The patch below makes it so the proper /proc/ide files are created when
this driver is loaded. Previously, /proc/ide would be empty even though
the devices could be used, etc. This is creating problems for us because
some installers and daemons expect the proc files to be in place.
We're open to comments but would appreciate this being accepted ASAP
as this problem is affecting our progress on some projects.
I verified this exact patch works on 2.6.9-rc1 (and it applies without
fuzz).
One thing I'll probably hear about is that we create /proc/ide/sgiioc4
but don't produce info and metrics for it. We may elect to produce some
metrics later but this driver is used primarily for the system CDROM and
having metrics there isn't a priority at this moment. So I create the proc
file anyway but it just reports the type of driver it is.
diffstat:
drivers/ide/pci/sgiioc4.c | 33 +++++++++++++++++++++++++++++++++
1 files changed, 33 insertions(+)
Signed-off-by: Erik Jacobson <[email protected]>
diff -Naru linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c linux-2.6.8/drivers/ide/pci/sgiioc4.c
--- linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c 2004-08-14 01:36:58.000000000 -0400
+++ linux-2.6.8/drivers/ide/pci/sgiioc4.c 2004-09-08 08:40:39.624159397 -0400
@@ -34,6 +34,7 @@
#include <linux/mm.h>
#include <linux/ioport.h>
#include <linux/blkdev.h>
+#include <linux/proc_fs.h>
#include <asm/io.h>
#include <linux/ide.h>
@@ -92,6 +93,27 @@
#define IOC4_PRD_BYTES 16
#define IOC4_PRD_ENTRIES (PAGE_SIZE /(4*IOC4_PRD_BYTES))
+/* Used so we only try to create the /proc/ide/sgiioc4 entry once. */
+static u8 sgiioc4_proc;
+
+/* Produce output for /proc/ide/sgiioc4:
+ * At this point, we don't produce any metrics information.
+ * Still, it seems we should create this so we behave like other IDE
+ * drivers do. Perhaps someone will write in metrics information some day
+ * and that can be hooked in here.
+ */
+static int sgiioc4_get_info(char *buffer, char **addr, off_t offset, int count)
+{
+ char *p = buffer;
+ int len;
+
+ p += sprintf(p, "\nSGI IOC4 IDE Driver\n");
+ p += sprintf(p, "This space may be used in the future for metrics.\n");
+
+ len = (p - buffer) - offset;
+ *addr = buffer + offset;
+ return len > count ? count : len;
+}
static void
sgiioc4_init_hwif_ports(hw_regs_t * hw, unsigned long data_port,
@@ -702,6 +724,10 @@
hwif->name, d->name);
probe_hwif_init(hwif);
+
+ /* Create /proc/ide entries */
+ create_proc_ide_interfaces();
+
return 0;
}
@@ -732,6 +758,13 @@
pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
class_rev &= 0xff;
+
+ /* create /proc/ide/sgiioc4 entry */
+ if (!sgiioc4_proc) {
+ ide_pci_create_host_proc("sgiioc4", sgiioc4_get_info);
+ sgiioc4_proc = 1;
+ }
+
printk(KERN_INFO "%s: IDE controller at PCI slot %s, revision %d\n",
d->name, dev->slot_name, class_rev);
if (class_rev < IOC4_SUPPORTED_FIRMWARE_REV) {
On Wednesday 08 September 2004 18:17, Erik Jacobson wrote:
> The patch below makes it so the proper /proc/ide files are created when
> this driver is loaded. Previously, /proc/ide would be empty even though
This patch only adds /proc/ide/sgiioc4 entry.
/proc/ide should be properly populated without it.
> the devices could be used, etc. This is creating problems for us because
> some installers and daemons expect the proc files to be in place.
Please name these installers and daemons.
If your application is checking for /proc/ide/sgiioc4 this is WRONG.
Please tell us what do you need this information for and we can
see what can be done.
> We're open to comments but would appreciate this being accepted ASAP
> as this problem is affecting our progress on some projects.
