[Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Thomas Wiedemann]

[-- Attachment #1: Type: text/plain, Size: 304 bytes --]

Hi,

we had a problem compiling the CAN-Driver for the last release candidate 
when
shared interrupts had been enabled, because of the re-named option
"CONFIG_XENO_OPT_SHIRQ_LEVEL.". After a quick look, version 2.4.0 still
doesn't fix this. A patch is included (for ksrc/drivers/can).


Greets,
Thomas



[-- Attachment #2: patch.txt --]
[-- Type: text/plain, Size: 1507 bytes --]

diff -urN sja1000.orig/rtcan_ems_pci.c sja1000/rtcan_ems_pci.c
--- sja1000.orig/rtcan_ems_pci.c	2007-09-16 19:20:45.000000000 +0200
+++ sja1000/rtcan_ems_pci.c	2007-12-10 10:14:58.000000000 +0100
@@ -272,7 +272,7 @@
 	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_MASTER,
 					  &master_dev)))
 		goto failure_cleanup;
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
+#ifdef CONFIG_XENO_OPT_SHIRQ
 	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_SLAVE,
 					  &master_dev)))
 		goto failure_cleanup;
diff -urN sja1000.orig/rtcan_ixxat_pci.c sja1000/rtcan_ixxat_pci.c
--- sja1000.orig/rtcan_ixxat_pci.c	2007-09-16 19:20:45.000000000 +0200
+++ sja1000/rtcan_ixxat_pci.c	2007-12-10 10:14:54.000000000 +0100
@@ -254,7 +254,7 @@
 	goto failure_iounmap;
 
     if (channel != CHANNEL_SINGLE) {
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
+#ifdef CONFIG_XENO_OPT_SHIRQ
 	channel = CHANNEL_SLAVE;
 	if ((ret = rtcan_ixxat_pci_add_chan(pdev, channel,
 					    &master_dev, conf_addr,
diff -urN sja1000.orig/rtcan_peak_pci.c sja1000/rtcan_peak_pci.c
--- sja1000.orig/rtcan_peak_pci.c	2007-09-16 19:20:45.000000000 +0200
+++ sja1000/rtcan_peak_pci.c	2007-12-10 10:14:51.000000000 +0100
@@ -292,7 +292,7 @@
 	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_MASTER, 
 					   &master_dev)))
 	    goto failure_cleanup;
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
+#ifdef CONFIG_XENO_OPT_SHIRQ
 	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_SLAVE, 
 					   &master_dev)))
 	    goto failure_cleanup;
=

Re: [Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Wolfgang Grandegger]

Thomas Wiedemann wrote:
> Hi,
> 
> we had a problem compiling the CAN-Driver for the last release candidate 
> when
> shared interrupts had been enabled, because of the re-named option
> "CONFIG_XENO_OPT_SHIRQ_LEVEL.". After a quick look, version 2.4.0 still
> doesn't fix this. A patch is included (for ksrc/drivers/can).

Ah, thanks for reporting. I'm going to remove these #ifdefs a.s.a.p.
It's sufficient to print an appropriate error messsage in case interrupt
registration fails with -EBUSY.

Wolfgang.

> ------------------------------------------------------------------------
> 
> diff -urN sja1000.orig/rtcan_ems_pci.c sja1000/rtcan_ems_pci.c
> --- sja1000.orig/rtcan_ems_pci.c	2007-09-16 19:20:45.000000000 +0200
> +++ sja1000/rtcan_ems_pci.c	2007-12-10 10:14:58.000000000 +0100
> @@ -272,7 +272,7 @@
>  	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_MASTER,
>  					  &master_dev)))
>  		goto failure_cleanup;
> -#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
> +#ifdef CONFIG_XENO_OPT_SHIRQ
>  	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_SLAVE,
>  					  &master_dev)))
>  		goto failure_cleanup;
> diff -urN sja1000.orig/rtcan_ixxat_pci.c sja1000/rtcan_ixxat_pci.c
> --- sja1000.orig/rtcan_ixxat_pci.c	2007-09-16 19:20:45.000000000 +0200
> +++ sja1000/rtcan_ixxat_pci.c	2007-12-10 10:14:54.000000000 +0100
> @@ -254,7 +254,7 @@
>  	goto failure_iounmap;
>  
>      if (channel != CHANNEL_SINGLE) {
> -#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
> +#ifdef CONFIG_XENO_OPT_SHIRQ
>  	channel = CHANNEL_SLAVE;
>  	if ((ret = rtcan_ixxat_pci_add_chan(pdev, channel,
>  					    &master_dev, conf_addr,
> diff -urN sja1000.orig/rtcan_peak_pci.c sja1000/rtcan_peak_pci.c
> --- sja1000.orig/rtcan_peak_pci.c	2007-09-16 19:20:45.000000000 +0200
> +++ sja1000/rtcan_peak_pci.c	2007-12-10 10:14:51.000000000 +0100
> @@ -292,7 +292,7 @@
>  	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_MASTER, 
>  					   &master_dev)))
>  	    goto failure_cleanup;
> -#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
> +#ifdef CONFIG_XENO_OPT_SHIRQ
>  	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_SLAVE, 
>  					   &master_dev)))
>  	    goto failure_cleanup;
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Xenomai-core mailing list
> Xenomai-core@domain.hid
> https://mail.gna.org/listinfo/xenomai-core

