[RFC PATCH 1/3] misra: address 5.5 pirq_cleanup_check

Dmytro Prokopchuk1 posted 3 patches 3 months ago
[RFC PATCH 1/3] misra: address 5.5 pirq_cleanup_check
Posted by Dmytro Prokopchuk1 3 months ago
Signed-off-by: Dmytro Prokopchuk <dmytro_prokopchuk1@epam.com>
---
 xen/arch/x86/irq.c                | 8 ++++----
 xen/common/event_channel.c        | 6 +++---
 xen/drivers/passthrough/x86/hvm.c | 8 ++++----
 xen/include/xen/irq.h             | 2 +-
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/xen/arch/x86/irq.c b/xen/arch/x86/irq.c
index 556134f85a..e70c7829b4 100644
--- a/xen/arch/x86/irq.c
+++ b/xen/arch/x86/irq.c
@@ -1325,7 +1325,7 @@ static void clear_domain_irq_pirq(struct domain *d, int irq, struct pirq *pirq)
 static void cleanup_domain_irq_pirq(struct domain *d, int irq,
                                     struct pirq *pirq)
 {
-    pirq_cleanup_check(pirq, d);
+    PIRQ_CLEANUP_CHECK(pirq, d);
     radix_tree_delete(&d->arch.irq_pirq, irq);
 }
 
@@ -1383,7 +1383,7 @@ struct pirq *alloc_pirq_struct(struct domain *d)
     return pirq;
 }
 
-void (pirq_cleanup_check)(struct pirq *pirq, struct domain *d)
+void pirq_cleanup_check(struct pirq *pirq, struct domain *d)
 {
     /*
      * Check whether all fields have their default values, and delete
@@ -2823,7 +2823,7 @@ int map_domain_emuirq_pirq(struct domain *d, int pirq, int emuirq)
                 radix_tree_int_to_ptr(pirq));
             break;
         default:
-            pirq_cleanup_check(info, d);
+            PIRQ_CLEANUP_CHECK(info, d);
             return err;
         }
     }
@@ -2858,7 +2858,7 @@ int unmap_domain_pirq_emuirq(struct domain *d, int pirq)
     if ( info )
     {
         info->arch.hvm.emuirq = IRQ_UNBOUND;
-        pirq_cleanup_check(info, d);
+        PIRQ_CLEANUP_CHECK(info, d);
     }
     if ( emuirq != IRQ_PT )
         radix_tree_delete(&d->arch.hvm.emuirq_pirq, emuirq);
diff --git a/xen/common/event_channel.c b/xen/common/event_channel.c
index c8c1bfa615..2efb5f5c78 100644
--- a/xen/common/event_channel.c
+++ b/xen/common/event_channel.c
@@ -672,7 +672,7 @@ static int evtchn_bind_pirq(evtchn_bind_pirq_t *bind)
     if ( rc != 0 )
     {
         info->evtchn = 0;
-        pirq_cleanup_check(info, d);
+        PIRQ_CLEANUP_CHECK(info, d);
         goto out;
     }
 
@@ -743,9 +743,9 @@ int evtchn_close(struct domain *d1, int port1, bool guest)
                  unmap_domain_pirq_emuirq(d1, pirq->pirq) < 0 )
                 /*
                  * The successful path of unmap_domain_pirq_emuirq() will have
-                 * called pirq_cleanup_check() already.
+                 * called PIRQ_CLEANUP_CHECK() already.
                  */
-                pirq_cleanup_check(pirq, d1);
+                PIRQ_CLEANUP_CHECK(pirq, d1);
         }
         unlink_pirq_port(chn1, d1->vcpu[chn1->notify_vcpu_id]);
         break;
diff --git a/xen/drivers/passthrough/x86/hvm.c b/xen/drivers/passthrough/x86/hvm.c
index a2ca7e0e57..1c545ed89d 100644
--- a/xen/drivers/passthrough/x86/hvm.c
+++ b/xen/drivers/passthrough/x86/hvm.c
@@ -296,7 +296,7 @@ int pt_irq_create_bind(
             pirq_dpci->gmsi.gflags = gflags;
             /*
              * 'pt_irq_create_bind' can be called after 'pt_irq_destroy_bind'.
-             * The 'pirq_cleanup_check' which would free the structure is only
+             * The 'PIRQ_CLEANUP_CHECK' which would free the structure is only
              * called if the event channel for the PIRQ is active. However
              * OS-es that use event channels usually bind PIRQs to eventds
              * and unbind them before calling 'pt_irq_destroy_bind' - with the
@@ -329,7 +329,7 @@ int pt_irq_create_bind(
                 pirq_dpci->gmsi.gvec = 0;
                 pirq_dpci->dom = NULL;
                 pirq_dpci->flags = 0;
-                pirq_cleanup_check(info, d);
+                PIRQ_CLEANUP_CHECK(info, d);
                 write_unlock(&d->event_lock);
                 return rc;
             }
@@ -536,7 +536,7 @@ int pt_irq_create_bind(
                     hvm_irq_dpci->link_cnt[link]--;
                 }
                 pirq_dpci->flags = 0;
-                pirq_cleanup_check(info, d);
+                PIRQ_CLEANUP_CHECK(info, d);
                 write_unlock(&d->event_lock);
                 xfree(girq);
                 xfree(digl);
@@ -693,7 +693,7 @@ int pt_irq_destroy_bind(
          */
         pt_pirq_softirq_reset(pirq_dpci);
 
