowo

NIP: 9570976234


pon., 9 mar 2026 o 21:27 David Hildenbrand (Arm) <david@kernel.org>
napisał(a):

> On 3/5/26 12:40, David Hildenbrand (Arm) wrote:
> > On 3/5/26 12:39, Tytus Rogalewski wrote:
> >> I had this problem on 6.19-rc4 yes. I saw it. But i cant reproduce it
> >> NOW - this is what i ment  :)
> >
> > Ah, thanks for clarifying, makes sense!
> >
>
> The following issue might explain the problem:
>
> https://lore.kernel.org/r/20260309174949.2514565-1-mboone@akamai.com
>
> It would be related to what I recently said: "vfio recently gained
> support for installing huge mappings into user page tables."
>
> smaps_walk_ops() does not have a pud_entry() callback, so we'd try to
> split (unmap). If that happens concurrently with a refault, we might be
> able to trigger a BUG.
>
>
> But it wouldn't explain any memory leak.
>
> --
> Cheers,
>
> David
>

[-- Attachment #2: Type: text/html, Size: 3278 bytes --]