Re: [Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Wolfgang Grandegger]

[-- Attachment #1: Type: text/plain, Size: 1169 bytes --]

Wolfgang Grandegger wrote:
> Thomas Wiedemann wrote:
>> Hi,
>>
>> we had a problem compiling the CAN-Driver for the last release candidate 
>> when
>> shared interrupts had been enabled, because of the re-named option
>> "CONFIG_XENO_OPT_SHIRQ_LEVEL.". After a quick look, version 2.4.0 still
>> doesn't fix this. A patch is included (for ksrc/drivers/can).
> 
> Ah, thanks for reporting. I'm going to remove these #ifdefs a.s.a.p.
> It's sufficient to print an appropriate error messsage in case interrupt
> registration fails with -EBUSY.

The attached patch fixes this issue. Here is the ChangeLog entry:

2007-12-17  Wolfgang Grandegger  <wg@domain.hid>

  * ksrc/drivers/can/sja1000/{rtcan_sja1000.c, rtcan_*_pci.c}: Remove
  CONFIG_XENO_OPT_SHIRQ* ifdefs to fix build problems with recent
  versions of Xenomai. If rtdm_irq_request() now fails with -EBUSY, an
  error message reminds the user to check for shared interrupt support.

  * ksrc/drivers/can/rtcan_dev.c (rtcan_dev_unregister): unregister
  devices properly in case refcounts are disabled.

If there are no objections, I'm going to apply this patch to Xenomai
v2.3.x, v2.4.x and the trunk.

Wolfgang.

[-- Attachment #2: xenomai-rtcan-shirq-fix.patch --]
[-- Type: text/x-patch, Size: 6263 bytes --]

Index: ChangeLog
===================================================================
--- ChangeLog	(revision 3274)
+++ ChangeLog	(working copy)
@@ -1,3 +1,13 @@
+2007-12-17  Wolfgang Grandegger  <wg@domain.hid>
+
+	* ksrc/drivers/can/sja1000/{rtcan_sja1000.c, rtcan_*_pci.c}: Remove
+	CONFIG_XENO_OPT_SHIRQ* ifdefs to fix build problems with recent
+	versions of Xenomai. If rtdm_irq_request() now fails with -EBUSY, an
+	error message reminds the user to check for shared interrupt support.
+
+	* ksrc/drivers/can/rtcan_dev.c (rtcan_dev_unregister): unregister
+	devices properly in case refcounts are disabled.
+
 2007-12-09  Philippe Gerum  <rpm@xenomai.org>
 
 	* include/native/misc.h: Do not run the auto-cleanup code for I/O
Index: ksrc/drivers/can/rtcan_dev.c
===================================================================
--- ksrc/drivers/can/rtcan_dev.c	(revision 3274)
+++ ksrc/drivers/can/rtcan_dev.c	(working copy)
@@ -260,7 +260,6 @@ int rtcan_dev_register(struct rtcan_devi
 
 int rtcan_dev_unregister(struct rtcan_device *dev)
 {
-#ifdef RTCAN_USE_REFCOUNT
     rtdm_lockctx_t context;
 
 
@@ -278,6 +277,7 @@ int rtcan_dev_unregister(struct rtcan_de
 
     rtdm_lock_get_irqsave(&rtcan_devices_rt_lock, context);
 
+#ifdef RTCAN_USE_REFCOUNT
     while (atomic_read(&dev->refcount) > 0) {
         rtdm_lock_put_irqrestore(&rtcan_devices_rt_lock, context);
         up(&rtcan_devices_nrt_lock);
@@ -290,21 +290,20 @@ int rtcan_dev_unregister(struct rtcan_de
         down(&rtcan_devices_nrt_lock);
         rtdm_lock_get_irqsave(&rtcan_devices_rt_lock, context);
     }
+#endif
     rtcan_devices[dev->ifindex - 1] = NULL;
 
     rtdm_lock_put_irqrestore(&rtcan_devices_rt_lock, context);
     up(&rtcan_devices_nrt_lock);
 
