2017-12-19 08:36:55

by Kishon Vijay Abraham I

[permalink] [raw]
Subject: [PATCH v2] PCI: endpoint: Use EPC's device in dma_alloc_coherent/dma_free_coherent

After commit 723288836628bc1c08 ("of: restrict DMA configuration"),
of_dma_configure doesn't configure the coherent_dma_mask/dma_mask
of endpoint function device (since it doesn't have a dt node associated
with and hence no dma-ranges property), resulting in dma_alloc_coherent
(used in pci_epf_alloc_space) to fail.

Fix it by making dma_alloc_coherent use EPC's device for allocating
memory address as per discussion in [1]

[1] https://lkml.org/lkml/2017/10/24/26

Signed-off-by: Kishon Vijay Abraham I <[email protected]>
Cc: Robin Murphy <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Tested-by: Cyrille Pitchen <[email protected]>
---
No change in patch. Only added "Tested-by" of Cyrille Pitchen.

drivers/pci/endpoint/pci-epc-core.c | 10 ----------
drivers/pci/endpoint/pci-epf-core.c | 4 ++--
2 files changed, 2 insertions(+), 12 deletions(-)

diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
index 42c2a1156325..cd7d4788b94d 100644
--- a/drivers/pci/endpoint/pci-epc-core.c
+++ b/drivers/pci/endpoint/pci-epc-core.c
@@ -18,7 +18,6 @@
*/

#include <linux/device.h>
-#include <linux/dma-mapping.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/of_device.h>
@@ -371,7 +370,6 @@ EXPORT_SYMBOL_GPL(pci_epc_write_header);
int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
{
unsigned long flags;
- struct device *dev = epc->dev.parent;

if (epf->epc)
return -EBUSY;
@@ -383,12 +381,6 @@ int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
return -EINVAL;

epf->epc = epc;
- if (dev->of_node) {
- of_dma_configure(&epf->dev, dev->of_node);
- } else {
- dma_set_coherent_mask(&epf->dev, epc->dev.coherent_dma_mask);
- epf->dev.dma_mask = epc->dev.dma_mask;
- }

spin_lock_irqsave(&epc->lock, flags);
list_add_tail(&epf->list, &epc->pci_epf);
@@ -503,9 +495,7 @@ __pci_epc_create(struct device *dev, const struct pci_epc_ops *ops,
INIT_LIST_HEAD(&epc->pci_epf);

device_initialize(&epc->dev);
- dma_set_coherent_mask(&epc->dev, dev->coherent_dma_mask);
epc->dev.class = pci_epc_class;
- epc->dev.dma_mask = dev->dma_mask;
epc->dev.parent = dev;
epc->ops = ops;

diff --git a/drivers/pci/endpoint/pci-epf-core.c b/drivers/pci/endpoint/pci-epf-core.c
index ae1611a62808..95ccc4b8a0a2 100644
--- a/drivers/pci/endpoint/pci-epf-core.c
+++ b/drivers/pci/endpoint/pci-epf-core.c
@@ -99,7 +99,7 @@ EXPORT_SYMBOL_GPL(pci_epf_bind);
*/
void pci_epf_free_space(struct pci_epf *epf, void *addr, enum pci_barno bar)
{
- struct device *dev = &epf->dev;
+ struct device *dev = epf->epc->dev.parent;

if (!addr)
return;
@@ -122,7 +122,7 @@ EXPORT_SYMBOL_GPL(pci_epf_free_space);
void *pci_epf_alloc_space(struct pci_epf *epf, size_t size, enum pci_barno bar)
{
void *space;
- struct device *dev = &epf->dev;
+ struct device *dev = epf->epc->dev.parent;
dma_addr_t phys_addr;

if (size < 128)
--
2.11.0


2018-01-09 12:40:33

by Niklas Cassel

[permalink] [raw]
Subject: Re: [PATCH v2] PCI: endpoint: Use EPC's device in dma_alloc_coherent/dma_free_coherent

This is needed to get PCIe endpoint support working
on ARTPEC-6 again.

