[PATCH 00/11] opencapi: enable card reset and link retraining

This is the linux part of the work to use the PCI hotplug framework to
control an opencapi card so that it can be reset and re-read after
flashing a new FPGA image. I had posted it earlier as an RFC and this
version is mostly similar, with just some minor editing.

It needs support in skiboot:
http://patchwork.ozlabs.org/project/skiboot/list/?series=129724
On an old skiboot, it will do nothing.

A virtual PCI slot is created for the opencapi adapter, and its state
can be controlled through the pnv-php hotplug driver:

  echo 0|1 > /sys/bus/pci/slots/OPENCAPI-<...>/power

Note that the power to the card is not really turned off, as the card
needs to stay on to be flashed with a new image. Instead the card is
in reset.

The first part of the series mostly deals with the pci/ioda state, as
the opencapi devices can now go away and the state needs to be cleaned
up.

The second part is modifications to the PCI hotplug driver on powernv,
so that a virtual slot is created for the opencapi adapters found in
the device tree.



Frederic Barrat (11):
  powerpc/powernv/ioda: Fix ref count for devices with their own PE
  powerpc/powernv/ioda: Protect PE list
  powerpc/powernv/ioda: set up PE on opencapi device when enabling
  powerpc/powernv/ioda: Release opencapi device
  powerpc/powernv/ioda: Find opencapi slot for a device node
  pci/hotplug/pnv-php: Remove erroneous warning
  pci/hotplug/pnv-php: Improve error msg on power state change failure
  pci/hotplug/pnv-php: Register opencapi slots
  pci/hotplug/pnv-php: Relax check when disabling slot
  pci/hotplug/pnv-php: Wrap warnings in macro
  ocxl: Add PCI hotplug dependency to Kconfig

 arch/powerpc/include/asm/pnv-pci.h        |   1 +
 arch/powerpc/platforms/powernv/pci-ioda.c | 107 ++++++++++++++--------
 arch/powerpc/platforms/powernv/pci.c      |  10 +-
 drivers/misc/ocxl/Kconfig                 |   1 +
 drivers/pci/hotplug/pnv_php.c             |  82 ++++++++++-------
 5 files changed, 125 insertions(+), 76 deletions(-)

-- 
2.21.0

Taking a reference on the pci_dev structure was required with initial
commit 184cd4a3b962 ("powerpc/powernv: PCI support for p7IOC under
OPAL v2"), where we were storing the pci_dev in the pci_dn structure.

However, the pci_dev was later removed from the pci_dn structure, but
the reference was kept. See commit 902bdc57451c ("powerpc/powernv/idoa:
Remove unnecessary pcidev from pci_dn").

The pnv_ioda_pe structure life cycle is the same as the pci_dev
structure, the PE is freed when the device is released. So we don't
need a reference for the pci_dev stored in the PE, otherwise the
pci_dev will never be released. Which is not really a surprise as the
comment (removed here as no longer needed) was stating as much.

Fixes: 902bdc57451c ("powerpc/powernv/idoa: Remove unnecessary pcidev from pci_dn")
Signed-off-by: Frederic Barrat <redacted>
---
 arch/powerpc/platforms/powernv/pci-ioda.c | 11 +----------
 1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index d8080558d020..92767f006f20 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1062,14 +1062,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)
 		return NULL;
 	}
 
-	/* NOTE: We get only one ref to the pci_dev for the pdn, not for the
-	 * pointer in the PE data structure, both should be destroyed at the
-	 * same time. However, this needs to be looked at more closely again
-	 * once we actually start removing things (Hotplug, SR-IOV, ...)
-	 *
-	 * At some point we want to remove the PDN completely anyways
-	 */
-	pci_dev_get(dev);
 	pdn->pe_number = pe->pe_number;
 	pe->flags = PNV_IODA_PE_DEV;
 	pe->pdev = dev;

@@ -1084,7 +1076,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)
 		pnv_ioda_free_pe(pe);
 		pdn->pe_number = IODA_INVALID_PE;
 		pe->pdev = NULL;
-		pci_dev_put(dev);
 		return NULL;
 	}
 

@@ -1228,7 +1219,7 @@ static struct pnv_ioda_pe *pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)
 			 */
 			dev_info(&npu_pdev->dev,
 				"Associating to existing PE %x\n", pe_num);
-			pci_dev_get(npu_pdev);
+			pci_dev_get(npu_pdev); // still needed after 902bdc57451c ?
 			npu_pdn = pci_get_pdn(npu_pdev);
 			rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
 			npu_pdn->pe_number = pe_num;

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 9/9/19 5:45 pm, Frederic Barrat wrote:

Did you mean to leave that comment in?

  			npu_pdn = pci_get_pdn(npu_pdev);
  			rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
  			npu_pdn->pe_number = pe_num;

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Le 26/09/2019 à 18:44, Andrew Donnellan a écrit :

On 9/9/19 5:45 pm, Frederic Barrat wrote:

quoted

Taking a reference on the pci_dev structure was required with initial
commit 184cd4a3b962 ("powerpc/powernv: PCI support for p7IOC under
OPAL v2"), where we were storing the pci_dev in the pci_dn structure.

However, the pci_dev was later removed from the pci_dn structure, but
the reference was kept. See commit 902bdc57451c ("powerpc/powernv/idoa:
Remove unnecessary pcidev from pci_dn").

The pnv_ioda_pe structure life cycle is the same as the pci_dev
structure, the PE is freed when the device is released. So we don't
need a reference for the pci_dev stored in the PE, otherwise the
pci_dev will never be released. Which is not really a surprise as the
comment (removed here as no longer needed) was stating as much.

Fixes: 902bdc57451c ("powerpc/powernv/idoa: Remove unnecessary pcidev 
from pci_dn")
Signed-off-by: Frederic Barrat <redacted>
---
  arch/powerpc/platforms/powernv/pci-ioda.c | 11 +----------
  1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c 

b/arch/powerpc/platforms/powernv/pci-ioda.c
index d8080558d020..92767f006f20 100644

--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1062,14 +1062,6 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_dev_PE(struct pci_dev *dev)
          return NULL;
      }
-    /* NOTE: We get only one ref to the pci_dev for the pdn, not for the
-     * pointer in the PE data structure, both should be destroyed at the
-     * same time. However, this needs to be looked at more closely again
-     * once we actually start removing things (Hotplug, SR-IOV, ...)
-     *
-     * At some point we want to remove the PDN completely anyways
-     */
-    pci_dev_get(dev);
      pdn->pe_number = pe->pe_number;
      pe->flags = PNV_IODA_PE_DEV;
      pe->pdev = dev;

@@ -1084,7 +1076,6 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_dev_PE(struct pci_dev *dev)
          pnv_ioda_free_pe(pe);
          pdn->pe_number = IODA_INVALID_PE;
          pe->pdev = NULL;
-        pci_dev_put(dev);
          return NULL;
      }

@@ -1228,7 +1219,7 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)
               */
              dev_info(&npu_pdev->dev,
                  "Associating to existing PE %x\n", pe_num);
-            pci_dev_get(npu_pdev);
+            pci_dev_get(npu_pdev); // still needed after 902bdc57451c ?

Did you mean to leave that comment in?

Yes, I assumed the series wouldn't get in on the first try and a 
nvlink-minded developer would weigh in :)

   Fred

quoted

              npu_pdn = pci_get_pdn(npu_pdev);
              rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
              npu_pdn->pe_number = pe_num;

On 27/09/2019 03:15, Frederic Barrat wrote:

Le 26/09/2019 à 18:44, Andrew Donnellan a écrit :

quoted

On 9/9/19 5:45 pm, Frederic Barrat wrote:

quoted

Taking a reference on the pci_dev structure was required with initial
commit 184cd4a3b962 ("powerpc/powernv: PCI support for p7IOC under
OPAL v2"), where we were storing the pci_dev in the pci_dn structure.

However, the pci_dev was later removed from the pci_dn structure, but
the reference was kept. See commit 902bdc57451c ("powerpc/powernv/idoa:
Remove unnecessary pcidev from pci_dn").

The pnv_ioda_pe structure life cycle is the same as the pci_dev
structure, the PE is freed when the device is released. So we don't
need a reference for the pci_dev stored in the PE, otherwise the
pci_dev will never be released. Which is not really a surprise as the
comment (removed here as no longer needed) was stating as much.

Fixes: 902bdc57451c ("powerpc/powernv/idoa: Remove unnecessary pcidev from pci_dn")
Signed-off-by: Frederic Barrat <redacted>
---
  arch/powerpc/platforms/powernv/pci-ioda.c | 11 +----------
  1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index d8080558d020..92767f006f20 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1062,14 +1062,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)

          return NULL;
      }
-    /* NOTE: We get only one ref to the pci_dev for the pdn, not for the
-     * pointer in the PE data structure, both should be destroyed at the
-     * same time. However, this needs to be looked at more closely again
-     * once we actually start removing things (Hotplug, SR-IOV, ...)
-     *
-     * At some point we want to remove the PDN completely anyways
-     */
-    pci_dev_get(dev);
      pdn->pe_number = pe->pe_number;
      pe->flags = PNV_IODA_PE_DEV;
      pe->pdev = dev;

@@ -1084,7 +1076,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)

          pnv_ioda_free_pe(pe);
          pdn->pe_number = IODA_INVALID_PE;
          pe->pdev = NULL;
-        pci_dev_put(dev);
          return NULL;
      }

@@ -1228,7 +1219,7 @@ static struct pnv_ioda_pe *pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)

               */
              dev_info(&npu_pdev->dev,
                  "Associating to existing PE %x\n", pe_num);
-            pci_dev_get(npu_pdev);
+            pci_dev_get(npu_pdev); // still needed after 902bdc57451c ?