>
> I verified this exact patch works on 2.6.9-rc1 (and it applies without
> fuzz).
>
> One thing I'll probably hear about is that we create /proc/ide/sgiioc4
> but don't produce info and metrics for it. We may elect to produce some
Actually no, these /proc/ide/<chipset> entries are just bloat and should
be removed - basic info should be always printed on the console (i.e. while
changing transfer mode) and more advanced info should be only printed
while debugging driver.
> metrics later but this driver is used primarily for the system CDROM and
> having metrics there isn't a priority at this moment. So I create the proc
> file anyway but it just reports the type of driver it is.
>
> diffstat:
>
> drivers/ide/pci/sgiioc4.c | 33 +++++++++++++++++++++++++++++++++
> 1 files changed, 33 insertions(+)
>
> Signed-off-by: Erik Jacobson <[email protected]>
>
>
> diff -Naru linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c linux-2.6.8/drivers/ide/pci/sgiioc4.c
> --- linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c 2004-08-14 01:36:58.000000000 -0400
> +++ linux-2.6.8/drivers/ide/pci/sgiioc4.c 2004-09-08 08:40:39.624159397 -0400
> @@ -34,6 +34,7 @@
> #include <linux/mm.h>
> #include <linux/ioport.h>
> #include <linux/blkdev.h>
> +#include <linux/proc_fs.h>
> #include <asm/io.h>
>
> #include <linux/ide.h>
> @@ -92,6 +93,27 @@
> #define IOC4_PRD_BYTES 16
> #define IOC4_PRD_ENTRIES (PAGE_SIZE /(4*IOC4_PRD_BYTES))
>
> +/* Used so we only try to create the /proc/ide/sgiioc4 entry once. */
> +static u8 sgiioc4_proc;
> +
> +/* Produce output for /proc/ide/sgiioc4:
> + * At this point, we don't produce any metrics information.
> + * Still, it seems we should create this so we behave like other IDE
> + * drivers do. Perhaps someone will write in metrics information some day
> + * and that can be hooked in here.
> + */
> +static int sgiioc4_get_info(char *buffer, char **addr, off_t offset, int count)
> +{
> + char *p = buffer;
> + int len;
> +
> + p += sprintf(p, "\nSGI IOC4 IDE Driver\n");
> + p += sprintf(p, "This space may be used in the future for metrics.\n");
> +
> + len = (p - buffer) - offset;
> + *addr = buffer + offset;
> + return len > count ? count : len;
> +}
>
> static void
> sgiioc4_init_hwif_ports(hw_regs_t * hw, unsigned long data_port,
> @@ -702,6 +724,10 @@
> hwif->name, d->name);
>
> probe_hwif_init(hwif);
> +
> + /* Create /proc/ide entries */
> + create_proc_ide_interfaces();
> +
> return 0;
> }
>
> @@ -732,6 +758,13 @@
>
> pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
> class_rev &= 0xff;
> +
> + /* create /proc/ide/sgiioc4 entry */
> + if (!sgiioc4_proc) {
> + ide_pci_create_host_proc("sgiioc4", sgiioc4_get_info);
> + sgiioc4_proc = 1;
> + }
> +
> printk(KERN_INFO "%s: IDE controller at PCI slot %s, revision %d\n",
> d->name, dev->slot_name, class_rev);
> if (class_rev < IOC4_SUPPORTED_FIRMWARE_REV) {
>
>
Hi. Thanks for the response.
I'm fine zapping the /proc/ide/sgiioc4 part of the patch. I didn't want
it anyway - I just put it there because it was what everybody else did.
The important part of the patch is this added line:
create_proc_ide_interfaces();
Without this line, /proc/ide contains only the file "drivers".
I believe this is because the sgiioc4_ide_setup_pci_device function
is our own version of ide_setup_pci_device. Since we don't call
ide_setup_pci_device, create_proc_ide_interfaces is never called
otherwise - so /proc/ide isn't populated properly.
