| Message ID | 112e8f3d3e7c054ecf5e12b5ac0aa5596ec00681.1664455433.git.robin.murphy@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v2] of: Fix "dma-ranges" handling for bus controllers | expand |
On Thu, 29 Sep 2022 13:48:38 +0100, Robin Murphy wrote: > Commit 951d48855d86 ("of: Make of_dma_get_range() work on bus nodes") > relaxed the handling of "dma-ranges" for any leaf node on the assumption > that it would still represent a usage error for the property to be > present on a non-bus leaf node. However there turns out to be a fiddly > case where a bus also represents a DMA-capable device in its own right, > such as a PCIe root complex with an integrated DMA engine on its > platform side. In such cases, "dma-ranges" translation is entirely valid > for devices discovered behind the bus, but should not be erroneously > applied to the bus controller device itself which operates in its > parent's address space. Fix this by restoring the previous behaviour for > the specific case where a device is configured via its own OF node, > since it is logical to assume that a device should never represent its > own parent bus. > > Reported-by: Serge Semin <Sergey.Semin@baikalelectronics.ru> > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > --- > > v2: Fix !HAS_DMA build error > > drivers/of/address.c | 4 +++- > drivers/of/device.c | 9 ++++++++- > drivers/of/of_private.h | 5 +++++ > 3 files changed, 16 insertions(+), 2 deletions(-) > Applied, thanks! I assume this was not tagged with Fixes or stable because there is not yet a user that needs it? I didn't add it either because I'm a bit worried about regressions and applying this just before the merge window. So send it to stable later if anyone cares. Rob
On 2022-09-30 20:39, Rob Herring wrote: > On Thu, 29 Sep 2022 13:48:38 +0100, Robin Murphy wrote: >> Commit 951d48855d86 ("of: Make of_dma_get_range() work on bus nodes") >> relaxed the handling of "dma-ranges" for any leaf node on the assumption >> that it would still represent a usage error for the property to be >> present on a non-bus leaf node. However there turns out to be a fiddly >> case where a bus also represents a DMA-capable device in its own right, >> such as a PCIe root complex with an integrated DMA engine on its >> platform side. In such cases, "dma-ranges" translation is entirely valid >> for devices discovered behind the bus, but should not be erroneously >> applied to the bus controller device itself which operates in its >> parent's address space. Fix this by restoring the previous behaviour for >> the specific case where a device is configured via its own OF node, >> since it is logical to assume that a device should never represent its >> own parent bus. >> >> Reported-by: Serge Semin <Sergey.Semin@baikalelectronics.ru> >> Signed-off-by: Robin Murphy <robin.murphy@arm.com> >> --- >> >> v2: Fix !HAS_DMA build error >> >> drivers/of/address.c | 4 +++- >> drivers/of/device.c | 9 ++++++++- >> drivers/of/of_private.h | 5 +++++ >> 3 files changed, 16 insertions(+), 2 deletions(-) >> > > Applied, thanks! > > I assume this was not tagged with Fixes or stable because there is not > yet a user that needs it? I didn't add it either because I'm a bit > worried about regressions and applying this just before the merge > window. So send it to stable later if anyone cares. Indeed this was only brought to light by a patch series that Serge is working on, so although it's technically a fix I don't believe it's affecting any in-tree users yet, thus doesn't warrant backporting. Cheers, Robin.
diff --git a/drivers/of/address.c b/drivers/of/address.c index 96f0a12e507c..c34ac33b7338 100644 --- a/drivers/of/address.c +++ b/drivers/of/address.c @@ -579,7 +579,8 @@ u64 of_translate_address(struct device_node *dev, const __be32 *in_addr) } EXPORT_SYMBOL(of_translate_address); -static struct device_node *__of_get_dma_parent(const struct device_node *np) +#ifdef CONFIG_HAS_DMA +struct device_node *__of_get_dma_parent(const struct device_node *np) { struct of_phandle_args args; int ret, index; @@ -596,6 +597,7 @@ static struct device_node *__of_get_dma_parent(const struct device_node *np) return of_node_get(args.np); } +#endif static struct device_node *of_get_next_dma_parent(struct device_node *np) { diff --git a/drivers/of/device.c b/drivers/of/device.c index 75b6cbffa755..8cefe5a7d04e 100644 --- a/drivers/of/device.c +++ b/drivers/of/device.c @@ -116,12 +116,19 @@ int of_dma_configure_id(struct device *dev, struct device_node *np, { const struct iommu_ops *iommu; const struct bus_dma_region *map = NULL; + struct device_node *bus_np; u64 dma_start = 0; u64 mask, end, size = 0; bool coherent; int ret; - ret = of_dma_get_range(np, &map); + if (np == dev->of_node) + bus_np = __of_get_dma_parent(np); + else + bus_np = of_node_get(np); + + ret = of_dma_get_range(bus_np, &map); + of_node_put(bus_np); if (ret < 0) { /* * For legacy reasons, we have to assume some devices need diff --git a/drivers/of/of_private.h b/drivers/of/of_private.h index 9324483397f6..fb6792d381a6 100644 --- a/drivers/of/of_private.h +++ b/drivers/of/of_private.h @@ -155,12 +155,17 @@ struct bus_dma_region; #if defined(CONFIG_OF_ADDRESS) && defined(CONFIG_HAS_DMA) int of_dma_get_range(struct device_node *np, const struct bus_dma_region **map); +struct device_node *__of_get_dma_parent(const struct device_node *np); #else static inline int of_dma_get_range(struct device_node *np, const struct bus_dma_region **map) { return -ENODEV; } +static inline struct device_node *__of_get_dma_parent(const struct device_node *np) +{ + return of_get_parent(np); +} #endif void fdt_init_reserved_mem(void);
Commit 951d48855d86 ("of: Make of_dma_get_range() work on bus nodes") relaxed the handling of "dma-ranges" for any leaf node on the assumption that it would still represent a usage error for the property to be present on a non-bus leaf node. However there turns out to be a fiddly case where a bus also represents a DMA-capable device in its own right, such as a PCIe root complex with an integrated DMA engine on its platform side. In such cases, "dma-ranges" translation is entirely valid for devices discovered behind the bus, but should not be erroneously applied to the bus controller device itself which operates in its parent's address space. Fix this by restoring the previous behaviour for the specific case where a device is configured via its own OF node, since it is logical to assume that a device should never represent its own parent bus. Reported-by: Serge Semin <Sergey.Semin@baikalelectronics.ru> Signed-off-by: Robin Murphy <robin.murphy@arm.com> --- v2: Fix !HAS_DMA build error drivers/of/address.c | 4 +++- drivers/of/device.c | 9 ++++++++- drivers/of/of_private.h | 5 +++++ 3 files changed, 16 insertions(+), 2 deletions(-)