Did you mean to leave that comment in?

Yes, I assumed the series wouldn't get in on the first try and a nvlink-minded developer would weigh in :)

I am looking at it and I am still not so sure what exactly guarantees that lifetime(@dev) == lifetime(@pe). For example,
sriov_disable() removes VFs first, and only then releases PEs so there is a window when we may have a stale pdev. Not sure.

  Fred

quoted

quoted

              npu_pdn = pci_get_pdn(npu_pdev);
              rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
              npu_pdn->pe_number = pe_num;

-- 
Alexey

Le 27/09/2019 à 08:54, Alexey Kardashevskiy a écrit :

On 27/09/2019 03:15, Frederic Barrat wrote:

quoted

Le 26/09/2019 à 18:44, Andrew Donnellan a écrit :

quoted

On 9/9/19 5:45 pm, Frederic Barrat wrote:

quoted

Taking a reference on the pci_dev structure was required with initial
commit 184cd4a3b962 ("powerpc/powernv: PCI support for p7IOC under
OPAL v2"), where we were storing the pci_dev in the pci_dn structure.

However, the pci_dev was later removed from the pci_dn structure, but
the reference was kept. See commit 902bdc57451c ("powerpc/powernv/idoa:
Remove unnecessary pcidev from pci_dn").

The pnv_ioda_pe structure life cycle is the same as the pci_dev
structure, the PE is freed when the device is released. So we don't
need a reference for the pci_dev stored in the PE, otherwise the
pci_dev will never be released. Which is not really a surprise as the
comment (removed here as no longer needed) was stating as much.

Fixes: 902bdc57451c ("powerpc/powernv/idoa: Remove unnecessary pcidev from pci_dn")
Signed-off-by: Frederic Barrat <redacted>
---
   arch/powerpc/platforms/powernv/pci-ioda.c | 11 +----------
   1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index d8080558d020..92767f006f20 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1062,14 +1062,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)
           return NULL;
       }
-    /* NOTE: We get only one ref to the pci_dev for the pdn, not for the
-     * pointer in the PE data structure, both should be destroyed at the
-     * same time. However, this needs to be looked at more closely again
-     * once we actually start removing things (Hotplug, SR-IOV, ...)
-     *
-     * At some point we want to remove the PDN completely anyways
-     */
-    pci_dev_get(dev);
       pdn->pe_number = pe->pe_number;
       pe->flags = PNV_IODA_PE_DEV;
       pe->pdev = dev;

@@ -1084,7 +1076,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)
           pnv_ioda_free_pe(pe);
           pdn->pe_number = IODA_INVALID_PE;
           pe->pdev = NULL;
-        pci_dev_put(dev);
           return NULL;
       }

@@ -1228,7 +1219,7 @@ static struct pnv_ioda_pe *pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)
                */
               dev_info(&npu_pdev->dev,
                   "Associating to existing PE %x\n", pe_num);
-            pci_dev_get(npu_pdev);
+            pci_dev_get(npu_pdev); // still needed after 902bdc57451c ?

Did you mean to leave that comment in?

Yes, I assumed the series wouldn't get in on the first try and a nvlink-minded developer would weigh in :)

I am looking at it and I am still not so sure what exactly guarantees that lifetime(@dev) == lifetime(@pe). For example,
sriov_disable() removes VFs first, and only then releases PEs so there is a window when we may have a stale pdev. Not sure.

Indeed, for SRIOV, PE life-cycle is handled differently. And hopefully 
correctly, but I don’t think this patch alters it.

I was discussing with Greg about it, as he had to look in that area in 
the past. My understanding is that this patch is ok as it follows the 
initial comment in pnv_ioda_setup_dev_PE() that we don’t need to get a 
reference for the PE, it makes sense and I could trace the origin of the 
pci_dev_get() and it seemed like it should have been removed in a 
previous patch (902bdc57451c).

However, one question is whether this patch breaks nvlink and if nvlink 
assumes the devices won’t go away because we explicitly take a reference 
forever. In npu_dma.c, there are 2 functions which allow to find the GPU 
associated to a npu device, and vice-versa. Both functions return a 
pointer to a struct pci_dev, but they don’t take a reference on the 
device being returned. So that seems dangerous. I’m probably missing 
something.

Alexey, Alistair: what, if anything, guarantees, that the npu or gpu 
devices stay valid. Is it because we simply don’t provide any means to 
get rid of them ? Otherwise, don’t we need the callers of 
pnv_pci_get_gpu_dev() and pnv_pci_get_npu_dev() to worry about reference 
counting ? I’ve started looking into it and the changes are scary, which 
explains Greg’s related commit 02c5f5394918b.

   Fred

quoted

   Fred

quoted

quoted

               npu_pdn = pci_get_pdn(npu_pdev);
               rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
               npu_pdn->pe_number = pe_num;

On Wednesday, 13 November 2019 4:38:21 AM AEDT Frederic Barrat wrote:

Le 27/09/2019 à 08:54, Alexey Kardashevskiy a écrit :

quoted

On 27/09/2019 03:15, Frederic Barrat wrote:

quoted

Le 26/09/2019 à 18:44, Andrew Donnellan a écrit :

quoted

On 9/9/19 5:45 pm, Frederic Barrat wrote:

quoted

Taking a reference on the pci_dev structure was required with initial
commit 184cd4a3b962 ("powerpc/powernv: PCI support for p7IOC under
OPAL v2"), where we were storing the pci_dev in the pci_dn structure.

However, the pci_dev was later removed from the pci_dn structure, but
the reference was kept. See commit 902bdc57451c ("powerpc/powernv/idoa:
Remove unnecessary pcidev from pci_dn").

The pnv_ioda_pe structure life cycle is the same as the pci_dev
structure, the PE is freed when the device is released. So we don't
need a reference for the pci_dev stored in the PE, otherwise the
pci_dev will never be released. Which is not really a surprise as the
comment (removed here as no longer needed) was stating as much.

Fixes: 902bdc57451c ("powerpc/powernv/idoa: Remove unnecessary pcidev 

from pci_dn")

quoted

quoted

quoted

quoted

Signed-off-by: Frederic Barrat <redacted>
---
   arch/powerpc/platforms/powernv/pci-ioda.c | 11 +----------
   1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/

platforms/powernv/pci-ioda.c

quoted

quoted

quoted

quoted

index d8080558d020..92767f006f20 100644

--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1062,14 +1062,6 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_dev_PE(struct pci_dev *dev)

quoted

quoted

quoted

quoted

           return NULL;
       }
-    /* NOTE: We get only one ref to the pci_dev for the pdn, not for 

the

quoted

quoted

quoted

quoted

-     * pointer in the PE data structure, both should be destroyed at 

the

quoted

quoted

quoted

quoted

-     * same time. However, this needs to be looked at more closely 

again

quoted

quoted

quoted

quoted

-     * once we actually start removing things (Hotplug, SR-IOV, ...)
-     *
-     * At some point we want to remove the PDN completely anyways
-     */
-    pci_dev_get(dev);
       pdn->pe_number = pe->pe_number;
       pe->flags = PNV_IODA_PE_DEV;
       pe->pdev = dev;

@@ -1084,7 +1076,6 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_dev_PE(struct pci_dev *dev)

quoted

quoted

quoted

quoted

           pnv_ioda_free_pe(pe);
           pdn->pe_number = IODA_INVALID_PE;
           pe->pdev = NULL;
-        pci_dev_put(dev);
           return NULL;
       }

@@ -1228,7 +1219,7 @@ static struct pnv_ioda_pe 

*pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)

quoted

quoted

quoted

quoted

                */
               dev_info(&npu_pdev->dev,
                   "Associating to existing PE %x\n", pe_num);
-            pci_dev_get(npu_pdev);
+            pci_dev_get(npu_pdev); // still needed after 902bdc57451c 

?

quoted

quoted

quoted

Did you mean to leave that comment in?

Yes, I assumed the series wouldn't get in on the first try and a nvlink-

minded developer would weigh in :)

quoted

I am looking at it and I am still not so sure what exactly guarantees that 

lifetime(@dev) == lifetime(@pe). For example,

quoted

sriov_disable() removes VFs first, and only then releases PEs so there is a 

window when we may have a stale pdev. Not sure.

Indeed, for SRIOV, PE life-cycle is handled differently. And hopefully 
correctly, but I don’t think this patch alters it.

I was discussing with Greg about it, as he had to look in that area in 
the past. My understanding is that this patch is ok as it follows the 
initial comment in pnv_ioda_setup_dev_PE() that we don’t need to get a 
reference for the PE, it makes sense and I could trace the origin of the 
pci_dev_get() and it seemed like it should have been removed in a 
previous patch (902bdc57451c).