-        pirq_cleanup_check(pirq, d);
+        PIRQ_CLEANUP_CHECK(pirq, d);
     }
 
     write_unlock(&d->event_lock);
diff --git a/xen/include/xen/irq.h b/xen/include/xen/irq.h
index 95034c0d6b..958d0b1aca 100644
--- a/xen/include/xen/irq.h
+++ b/xen/include/xen/irq.h
@@ -185,7 +185,7 @@ extern struct pirq *pirq_get_info(struct domain *d, int pirq);
 
 void pirq_cleanup_check(struct pirq *pirq, struct domain *d);
 
-#define pirq_cleanup_check(pirq, d) \
+#define PIRQ_CLEANUP_CHECK(pirq, d) \
     (!(pirq)->evtchn ? pirq_cleanup_check(pirq, d) : (void)0)
 
 extern void pirq_guest_eoi(struct pirq *pirq);
-- 
2.43.0
Re: [RFC PATCH 1/3] misra: address 5.5 pirq_cleanup_check
Posted by Andrew Cooper 3 months ago
On 29/07/2025 10:24 pm, Dmytro Prokopchuk1 wrote:
> diff --git a/xen/common/event_channel.c b/xen/common/event_channel.c
> index c8c1bfa615..2efb5f5c78 100644
> --- a/xen/common/event_channel.c
> +++ b/xen/common/event_channel.c
> @@ -672,7 +672,7 @@ static int evtchn_bind_pirq(evtchn_bind_pirq_t *bind)
>      if ( rc != 0 )
>      {
>          info->evtchn = 0;
> -        pirq_cleanup_check(info, d);
> +        PIRQ_CLEANUP_CHECK(info, d);

Well, this is awkward.  This is dead code, but only when you realise ...

> diff --git a/xen/include/xen/irq.h b/xen/include/xen/irq.h
> index 95034c0d6b..958d0b1aca 100644
> --- a/xen/include/xen/irq.h
> +++ b/xen/include/xen/irq.h
> @@ -185,7 +185,7 @@ extern struct pirq *pirq_get_info(struct domain *d, int pirq);
>  
>  void pirq_cleanup_check(struct pirq *pirq, struct domain *d);
>  
> -#define pirq_cleanup_check(pirq, d) \
> +#define PIRQ_CLEANUP_CHECK(pirq, d) \
>      (!(pirq)->evtchn ? pirq_cleanup_check(pirq, d) : (void)0)
>  

... what this is really doing.

Looking at this overall diff, it really is outrageous that we're hiding
a conditional call like this.

We should just remove the macro, and expand

    if ( !pirq->evtchn )
        pirq_cleanup_check(pirq, d);

in most of the callsites.  The overall diff will be smaller (no need to
change the comments), and the end result is proper regular normal C.

I can draft a patch to that effect.

~Andrew

Re: [RFC PATCH 1/3] misra: address 5.5 pirq_cleanup_check
Posted by Stefano Stabellini 3 months ago
On Tue, 29 Jul 2025, Andrew Cooper wrote:
> On 29/07/2025 10:24 pm, Dmytro Prokopchuk1 wrote:
> > diff --git a/xen/common/event_channel.c b/xen/common/event_channel.c
> > index c8c1bfa615..2efb5f5c78 100644
> > --- a/xen/common/event_channel.c
> > +++ b/xen/common/event_channel.c
> > @@ -672,7 +672,7 @@ static int evtchn_bind_pirq(evtchn_bind_pirq_t *bind)
> >      if ( rc != 0 )
> >      {
> >          info->evtchn = 0;
> > -        pirq_cleanup_check(info, d);
> > +        PIRQ_CLEANUP_CHECK(info, d);
> 
> Well, this is awkward.  This is dead code, but only when you realise ...
> 
> > diff --git a/xen/include/xen/irq.h b/xen/include/xen/irq.h
> > index 95034c0d6b..958d0b1aca 100644
> > --- a/xen/include/xen/irq.h
> > +++ b/xen/include/xen/irq.h
> > @@ -185,7 +185,7 @@ extern struct pirq *pirq_get_info(struct domain *d, int pirq);
> >  
> >  void pirq_cleanup_check(struct pirq *pirq, struct domain *d);
> >  
> > -#define pirq_cleanup_check(pirq, d) \
> > +#define PIRQ_CLEANUP_CHECK(pirq, d) \
> >      (!(pirq)->evtchn ? pirq_cleanup_check(pirq, d) : (void)0)
> >  
> 
> ... what this is really doing.
> 
> Looking at this overall diff, it really is outrageous that we're hiding
> a conditional call like this.
> 
> We should just remove the macro, and expand
> 
>     if ( !pirq->evtchn )
>         pirq_cleanup_check(pirq, d);
> 
> in most of the callsites.  The overall diff will be smaller (no need to
> change the comments), and the end result is proper regular normal C.
 
Yes, would look much better. +1 from me.