XPost: linux.debian.kernel

Control: tags -1 moreinfo

On Thu, Jul 24, 2025 at 12:45:39AM +0900, Simon Richter wrote:

this is a set of backported patches to make the Xe driver in the stable kernel not Oops when run on a machine with a page size that isn't 4k.

I see that patches on stable@, we won't take patches that can go this
way usually.

Bastian

--
Not one hundred percent efficient, of course ... but nothing ever is.
-- Kirk, "Metamorphosis", stardate 3219.8

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

XPost: linux.debian.kernel

Hi,

On 7/24/25 00:59, Bastian Blank wrote:

this is a set of backported patches to make the Xe driver in the stable
kernel not Oops when run on a machine with a page size that isn't 4k.

I see that patches on stable@, we won't take patches that can go this
way usually.

The full patchset[1] for later kernels has a bunch of extra changes, but
the stable kernel is old enough that we can get by with a number of
simple substitutions:
- replace SZ_4K with PAGE_SIZE when finding an alignment palatable to
the CPU
- replace PAGE_SIZE with SZ_4K when describing a memory layout in GPU
memory that uses 4k alignment regardless of host page size.

So on 4k, where PAGE_SIZE *is* SZ_4K, this does nothing, but it fixes
the Oops on boot for 16k (aarch64, loong64) and 64k (ppc64le) users.

Since these are derived from the proper fix, you can easily see when
they are no longer required.

Simon

[1] https://lore.kernel.org/all/[email protected]/

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

XPost: linux.debian.kernel

--3521eccfac821a72d823baf95196e96e1b69bea7d6466edb5cada90ea00a Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset=UTF-8

Control: Forwarded -1 https://lore.kernel.org/all/[email protected]/

On Wed Jul 23, 2025 at 5:45 PM CEST, Simon Richter wrote:

Package: src:linux
Version: 6.12.35-1
Severity: normal
Tags: patch
X-Debbugs-Cc: [email protected], [email protected], [email protected]

this is a set of backported patches to make the Xe driver in the stable kernel not Oops when run on a machine with a page size that isn't 4k.

You already did the right thing: send those fixes upstream

You could've mentioned that ...


--3521eccfac821a72d823baf95196e96e1b69bea7d6466edb5cada90ea00a
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----

iHUEABYKAB0WIQT1sUPBYsyGmi4usy/XblvOeH7bbgUCaIE59QAKCRDXblvOeH7b bqcCAQDygoktGAy1RMBkYUnObMPdM70RClRV1NUT+dcGaGhxgQD+IPvM03hRIMbF 2z83ZKm7roBYhl33RfTMgR9medhmFQg!uT
-----END PGP SIGNATURE-----

--3521eccfac821a72d823baf95196e96e1b69bea7d6466edb5cada90ea00a--

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

XPost: linux.debian.kernel

Hi,

On 7/24/25 04:37, Diederik de Haas wrote:

this is a set of backported patches to make the Xe driver in the stable
kernel not Oops when run on a machine with a page size that isn't 4k.

You already did the right thing: send those fixes upstream

They're not my fixes, all I did upstream was retrigger CI because it
showed a conflict.

But yes, these patches are already queued upstream (although they are
larger there, because some of the places near the "FIXME" actually got
fixed in the meantime). Sorry for not making that clear.

Simon

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

XPost: linux.debian.kernel

On Thu, 2025-07-24 at 10:39 +0900, Simon Richter wrote:

Hi,

On 7/24/25 04:37, Diederik de Haas wrote:

this is a set of backported patches to make the Xe driver in the stable kernel not Oops when run on a machine with a page size that isn't 4k.

You already did the right thing: send those fixes upstream

They're not my fixes, all I did upstream was retrigger CI because it
showed a conflict.

But yes, these patches are already queued upstream (although they are
larger there, because some of the places near the "FIXME" actually got
fixed in the meantime). Sorry for not making that clear.

I don't see them in drm-misc. Once the corresponding patches are
applied upstream, please send the backported patches here with proper
headers:

- "Bug-Debian" header pointing to this bug report
- "Origin" header pointing to the source of the backport (Git forge or
mail archive)
- Include the upstream commit messages and commit hashes in the commit
meesage, in addition to the current notes about backporting

Ben.

--
Ben Hutchings
Reality is just a crutch for people who can't handle science fiction.

-----BEGIN PGP SIGNATURE-----

iQIzBAABCgAdFiEErCspvTSmr92z9o8157/I7JWGEQkFAmiCkC0ACgkQ57/I7JWG EQmkEQ//Zf6GgFMmmtyPd5R5GXNOl+HGHeTXedP03Pptv4cFyidtugy0wiOvUH8K eKVzGCZi67olLI4oa/9tRn6tYz3T9DSvrzvGA59pLKtAvrROgbB0RsXPkFCH05qb eVtF+YJGE8IumjSFoLPVURTqDQquc83XxhesYaSAHVhwmsIRUS164emXl8fqhhei WTzEnU3U4Km5ZUyczVCw6cb4dWNsH1BgFT1pKR9oKBzM37QurwcNaPd3hfOxt9VQ ylyxEVKB7rZWZwsNAvdmv4tj/JOvYoiDD4U37AzvUDHTkDiRhZXUkvbGKSBDyGEW 6KBe87ArPijmdSZbwhf/vChxCIUvJMYB5xXZCUaKQpKSPqlmJELsrErP3NXhGfAa JY0r6IOfQ1X50qJJj15lOBg6gOef22CLvgxR3p3gg++nF5XZoh5e/Xoj2XWNsFyK lYabI9nqatd/jelEtYa4dzcwyWhJ58/h7Dka+N/ZE56cDqGoLfDp0xRDUcCb6fQo 9/lJvacI6/8n+Mz5S+oswH1MWU5PRMhJkVO2T7y2fjbBfEfJxg7PjwWWbLJ/6eFP 5uWeoJfG2SoNnRFX5HB0XYTE7acjyt0cMBReSJBcZyBxw+wvof6N6pYptAbR3DwS QpYCQZhtjTGHLWBAvaWzOWxMd6egX9WcD5iVR+p+j4qTfwnRckk=
=VS5C
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)