However, one question is whether this patch breaks nvlink and if nvlink 
assumes the devices won’t go away because we explicitly take a reference 
forever. In npu_dma.c, there are 2 functions which allow to find the GPU 
associated to a npu device, and vice-versa. Both functions return a 
pointer to a struct pci_dev, but they don’t take a reference on the 
device being returned. So that seems dangerous. I’m probably missing 
something.

Alexey, Alistair: what, if anything, guarantees, that the npu or gpu 
devices stay valid. Is it because we simply don’t provide any means to 
get rid of them ? Otherwise, don’t we need the callers of 
pnv_pci_get_gpu_dev() and pnv_pci_get_npu_dev() to worry about reference 
counting ? I’ve started looking into it and the changes are scary, which 
explains Greg’s related commit 02c5f5394918b.

To be honest the reference counting looks like it has evolved into something 
quite suspect and I don't think you're missing anything. In practice though we 
likely haven't hit any issues because the original callers didn't store 
references to the pdev which would make the window quite small (although the 
pass through work may have changed that). And as you say there simply wasn't 
any means to get rid of them anyway (EEH, hotplug, etc. has never been 
implemented or supported for GPUs and all sorts of terrible things happen if 
you try).

The best solution would likely be to review the reference counting and to 
teach callers of get_*_dev() to call pci_put_dev(), etc.

- Alistair

   Fred

quoted

quoted

   Fred

quoted

quoted

               npu_pdn = pci_get_pdn(npu_pdev);
               rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
               npu_pdn->pe_number = pe_num;

On Mon, Nov 18, 2019 at 12:06 PM Alistair Popple [off-list ref] wrote:

On Wednesday, 13 November 2019 4:38:21 AM AEDT Frederic Barrat wrote:

quoted

However, one question is whether this patch breaks nvlink and if nvlink
assumes the devices won’t go away because we explicitly take a reference
forever. In npu_dma.c, there are 2 functions which allow to find the GPU
associated to a npu device, and vice-versa. Both functions return a
pointer to a struct pci_dev, but they don’t take a reference on the
device being returned. So that seems dangerous. I’m probably missing
something.

Alexey, Alistair: what, if anything, guarantees, that the npu or gpu
devices stay valid. Is it because we simply don’t provide any means to
get rid of them ? Otherwise, don’t we need the callers of
pnv_pci_get_gpu_dev() and pnv_pci_get_npu_dev() to worry about reference
counting ? I’ve started looking into it and the changes are scary, which
explains Greg’s related commit 02c5f5394918b.

To be honest the reference counting looks like it has evolved into something
quite suspect and I don't think you're missing anything. In practice though we
likely haven't hit any issues because the original callers didn't store
references to the pdev which would make the window quite small (although the
pass through work may have changed that). And as you say there simply wasn't
any means to get rid of them anyway (EEH, hotplug, etc. has never been
implemented or supported for GPUs and all sorts of terrible things happen if
you try).

In other words: leaking a ref is the only safe thing to do.

The best solution would likely be to review the reference counting and to
teach callers of get_*_dev() to call pci_put_dev(), etc.

The issue is that the two callers of get_pci_dev() are non-GPL
exported symbols so we don't know what's calling them or what their
expectations are. We be doing whatever makes sense for OpenCAPI and if
that happens to cause a problem for someone else, then they can deal
with it.

Oliver

On Monday, 18 November 2019 12:24:24 PM AEDT Oliver O'Halloran wrote:

On Mon, Nov 18, 2019 at 12:06 PM Alistair Popple [off-list ref] 

wrote:

quoted

On Wednesday, 13 November 2019 4:38:21 AM AEDT Frederic Barrat wrote:

quoted

However, one question is whether this patch breaks nvlink and if nvlink
assumes the devices won’t go away because we explicitly take a reference
forever. In npu_dma.c, there are 2 functions which allow to find the GPU
associated to a npu device, and vice-versa. Both functions return a
pointer to a struct pci_dev, but they don’t take a reference on the
device being returned. So that seems dangerous. I’m probably missing
something.

Alexey, Alistair: what, if anything, guarantees, that the npu or gpu
devices stay valid. Is it because we simply don’t provide any means to
get rid of them ? Otherwise, don’t we need the callers of
pnv_pci_get_gpu_dev() and pnv_pci_get_npu_dev() to worry about reference
counting ? I’ve started looking into it and the changes are scary, which
explains Greg’s related commit 02c5f5394918b.

To be honest the reference counting looks like it has evolved into 

something

quoted

quite suspect and I don't think you're missing anything. In practice 

though we

quoted

likely haven't hit any issues because the original callers didn't store
references to the pdev which would make the window quite small (although 

the

quoted

pass through work may have changed that). And as you say there simply 

wasn't

quoted

any means to get rid of them anyway (EEH, hotplug, etc. has never been
implemented or supported for GPUs and all sorts of terrible things happen 

if

quoted

you try).

In other words: leaking a ref is the only safe thing to do.

Correct.

quoted

The best solution would likely be to review the reference counting and to
teach callers of get_*_dev() to call pci_put_dev(), etc.

The issue is that the two callers of get_pci_dev() are non-GPL
exported symbols so we don't know what's calling them or what their
expectations are. We be doing whatever makes sense for OpenCAPI and if
that happens to cause a problem for someone else, then they can deal
with it.

The issue isn't that it's exported non-GPL vs. GPL, rather that there are 
probably out-of-tree modules like the NVIDIA driver using it. However as you 
say supporting out-of-tree modules is not generally a concern for kernel 
developers so we certainly shouldn't let that prevent us doing a fix.

- Alistair

Oliver

Le 18/11/2019 à 03:36, Alistair Popple a écrit :

On Monday, 18 November 2019 12:24:24 PM AEDT Oliver O'Halloran wrote:

quoted

On Mon, Nov 18, 2019 at 12:06 PM Alistair Popple [off-list ref]

wrote:

quoted

quoted

On Wednesday, 13 November 2019 4:38:21 AM AEDT Frederic Barrat wrote:

quoted

However, one question is whether this patch breaks nvlink and if nvlink
assumes the devices won’t go away because we explicitly take a reference
forever. In npu_dma.c, there are 2 functions which allow to find the GPU
associated to a npu device, and vice-versa. Both functions return a
pointer to a struct pci_dev, but they don’t take a reference on the
device being returned. So that seems dangerous. I’m probably missing
something.

Alexey, Alistair: what, if anything, guarantees, that the npu or gpu
devices stay valid. Is it because we simply don’t provide any means to
get rid of them ? Otherwise, don’t we need the callers of
pnv_pci_get_gpu_dev() and pnv_pci_get_npu_dev() to worry about reference
counting ? I’ve started looking into it and the changes are scary, which
explains Greg’s related commit 02c5f5394918b.

To be honest the reference counting looks like it has evolved into

something

quoted

quoted

quite suspect and I don't think you're missing anything. In practice

though we

quoted

quoted

likely haven't hit any issues because the original callers didn't store
references to the pdev which would make the window quite small (although

the

quoted

quoted

pass through work may have changed that). And as you say there simply

wasn't

quoted

quoted

any means to get rid of them anyway (EEH, hotplug, etc. has never been
implemented or supported for GPUs and all sorts of terrible things happen

if

quoted

quoted

you try).

In other words: leaking a ref is the only safe thing to do.

Correct.

quoted

quoted

The best solution would likely be to review the reference counting and to
teach callers of get_*_dev() to call pci_put_dev(), etc.

The issue is that the two callers of get_pci_dev() are non-GPL
exported symbols so we don't know what's calling them or what their
expectations are. We be doing whatever makes sense for OpenCAPI and if
that happens to cause a problem for someone else, then they can deal
with it.

The issue isn't that it's exported non-GPL vs. GPL, rather that there are
probably out-of-tree modules like the NVIDIA driver using it. However as you
say supporting out-of-tree modules is not generally a concern for kernel
developers so we certainly shouldn't let that prevent us doing a fix.

Thanks Alistair and Oliver. I'll bite the bullet and try do the right 
thing wrt ref counting in npu-dma.c

   Fred

- Alistair

quoted

Oliver

The PE for an opencapi device was set as part of a late PHB fixup
operation, when creating the PHB. To use the PCI hotplug framework,
this is not going to work, as the PHB stays the same, it's only the
devices underneath which are updated. For regular PCI devices, it is
done as part of the reconfiguration of the bridge, but for opencapi
PHBs, we don't have an intermediate bridge. So let's define the PE
when the device is enabled. PEs are meaningless for opencapi, the NPU
doesn't define them and opal is not doing anything with them.