On Wed, 8 Sep 2004, Bartlomiej Zolnierkiewicz wrote:
>
> On Wednesday 08 September 2004 18:17, Erik Jacobson wrote:
> > The patch below makes it so the proper /proc/ide files are created when
> > this driver is loaded. Previously, /proc/ide would be empty even though
>
> This patch only adds /proc/ide/sgiioc4 entry.
> /proc/ide should be properly populated without it.
>
> > the devices could be used, etc. This is creating problems for us because
> > some installers and daemons expect the proc files to be in place.
>
> Please name these installers and daemons.
>
> If your application is checking for /proc/ide/sgiioc4 this is WRONG.
>
> Please tell us what do you need this information for and we can
> see what can be done.
>
> > We're open to comments but would appreciate this being accepted ASAP
> > as this problem is affecting our progress on some projects.
> >
> > I verified this exact patch works on 2.6.9-rc1 (and it applies without
> > fuzz).
> >
> > One thing I'll probably hear about is that we create /proc/ide/sgiioc4
> > but don't produce info and metrics for it. We may elect to produce some
>
> Actually no, these /proc/ide/<chipset> entries are just bloat and should
> be removed - basic info should be always printed on the console (i.e. while
> changing transfer mode) and more advanced info should be only printed
> while debugging driver.
>
> > metrics later but this driver is used primarily for the system CDROM and
> > having metrics there isn't a priority at this moment. So I create the proc
> > file anyway but it just reports the type of driver it is.
> >
> > diffstat:
> >
> > drivers/ide/pci/sgiioc4.c | 33 +++++++++++++++++++++++++++++++++
> > 1 files changed, 33 insertions(+)
> >
> > Signed-off-by: Erik Jacobson <[email protected]>
> >
> >
> > diff -Naru linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c linux-2.6.8/drivers/ide/pci/sgiioc4.c
> > --- linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c 2004-08-14 01:36:58.000000000 -0400
> > +++ linux-2.6.8/drivers/ide/pci/sgiioc4.c 2004-09-08 08:40:39.624159397 -0400
> > @@ -34,6 +34,7 @@
> > #include <linux/mm.h>
> > #include <linux/ioport.h>
> > #include <linux/blkdev.h>
> > +#include <linux/proc_fs.h>
> > #include <asm/io.h>
> >
> > #include <linux/ide.h>
> > @@ -92,6 +93,27 @@
> > #define IOC4_PRD_BYTES 16
> > #define IOC4_PRD_ENTRIES (PAGE_SIZE /(4*IOC4_PRD_BYTES))
> >
> > +/* Used so we only try to create the /proc/ide/sgiioc4 entry once. */
> > +static u8 sgiioc4_proc;
> > +
> > +/* Produce output for /proc/ide/sgiioc4:
> > + * At this point, we don't produce any metrics information.
> > + * Still, it seems we should create this so we behave like other IDE
> > + * drivers do. Perhaps someone will write in metrics information some day
> > + * and that can be hooked in here.
> > + */
> > +static int sgiioc4_get_info(char *buffer, char **addr, off_t offset, int count)
> > +{
> > + char *p = buffer;
> > + int len;
> > +
> > + p += sprintf(p, "\nSGI IOC4 IDE Driver\n");
> > + p += sprintf(p, "This space may be used in the future for metrics.\n");
> > +
> > + len = (p - buffer) - offset;
> > + *addr = buffer + offset;
> > + return len > count ? count : len;
> > +}
> >
> > static void
> > sgiioc4_init_hwif_ports(hw_regs_t * hw, unsigned long data_port,
> > @@ -702,6 +724,10 @@
> > hwif->name, d->name);
> >
> > probe_hwif_init(hwif);
> > +
> > + /* Create /proc/ide entries */
> > + create_proc_ide_interfaces();
> > +
> > return 0;
> > }
> >
> > @@ -732,6 +758,13 @@
> >
> > pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
> > class_rev &= 0xff;
> > +
> > + /* create /proc/ide/sgiioc4 entry */
> > + if (!sgiioc4_proc) {
> > + ide_pci_create_host_proc("sgiioc4", sgiioc4_get_info);
> > + sgiioc4_proc = 1;
> > + }
> > +
> > printk(KERN_INFO "%s: IDE controller at PCI slot %s, revision %d\n",
> > d->name, dev->slot_name, class_rev);
> > if (class_rev < IOC4_SUPPORTED_FIRMWARE_REV) {
> >
> >
>
--
Erik Jacobson - Linux System Software - Silicon Graphics - Eagan, Minnesota
Now that I understand we don't need to create /proc/ide/sgiioc4, the patch
is very simple.
