Re: [PATCH v2 3/4] watchdog: iTCO_wdt: Fix PMC GCR memory mapping failure

[Rajneesh Bhardwaj]

On Fri, Mar 17, 2017 at 06:40:22AM -0700, Guenter Roeck wrote:
> On 03/17/2017 04:43 AM, Rajneesh Bhardwaj wrote:
> >On Thu, Mar 16, 2017 at 05:41:35PM -0700, Kuppuswamy Sathyanarayanan wrote:
> >>Currently, iTCO watchdog driver uses memory map to access
> >>PMC_CFG GCR register. But the entire GCR address space is
> >>already mapped in intel_scu_ipc driver. So remapping the
> >
> >intel_pmc_ipc driver.
> >
> >>GCR register in this driver causes the mem request failure in
> >>iTCO_wdt probe function. This patch fixes this issue by
> >>using PMC GCR read/write API's to access PMC_CFG register.
> >>
> >>Signed-off-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@xxxxxxxxxxxxxxx>
> >>---
> >> drivers/watchdog/iTCO_wdt.c | 31 +++++++------------------------
> >> 1 file changed, 7 insertions(+), 24 deletions(-)
> >>
> >>diff --git a/drivers/watchdog/iTCO_wdt.c b/drivers/watchdog/iTCO_wdt.c
> >>index 3d0abc0..31abfc5 100644
> >>--- a/drivers/watchdog/iTCO_wdt.c
> >>+++ b/drivers/watchdog/iTCO_wdt.c
> >>@@ -68,6 +68,8 @@
> >> #include <linux/io.h>			/* For inb/outb/... */
> >> #include <linux/platform_data/itco_wdt.h>
> >>
> >>+#include <asm/intel_pmc_ipc.h>
> >>+
> >> #include "iTCO_vendor.h"
> >>
> >> /* Address definitions for the TCO */
> >>@@ -94,12 +96,6 @@ struct iTCO_wdt_private {
> >> 	unsigned int iTCO_version;
> >> 	struct resource *tco_res;
> >> 	struct resource *smi_res;
> >>-	/*
> >>-	 * NO_REBOOT flag is Memory-Mapped GCS register bit 5 (TCO version 2),
> >>-	 * or memory-mapped PMC register bit 4 (TCO version 3).
> >>-	 */
> >
> >Better to retain this comment elsewhere.
> >
> >>-	struct resource *gcs_pmc_res;
> >>-	unsigned long __iomem *gcs_pmc;
> >> 	/* the lock for io operations */
> >> 	spinlock_t io_lock;
> >> 	/* the PCI-device */
> >>@@ -176,9 +172,9 @@ static void iTCO_wdt_set_NO_REBOOT_bit(struct iTCO_wdt_private *p)
> >>
> >> 	/* Set the NO_REBOOT bit: this disables reboots */
> >> 	if (p->iTCO_version >= 2) {
> >>-		val32 = readl(p->gcs_pmc);
> >>+		val32 = intel_pmc_gcr_read(PMC_GCR_PMC_CFG_REG);
> >
> >better to have protection and error handling, discussed in v2, 2/4.
> >
> >compiled and tested this on APL and i see iTCO_WDT driver loads fine. Since
> >it impacts core WDT functionality, need to be thoroughly tested on various
> >platforms.
> >
> 
> I don't think I (or the watchdog mailing list) was copied on the original patch.
> Major immediate concern is that this introduces a dependency on external code.
> The pmc_ipc driver's Kconfig entry states "This is not needed for PC-type
> machines". I don't know where the function is introduced, but I hope this change
> does not require the pmc_ipc code to be present on such machines for the watchdog
> to work. It would be bad if it does. If it doesn't, it appears that the function
> should not be declared in asm/intel_pmc_ipc.h.

Yes i added you and the wdt mailing list since you were missed on original
patch. Apparently, iTCO_wdt driver always had a dependency over PMC
resources when iTCO_version >=2. The below threads should help you give some
background on this.
https://lkml.org/lkml/2017/3/16/504
https://lkml.org/lkml/2017/2/13/282

WDT driver fails to laod since the resources its requesting were reserved by
the PMC driver since they belong to PMC GCR region. WDT driver remapping those
resources is still ok but it cant reclaim them hence the problem.

> 
> Guenter
> 

-- 
Best Regards,
Rajneesh