Signed-off-by: Frederic Barrat <redacted>
---
 arch/powerpc/platforms/powernv/pci-ioda.c | 31 +++++++++++++++++------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index 3dbbf5365c1c..06ce7ddaa0cf 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1260,8 +1260,6 @@ static void pnv_pci_ioda_setup_PEs(void)
 {
 	struct pci_controller *hose;
 	struct pnv_phb *phb;
-	struct pci_bus *bus;
-	struct pci_dev *pdev;
 	struct pnv_ioda_pe *pe;
 
 	list_for_each_entry(hose, &hose_list, list_node) {

@@ -1273,11 +1271,6 @@ static void pnv_pci_ioda_setup_PEs(void)
 			if (phb->model == PNV_PHB_MODEL_NPU2)
 				WARN_ON_ONCE(pnv_npu2_init(hose));
 		}
-		if (phb->type == PNV_PHB_NPU_OCAPI) {
-			bus = hose->bus;
-			list_for_each_entry(pdev, &bus->devices, bus_list)
-				pnv_ioda_setup_dev_PE(pdev);
-		}
 	}
 	list_for_each_entry(hose, &hose_list, list_node) {
 		phb = hose->private_data;

@@ -3385,6 +3378,28 @@ static bool pnv_pci_enable_device_hook(struct pci_dev *dev)
 	return true;
 }
 
+static bool pnv_ocapi_enable_device_hook(struct pci_dev *dev)
+{
+	struct pci_controller *hose = pci_bus_to_host(dev->bus);
+	struct pnv_phb *phb = hose->private_data;
+	struct pci_dn *pdn;
+	struct pnv_ioda_pe *pe;
+
+	if (!phb->initialized)
+		return true;
+
+	pdn = pci_get_pdn(dev);
+	if (!pdn)
+		return false;
+
+	if (pdn->pe_number == IODA_INVALID_PE) {
+		pe = pnv_ioda_setup_dev_PE(dev);
+		if (!pe)
+			return false;
+	}
+	return true;
+}
+
 static long pnv_pci_ioda1_unset_window(struct iommu_table_group *table_group,
 				       int num)
 {

@@ -3625,7 +3640,7 @@ static const struct pci_controller_ops pnv_npu_ioda_controller_ops = {
 };
 
 static const struct pci_controller_ops pnv_npu_ocapi_ioda_controller_ops = {
-	.enable_device_hook	= pnv_pci_enable_device_hook,
+	.enable_device_hook	= pnv_ocapi_enable_device_hook,
 	.window_alignment	= pnv_pci_window_alignment,
 	.reset_secondary_bus	= pnv_pci_reset_secondary_bus,
 	.shutdown		= pnv_pci_ioda_shutdown,

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 9/9/19 5:45 pm, Frederic Barrat wrote:

The PE for an opencapi device was set as part of a late PHB fixup
operation, when creating the PHB. To use the PCI hotplug framework,
this is not going to work, as the PHB stays the same, it's only the
devices underneath which are updated. For regular PCI devices, it is
done as part of the reconfiguration of the bridge, but for opencapi
PHBs, we don't have an intermediate bridge. So let's define the PE
when the device is enabled. PEs are meaningless for opencapi, the NPU
doesn't define them and opal is not doing anything with them.

Signed-off-by: Frederic Barrat <redacted>

Reviewed-by: Andrew Donnellan <redacted>

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

With hotplug, an opencapi device can now go away. It needs to be
released, mostly to clean up its PE state. We were previously not
defining any device callback. We can reuse the standard PCI release
callback, it does a bit too much for an opencapi device, but it's
harmless, and only needs minor tuning.

Also separate the undo of the PELT-V code in a separate function, it
is not needed for NPU devices and it improves a bit the readability of
the code.

Signed-off-by: Frederic Barrat <redacted>
---
 arch/powerpc/platforms/powernv/pci-ioda.c | 59 +++++++++++++++--------
 1 file changed, 39 insertions(+), 20 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index 06ce7ddaa0cf..e5895c05efae 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -188,7 +188,7 @@ static void pnv_ioda_free_pe(struct pnv_ioda_pe *pe)
 	unsigned int pe_num = pe->pe_number;
 
 	WARN_ON(pe->pdev);
-	WARN_ON(pe->npucomp); /* NPUs are not supposed to be freed */
+	WARN_ON(pe->npucomp); /* NPUs for nvlink are not supposed to be freed */
 	kfree(pe->npucomp);
 	memset(pe, 0, sizeof(struct pnv_ioda_pe));
 	clear_bit(pe_num, phb->ioda.pe_alloc);

@@ -777,6 +777,34 @@ static int pnv_ioda_set_peltv(struct pnv_phb *phb,
 	return 0;
 }
 
+static void pnv_ioda_unset_peltv(struct pnv_phb *phb,
+				 struct pnv_ioda_pe *pe,
+				 struct pci_dev *parent)
+{
+	int64_t rc;
+
+	while (parent) {
+		struct pci_dn *pdn = pci_get_pdn(parent);
+
+		if (pdn && pdn->pe_number != IODA_INVALID_PE) {
+			rc = opal_pci_set_peltv(phb->opal_id, pdn->pe_number,
+						pe->pe_number,
+						OPAL_REMOVE_PE_FROM_DOMAIN);
+			/* XXX What to do in case of error ? */
+		}
+		parent = parent->bus->self;
+	}
+
+	opal_pci_eeh_freeze_clear(phb->opal_id, pe->pe_number,
+				  OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+
+	/* Disassociate PE in PELT */
+	rc = opal_pci_set_peltv(phb->opal_id, pe->pe_number,
+				pe->pe_number, OPAL_REMOVE_PE_FROM_DOMAIN);
+	if (rc)
+		pe_warn(pe, "OPAL error %lld remove self from PELTV\n", rc);
+}
+
 static int pnv_ioda_deconfigure_pe(struct pnv_phb *phb, struct pnv_ioda_pe *pe)
 {
 	struct pci_dev *parent;

@@ -827,25 +855,13 @@ static int pnv_ioda_deconfigure_pe(struct pnv_phb *phb, struct pnv_ioda_pe *pe)
 	for (rid = pe->rid; rid < rid_end; rid++)
 		phb->ioda.pe_rmap[rid] = IODA_INVALID_PE;
 
-	/* Release from all parents PELT-V */
-	while (parent) {
-		struct pci_dn *pdn = pci_get_pdn(parent);
-		if (pdn && pdn->pe_number != IODA_INVALID_PE) {
-			rc = opal_pci_set_peltv(phb->opal_id, pdn->pe_number,
-						pe->pe_number, OPAL_REMOVE_PE_FROM_DOMAIN);
-			/* XXX What to do in case of error ? */
-		}
-		parent = parent->bus->self;
-	}
-
-	opal_pci_eeh_freeze_clear(phb->opal_id, pe->pe_number,
-				  OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+	/*
+	 * Release from all parents PELT-V. NPUs don't have a PELTV
+	 * table
+	 */
+	if (phb->type != PNV_PHB_NPU_NVLINK && phb->type != PNV_PHB_NPU_OCAPI)
+		pnv_ioda_unset_peltv(phb, pe, parent);
 
-	/* Disassociate PE in PELT */
-	rc = opal_pci_set_peltv(phb->opal_id, pe->pe_number,
-				pe->pe_number, OPAL_REMOVE_PE_FROM_DOMAIN);
-	if (rc)
-		pe_warn(pe, "OPAL error %lld remove self from PELTV\n", rc);
 	rc = opal_pci_set_pe(phb->opal_id, pe->pe_number, pe->rid,
 			     bcomp, dcomp, fcomp, OPAL_UNMAP_PE);
 	if (rc)

@@ -3540,6 +3556,8 @@ static void pnv_ioda_release_pe(struct pnv_ioda_pe *pe)
 	case PNV_PHB_IODA2:
 		pnv_pci_ioda2_release_pe_dma(pe);
 		break;
+	case PNV_PHB_NPU_OCAPI:
+		break;
 	default:
 		WARN_ON(1);
 	}

@@ -3592,7 +3610,7 @@ static void pnv_pci_release_device(struct pci_dev *pdev)
 	pe = &phb->ioda.pe_array[pdn->pe_number];
 	pdn->pe_number = IODA_INVALID_PE;
 
-	WARN_ON(--pe->device_count < 0);
+	WARN_ON((pe->flags != PNV_IODA_PE_DEV) && (--pe->device_count < 0));
 	if (pe->device_count == 0)
 		pnv_ioda_release_pe(pe);
 }

@@ -3641,6 +3659,7 @@ static const struct pci_controller_ops pnv_npu_ioda_controller_ops = {
 
 static const struct pci_controller_ops pnv_npu_ocapi_ioda_controller_ops = {
 	.enable_device_hook	= pnv_ocapi_enable_device_hook,
+	.release_device		= pnv_pci_release_device,
 	.window_alignment	= pnv_pci_window_alignment,
 	.reset_secondary_bus	= pnv_pci_reset_secondary_bus,
 	.shutdown		= pnv_pci_ioda_shutdown,

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Can we take the opportunity to address this comment?

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

Le 10/09/2019 à 02:56, Alastair D'Silva a écrit :

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

quoted

With hotplug, an opencapi device can now go away. It needs to be
released, mostly to clean up its PE state. We were previously not
defining any device callback. We can reuse the standard PCI release
callback, it does a bit too much for an opencapi device, but it's
harmless, and only needs minor tuning.

Also separate the undo of the PELT-V code in a separate function, it
is not needed for NPU devices and it improves a bit the readability
of
the code.