Below you will find the new patch. I just tested it now and it works
properly - /proc/ide is properly populated with this patch (and not
populated on altix without it).
Add create_proc_ide_interfaces() call to sgiioc4_ide_setup_pci_device()
so /proc/ide gets populated properly.
Signed-off-by: Erik Jacobson <[email protected]>
---
drivers/ide/pci/sgiioc4.c | 4 ++++
1 files changed, 4 insertions(+)
diff -Naru linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c linux-2.6.8/drivers/ide/pci/sgiioc4.c
--- linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c 2004-08-14 01:36:58.000000000 -0400
+++ linux-2.6.8/drivers/ide/pci/sgiioc4.c 2004-09-08 16:12:29.677605262 -0400
@@ -702,6 +702,10 @@
hwif->name, d->name);
probe_hwif_init(hwif);
+
+ /* Create /proc/ide entries */
+ create_proc_ide_interfaces();
+
return 0;
}
On Wednesday 08 September 2004 20:28, Erik Jacobson wrote:
> Hi. Thanks for the response.
>
> I'm fine zapping the /proc/ide/sgiioc4 part of the patch. I didn't want
> it anyway - I just put it there because it was what everybody else did.
>
> The important part of the patch is this added line:
> create_proc_ide_interfaces();
>
> Without this line, /proc/ide contains only the file "drivers".
>
> I believe this is because the sgiioc4_ide_setup_pci_device function
> is our own version of ide_setup_pci_device. Since we don't call
> ide_setup_pci_device, create_proc_ide_interfaces is never called
> otherwise - so /proc/ide isn't populated properly.
Oh yes, you are right. I completely forgot about this fact.
Could you resend minimal patch (without /proc/ide/sgiioc4)?
Thanks.
> On Wed, 8 Sep 2004, Bartlomiej Zolnierkiewicz wrote:
>
> >
> > On Wednesday 08 September 2004 18:17, Erik Jacobson wrote:
> > > The patch below makes it so the proper /proc/ide files are created when
> > > this driver is loaded. Previously, /proc/ide would be empty even though
> >
> > This patch only adds /proc/ide/sgiioc4 entry.
> > /proc/ide should be properly populated without it.
> >
> > > the devices could be used, etc. This is creating problems for us because
> > > some installers and daemons expect the proc files to be in place.
> >
> > Please name these installers and daemons.
> >
> > If your application is checking for /proc/ide/sgiioc4 this is WRONG.
> >
> > Please tell us what do you need this information for and we can
> > see what can be done.
> >
> > > We're open to comments but would appreciate this being accepted ASAP
> > > as this problem is affecting our progress on some projects.
> > >
> > > I verified this exact patch works on 2.6.9-rc1 (and it applies without
> > > fuzz).
> > >
> > > One thing I'll probably hear about is that we create /proc/ide/sgiioc4
> > > but don't produce info and metrics for it. We may elect to produce some
> >
> > Actually no, these /proc/ide/<chipset> entries are just bloat and should
> > be removed - basic info should be always printed on the console (i.e. while
> > changing transfer mode) and more advanced info should be only printed
> > while debugging driver.
