| Message ID | 01a58dcc9c70836b7556d54dbdac6a340f206781.1650890638.git.robin.murphy@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | iommu: Retire bus_set_iommu() | expand |
On 28/04/2022 23:18, Robin Murphy wrote: > Stop calling bus_set_iommu() since it's now unnecessary. This also > leaves the custom initcall effectively doing nothing but register > the driver, which no longer needs to happen early either, so convert > it to builtin_platform_driver(). > > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > --- > drivers/iommu/ipmmu-vmsa.c | 35 +---------------------------------- > 1 file changed, 1 insertion(+), 34 deletions(-) > > diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c > index 8fdb84b3642b..2549d32f0ddd 100644 > --- a/drivers/iommu/ipmmu-vmsa.c > +++ b/drivers/iommu/ipmmu-vmsa.c > @@ -1090,11 +1090,6 @@ static int ipmmu_probe(struct platform_device *pdev) > ret = iommu_device_register(&mmu->iommu, &ipmmu_ops, &pdev->dev); > if (ret) > return ret; > - > -#if defined(CONFIG_IOMMU_DMA) > - if (!iommu_present(&platform_bus_type)) > - bus_set_iommu(&platform_bus_type, &ipmmu_ops); > -#endif > } > > /* The comment which starts here did not make it to the patch but it should have as it mentions bus_set_iommu() which is gone by the end of the series. More general question/request - could you please include the exact sha1 the patchset is based on? It did not apply to any current trees and while it was trivial, it was slightly annoying to resolve the conflicts :) Thanks, > @@ -1168,32 +1163,4 @@ static struct platform_driver ipmmu_driver = { > .probe = ipmmu_probe, > .remove = ipmmu_remove, > }; > - > -static int __init ipmmu_init(void) > -{ > - struct device_node *np; > - static bool setup_done; > - int ret; > - > - if (setup_done) > - return 0; > - > - np = of_find_matching_node(NULL, ipmmu_of_ids); > - if (!np) > - return 0; > - > - of_node_put(np); > - > - ret = platform_driver_register(&ipmmu_driver); > - if (ret < 0) > - return ret; > - > -#if defined(CONFIG_ARM) && !defined(CONFIG_IOMMU_DMA) > - if (!iommu_present(&platform_bus_type)) > - bus_set_iommu(&platform_bus_type, &ipmmu_ops); > -#endif > - > - setup_done = true; > - return 0; > -} > -subsys_initcall(ipmmu_init); > +builtin_platform_driver(ipmmu_driver);
On 2022-07-06 09:38, Alexey Kardashevskiy wrote: > > > On 28/04/2022 23:18, Robin Murphy wrote: >> Stop calling bus_set_iommu() since it's now unnecessary. This also >> leaves the custom initcall effectively doing nothing but register >> the driver, which no longer needs to happen early either, so convert >> it to builtin_platform_driver(). >> >> Signed-off-by: Robin Murphy <robin.murphy@arm.com> >> --- >> drivers/iommu/ipmmu-vmsa.c | 35 +---------------------------------- >> 1 file changed, 1 insertion(+), 34 deletions(-) >> >> diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c >> index 8fdb84b3642b..2549d32f0ddd 100644 >> --- a/drivers/iommu/ipmmu-vmsa.c >> +++ b/drivers/iommu/ipmmu-vmsa.c >> @@ -1090,11 +1090,6 @@ static int ipmmu_probe(struct platform_device >> *pdev) >> ret = iommu_device_register(&mmu->iommu, &ipmmu_ops, >> &pdev->dev); >> if (ret) >> return ret; >> - >> -#if defined(CONFIG_IOMMU_DMA) >> - if (!iommu_present(&platform_bus_type)) >> - bus_set_iommu(&platform_bus_type, &ipmmu_ops); >> -#endif >> } >> /* > > The comment which starts here did not make it to the patch but it should > have as it mentions bus_set_iommu() which is gone by the end of the series. Heh, busted! In fact I think the whole point of that comment stops being true, but I couldn't be bothered to reason about it since one of the next steps after this is to start ripping all the arm_iommu_* stuff out anyway. > More general question/request - could you please include the exact sha1 > the patchset is based on? It did not apply to any current trees and > while it was trivial, it was slightly annoying to resolve the conflicts > :) Thanks, v3 is based directly on 5.19-rc3: https://lore.kernel.org/lkml/cover.1657034827.git.robin.murphy@arm.com/ And if it helps I have it on a branch here as well: https://gitlab.arm.com/linux-arm/linux-rm/-/tree/bus-set-iommu-v3 Robin.
diff --git a/drivers/iommu/ipmmu-vmsa.c b/drivers/iommu/ipmmu-vmsa.c index 8fdb84b3642b..2549d32f0ddd 100644 --- a/drivers/iommu/ipmmu-vmsa.c +++ b/drivers/iommu/ipmmu-vmsa.c @@ -1090,11 +1090,6 @@ static int ipmmu_probe(struct platform_device *pdev) ret = iommu_device_register(&mmu->iommu, &ipmmu_ops, &pdev->dev); if (ret) return ret; - -#if defined(CONFIG_IOMMU_DMA) - if (!iommu_present(&platform_bus_type)) - bus_set_iommu(&platform_bus_type, &ipmmu_ops); -#endif } /* @@ -1168,32 +1163,4 @@ static struct platform_driver ipmmu_driver = { .probe = ipmmu_probe, .remove = ipmmu_remove, }; - -static int __init ipmmu_init(void) -{ - struct device_node *np; - static bool setup_done; - int ret; - - if (setup_done) - return 0; - - np = of_find_matching_node(NULL, ipmmu_of_ids); - if (!np) - return 0; - - of_node_put(np); - - ret = platform_driver_register(&ipmmu_driver); - if (ret < 0) - return ret; - -#if defined(CONFIG_ARM) && !defined(CONFIG_IOMMU_DMA) - if (!iommu_present(&platform_bus_type)) - bus_set_iommu(&platform_bus_type, &ipmmu_ops); -#endif - - setup_done = true; - return 0; -} -subsys_initcall(ipmmu_init); +builtin_platform_driver(ipmmu_driver);
Stop calling bus_set_iommu() since it's now unnecessary. This also leaves the custom initcall effectively doing nothing but register the driver, which no longer needs to happen early either, so convert it to builtin_platform_driver(). Signed-off-by: Robin Murphy <robin.murphy@arm.com> --- drivers/iommu/ipmmu-vmsa.c | 35 +---------------------------------- 1 file changed, 1 insertion(+), 34 deletions(-)