linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] powerpc/pseries/iommu: enable_ddw incorrectly returns direct mapping for SR-IOV device.
@ 2023-10-02 21:46 Gaurav Batra
  2023-10-03  2:22 ` Madhavan Srinivasan
  0 siblings, 1 reply; 2+ messages in thread
From: Gaurav Batra @ 2023-10-02 21:46 UTC (permalink / raw)
  To: mpe; +Cc: brking, linuxppc-dev, Gaurav Batra, gjoyce

When a device is initialized, the driver invokes dma_supported() twice - first
for streaming mappings followed by coherent mappings. For an SR-IOV device,
default window is deleted and DDW created. With vPMEM enabled, TCE mappings
are dynamically created for both vPMEM and SR-IOV device. There are no direct
mappings.

First time when dma_supported() is called with 64 bit mask, DDW is created and
marked as dynamic window. The second time dma_supported() is called, enable_ddw()
finds existing window for the device and incorrectly returns it as "direct mapping".

This only happens when size of DDW is capable of mapping max LPAR memory.

This results in streaming TCEs to not get dynamically mapped, since code incorrently
assumes these are already pre-mapped. The adapter initially comes up but goes down
due to EEH.
---
 arch/powerpc/platforms/pseries/iommu.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c
index 16d93b580f61..d8b4adcef1ad 100644
--- a/arch/powerpc/platforms/pseries/iommu.c
+++ b/arch/powerpc/platforms/pseries/iommu.c
@@ -914,7 +914,8 @@ static int remove_ddw(struct device_node *np, bool remove_prop, const char *win_
 	return 0;
 }
 
-static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *window_shift)
+static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *window_shift,
+			      bool *direct_mapping)
 {
 	struct dma_win *window;
 	const struct dynamic_dma_window_prop *dma64;
@@ -927,6 +928,7 @@ static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *windo
 			dma64 = window->prop;
 			*dma_addr = be64_to_cpu(dma64->dma_base);
 			*window_shift = be32_to_cpu(dma64->window_shift);
+			*direct_mapping = window->direct;
 			found = true;
 			break;
 		}
@@ -1270,8 +1272,7 @@ static bool enable_ddw(struct pci_dev *dev, struct device_node *pdn)
 
 	mutex_lock(&dma_win_init_mutex);
 
-	if (find_existing_ddw(pdn, &dev->dev.archdata.dma_offset, &len)) {
-		direct_mapping = (len >= max_ram_len);
+	if (find_existing_ddw(pdn, &dev->dev.archdata.dma_offset, &len, &direct_mapping)) {
 		goto out_unlock;
 	}
 
-- 
2.39.2 (Apple Git-143)


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] powerpc/pseries/iommu: enable_ddw incorrectly returns direct mapping for SR-IOV device.
  2023-10-02 21:46 [PATCH] powerpc/pseries/iommu: enable_ddw incorrectly returns direct mapping for SR-IOV device Gaurav Batra
@ 2023-10-03  2:22 ` Madhavan Srinivasan
  0 siblings, 0 replies; 2+ messages in thread
From: Madhavan Srinivasan @ 2023-10-03  2:22 UTC (permalink / raw)
  To: Gaurav Batra, mpe; +Cc: brking, linuxppc-dev, gjoyce


On 10/3/23 3:16 AM, Gaurav Batra wrote:
> When a device is initialized, the driver invokes dma_supported() twice - first
> for streaming mappings followed by coherent mappings. For an SR-IOV device,
> default window is deleted and DDW created. With vPMEM enabled, TCE mappings
> are dynamically created for both vPMEM and SR-IOV device. There are no direct
> mappings.
>
> First time when dma_supported() is called with 64 bit mask, DDW is created and
> marked as dynamic window. The second time dma_supported() is called, enable_ddw()
> finds existing window for the device and incorrectly returns it as "direct mapping".
>
> This only happens when size of DDW is capable of mapping max LPAR memory.
>
> This results in streaming TCEs to not get dynamically mapped, since code incorrently
> assumes these are already pre-mapped. The adapter initially comes up but goes down
> due to EEH.

Just checked for patch submission and it is missing "Signed-off-by:"
It is good to run once with checkpatch.pl before posting to mailing list

maddy

> ---
>   arch/powerpc/platforms/pseries/iommu.c | 7 ++++---
>   1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/arch/powerpc/platforms/pseries/iommu.c b/arch/powerpc/platforms/pseries/iommu.c
> index 16d93b580f61..d8b4adcef1ad 100644
> --- a/arch/powerpc/platforms/pseries/iommu.c
> +++ b/arch/powerpc/platforms/pseries/iommu.c
> @@ -914,7 +914,8 @@ static int remove_ddw(struct device_node *np, bool remove_prop, const char *win_
>   	return 0;
>   }
>
> -static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *window_shift)
> +static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *window_shift,
> +			      bool *direct_mapping)
>   {
>   	struct dma_win *window;
>   	const struct dynamic_dma_window_prop *dma64;
> @@ -927,6 +928,7 @@ static bool find_existing_ddw(struct device_node *pdn, u64 *dma_addr, int *windo
>   			dma64 = window->prop;
>   			*dma_addr = be64_to_cpu(dma64->dma_base);
>   			*window_shift = be32_to_cpu(dma64->window_shift);
> +			*direct_mapping = window->direct;
>   			found = true;
>   			break;
>   		}
> @@ -1270,8 +1272,7 @@ static bool enable_ddw(struct pci_dev *dev, struct device_node *pdn)
>
>   	mutex_lock(&dma_win_init_mutex);
>
> -	if (find_existing_ddw(pdn, &dev->dev.archdata.dma_offset, &len)) {
> -		direct_mapping = (len >= max_ram_len);
> +	if (find_existing_ddw(pdn, &dev->dev.archdata.dma_offset, &len, &direct_mapping)) {
>   		goto out_unlock;
>   	}
>

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2023-10-03  2:23 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2023-10-02 21:46 [PATCH] powerpc/pseries/iommu: enable_ddw incorrectly returns direct mapping for SR-IOV device Gaurav Batra
2023-10-03  2:22 ` Madhavan Srinivasan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).