2009-06-05 11:30:08

by Denis V. Lunev

Subject: [PATCH 1/1] [WATCHDOG] fix memory corruption when RCBA is disabled by hardware

According to 9.1.33 on p.343 of ICH8.pdf RCBA can be disabled by
hardware if bit 0 of RCBA register is not set.

Perform correct check for this to prevent memory corruption under
some virtual machines where this feature is disabled.

Signed-off-by: Denis V. Lunev <[email protected]>
CC: Vasily Averin <[email protected]>
CC: Wim Van Sebroeck <[email protected]>
---
drivers/watchdog/iTCO_wdt.c | 7 +++++++
1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/drivers/watchdog/iTCO_wdt.c b/drivers/watchdog/iTCO_wdt.c
index 648250b..c35c1f5 100644
--- a/drivers/watchdog/iTCO_wdt.c
+++ b/drivers/watchdog/iTCO_wdt.c
@@ -666,6 +666,12 @@ static int __devinit iTCO_wdt_init(struct pci_dev *pdev,
GCS = RCBA + ICH6_GCS(0x3410). */
if (iTCO_wdt_private.iTCO_version == 2) {
pci_read_config_dword(pdev, 0xf0, &base_address);
+ if ((base_address & 1) == 0) {
+ printk(KERN_ERR PFX "failed to reset NO_REBOOT flag, "
+ "RCBA is disabled by harddware\n");
+ ret = -ENODEV;
+ goto out_put;
+ }
RCBA = base_address & 0xffffc000;
iTCO_wdt_private.gcs = ioremap((RCBA + 0x3410), 4);
}
@@ -745,6 +751,7 @@ unreg_smi_en:
out:
if (iTCO_wdt_private.iTCO_version == 2)
iounmap(iTCO_wdt_private.gcs);
+out_put:
pci_dev_put(iTCO_wdt_private.pdev);
iTCO_wdt_private.ACPIBASE = 0;
return ret;
--
1.6.0.6

2009-06-10 17:21:59

by Wim Van Sebroeck

Subject: Re: [PATCH 1/1] [WATCHDOG] fix memory corruption when RCBA is disabled by hardware

Hi Denis,

> According to 9.1.33 on p.343 of ICH8.pdf RCBA can be disabled by
> hardware if bit 0 of RCBA register is not set.
>
> Perform correct check for this to prevent memory corruption under
> some virtual machines where this feature is disabled.

Patch went in the watchdog-next tree.

Thanks,
Wim.