[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <778e28df-3bb6-4428-b2ad-3355ddaa12e6@lucifer.local>
Date: Wed, 2 Oct 2024 22:13:44 +0100
From: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
To: Bert Karwatzki <spasswolf@....de>
Cc: "Liam R . Howlett" <Liam.Howlett@...cle.com>,
Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v8 14/21] mm/mmap: Avoid zeroing vma tree in mmap_region()
On Wed, Oct 02, 2024 at 09:44:42PM +0100, Lorenzo Stoakes wrote:
> On Wed, Oct 02, 2024 at 10:39:06PM GMT, Bert Karwatzki wrote:
>
> [snip]
>
> > I just figured out what was wrong for me in "Ruins of Tasos", I didn't have the
> > 32bit vulkan drivers installed, so perhaps if you remove your 32bit vulkan
> > driver package you can get the crash, too.
>
> Thanks, unfortunately I use arch and it has a hard dependency on that. I might
> try to force it uninstalled anyway.
>
> Sorry to have to keep asking you to repro but for now the only way sadly :(
>
> Hopefully this final patch that I set should give us the final bits of
> information we need, can't be sure though.
>
> This is certainly a very strange thing! I wonder whether a driver is
> relying on old mmap() MAP_FIXED behaviour...
I realise (being tired) I missed one thing... if you have a sec could you add
this to the previously included patch?
Thanks!
Rather important, as this is the main munmap() entry point...
Again, hugely appreciate all your help!
----8<----
>From 200301df3dbd1ca5b3ab2e9238af15602d72e839 Mon Sep 17 00:00:00 2001
From: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
Date: Wed, 2 Oct 2024 22:09:21 +0100
Subject: [PATCH] munmap fix
---
mm/mmap.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/mm/mmap.c b/mm/mmap.c
index 8a1d5c0da86f..68817f83263c 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1658,6 +1658,8 @@ static int __vm_munmap(unsigned long start, size_t len, bool unlock)
LIST_HEAD(uf);
VMA_ITERATOR(vmi, mm, start);
+ ljs_dump(mm, start, start + len, 0, true);
+
if (mmap_write_lock_killable(mm))
return -EINTR;
--
2.46.2
Powered by blists - more mailing lists