From mboxrd@z Thu Jan 1 00:00:00 1970 From: arnd@arndb.de (Arnd Bergmann) Date: Wed, 06 May 2015 11:24:58 +0200 Subject: [PATCH v3 3/5] PCI: st: Provide support for the sti PCIe controller In-Reply-To: References: <1428657168-12495-1-git-send-email-gabriel.fernandez@linaro.org> <20150505221646.GE24643@google.com> Message-ID: <2236679.A9NfANRyhD@wuerfel> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Wednesday 06 May 2015 11:14:25 Gabriel Fernandez wrote: > > >> +static int __init pcie_init(void) > >> +{ > >> + return platform_driver_probe(&st_pcie_driver, st_pcie_probe); > >> +} > >> +device_initcall(pcie_init); > > > > Can you use module_platform_driver_probe() or module_init() here? > > > > Yes we can use module_init() here. > > By the way if figure out i removed __init attribute on st_pcie_probe() > in previous version to follow Arnd's remark. > But st_pcie_probe calls hook_fault_code() that has __init attribute. > So I think we need to keep __init for probe routine ? > Also, we have to restrict bind/unbind with "suppress_bind_attrs" in > platform_driver structure. > This is the main reason to not allow module unload/reload. > > If you are ok i will send a v4 ? > If you mark the probe function as __init, you have to use module_platform_driver_probe() or platform_driver_probe() instead, to prevent the probe function from being deferred. Calling hook_fault_code() also means you need to prevent module unloading, by providing only a module_init but not a module_exit function, so module_platform_driver_probe() also won't work. It may be a good idea to work on improving the hook_fault_code() infrastructure to make it more usable with loadable device drivers. Arnd