Re: [PATCH -next] scsi: ufs: ufs-mediatek: Fix build error

From: Paul Cercueil
Date: Mon Jul 25 2022 - 04:47:40 EST


Hi Arnd,

Le lun., juil. 25 2022 at 08:34:32 +0200, Arnd Bergmann <arnd@xxxxxxxx> a écrit :
On Mon, Jul 25, 2022 at 5:38 AM Ren Zhijie <renzhijie2@xxxxxxxxxx> wrote:
在 2022/7/13 16:48, Arnd Bergmann 写道:

I try to use the new marcos SYSTEM_SLEEP_PM_OPS and RUNTIME_PM_OPS to
replace the old ones, and remove #ifdef around the declarations in the
header, my local changes attach below.

But it seems that doesn't work, which has ld errors:

aarch64-linux-gnu-ld: Unexpected GOT/PLT entries detected!
aarch64-linux-gnu-ld: Unexpected run-time procedure linkages detected!
drivers/ufs/host/ufs-mediatek.o: In function `ufs_mtk_runtime_resume':
ufs-mediatek.c:(.text+0x1d0c): undefined reference to
`ufshcd_runtime_resume'
drivers/ufs/host/ufs-mediatek.o: In function `ufs_mtk_runtime_suspend':
ufs-mediatek.c:(.text+0x1d64): undefined reference to
`ufshcd_runtime_suspend'
Makefile:1255: recipe for target 'vmlinux' failed
make: *** [vmlinux] Error 1

(CONFIG_PM and CONFIG_PM_SLEEP are both not set, and


It appears that there is a mistake in the RUNTIME_PM_OPS()
macro definition, can you try this patch on top?

diff --git a/include/linux/pm.h b/include/linux/pm.h
index 871c9c49ec9d..84592229d754 100644
--- a/include/linux/pm.h
+++ b/include/linux/pm.h
@@ -334,9 +334,9 @@ struct dev_pm_ops {
.restore_noirq = pm_sleep_ptr(resume_fn),

#define RUNTIME_PM_OPS(suspend_fn, resume_fn, idle_fn) \
- .runtime_suspend = suspend_fn, \
- .runtime_resume = resume_fn, \
- .runtime_idle = idle_fn,
+ .runtime_suspend = pm_ptr(suspend_fn), \
+ .runtime_resume = pm_ptr(resume_fn), \
+ .runtime_idle = pm_ptr(idle_fn),

Not a mistake, the pm_ptr() macros are not needed here, because the dev_pm_ops pointer must always be wrapped with pm_ptr().

The proper fix is:

- .pm = &ufs_mtk_pm_ops,
+ .pm = pm_ptr(&ufs_mtk_pm_ops),

Cheers,
-Paul


#ifdef CONFIG_PM_SLEEP
#define SET_SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn) \

Commit 1a3c7bb08826 ("PM: core: Add new *_PM_OPS macros, deprecate
old ones") introduced all the macros with the intent of using pm_ptr()
and pm_sleep_ptr() in them, and I think Paul accidentally forgot to add
those in this instance.

Arnd