Skip to contentSkip to conversation

xen/arm: Black list everything with a PPI

PATCH Superseded Amit Singh Tomar 4 participants 7 messages May 3, 2019 – May 29, 2019

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xen/arm: Black list everything with a PPI

XEN should not forward PPIs to Dom0 as it only support SPIs. One of solution to this problem is to skip any device that uses PPI source completely while building domain itself.

This patch goes through all the interrupt sources of device and skip it if one of interrupt source is PPI. It fixes XEN boot on i.MX8MQ by skipping PMU node.

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
 xen/arch/arm/domain_build.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +15 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
         { /* sentinel */ },
     };
     struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
     const char *name;
     const char *path;
 
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
         return 0;
     }
 
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);
+
+        if ( irq_id >= 16 && irq_id < 32 )
+        {
+            dt_dprintk(" Skip node with (PPI source)\n");
+            return 0;
+        }
+    }
+
     /*
      * Xen is using some path for its own purpose. Warn if a node
      * already exists with the same path.
Signature · 7 lines
2.7.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#1 lore raw

On 03.05.19 20:02, Amit Singh Tomar wrote:

Hi, Amit

Quoted 31 lines

XEN should not forward PPIs to Dom0 as it only support SPIs. One of solution to this problem is to skip any device that uses PPI source completely while building domain itself.

This patch goes through all the interrupt sources of device and skip it if one of interrupt source is PPI. It fixes XEN boot on i.MX8MQ by skipping PMU node.

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
  xen/arch/arm/domain_build.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +1 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          { /* sentinel */ },
      };
      struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
      const char *name;
      const char *path;
 

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1407

Quoted 9 lines
xen/arch/arm/domain_build.c +6 -0
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          return 0;
      }
 
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);

Do we need to do something here if platform_get_irq() returns -1?

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1418

Quoted 11 lines
xen/arch/arm/domain_build.c +8 -0
+
+        if ( irq_id >= 16 && irq_id < 32 )
+        {
+            dt_dprintk(" Skip node with (PPI source)\n");
+            return 0;
+        }
+    }
+
      /*
       * Xen is using some path for its own purpose. Warn if a node
       * already exists with the same path.

Patch looks good. Although R-Car H3 seems to not use PPIs for PMU, I have tested your patch just to be sure it wouldn't skip anything by a mistake)

Signature · 9 lines
Regards,

Oleksandr Tyshchenko


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
author
#2 lore raw

Hello,

Thanks for having a look at it.

Quoted 40 lines · Oleksandr

On 03.05.19 20:02, Amit Singh Tomar wrote:

Hi, Amit

Quoted 34 lines

XEN should not forward PPIs to Dom0 as it only support SPIs. One of solution to this problem is to skip any device that uses PPI source completely while building domain itself.

This patch goes through all the interrupt sources of device and skip it if one of interrupt source is PPI. It fixes XEN boot on i.MX8MQ by skipping PMU node.

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
  xen/arch/arm/domain_build.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +1 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          { /* sentinel */ },
      };
      struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
      const char *name;
      const char *path;
 
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          return 0;
      }
 
Quoted 8 lines

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1407

Quoted 6 lines
xen/arch/arm/domain_build.c +6 -0
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);

Do we need to do something here if platform_get_irq() returns -1?

Yeah, I should have done it. some think like following:
https://xenbits.xen.org/gitweb/?p=xen.git;a=blob;f=xen/arch/arm/domain_build.c;h=d9836779d17c90e84b94ba32e4a20f028189fc5b;hb=HEAD#l1284

Quoted 7 lines
Quoted 7 lines
xen/arch/arm/domain_build.c +7 -0
+
+        if ( irq_id >= 16 && irq_id < 32 )
+        {
+            dt_dprintk(" Skip node with (PPI source)\n");
+            return 0;
+        }
+    }
Quoted 8 lines

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1418

Quoted 4 lines
xen/arch/arm/domain_build.c +1 -0
+
      /*
       * Xen is using some path for its own purpose. Warn if a node
       * already exists with the same path.

Patch looks good. Although R-Car H3 seems to not use PPIs for PMU, I have tested your patch just to be sure it wouldn't skip anything by a mistake)

Ok, Thanks for testing it. I would resend it with error condition check.

-Thanks
Amit.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw

On Thu, 16 May 2019 17:15:36 +0530
Amit Tomer <[email protected]> wrote:

Hi,

Quoted 52 lines

Thanks for having a look at it.

Quoted 40 lines · Oleksandr

On 03.05.19 20:02, Amit Singh Tomar wrote:

Hi, Amit

Quoted 34 lines

XEN should not forward PPIs to Dom0 as it only support SPIs. One of solution to this problem is to skip any device that uses PPI source completely while building domain itself.

This patch goes through all the interrupt sources of device and skip it if one of interrupt source is PPI. It fixes XEN boot on i.MX8MQ by skipping PMU node.