Signed-off-by: Frederic Barrat <redacted>
---
  arch/powerpc/platforms/powernv/pci-ioda.c | 59 +++++++++++++++----
----
  1 file changed, 39 insertions(+), 20 deletions(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c

b/arch/powerpc/platforms/powernv/pci-ioda.c
index 06ce7ddaa0cf..e5895c05efae 100644

--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -188,7 +188,7 @@ static void pnv_ioda_free_pe(struct pnv_ioda_pe

*pe)
  	unsigned int pe_num = pe->pe_number;
  
  	WARN_ON(pe->pdev);
-	WARN_ON(pe->npucomp); /* NPUs are not supposed to be freed */
+	WARN_ON(pe->npucomp); /* NPUs for nvlink are not supposed to be
freed */
  	kfree(pe->npucomp);
  	memset(pe, 0, sizeof(struct pnv_ioda_pe));
  	clear_bit(pe_num, phb->ioda.pe_alloc);

@@ -777,6 +777,34 @@ static int pnv_ioda_set_peltv(struct pnv_phb

*phb,
  	return 0;
  }
  
+static void pnv_ioda_unset_peltv(struct pnv_phb *phb,
+				 struct pnv_ioda_pe *pe,
+				 struct pci_dev *parent)
+{
+	int64_t rc;
+
+	while (parent) {
+		struct pci_dn *pdn = pci_get_pdn(parent);
+
+		if (pdn && pdn->pe_number != IODA_INVALID_PE) {
+			rc = opal_pci_set_peltv(phb->opal_id, pdn-

quoted

pe_number,

+						pe->pe_number,
+						OPAL_REMOVE_PE_FROM_DOM
AIN);
+			/* XXX What to do in case of error ? */

Can we take the opportunity to address this comment?

Probably like the original author, I'm not sure what we could do better 
here. We already log a warning below if opal returns an error when 
dissociating the PE (admittedly not the same thing as the parent). Note 
that this code is not executed for an opencapi device, so I'm just 
keeping thing as is.

  Fred

quoted

+		}
+		parent = parent->bus->self;
+	}
+
+	opal_pci_eeh_freeze_clear(phb->opal_id, pe->pe_number,
+				  OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+
+	/* Disassociate PE in PELT */
+	rc = opal_pci_set_peltv(phb->opal_id, pe->pe_number,
+				pe->pe_number,
OPAL_REMOVE_PE_FROM_DOMAIN);
+	if (rc)
+		pe_warn(pe, "OPAL error %lld remove self from PELTV\n",
rc);
+}
+
  static int pnv_ioda_deconfigure_pe(struct pnv_phb *phb, struct
pnv_ioda_pe *pe)
  {
  	struct pci_dev *parent;

@@ -827,25 +855,13 @@ static int pnv_ioda_deconfigure_pe(struct

pnv_phb *phb, struct pnv_ioda_pe *pe)
  	for (rid = pe->rid; rid < rid_end; rid++)
  		phb->ioda.pe_rmap[rid] = IODA_INVALID_PE;
  
-	/* Release from all parents PELT-V */
-	while (parent) {
-		struct pci_dn *pdn = pci_get_pdn(parent);
-		if (pdn && pdn->pe_number != IODA_INVALID_PE) {
-			rc = opal_pci_set_peltv(phb->opal_id, pdn-

quoted

pe_number,

-						pe->pe_number,
OPAL_REMOVE_PE_FROM_DOMAIN);
-			/* XXX What to do in case of error ? */
-		}
-		parent = parent->bus->self;
-	}
-
-	opal_pci_eeh_freeze_clear(phb->opal_id, pe->pe_number,
-				  OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+	/*
+	 * Release from all parents PELT-V. NPUs don't have a PELTV
+	 * table
+	 */
+	if (phb->type != PNV_PHB_NPU_NVLINK && phb->type !=
PNV_PHB_NPU_OCAPI)
+		pnv_ioda_unset_peltv(phb, pe, parent);
  
-	/* Disassociate PE in PELT */
-	rc = opal_pci_set_peltv(phb->opal_id, pe->pe_number,
-				pe->pe_number,
OPAL_REMOVE_PE_FROM_DOMAIN);
-	if (rc)
-		pe_warn(pe, "OPAL error %lld remove self from PELTV\n",
rc);
  	rc = opal_pci_set_pe(phb->opal_id, pe->pe_number, pe->rid,
  			     bcomp, dcomp, fcomp, OPAL_UNMAP_PE);
  	if (rc)

@@ -3540,6 +3556,8 @@ static void pnv_ioda_release_pe(struct

pnv_ioda_pe *pe)
  	case PNV_PHB_IODA2:
  		pnv_pci_ioda2_release_pe_dma(pe);
  		break;
+	case PNV_PHB_NPU_OCAPI:
+		break;
  	default:
  		WARN_ON(1);
  	}

@@ -3592,7 +3610,7 @@ static void pnv_pci_release_device(struct

pci_dev *pdev)
  	pe = &phb->ioda.pe_array[pdn->pe_number];
  	pdn->pe_number = IODA_INVALID_PE;
  
-	WARN_ON(--pe->device_count < 0);
+	WARN_ON((pe->flags != PNV_IODA_PE_DEV) && (--pe->device_count <
0));
  	if (pe->device_count == 0)
  		pnv_ioda_release_pe(pe);
  }

@@ -3641,6 +3659,7 @@ static const struct pci_controller_ops

pnv_npu_ioda_controller_ops = {
  
  static const struct pci_controller_ops
pnv_npu_ocapi_ioda_controller_ops = {
  	.enable_device_hook	= pnv_ocapi_enable_device_hook,
+	.release_device		= pnv_pci_release_device,
  	.window_alignment	= pnv_pci_window_alignment,
  	.reset_secondary_bus	= pnv_pci_reset_secondary_bus,
  	.shutdown		= pnv_pci_ioda_shutdown,

On 10/9/19 1:45 am, Frederic Barrat wrote:

With hotplug, an opencapi device can now go away. It needs to be
released, mostly to clean up its PE state. We were previously not
defining any device callback. We can reuse the standard PCI release
callback, it does a bit too much for an opencapi device, but it's
harmless, and only needs minor tuning.

Also separate the undo of the PELT-V code in a separate function, it
is not needed for NPU devices and it improves a bit the readability of
the code.

Signed-off-by: Frederic Barrat <redacted>

Reviewed-by: Andrew Donnellan <redacted>

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Protect the PHB's list of PE. Probably not needed as long as it was
populated during PHB creation, but it feels right and will become
required once we can add/remove opencapi devices on hotplug.

Signed-off-by: Frederic Barrat <redacted>
---
 arch/powerpc/platforms/powernv/pci-ioda.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
index 92767f006f20..3dbbf5365c1c 100644
--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1080,8 +1080,9 @@ static struct pnv_ioda_pe *pnv_ioda_setup_dev_PE(struct pci_dev *dev)
 	}
 
 	/* Put PE to the list */
+	mutex_lock(&phb->ioda.pe_list_mutex);
 	list_add_tail(&pe->list, &phb->ioda.pe_list);
-
+	mutex_unlock(&phb->ioda.pe_list_mutex);
 	return pe;
 }
 

@@ -3513,7 +3514,10 @@ static void pnv_ioda_release_pe(struct pnv_ioda_pe *pe)
 	struct pnv_phb *phb = pe->phb;
 	struct pnv_ioda_pe *slave, *tmp;
 
+	mutex_lock(&phb->ioda.pe_list_mutex);
 	list_del(&pe->list);
+	mutex_unlock(&phb->ioda.pe_list_mutex);
+
 	switch (phb->type) {
 	case PNV_PHB_IODA1:
 		pnv_pci_ioda1_release_pe_dma(pe);

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Hmm, the ioda.pe_list_mutex muxtex exists, and is inited, but there are
no other users. It's position & naming in the struct suggests it
belongs to ioda.pe_list, rather than pnv_ioda_pe.list (as suggested by
the lock/unlock around the list del).

Do the other accessors of ioda.pe_list also need mutex protection?
pnv_ioda_setup_bus_PE()
pnv_pci_dma_bus_setup()
pnv_pci_init_ioda_phb()
pnv_pci_ioda_setup_PEs()

If not, perhaps the metux should be removed from ioda and replaced with
pe.list_mutex instead.

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

Le 10/09/2019 à 02:34, Alastair D'Silva a écrit :

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

quoted

Protect the PHB's list of PE. Probably not needed as long as it was
populated during PHB creation, but it feels right and will become
required once we can add/remove opencapi devices on hotplug.