Tested-by: Niklas Cassel <[email protected]>

On 19/12/17 09:36, Kishon Vijay Abraham I wrote:
> After commit 723288836628bc1c08 ("of: restrict DMA configuration"),
> of_dma_configure doesn't configure the coherent_dma_mask/dma_mask
> of endpoint function device (since it doesn't have a dt node associated
> with and hence no dma-ranges property), resulting in dma_alloc_coherent
> (used in pci_epf_alloc_space) to fail.
>
> Fix it by making dma_alloc_coherent use EPC's device for allocating
> memory address as per discussion in [1]
>
> [1] https://lkml.org/lkml/2017/10/24/26
>
> Signed-off-by: Kishon Vijay Abraham I <[email protected]>
> Cc: Robin Murphy <[email protected]>
> Cc: Rob Herring <[email protected]>
> Cc: Christoph Hellwig <[email protected]>
> Tested-by: Cyrille Pitchen <[email protected]>
> ---
> No change in patch. Only added "Tested-by" of Cyrille Pitchen.
>
> drivers/pci/endpoint/pci-epc-core.c | 10 ----------
> drivers/pci/endpoint/pci-epf-core.c | 4 ++--
> 2 files changed, 2 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> index 42c2a1156325..cd7d4788b94d 100644
> --- a/drivers/pci/endpoint/pci-epc-core.c
> +++ b/drivers/pci/endpoint/pci-epc-core.c
> @@ -18,7 +18,6 @@
> */
>
> #include <linux/device.h>
> -#include <linux/dma-mapping.h>
> #include <linux/slab.h>
> #include <linux/module.h>
> #include <linux/of_device.h>
> @@ -371,7 +370,6 @@ EXPORT_SYMBOL_GPL(pci_epc_write_header);
> int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> {
> unsigned long flags;
> - struct device *dev = epc->dev.parent;
>
> if (epf->epc)
> return -EBUSY;
> @@ -383,12 +381,6 @@ int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> return -EINVAL;
>
> epf->epc = epc;
> - if (dev->of_node) {
> - of_dma_configure(&epf->dev, dev->of_node);
> - } else {
> - dma_set_coherent_mask(&epf->dev, epc->dev.coherent_dma_mask);
> - epf->dev.dma_mask = epc->dev.dma_mask;
> - }
>
> spin_lock_irqsave(&epc->lock, flags);
> list_add_tail(&epf->list, &epc->pci_epf);
> @@ -503,9 +495,7 @@ __pci_epc_create(struct device *dev, const struct pci_epc_ops *ops,
> INIT_LIST_HEAD(&epc->pci_epf);
>
> device_initialize(&epc->dev);
> - dma_set_coherent_mask(&epc->dev, dev->coherent_dma_mask);
> epc->dev.class = pci_epc_class;
> - epc->dev.dma_mask = dev->dma_mask;
> epc->dev.parent = dev;
> epc->ops = ops;
>
> diff --git a/drivers/pci/endpoint/pci-epf-core.c b/drivers/pci/endpoint/pci-epf-core.c
> index ae1611a62808..95ccc4b8a0a2 100644
> --- a/drivers/pci/endpoint/pci-epf-core.c
> +++ b/drivers/pci/endpoint/pci-epf-core.c
> @@ -99,7 +99,7 @@ EXPORT_SYMBOL_GPL(pci_epf_bind);
> */
> void pci_epf_free_space(struct pci_epf *epf, void *addr, enum pci_barno bar)
> {
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
>
> if (!addr)
> return;
> @@ -122,7 +122,7 @@ EXPORT_SYMBOL_GPL(pci_epf_free_space);
> void *pci_epf_alloc_space(struct pci_epf *epf, size_t size, enum pci_barno bar)
> {
> void *space;
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
> dma_addr_t phys_addr;
>
> if (size < 128)
>

2018-01-09 12:46:33

by Robin Murphy

[permalink] [raw]
Subject: Re: [PATCH v2] PCI: endpoint: Use EPC's device in dma_alloc_coherent/dma_free_coherent

On 19/12/17 08:36, Kishon Vijay Abraham I wrote:
> After commit 723288836628bc1c08 ("of: restrict DMA configuration"),
> of_dma_configure doesn't configure the coherent_dma_mask/dma_mask
> of endpoint function device (since it doesn't have a dt node associated
> with and hence no dma-ranges property), resulting in dma_alloc_coherent
> (used in pci_epf_alloc_space) to fail.
>
> Fix it by making dma_alloc_coherent use EPC's device for allocating
> memory address as per discussion in [1]

Looks like cleaning up the API usage also cleans up the code nicely too,
yay :)

