| Message ID | 58fc8553bdc3cbad8d107a905947b7e4d9c12b92.1522054975.git.sean.wang@mediatek.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, Mar 26, 2018 at 06:04:18PM +0800, sean.wang@mediatek.com wrote: > From: Sean Wang <sean.wang@mediatek.com> > > MODULE_ALIAS exports information to allow the module to be auto-loaded at > boot for the drivers registered using legacy platform registration. > > However, currently the driver is always used by DT-only platform, > MODULE_ALIAS is redundant and should be removed properly. There is no meaningful cost to having the alias there, why remove it?
On Wed, 2018-03-28 at 10:14 +0800, Mark Brown wrote: > On Mon, Mar 26, 2018 at 06:04:18PM +0800, sean.wang@mediatek.com wrote: > > From: Sean Wang <sean.wang@mediatek.com> > > > > MODULE_ALIAS exports information to allow the module to be auto-loaded at > > boot for the drivers registered using legacy platform registration. > > > > However, currently the driver is always used by DT-only platform, > > MODULE_ALIAS is redundant and should be removed properly. > > There is no meaningful cost to having the alias there, why remove it? It's no explicitly improved to the driver, it only takes less bytes in .secion modinfo from 644 bytes to 620 bytes in the observation with objdump. More importantly, it seems the driver should probably refer to another qualified driver at the initial development, but MODULE_ALIAS is forgetting to be removed to meet real usage for the module autoload. Even I was curious whether people for the driver know the actual purpose the macro it's for. Removing MODULE_ALIAS is just to correct the driver to where it should be.
On Wed, Mar 28, 2018 at 11:09:03AM +0800, Sean Wang wrote: > On Wed, 2018-03-28 at 10:14 +0800, Mark Brown wrote: > > On Mon, Mar 26, 2018 at 06:04:18PM +0800, sean.wang@mediatek.com wrote: > > > From: Sean Wang <sean.wang@mediatek.com> > > > MODULE_ALIAS exports information to allow the module to be auto-loaded at > > > boot for the drivers registered using legacy platform registration. > > > However, currently the driver is always used by DT-only platform, > > > MODULE_ALIAS is redundant and should be removed properly. > > There is no meaningful cost to having the alias there, why remove it? > It's no explicitly improved to the driver, it only takes less bytes > in .secion modinfo from 644 bytes to 620 bytes in the observation with > objdump. OK, that's a *very* small saving though especially given the block sizes that filesystems tend to use. > More importantly, it seems the driver should probably refer to another > qualified driver at the initial development, but MODULE_ALIAS is > forgetting to be removed to meet real usage for the module autoload. I can't parse the above, sorry. > Even I was curious whether people for the driver know the actual purpose > the macro it's for. > Removing MODULE_ALIAS is just to correct the driver to where it should > be. To me it's the other way around - ideally platform drivers should remember to set up the MODULE_ALIAS, just as a joining the dots thing to make sure that things that work when the driver is built in work when built as a module.
diff --git a/drivers/spi/spi-mt65xx.c b/drivers/spi/spi-mt65xx.c index 86bf456..9fdc2e5 100644 --- a/drivers/spi/spi-mt65xx.c +++ b/drivers/spi/spi-mt65xx.c @@ -856,4 +856,3 @@ module_platform_driver(mtk_spi_driver); MODULE_DESCRIPTION("MTK SPI Controller driver"); MODULE_AUTHOR("Leilk Liu <leilk.liu@mediatek.com>"); MODULE_LICENSE("GPL v2"); -MODULE_ALIAS("platform:mtk-spi");