Signed-off-by: Frederic Barrat <redacted>
---
  arch/powerpc/platforms/powernv/pci-ioda.c | 6 +++++-
  1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c

b/arch/powerpc/platforms/powernv/pci-ioda.c
index 92767f006f20..3dbbf5365c1c 100644

--- a/arch/powerpc/platforms/powernv/pci-ioda.c
+++ b/arch/powerpc/platforms/powernv/pci-ioda.c

@@ -1080,8 +1080,9 @@ static struct pnv_ioda_pe

*pnv_ioda_setup_dev_PE(struct pci_dev *dev)
  	}
  
  	/* Put PE to the list */
+	mutex_lock(&phb->ioda.pe_list_mutex);
  	list_add_tail(&pe->list, &phb->ioda.pe_list);
-
+	mutex_unlock(&phb->ioda.pe_list_mutex);
  	return pe;
  }
  

@@ -3513,7 +3514,10 @@ static void pnv_ioda_release_pe(struct

pnv_ioda_pe *pe)
  	struct pnv_phb *phb = pe->phb;
  	struct pnv_ioda_pe *slave, *tmp;
  
+	mutex_lock(&phb->ioda.pe_list_mutex);
  	list_del(&pe->list);
+	mutex_unlock(&phb->ioda.pe_list_mutex);
+
  	switch (phb->type) {
  	case PNV_PHB_IODA1:
  		pnv_pci_ioda1_release_pe_dma(pe);

Hmm, the ioda.pe_list_mutex muxtex exists, and is inited, but there are
no other users. It's position & naming in the struct suggests it
belongs to ioda.pe_list, rather than pnv_ioda_pe.list (as suggested by
the lock/unlock around the list del).

I don't quite understand the above. The mutex is already in use by the 
functions handling virtual functions, pnv_ioda_setup_vf_PE() and 
pnv_ioda_release_vf_PE(). The point of the list is to keep track of all 
the PEs used by the PHB, so it makes sense to have the mutex at that level.

Do the other accessors of ioda.pe_list also need mutex protection?
pnv_ioda_setup_bus_PE()
pnv_pci_dma_bus_setup()
pnv_pci_init_ioda_phb()
pnv_pci_ioda_setup_PEs()

I think we could also use it there, it wouldn't hurt. Those functions 
are called when the kernel is building part of the PCI topology, and 
devices are not really active yet, so I don't think it's absolutely 
required.

I'm actually not sure my patch is needed either. With hotplug, the 
devices can come and go, whereas the PHB remains. So it feels right to 
start protecting the list when adding/removing a device. But I don't 
think we can really have concurrency and have 2 different operations 
adding/removing devices at the same time under the same PHB, at least 
for opencapi. Maybe for PCI, if we have multiple slots under the same 
PHB. Not sure.

   Fred

If not, perhaps the metux should be removed from ioda and replaced with
pe.list_mutex instead.

On Tue, Nov 19, 2019 at 11:57 PM Frederic Barrat [off-list ref] wrote:

quoted

Do the other accessors of ioda.pe_list also need mutex protection?
pnv_ioda_setup_bus_PE()
pnv_pci_dma_bus_setup()
pnv_pci_init_ioda_phb()
pnv_pci_ioda_setup_PEs()

I think we could also use it there, it wouldn't hurt. Those functions
are called when the kernel is building part of the PCI topology, and
devices are not really active yet, so I don't think it's absolutely
required.

I'm actually not sure my patch is needed either. With hotplug, the
devices can come and go, whereas the PHB remains. So it feels right to
start protecting the list when adding/removing a device. But I don't
think we can really have concurrency and have 2 different operations
adding/removing devices at the same time under the same PHB, at least
for opencapi. Maybe for PCI, if we have multiple slots under the same
PHB. Not sure.

Creation of new pci_dev's is serialised by the global pci rescan and
remove lock so on the creation side it's not an issue. However, we can
release IODA PEs in the pci_dev's release function which might be
called without that lock held. It's pretty hard to hit that case
though since it require something to be holding a ref to the pci_dev
even after the driver's ->remove() function has been called.

Le 19/11/2019 à 14:22, Oliver O'Halloran a écrit :

On Tue, Nov 19, 2019 at 11:57 PM Frederic Barrat [off-list ref] wrote:

quoted

quoted

Do the other accessors of ioda.pe_list also need mutex protection?
pnv_ioda_setup_bus_PE()
pnv_pci_dma_bus_setup()
pnv_pci_init_ioda_phb()
pnv_pci_ioda_setup_PEs()

I think we could also use it there, it wouldn't hurt. Those functions
are called when the kernel is building part of the PCI topology, and
devices are not really active yet, so I don't think it's absolutely
required.

I'm actually not sure my patch is needed either. With hotplug, the
devices can come and go, whereas the PHB remains. So it feels right to
start protecting the list when adding/removing a device. But I don't
think we can really have concurrency and have 2 different operations
adding/removing devices at the same time under the same PHB, at least
for opencapi. Maybe for PCI, if we have multiple slots under the same
PHB. Not sure.

Creation of new pci_dev's is serialised by the global pci rescan and
remove lock so on the creation side it's not an issue. However, we can
release IODA PEs in the pci_dev's release function which might be
called without that lock held. It's pretty hard to hit that case
though since it require something to be holding a ref to the pci_dev
even after the driver's ->remove() function has been called.

Thanks for clarifying it! Indeed, the pci_rescan_remove_lock in hotplug 
helps for the PCI case. I guess we can keep this patch for that hard to 
hit case, just in case it doesn't blow up somewhere else.

   Fred

On 10/9/19 1:45 am, Frederic Barrat wrote:

Protect the PHB's list of PE. Probably not needed as long as it was
populated during PHB creation, but it feels right and will become
required once we can add/remove opencapi devices on hotplug.

Signed-off-by: Frederic Barrat <redacted>

Reviewed-by: Andrew Donnellan <redacted>


-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Unlike real PCI slots, opencapi slots are directly associated to
the (virtual) opencapi PHB, there's no intermediate bridge. So when
looking for a slot ID, we must start the search from the device node
itself and not its parent.

Also, the slot ID is not attached to a specific bdfn, so let's build
it from the PHB ID, like skiboot.

Signed-off-by: Frederic Barrat <redacted>
---
 arch/powerpc/include/asm/pnv-pci.h   |  1 +
 arch/powerpc/platforms/powernv/pci.c | 10 +++++++---
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/include/asm/pnv-pci.h b/arch/powerpc/include/asm/pnv-pci.h
index edcb1fc50aeb..d0ee0ede5767 100644
--- a/arch/powerpc/include/asm/pnv-pci.h
+++ b/arch/powerpc/include/asm/pnv-pci.h

@@ -15,6 +15,7 @@
 #define PCI_SLOT_ID_PREFIX	(1UL << 63)
 #define PCI_SLOT_ID(phb_id, bdfn)	\
 	(PCI_SLOT_ID_PREFIX | ((uint64_t)(bdfn) << 16) | (phb_id))
+#define PCI_PHB_SLOT_ID(phb_id)		(phb_id)
 
 extern int pnv_pci_get_slot_id(struct device_node *np, uint64_t *id);
 extern int pnv_pci_get_device_tree(uint32_t phandle, void *buf, uint64_t len);

diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c
index 6104418c9ad5..00a79f3c989f 100644
--- a/arch/powerpc/platforms/powernv/pci.c
+++ b/arch/powerpc/platforms/powernv/pci.c

@@ -48,13 +48,14 @@ int pnv_pci_get_slot_id(struct device_node *np, uint64_t *id)
 		return -ENXIO;
 
 	bdfn = ((bdfn & 0x00ffff00) >> 8);
-	while ((parent = of_get_parent(parent))) {
+	for (parent = np; parent; parent = of_get_parent(parent)) {
 		if (!PCI_DN(parent)) {
 			of_node_put(parent);
 			break;
 		}
 
-		if (!of_device_is_compatible(parent, "ibm,ioda2-phb")) {
+		if (!of_device_is_compatible(parent, "ibm,ioda2-phb") &&
+		    !of_device_is_compatible(parent, "ibm,ioda2-npu2-opencapi-phb")) {
 			of_node_put(parent);
 			continue;
 		}

@@ -65,7 +66,10 @@ int pnv_pci_get_slot_id(struct device_node *np, uint64_t *id)
 			return -ENXIO;
 		}
 
-		*id = PCI_SLOT_ID(phbid, bdfn);
+		if (of_device_is_compatible(parent, "ibm,ioda2-npu2-opencapi-phb"))
+			*id = PCI_PHB_SLOT_ID(phbid);
+		else
+			*id = PCI_SLOT_ID(phbid, bdfn);
 		return 0;
 	}
 

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 10/9/19 1:45 am, Frederic Barrat wrote:

I think we should rename this pointer from "parent" as it no longer 
means that.

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Le 19/11/2019 à 02:26, Andrew Donnellan a écrit :

On 10/9/19 1:45 am, Frederic Barrat wrote:

quoted

diff --git a/arch/powerpc/platforms/powernv/pci.c 

b/arch/powerpc/platforms/powernv/pci.c
index 6104418c9ad5..00a79f3c989f 100644

--- a/arch/powerpc/platforms/powernv/pci.c
+++ b/arch/powerpc/platforms/powernv/pci.c

@@ -48,13 +48,14 @@ int pnv_pci_get_slot_id(struct device_node *np, 

uint64_t *id)
          return -ENXIO;
      bdfn = ((bdfn & 0x00ffff00) >> 8);
-    while ((parent = of_get_parent(parent))) {
+    for (parent = np; parent; parent = of_get_parent(parent)) {

I think we should rename this pointer from "parent" as it no longer 
means that.

ok. We still walk the parent list, but we don't start from the first 
parent any more. I'll rename with a more generic "node" to avoid confusion.

   Fred

The PCI hotplug framework is used to update the devices when a new
image is written to the FPGA.