Reviewed-by: Robin Murphy <[email protected]>

> [1] https://lkml.org/lkml/2017/10/24/26
>
> Signed-off-by: Kishon Vijay Abraham I <[email protected]>
> Cc: Robin Murphy <[email protected]>
> Cc: Rob Herring <[email protected]>
> Cc: Christoph Hellwig <[email protected]>
> Tested-by: Cyrille Pitchen <[email protected]>
> ---
> No change in patch. Only added "Tested-by" of Cyrille Pitchen.
>
> drivers/pci/endpoint/pci-epc-core.c | 10 ----------
> drivers/pci/endpoint/pci-epf-core.c | 4 ++--
> 2 files changed, 2 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> index 42c2a1156325..cd7d4788b94d 100644
> --- a/drivers/pci/endpoint/pci-epc-core.c
> +++ b/drivers/pci/endpoint/pci-epc-core.c
> @@ -18,7 +18,6 @@
> */
>
> #include <linux/device.h>
> -#include <linux/dma-mapping.h>
> #include <linux/slab.h>
> #include <linux/module.h>
> #include <linux/of_device.h>
> @@ -371,7 +370,6 @@ EXPORT_SYMBOL_GPL(pci_epc_write_header);
> int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> {
> unsigned long flags;
> - struct device *dev = epc->dev.parent;
>
> if (epf->epc)
> return -EBUSY;
> @@ -383,12 +381,6 @@ int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> return -EINVAL;
>
> epf->epc = epc;
> - if (dev->of_node) {
> - of_dma_configure(&epf->dev, dev->of_node);
> - } else {
> - dma_set_coherent_mask(&epf->dev, epc->dev.coherent_dma_mask);
> - epf->dev.dma_mask = epc->dev.dma_mask;
> - }
>
> spin_lock_irqsave(&epc->lock, flags);
> list_add_tail(&epf->list, &epc->pci_epf);
> @@ -503,9 +495,7 @@ __pci_epc_create(struct device *dev, const struct pci_epc_ops *ops,
> INIT_LIST_HEAD(&epc->pci_epf);
>
> device_initialize(&epc->dev);
> - dma_set_coherent_mask(&epc->dev, dev->coherent_dma_mask);
> epc->dev.class = pci_epc_class;
> - epc->dev.dma_mask = dev->dma_mask;
> epc->dev.parent = dev;
> epc->ops = ops;
>
> diff --git a/drivers/pci/endpoint/pci-epf-core.c b/drivers/pci/endpoint/pci-epf-core.c
> index ae1611a62808..95ccc4b8a0a2 100644
> --- a/drivers/pci/endpoint/pci-epf-core.c
> +++ b/drivers/pci/endpoint/pci-epf-core.c
> @@ -99,7 +99,7 @@ EXPORT_SYMBOL_GPL(pci_epf_bind);
> */
> void pci_epf_free_space(struct pci_epf *epf, void *addr, enum pci_barno bar)
> {
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
>
> if (!addr)
> return;
> @@ -122,7 +122,7 @@ EXPORT_SYMBOL_GPL(pci_epf_free_space);
> void *pci_epf_alloc_space(struct pci_epf *epf, size_t size, enum pci_barno bar)
> {
> void *space;
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
> dma_addr_t phys_addr;
>
> if (size < 128)
>

2018-01-10 11:10:34

by Lorenzo Pieralisi

[permalink] [raw]
Subject: Re: [PATCH v2] PCI: endpoint: Use EPC's device in dma_alloc_coherent/dma_free_coherent

On Tue, Dec 19, 2017 at 02:06:27PM +0530, Kishon Vijay Abraham I wrote:
> After commit 723288836628bc1c08 ("of: restrict DMA configuration"),
> of_dma_configure doesn't configure the coherent_dma_mask/dma_mask
> of endpoint function device (since it doesn't have a dt node associated
> with and hence no dma-ranges property), resulting in dma_alloc_coherent
> (used in pci_epf_alloc_space) to fail.
>
> Fix it by making dma_alloc_coherent use EPC's device for allocating
> memory address as per discussion in [1]
>
> [1] https://lkml.org/lkml/2017/10/24/26

Hi Kishon,

small nit, mind embedding the link [1] above in a lkml redirector
format - as described in:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v4.15-rc7

put it in a Link: tag to be added below - I think that's important
for the commit log to have.

Either send me the formatted link or send a v3 with it, I will merge
then.