+#ifdef RTCAN_USE_REFCOUNT
     RTCAN_ASSERT(atomic_read(&dev->refcount) == 0,
 		 printk("RTCAN: dev reference counter < 0!\n"););
+#endif
 
     printk("RTCAN: unregistered %s\n", dev->name);
 
     return 0;
-#else
-    printk("RTCAN: Oops, unexpected call of rtcan_dev_unregister()\n");
-    return -EINVAL;
-#endif
 }
 
 
Index: ksrc/drivers/can/sja1000/rtcan_ixxat_pci.c
===================================================================
--- ksrc/drivers/can/sja1000/rtcan_ixxat_pci.c	(revision 3274)
+++ ksrc/drivers/can/sja1000/rtcan_ixxat_pci.c	(working copy)
@@ -186,7 +186,7 @@ static int rtcan_ixxat_pci_add_chan(stru
     /* Register SJA1000 device */
     ret = rtcan_sja1000_register(dev);
     if (ret) {
-	printk(KERN_ERR "ERROR while trying to register SJA1000 device %d!\n",
+	printk(KERN_ERR "ERROR %d while trying to register SJA1000 device!\n",
 	       ret);
 	goto failure;
     }
@@ -254,15 +254,11 @@ static int __devinit ixxat_pci_init_one 
 	goto failure_iounmap;
 
     if (channel != CHANNEL_SINGLE) {
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
 	channel = CHANNEL_SLAVE;
 	if ((ret = rtcan_ixxat_pci_add_chan(pdev, channel,
 					    &master_dev, conf_addr,
 					    base_addr + CHANNEL_OFFSET)))
 	    goto failure_iounmap;
-#else
-	printk("Shared interrupts not enabled, using single channel!\n");
-#endif
     }
 
     pci_set_drvdata(pdev, master_dev);
Index: ksrc/drivers/can/sja1000/rtcan_peak_pci.c
===================================================================
--- ksrc/drivers/can/sja1000/rtcan_peak_pci.c	(revision 3274)
+++ ksrc/drivers/can/sja1000/rtcan_peak_pci.c	(working copy)
@@ -247,10 +247,10 @@ static int rtcan_peak_pci_add_chan(struc
     /* Register SJA1000 device */
     ret = rtcan_sja1000_register(dev);
     if (ret) {
-	printk(KERN_ERR "ERROR while trying to register SJA1000 device %d!\n",
-	       ret);
+	printk(KERN_ERR
+	       "ERROR %d while trying to register SJA1000 device!\n", ret);
 	goto failure;
-    }    
+    }
 
     if (channel != CHANNEL_SLAVE)
 	*master_dev = dev;
@@ -292,13 +292,9 @@ static int __devinit peak_pci_init_one (
 	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_MASTER, 
 					   &master_dev)))
 	    goto failure_cleanup;
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
 	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_SLAVE, 
 					   &master_dev)))
 	    goto failure_cleanup;
-#else
-	printk("Shared interrupts not enabled, using single channel!\n");
-#endif
     } else {
 	if ((ret = rtcan_peak_pci_add_chan(pdev, CHANNEL_SINGLE,
 					   &master_dev)))
Index: ksrc/drivers/can/sja1000/rtcan_sja1000.c
===================================================================
--- ksrc/drivers/can/sja1000/rtcan_sja1000.c	(revision 3274)
+++ ksrc/drivers/can/sja1000/rtcan_sja1000.c	(working copy)
@@ -763,7 +763,9 @@ int rtcan_sja1000_register(struct rtcan_
 			   chip->irq_num, rtcan_sja_interrupt,
 			   chip->irq_flags, sja_ctrl_name, dev);
     if (ret) {
-	printk("ERROR! IRQ %d busy or invalid (code=%d)!\n", chip->irq_num, ret);
+	printk(KERN_ERR "ERROR %d: IRQ %d is %s!\n",
+	       ret, chip->irq_num, ret == -EBUSY ?
+	       "busy, check shared interrupt support" : "invalid");
 	return ret;
     }
 
@@ -772,7 +774,8 @@ int rtcan_sja1000_register(struct rtcan_
     /* Register RTDM device */
     ret = rtcan_dev_register(dev);
     if (ret) {
-	printk(KERN_ERR "ERROR while trying to register RTCAN device!\n");
+	    printk(KERN_ERR
+		   "ERROR %d while trying to register RTCAN device!\n", ret);
         goto out_irq_free;
     }
 