Signed-off-by: Frederic Barrat <redacted>
---
 drivers/misc/ocxl/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/misc/ocxl/Kconfig b/drivers/misc/ocxl/Kconfig
index 1916fa65f2f2..2d2266c1439e 100644
--- a/drivers/misc/ocxl/Kconfig
+++ b/drivers/misc/ocxl/Kconfig

@@ -11,6 +11,7 @@ config OCXL
 	tristate "OpenCAPI coherent accelerator support"
 	depends on PPC_POWERNV && PCI && EEH
 	select OCXL_BASE
+	select HOTPLUG_PCI_POWERNV
 	default m
 	help
 	  Select this option to enable the ocxl driver for Open

-- 
2.21.0

On Mon, 2019-09-09 at 17:46 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 9/9/19 5:46 pm, Frederic Barrat wrote:

The PCI hotplug framework is used to update the devices when a new
image is written to the FPGA.

Signed-off-by: Frederic Barrat <redacted>

Acked-by: Andrew Donnellan <redacted>

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

When changing the slot state, if opal hits an error and tells as such
in the asynchronous reply, the warning "Wrong msg" is logged, which is
rather confusing. Instead we can reuse the better message which is
already used when we couldn't submit the asynchronous opal request
initially.

Signed-off-by: Frederic Barrat <redacted>
---
 drivers/pci/hotplug/pnv_php.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/pci/hotplug/pnv_php.c b/drivers/pci/hotplug/pnv_php.c
index 5b5cbf1e636d..304bdbcdb77c 100644
--- a/drivers/pci/hotplug/pnv_php.c
+++ b/drivers/pci/hotplug/pnv_php.c

@@ -336,18 +336,19 @@ int pnv_php_set_slot_power_state(struct hotplug_slot *slot,
 	ret = pnv_pci_set_power_state(php_slot->id, state, &msg);
 	if (ret > 0) {
 		if (be64_to_cpu(msg.params[1]) != php_slot->dn->phandle	||
-		    be64_to_cpu(msg.params[2]) != state			||
-		    be64_to_cpu(msg.params[3]) != OPAL_SUCCESS) {
+		    be64_to_cpu(msg.params[2]) != state) {
 			pci_warn(php_slot->pdev, "Wrong msg (%lld, %lld, %lld)\n",
 				 be64_to_cpu(msg.params[1]),
 				 be64_to_cpu(msg.params[2]),
 				 be64_to_cpu(msg.params[3]));
 			return -ENOMSG;
 		}
+		if (be64_to_cpu(msg.params[3]) != OPAL_SUCCESS) {
+			ret = -ENODEV;
+			goto error;
+		}
 	} else if (ret < 0) {
-		pci_warn(php_slot->pdev, "Error %d powering %s\n",
-			 ret, (state == OPAL_PCI_SLOT_POWER_ON) ? "on" : "off");
-		return ret;
+		goto error;
 	}
 
 	if (state == OPAL_PCI_SLOT_POWER_OFF || state == OPAL_PCI_SLOT_OFFLINE)

@@ -356,6 +357,11 @@ int pnv_php_set_slot_power_state(struct hotplug_slot *slot,
 		ret = pnv_php_add_devtree(php_slot);
 
 	return ret;
+
+error:
+	pci_warn(php_slot->pdev, "Error %d powering %s\n",
+		 ret, (state == OPAL_PCI_SLOT_POWER_ON) ? "on" : "off");
+	return ret;
 }
 EXPORT_SYMBOL_GPL(pnv_php_set_slot_power_state);
 

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 10/9/19 1:45 am, Frederic Barrat wrote:

When changing the slot state, if opal hits an error and tells as such
in the asynchronous reply, the warning "Wrong msg" is logged, which is
rather confusing. Instead we can reuse the better message which is
already used when we couldn't submit the asynchronous opal request
initially.

Signed-off-by: Frederic Barrat <redacted>

Reviewed-by: Andrew Donnellan <redacted>

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

The driver only allows to disable a slot in the POPULATED
state. However, if an error occurs while enabling the slot, say
because the link couldn't be trained, then the POPULATED state may not
be reached, yet the power state of the slot is on. So allow to disable
a slot in the REGISTERED state. Removing the devices will do nothing
since it's not populated, and we'll set the power state of the slot
back to off.

Signed-off-by: Frederic Barrat <redacted>
---
 drivers/pci/hotplug/pnv_php.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/hotplug/pnv_php.c b/drivers/pci/hotplug/pnv_php.c
index f0a7360154e7..5ca51d67db4b 100644
--- a/drivers/pci/hotplug/pnv_php.c
+++ b/drivers/pci/hotplug/pnv_php.c

@@ -523,7 +523,13 @@ static int pnv_php_disable_slot(struct hotplug_slot *slot)
 	struct pnv_php_slot *php_slot = to_pnv_php_slot(slot);
 	int ret;
 
-	if (php_slot->state != PNV_PHP_STATE_POPULATED)
+	/*
+	 * Allow to disable a slot already in the registered state to
+	 * cover cases where the slot couldn't be enabled and never
+	 * reached the populated state
+	 */
+	if (php_slot->state != PNV_PHP_STATE_POPULATED &&
+	    php_slot->state != PNV_PHP_STATE_REGISTERED)
 		return 0;
 
 	/* Remove all devices behind the slot */

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 9/9/19 5:45 pm, Frederic Barrat wrote:

The driver only allows to disable a slot in the POPULATED
state. However, if an error occurs while enabling the slot, say
because the link couldn't be trained, then the POPULATED state may not
be reached, yet the power state of the slot is on. So allow to disable
a slot in the REGISTERED state. Removing the devices will do nothing
since it's not populated, and we'll set the power state of the slot
back to off.

Signed-off-by: Frederic Barrat <redacted>

Makes sense

Reviewed-by: Andrew Donnellan <redacted>

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Add the opencapi PHBs to the list of PHBs being scanned to look for
slots.

Signed-off-by: Frederic Barrat <redacted>
---
 drivers/pci/hotplug/pnv_php.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/hotplug/pnv_php.c b/drivers/pci/hotplug/pnv_php.c
index 304bdbcdb77c..f0a7360154e7 100644
--- a/drivers/pci/hotplug/pnv_php.c
+++ b/drivers/pci/hotplug/pnv_php.c

@@ -954,7 +954,8 @@ static int __init pnv_php_init(void)
 	pr_info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
 	for_each_compatible_node(dn, NULL, "ibm,ioda2-phb")
 		pnv_php_register(dn);
-
+	for_each_compatible_node(dn, NULL, "ibm,ioda2-npu2-opencapi-phb")
+		pnv_php_register_one(dn);
 	return 0;
 }
 

@@ -964,6 +965,8 @@ static void __exit pnv_php_exit(void)
 
 	for_each_compatible_node(dn, NULL, "ibm,ioda2-phb")
 		pnv_php_unregister(dn);
+	for_each_compatible_node(dn, NULL, "ibm,ioda2-npu2-opencapi-phb")
+		pnv_php_unregister(dn);
 }
 
 module_init(pnv_php_init);

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 10/9/19 1:45 am, Frederic Barrat wrote:

Why do we use register_one to register and unregister rather than 
unregister_one to unregister?

-- 
Andrew Donnellan              OzLabs, ADL Canberra
ajd@linux.ibm.com             IBM Australia Limited

Le 19/11/2019 à 06:18, Andrew Donnellan a écrit :

On 10/9/19 1:45 am, Frederic Barrat wrote:

quoted

Add the opencapi PHBs to the list of PHBs being scanned to look for
slots.

Signed-off-by: Frederic Barrat <redacted>
---
  drivers/pci/hotplug/pnv_php.c | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/pci/hotplug/pnv_php.c 

b/drivers/pci/hotplug/pnv_php.c
index 304bdbcdb77c..f0a7360154e7 100644

--- a/drivers/pci/hotplug/pnv_php.c
+++ b/drivers/pci/hotplug/pnv_php.c

@@ -954,7 +954,8 @@ static int __init pnv_php_init(void)

      pr_info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
      for_each_compatible_node(dn, NULL, "ibm,ioda2-phb")
          pnv_php_register(dn);
-
+    for_each_compatible_node(dn, NULL, "ibm,ioda2-npu2-opencapi-phb")
+        pnv_php_register_one(dn);
      return 0;
  }

@@ -964,6 +965,8 @@ static void __exit pnv_php_exit(void)

      for_each_compatible_node(dn, NULL, "ibm,ioda2-phb")
          pnv_php_unregister(dn);
+    for_each_compatible_node(dn, NULL, "ibm,ioda2-npu2-opencapi-phb")
+        pnv_php_unregister(dn);
  }

Why do we use register_one to register and unregister rather than 
unregister_one to unregister?

Good catch! With the above, the slot was not removed. 
pnv_php_unregister() looks at the children only and was missing the 
opencapi slot, since it's directly under the PHB.

   Fred

An opencapi slot doesn't have an associated bridge device. It's not
needed for operation, but any warning is displayed through pci_warn()
which uses the pci_dev struct of the assocated bridge device. So wrap
those warning so that a different trace mechanism can be used if it's
an opencapi slot.