Thanks,
Lorenzo

> Signed-off-by: Kishon Vijay Abraham I <[email protected]>
> Cc: Robin Murphy <[email protected]>
> Cc: Rob Herring <[email protected]>
> Cc: Christoph Hellwig <[email protected]>
> Tested-by: Cyrille Pitchen <[email protected]>
> ---
> No change in patch. Only added "Tested-by" of Cyrille Pitchen.
>
> drivers/pci/endpoint/pci-epc-core.c | 10 ----------
> drivers/pci/endpoint/pci-epf-core.c | 4 ++--
> 2 files changed, 2 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/pci/endpoint/pci-epc-core.c b/drivers/pci/endpoint/pci-epc-core.c
> index 42c2a1156325..cd7d4788b94d 100644
> --- a/drivers/pci/endpoint/pci-epc-core.c
> +++ b/drivers/pci/endpoint/pci-epc-core.c
> @@ -18,7 +18,6 @@
> */
>
> #include <linux/device.h>
> -#include <linux/dma-mapping.h>
> #include <linux/slab.h>
> #include <linux/module.h>
> #include <linux/of_device.h>
> @@ -371,7 +370,6 @@ EXPORT_SYMBOL_GPL(pci_epc_write_header);
> int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> {
> unsigned long flags;
> - struct device *dev = epc->dev.parent;
>
> if (epf->epc)
> return -EBUSY;
> @@ -383,12 +381,6 @@ int pci_epc_add_epf(struct pci_epc *epc, struct pci_epf *epf)
> return -EINVAL;
>
> epf->epc = epc;
> - if (dev->of_node) {
> - of_dma_configure(&epf->dev, dev->of_node);
> - } else {
> - dma_set_coherent_mask(&epf->dev, epc->dev.coherent_dma_mask);
> - epf->dev.dma_mask = epc->dev.dma_mask;
> - }
>
> spin_lock_irqsave(&epc->lock, flags);
> list_add_tail(&epf->list, &epc->pci_epf);
> @@ -503,9 +495,7 @@ __pci_epc_create(struct device *dev, const struct pci_epc_ops *ops,
> INIT_LIST_HEAD(&epc->pci_epf);
>
> device_initialize(&epc->dev);
> - dma_set_coherent_mask(&epc->dev, dev->coherent_dma_mask);
> epc->dev.class = pci_epc_class;
> - epc->dev.dma_mask = dev->dma_mask;
> epc->dev.parent = dev;
> epc->ops = ops;
>
> diff --git a/drivers/pci/endpoint/pci-epf-core.c b/drivers/pci/endpoint/pci-epf-core.c
> index ae1611a62808..95ccc4b8a0a2 100644
> --- a/drivers/pci/endpoint/pci-epf-core.c
> +++ b/drivers/pci/endpoint/pci-epf-core.c
> @@ -99,7 +99,7 @@ EXPORT_SYMBOL_GPL(pci_epf_bind);
> */
> void pci_epf_free_space(struct pci_epf *epf, void *addr, enum pci_barno bar)
> {
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
>
> if (!addr)
> return;
> @@ -122,7 +122,7 @@ EXPORT_SYMBOL_GPL(pci_epf_free_space);
> void *pci_epf_alloc_space(struct pci_epf *epf, size_t size, enum pci_barno bar)
> {
> void *space;
> - struct device *dev = &epf->dev;
> + struct device *dev = epf->epc->dev.parent;
> dma_addr_t phys_addr;
>
> if (size < 128)
> --
> 2.11.0
>

2018-01-11 08:32:10

by Kishon Vijay Abraham I

[permalink] [raw]
Subject: Re: [PATCH v2] PCI: endpoint: Use EPC's device in dma_alloc_coherent/dma_free_coherent



On Wednesday 10 January 2018 04:41 PM, Lorenzo Pieralisi wrote:
> On Tue, Dec 19, 2017 at 02:06:27PM +0530, Kishon Vijay Abraham I wrote:
>> After commit 723288836628bc1c08 ("of: restrict DMA configuration"),
>> of_dma_configure doesn't configure the coherent_dma_mask/dma_mask
>> of endpoint function device (since it doesn't have a dt node associated
>> with and hence no dma-ranges property), resulting in dma_alloc_coherent
>> (used in pci_epf_alloc_space) to fail.
>>
>> Fix it by making dma_alloc_coherent use EPC's device for allocating
>> memory address as per discussion in [1]
>>
>> [1] https://lkml.org/lkml/2017/10/24/26
>
> Hi Kishon,
>
> small nit, mind embedding the link [1] above in a lkml redirector
> format - as described in:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v4.15-rc7
>
> put it in a Link: tag to be added below - I think that's important
> for the commit log to have.
>
> Either send me the formatted link or send a v3 with it, I will merge
> then.

sure, sent you a v3.

Thanks
Kishon