> >
> > > metrics later but this driver is used primarily for the system CDROM and
> > > having metrics there isn't a priority at this moment. So I create the proc
> > > file anyway but it just reports the type of driver it is.
> > >
> > > diffstat:
> > >
> > > drivers/ide/pci/sgiioc4.c | 33 +++++++++++++++++++++++++++++++++
> > > 1 files changed, 33 insertions(+)
> > >
> > > Signed-off-by: Erik Jacobson <[email protected]>
> > >
> > >
> > > diff -Naru linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c linux-2.6.8/drivers/ide/pci/sgiioc4.c
> > > --- linux-2.6.8-orig/drivers/ide/pci/sgiioc4.c 2004-08-14 01:36:58.000000000 -0400
> > > +++ linux-2.6.8/drivers/ide/pci/sgiioc4.c 2004-09-08 08:40:39.624159397 -0400
> > > @@ -34,6 +34,7 @@
> > > #include <linux/mm.h>
> > > #include <linux/ioport.h>
> > > #include <linux/blkdev.h>
> > > +#include <linux/proc_fs.h>
> > > #include <asm/io.h>
> > >
> > > #include <linux/ide.h>
> > > @@ -92,6 +93,27 @@
> > > #define IOC4_PRD_BYTES 16
> > > #define IOC4_PRD_ENTRIES (PAGE_SIZE /(4*IOC4_PRD_BYTES))
> > >
> > > +/* Used so we only try to create the /proc/ide/sgiioc4 entry once. */
> > > +static u8 sgiioc4_proc;
> > > +
> > > +/* Produce output for /proc/ide/sgiioc4:
> > > + * At this point, we don't produce any metrics information.
> > > + * Still, it seems we should create this so we behave like other IDE
> > > + * drivers do. Perhaps someone will write in metrics information some day
> > > + * and that can be hooked in here.
> > > + */
> > > +static int sgiioc4_get_info(char *buffer, char **addr, off_t offset, int count)
> > > +{
> > > + char *p = buffer;
> > > + int len;
> > > +
> > > + p += sprintf(p, "\nSGI IOC4 IDE Driver\n");
> > > + p += sprintf(p, "This space may be used in the future for metrics.\n");
> > > +
> > > + len = (p - buffer) - offset;
> > > + *addr = buffer + offset;
> > > + return len > count ? count : len;
> > > +}
> > >
> > > static void
> > > sgiioc4_init_hwif_ports(hw_regs_t * hw, unsigned long data_port,
> > > @@ -702,6 +724,10 @@
> > > hwif->name, d->name);
> > >
> > > probe_hwif_init(hwif);
> > > +
> > > + /* Create /proc/ide entries */
> > > + create_proc_ide_interfaces();
> > > +
> > > return 0;
> > > }
> > >
> > > @@ -732,6 +758,13 @@
> > >
> > > pci_read_config_dword(dev, PCI_CLASS_REVISION, &class_rev);
> > > class_rev &= 0xff;
> > > +
> > > + /* create /proc/ide/sgiioc4 entry */
> > > + if (!sgiioc4_proc) {
> > > + ide_pci_create_host_proc("sgiioc4", sgiioc4_get_info);
> > > + sgiioc4_proc = 1;
> > > + }
> > > +
> > > printk(KERN_INFO "%s: IDE controller at PCI slot %s, revision %d\n",
> > > d->name, dev->slot_name, class_rev);
> > > if (class_rev < IOC4_SUPPORTED_FIRMWARE_REV) {
> > >
> > >
> >
>
> --
> Erik Jacobson - Linux System Software - Silicon Graphics - Eagan, Minnesota
>
>
On Wednesday 08 September 2004 22:50, Erik Jacobson wrote:
> Now that I understand we don't need to create /proc/ide/sgiioc4, the patch
> is very simple.
>
> Below you will find the new patch. I just tested it now and it works
> properly - /proc/ide is properly populated with this patch (and not
> populated on altix without it).
merged by Linus