Signed-off-by: Frederic Barrat <redacted>
---
 drivers/pci/hotplug/pnv_php.c | 51 +++++++++++++++++++----------------
 1 file changed, 28 insertions(+), 23 deletions(-)

diff --git a/drivers/pci/hotplug/pnv_php.c b/drivers/pci/hotplug/pnv_php.c
index 5ca51d67db4b..d01a8595bc5c 100644
--- a/drivers/pci/hotplug/pnv_php.c
+++ b/drivers/pci/hotplug/pnv_php.c

@@ -18,6 +18,9 @@
 #define DRIVER_AUTHOR	"Gavin Shan, IBM Corporation"
 #define DRIVER_DESC	"PowerPC PowerNV PCI Hotplug Driver"
 
+#define SLOT_WARN(sl, x...) \
+	((sl)->pdev ? pci_warn((sl)->pdev, x) : dev_warn(&(sl)->bus->dev, x))
+
 struct pnv_php_event {
 	bool			added;
 	struct pnv_php_slot	*php_slot;

@@ -265,7 +268,7 @@ static int pnv_php_add_devtree(struct pnv_php_slot *php_slot)
 
 	ret = pnv_pci_get_device_tree(php_slot->dn->phandle, fdt1, 0x10000);
 	if (ret) {
-		pci_warn(php_slot->pdev, "Error %d getting FDT blob\n", ret);
+		SLOT_WARN(php_slot, "Error %d getting FDT blob\n", ret);
 		goto free_fdt1;
 	}
 

@@ -279,7 +282,7 @@ static int pnv_php_add_devtree(struct pnv_php_slot *php_slot)
 	dt = of_fdt_unflatten_tree(fdt, php_slot->dn, NULL);
 	if (!dt) {
 		ret = -EINVAL;
-		pci_warn(php_slot->pdev, "Cannot unflatten FDT\n");
+		SLOT_WARN(php_slot, "Cannot unflatten FDT\n");
 		goto free_fdt;
 	}
 

@@ -289,15 +292,15 @@ static int pnv_php_add_devtree(struct pnv_php_slot *php_slot)
 	ret = pnv_php_populate_changeset(&php_slot->ocs, php_slot->dn);
 	if (ret) {
 		pnv_php_reverse_nodes(php_slot->dn);
-		pci_warn(php_slot->pdev, "Error %d populating changeset\n",
-			 ret);
+		SLOT_WARN(php_slot, "Error %d populating changeset\n",
+			  ret);
 		goto free_dt;
 	}
 
 	php_slot->dn->child = NULL;
 	ret = of_changeset_apply(&php_slot->ocs);
 	if (ret) {
-		pci_warn(php_slot->pdev, "Error %d applying changeset\n", ret);
+		SLOT_WARN(php_slot, "Error %d applying changeset\n", ret);
 		goto destroy_changeset;
 	}
 

@@ -337,10 +340,10 @@ int pnv_php_set_slot_power_state(struct hotplug_slot *slot,
 	if (ret > 0) {
 		if (be64_to_cpu(msg.params[1]) != php_slot->dn->phandle	||
 		    be64_to_cpu(msg.params[2]) != state) {
-			pci_warn(php_slot->pdev, "Wrong msg (%lld, %lld, %lld)\n",
-				 be64_to_cpu(msg.params[1]),
-				 be64_to_cpu(msg.params[2]),
-				 be64_to_cpu(msg.params[3]));
+			SLOT_WARN(php_slot, "Wrong msg (%lld, %lld, %lld)\n",
+				  be64_to_cpu(msg.params[1]),
+				  be64_to_cpu(msg.params[2]),
+				  be64_to_cpu(msg.params[3]));
 			return -ENOMSG;
 		}
 		if (be64_to_cpu(msg.params[3]) != OPAL_SUCCESS) {

@@ -359,8 +362,8 @@ int pnv_php_set_slot_power_state(struct hotplug_slot *slot,
 	return ret;
 
 error:
-	pci_warn(php_slot->pdev, "Error %d powering %s\n",
-		 ret, (state == OPAL_PCI_SLOT_POWER_ON) ? "on" : "off");
+	SLOT_WARN(php_slot, "Error %d powering %s\n",
+		  ret, (state == OPAL_PCI_SLOT_POWER_ON) ? "on" : "off");
 	return ret;
 }
 EXPORT_SYMBOL_GPL(pnv_php_set_slot_power_state);

@@ -378,8 +381,8 @@ static int pnv_php_get_power_state(struct hotplug_slot *slot, u8 *state)
 	 */
 	ret = pnv_pci_get_power_state(php_slot->id, &power_state);
 	if (ret) {
-		pci_warn(php_slot->pdev, "Error %d getting power status\n",
-			 ret);
+		SLOT_WARN(php_slot, "Error %d getting power status\n",
+			  ret);
 	} else {
 		*state = power_state;
 	}

@@ -402,7 +405,7 @@ static int pnv_php_get_adapter_state(struct hotplug_slot *slot, u8 *state)
 		*state = presence;
 		ret = 0;
 	} else {
-		pci_warn(php_slot->pdev, "Error %d getting presence\n", ret);
+		SLOT_WARN(php_slot, "Error %d getting presence\n", ret);
 	}
 
 	return ret;

@@ -637,7 +640,7 @@ static int pnv_php_register_slot(struct pnv_php_slot *php_slot)
 	ret = pci_hp_register(&php_slot->slot, php_slot->bus,
 			      php_slot->slot_no, php_slot->name);
 	if (ret) {
-		pci_warn(php_slot->pdev, "Error %d registering slot\n", ret);
+		SLOT_WARN(php_slot, "Error %d registering slot\n", ret);
 		return ret;
 	}
 

@@ -690,7 +693,7 @@ static int pnv_php_enable_msix(struct pnv_php_slot *php_slot)
 	/* Enable MSIx */
 	ret = pci_enable_msix_exact(pdev, &entry, 1);
 	if (ret) {
-		pci_warn(pdev, "Error %d enabling MSIx\n", ret);
+		SLOT_WARN(php_slot, "Error %d enabling MSIx\n", ret);
 		return ret;
 	}
 

@@ -734,8 +737,9 @@ static irqreturn_t pnv_php_interrupt(int irq, void *data)
 		   (sts & PCI_EXP_SLTSTA_PDC)) {
 		ret = pnv_pci_get_presence_state(php_slot->id, &presence);
 		if (ret) {
-			pci_warn(pdev, "PCI slot [%s] error %d getting presence (0x%04x), to retry the operation.\n",
-				 php_slot->name, ret, sts);
+			SLOT_WARN(php_slot,
+				  "PCI slot [%s] error %d getting presence (0x%04x), to retry the operation.\n",
+				  php_slot->name, ret, sts);
 			return IRQ_HANDLED;
 		}
 

@@ -764,8 +768,9 @@ static irqreturn_t pnv_php_interrupt(int irq, void *data)
 	 */
 	event = kzalloc(sizeof(*event), GFP_ATOMIC);
 	if (!event) {
-		pci_warn(pdev, "PCI slot [%s] missed hotplug event 0x%04x\n",
-			 php_slot->name, sts);
+		SLOT_WARN(php_slot,
+			  "PCI slot [%s] missed hotplug event 0x%04x\n",
+			  php_slot->name, sts);
 		return IRQ_HANDLED;
 	}
 

@@ -789,7 +794,7 @@ static void pnv_php_init_irq(struct pnv_php_slot *php_slot, int irq)
 	/* Allocate workqueue */
 	php_slot->wq = alloc_workqueue("pciehp-%s", 0, 0, php_slot->name);
 	if (!php_slot->wq) {
-		pci_warn(pdev, "Cannot alloc workqueue\n");
+		SLOT_WARN(php_slot, "Cannot alloc workqueue\n");
 		pnv_php_disable_irq(php_slot, true);
 		return;
 	}

@@ -813,7 +818,7 @@ static void pnv_php_init_irq(struct pnv_php_slot *php_slot, int irq)
 			  php_slot->name, php_slot);
 	if (ret) {
 		pnv_php_disable_irq(php_slot, true);
-		pci_warn(pdev, "Error %d enabling IRQ %d\n", ret, irq);
+		SLOT_WARN(php_slot, "Error %d enabling IRQ %d\n", ret, irq);
 		return;
 	}
 

@@ -849,7 +854,7 @@ static void pnv_php_enable_irq(struct pnv_php_slot *php_slot)
 
 	ret = pci_enable_device(pdev);
 	if (ret) {
-		pci_warn(pdev, "Error %d enabling device\n", ret);
+		SLOT_WARN(php_slot, "Error %d enabling device\n", ret);
 		return;
 	}
 

-- 
2.21.0

On Mon, 2019-09-09 at 17:45 +0200, Frederic Barrat wrote:

Reviewed-by: Alastair D'Silva <redacted>

-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australia
mob: 0423 762 819

On 9/9/19 5:45 pm, Frederic Barrat wrote:

An opencapi slot doesn't have an associated bridge device. It's not
needed for operation, but any warning is displayed through pci_warn()
which uses the pci_dev struct of the assocated bridge device. So wrap
those warning so that a different trace mechanism can be used if it's
an opencapi slot.

Signed-off-by: Frederic Barrat <redacted>

Reviewed-by: Andrew Donnellan <redacted>