Index: ksrc/drivers/can/sja1000/rtcan_ems_pci.c
===================================================================
--- ksrc/drivers/can/sja1000/rtcan_ems_pci.c	(revision 3274)
+++ ksrc/drivers/can/sja1000/rtcan_ems_pci.c	(working copy)
@@ -236,7 +236,7 @@ static int rtcan_ems_pci_add_chan(struct
 	err = rtcan_sja1000_register(dev);
 	if (err) {
 		printk(KERN_ERR
-		       "ERROR while trying to register SJA1000 device %d!\n",
+		       "ERROR %d while trying to register SJA1000 device!\n",
 		       err);
 		goto failure;
 	}
@@ -272,13 +272,9 @@ static int __devinit ems_pci_init_one (s
 	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_MASTER,
 					  &master_dev)))
 		goto failure_cleanup;
-#ifdef CONFIG_XENO_OPT_SHIRQ_LEVEL
 	if ((err = rtcan_ems_pci_add_chan(pdev, EMS_PCI_SLAVE,
 					  &master_dev)))
 		goto failure_cleanup;
-#else
-	printk("Shared interrupts not enabled, using single channel!\n");
-#endif
 
 	pci_set_drvdata(pdev, master_dev);
 	return 0;

Re: [Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Philippe Gerum]

Wolfgang Grandegger wrote:
> Wolfgang Grandegger wrote:
>> Thomas Wiedemann wrote:
>>> Hi,
>>>
>>> we had a problem compiling the CAN-Driver for the last release candidate 
>>> when
>>> shared interrupts had been enabled, because of the re-named option
>>> "CONFIG_XENO_OPT_SHIRQ_LEVEL.". After a quick look, version 2.4.0 still
>>> doesn't fix this. A patch is included (for ksrc/drivers/can).
>> Ah, thanks for reporting. I'm going to remove these #ifdefs a.s.a.p.
>> It's sufficient to print an appropriate error messsage in case interrupt
>> registration fails with -EBUSY.
> 
> The attached patch fixes this issue. Here is the ChangeLog entry:
> 
> 2007-12-17  Wolfgang Grandegger  <wg@domain.hid>
> 
>   * ksrc/drivers/can/sja1000/{rtcan_sja1000.c, rtcan_*_pci.c}: Remove
>   CONFIG_XENO_OPT_SHIRQ* ifdefs to fix build problems with recent
>   versions of Xenomai. If rtdm_irq_request() now fails with -EBUSY, an
>   error message reminds the user to check for shared interrupt support.
> 
>   * ksrc/drivers/can/rtcan_dev.c (rtcan_dev_unregister): unregister
>   devices properly in case refcounts are disabled.
> 
> If there are no objections, I'm going to apply this patch to Xenomai
> v2.3.x, v2.4.x and the trunk.
>

Please merge. TIA,

-- 
Philippe.

Re: [Xenomai-core] Antwort: Xenomai v2.4.0: CONFIG_XENO_OPT_SHIRQ_LEVEL vs. CONFIG_XENO_OPT_SHIRQ

[Wolfgang Grandegger]

Philippe Gerum wrote:
> Wolfgang Grandegger wrote:
>> Wolfgang Grandegger wrote:
>>> Thomas Wiedemann wrote:
>>>> Hi,
>>>>
>>>> we had a problem compiling the CAN-Driver for the last release candidate 
>>>> when
>>>> shared interrupts had been enabled, because of the re-named option
>>>> "CONFIG_XENO_OPT_SHIRQ_LEVEL.". After a quick look, version 2.4.0 still
>>>> doesn't fix this. A patch is included (for ksrc/drivers/can).
>>> Ah, thanks for reporting. I'm going to remove these #ifdefs a.s.a.p.
>>> It's sufficient to print an appropriate error messsage in case interrupt
>>> registration fails with -EBUSY.
>> The attached patch fixes this issue. Here is the ChangeLog entry:
>>
>> 2007-12-17  Wolfgang Grandegger  <wg@domain.hid>
>>
>>   * ksrc/drivers/can/sja1000/{rtcan_sja1000.c, rtcan_*_pci.c}: Remove
>>   CONFIG_XENO_OPT_SHIRQ* ifdefs to fix build problems with recent
>>   versions of Xenomai. If rtdm_irq_request() now fails with -EBUSY, an
>>   error message reminds the user to check for shared interrupt support.
>>
>>   * ksrc/drivers/can/rtcan_dev.c (rtcan_dev_unregister): unregister
>>   devices properly in case refcounts are disabled.
>>
>> If there are no objections, I'm going to apply this patch to Xenomai
>> v2.3.x, v2.4.x and the trunk.
>>
> 
> Please merge. TIA,

Done.

Wolfgang.