Re: [PATCH -next] fbdev: via: Fix section mismatch warning in via_core_init()
From: Helge Deller <deller@gmx.de>
Date: 2022-11-15 09:36:43
Also in:
dri-devel
On 11/15/22 08:09, shangxiaojing wrote:
On 2022/11/15 15:05, Helge Deller wrote:quoted
On 11/15/22 07:52, shangxiaojing wrote:quoted
On 2022/11/15 13:14, Helge Deller wrote:quoted
On 11/15/22 03:53, Shang XiaoJing wrote:quoted
Due to viafb_exit() with "__exit" tag, it should not be called by the __init function via_core_init().I wonder if you can't instead of this and your previous patch (ab885d8c7e15) turn the i2c and gpio drivers to proper platform drivers, e.g. adding to bottom of via/via_i2c.c: module_platform_driver(&via_i2c_driver) instead of viafb_i2c_init() and viafb_i2c_exit(). Shouldn't they then automatically be loaded/unloaded?I'm sorry that I have no idea how to change an i2c driver to a platform driver. As for module_platform_driver(), I have checked and it looks like just a helper macro to definite XXX_init() and XXX_exit() instead of automatically load/unload the driver.module_platform_driver() uses module_driver() which adds module_init() which then adds code to call the generated xxx_init() functions at startup and exit.quoted
Besides, the XXX_init() and XXX_exit() definited by the module_platform_driver() can only call platform_driver_register()/platform_driver_unregister(), perhaps not suitable for via_driver? (or just need to call platform_driver_register() after change to platform driver)platform_driver_register() will be called by the newly generated XXX_init(). Do you have such a viafb card and can try?I'm sorry that I run the test on qemu, and have no viafb card.
Ok, then we leave it like this (your 2 patches merged): https://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev.git/commit/?h=for-next&id=7ecee8104be8303f6b23380f14e06915d79ddd34 Helge