On Mon, Dec 22, 2025 at 02:03:55PM -0500, Rodrigo Vivi wrote:
> On Mon, Dec 22, 2025 at 06:50:06PM +0100, Michal Wajdeczko wrote:
> > On configs where Xe is built-in (DRM_XE=y) but configfs is defined
> > as a module (CONFIGFS=m), we were not enabling our configfs stubs,
> > which might lead to the following build errors:
> >
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_guc.o: in function
> > `xe_guc_init_post_hwconfig':
> > xe_guc.c:(.text+0x2f08): undefined reference to
> > `xe_configfs_get_psmi_enabled'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_hw_engine.o: in function
> > `xe_hw_engines_init_early':
> > xe_hw_engine.c:(.text+0xedc): undefined reference to
> > `xe_configfs_get_engines_allowed'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_lrc.o: in function
> > `setup_configfs_post_ctx_restore_bb':
> > xe_lrc.c:(.text+0xb30): undefined reference to
> > `xe_configfs_get_ctx_restore_post_bb'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_lrc.o: in function
> > `setup_configfs_mid_ctx_restore_bb':
> > xe_lrc.c:(.text+0xbc0): undefined reference to
> > `xe_configfs_get_ctx_restore_mid_bb'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_lrc.o: in function
> > `xe_lrc_init':
> > xe_lrc.c:(.text+0x32fc): undefined reference to
> > `xe_configfs_get_ctx_restore_mid_bb'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_module.o:(.data.rel.ro+0x10):
> > undefined reference to `xe_configfs_init'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_module.o:(.data.rel.ro+0x18):
> > undefined reference to `xe_configfs_exit'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_pci.o: in function
> > `xe_pci_probe':
> > xe_pci.c:(.text+0x1514): undefined reference to
> > `xe_configfs_check_device'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_psmi.o: in function
> > `xe_psmi_debugfs_register':
> > xe_psmi.c:(.text+0x508): undefined reference to
> > `xe_configfs_get_psmi_enabled'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_psmi.o: in function
> > `xe_psmi_init':
> > xe_psmi.c:(.text+0x5c4): undefined reference to
> > `xe_configfs_get_psmi_enabled'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_rtp.o: in function
> > `xe_rtp_match_psmi_enabled':
> > xe_rtp.c:(.text+0xba0): undefined reference to
> > `xe_configfs_get_psmi_enabled'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_survivability_mode.o: in
> > function `xe_survivability_mode_is_requested':
> > xe_survivability_mode.c:(.text+0x434): undefined reference to
> > `xe_configfs_get_survivability_mode'
> > powerpc64-linux-ld: drivers/gpu/drm/xe/xe_sriov_pf.o: in function
> > `xe_sriov_pf_readiness':
> > xe_sriov_pf.c:(.text+0x2a0): undefined reference to
> > `xe_configfs_get_max_vfs'
> >
> > Fix that by using IS_REACHABLE check instead of IS_ENABLED when
> > deciding whether to stub variant or not.
> >
> > Fixes: 16280ded45fb ("drm/xe: Add configfs to enable survivability mode")
> > Reported-by: kernel test robot <[email protected]>
> > Closes:
> > https://lore.kernel.org/oe-kbuild-all/[email protected]/
> > Signed-off-by: Michal Wajdeczko <[email protected]>
> > Cc: Rodrigo Vivi <[email protected]>
> > Cc: Matt Roper <[email protected]>
> > Cc: Riana Tauro <[email protected]>
> > ---
> > drivers/gpu/drm/xe/xe_configfs.h | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_configfs.h
> > b/drivers/gpu/drm/xe/xe_configfs.h
> > index 9998ec7c9956..ab75485b08b6 100644
> > --- a/drivers/gpu/drm/xe/xe_configfs.h
> > +++ b/drivers/gpu/drm/xe/xe_configfs.h
> > @@ -12,7 +12,7 @@
> >
> > struct pci_dev;
> >
> > -#if IS_ENABLED(CONFIG_CONFIGFS_FS)
> > +#if IS_REACHABLE(CONFIG_CONFIGFS_FS)
>
> Well, I'm not 100% confident with this. And probably the Jani's suggestion is
> the safest one.
>
> But perhaps this is acceptable in this case, only because if
> we are built-in (y), we are not able to use configfs at all anyway...
> our configfs usage already depends on us as module...
Is it not possible to make this work? Even if Xe is built into the
kernel, we're still free to unbind and rebind PCI devices from it.
If possible we'd want configfs to function properly if its module has
been loaded by the time we re-bind a PCI device.
Matt
>
> > int xe_configfs_init(void);
> > void xe_configfs_exit(void);
> > void xe_configfs_check_device(struct pci_dev *pdev);
> > --
> > 2.47.1
> >
--
Matt Roper
Graphics Software Engineer
Linux GPU Platform Enablement
Intel Corporation