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
  xen/arch/arm/domain_build.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +1 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          { /* sentinel */ },
      };
      struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
      const char *name;
      const char *path;
 
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          return 0;
      }
 
Quoted 8 lines

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1407

Quoted 6 lines
xen/arch/arm/domain_build.c +6 -0
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);  

Do we need to do something here if platform_get_irq() returns -1?

Quoted 2 lines

Yeah, I should have done it. some think like following:
https://xenbits.xen.org/gitweb/?p=xen.git;a=blob;f=xen/arch/arm/domain_build.c;h=d9836779d17c90e84b94ba32e4a20f028189fc5b;hb=HEAD#l1284

Why would that (or any actual check against -1) be necessary?
If irq_id is < 0, then it would surely not match the condition below and *nothing* would happen. So I'd say: Keep it like it is, no action necessary.

Quoted 2 lines
Quoted 2 lines

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1409

Quoted 2 lines
xen/arch/arm/domain_build.c +2 -0
+
+        if ( irq_id >= 16 && irq_id < 32 )

Any chance you can put names there? Or at least add a comment that PPIs range from 16 to 31?

Quoted 10 lines
Quoted 10 lines
Quoted 9 lines
xen/arch/arm/domain_build.c +6 -0
+        {
+            dt_dprintk(" Skip node with (PPI source)\n");
+            return 0;
+        }
+    }
+
      /*
       * Xen is using some path for its own purpose. Warn if a node
       * already exists with the same path.  
Quoted 5 lines
Quoted 3 lines

Patch looks good. Although R-Car H3 seems to not use PPIs for PMU, I have tested your patch just to be sure it wouldn't skip anything by a mistake)

Ok, Thanks for testing it. I would resend it with error condition check.

Please don't ;-)

Cheers,
Andre

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#4 lore raw
Quoted 39 lines · Andre Przywara
Quoted 37 lines · On Thu, 16 May 2019 17:15:36 +0530 Amit Tomer <[email protected]> wrote:
Quoted 35 lines · Oleksandr
Quoted 23 lines · Amit Singh Tomar

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
   xen/arch/arm/domain_build.c | 16 +++++++++++++++-
   1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +1 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
           { /* sentinel */ },
       };
       struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
       const char *name;
       const char *path;
 

xen/arch/arm/domain_build.c static int __init handle_node(struct domain *d, struct kernel_info *kinfo, · line 1407

Quoted 9 lines
xen/arch/arm/domain_build.c +6 -0
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
           return 0;
       }
 
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);

Do we need to do something here if platform_get_irq() returns -1?

Quoted 6 lines
Quoted 2 lines

Yeah, I should have done it. some think like following:
https://xenbits.xen.org/gitweb/?p=xen.git;a=blob;f=xen/arch/arm/domain_build.c;h=d9836779d17c90e84b94ba32e4a20f028189fc5b;hb=HEAD#l1284

Why would that (or any actual check against -1) be necessary?
If irq_id is < 0, then it would surely not match the condition below and *nothing* would happen. So I'd say: Keep it like it is, no action necessary.

Worst, depending on the action done with check, you could actively break support for platform with multiple interrupt controllers. That's why in handle_device(), the interrupt controller is checked before calling platform_get_irq().

Cheers,

Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#5 lore raw
In reply to Amit Singh Tomar

Hi Amit,

Just a quick follow-up. Is there any plan to send a new version of this patch?

Cheers,

Quoted 52 lines · Amit Singh Tomar

XEN should not forward PPIs to Dom0 as it only support SPIs. One of solution to this problem is to skip any device that uses PPI source completely while building domain itself.

This patch goes through all the interrupt sources of device and skip it if one of interrupt source is PPI. It fixes XEN boot on i.MX8MQ by skipping PMU node.

---
* This replaces following patch.
https://patchwork.kernel.org/patch/10899881/

---
  xen/arch/arm/domain_build.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)
xen/arch/arm/domain_build.c +15 -1
index d983677..8f54472 100644
@@ -1353,7 +1353,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          { /* sentinel */ },
      };
      struct dt_device_node *child;
-    int res;
+    int res, i, nirq, irq_id;
      const char *name;
      const char *path;
 
@@ -1399,6 +1399,20 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo,
          return 0;
      }
 
+    /* Skip the node, using PPI source */
+    nirq = dt_number_of_irq(node);
+
+    for ( i = 0 ; i < nirq ; i++ )
+    {
+        irq_id = platform_get_irq(node, i);
+
+        if ( irq_id >= 16 && irq_id < 32 )
+        {
+            dt_dprintk(" Skip node with (PPI source)\n");
+            return 0;
+        }
+    }
+
      /*
       * Xen is using some path for its own purpose. Warn if a node
       * already exists with the same path.
Signature · 6 lines
Julien Grall

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel